Datasets:

Galahad3x

/

QADatasetForPatho

like 0

1500

Value of syndromic surveillance in monitoring a focal waterborne outbreak due to an unusual Cryptosporidium genotype in Northamptonshire

The United Kingdom (UK) has several national syndromic surveillance systems. The Health Protection Agency (HPA)/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data from a national telephone helpline, while the HPA/ QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems. Data from both of these systems were used to monitor a local outbreak of cryptosporidiosis that occurred following Cryptosporidium oocyst contamination of drinking water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, in June 2008. There was a peak in the number of calls to NHS Direct concerning diarrhoea that coincided with the incident. QSurveillance data for the local areas affected by the outbreak showed a significant increase in GP consultations for diarrhoea and gastroenteritis in the week of the incident but there was no increase in consultations for vomiting. A total of 33 clinical cases of cryptosporidiosis were identified in the outbreak investigation, of which 23 were confirmed as infected with the outbreak strain. However, QSurveillance data suggest that there were an estimated 422 excess diarrhoea cases during the outbreak, an increase of about 25% over baseline weekly levels. To our knowledge, this is the first time that data from a syndromic surveillance system, the HPA/QSurveillance national surveillance system, have been able to show the extent of such a small outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide data at local health district (primary care trust) level. The Cryptosporidium contamination incident described demonstrates the potential usefulness of this information, as it is unusual for syndromic surveillance systems to be able to help monitor such a small-scale outbreak. Introduction As syndromic surveillance systems usually capture data already collected for other purposes, and monitor generic symptoms and/or clinically diagnosed disease, they provide information at an earlier stage of illness (compared with laboratory-confirmed diagnoses), so that action can be taken in time to substantially reduce the impact of disease. Some systems, for example, the Royal College of General Practitioners Weekly Returns Service, are now well established, with many years of historical data that allow monitoring of longer-term disease trends [1]. They have the ability to provide early warning of, for example, seasonal rises in influenza and can trigger public health action, such as a recommendation to prescribe antiviral drugs in line with national guidance [2-4]. They can also provide reassurance to incident response teams and the general public that an incident has not caused adverse health effects – for example, following an explosion at the Buncefield oil storage depot in Hemel Hempstead, United Kingdom (UK), in 2005, syndromic surveillance confirmed that there were no unusual rises in community-based morbidity linked to the incident [5]; following the eruption of the Eyjafjallajökull volcano in Iceland in April 2010 similar assurance was given about lack of impact on community morbidity [6]. Health departments are increasingly expected to monitor health effects of natural events such as heat wave or flooding, or implement surveillance – of which syndromic surveillance plays a major role – for mass gatherings such as the Olympics or football World Cup [7-9]. Systems in France, Australia and Taiwan use data from emergency departments [10-12], a Canadian system uses over-the-counter pharmacy sales [13,14], and in the Netherlands data from both syndromic and surrogate data sources, such as employee absence records and prescription medications dispensed by pharmacies, are included in surveillance systems [15,16]. Currently systems based on Internet searches via search engines or on queries submitted to medical websites are being developed [17,18]. In the UK, the HPA/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data collected from the NHS Direct telephone helpline [19], while the HPA/QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems [20]. The HPA Real-time Syndromic Surveillance Team is a small team that coordinates a number of syndromic surveillance systems within the HPA and takes a lead for syndromic surveillance in England [21]. This paper describes the support provided by the team to the local incident management team during a local cryptosporidiosis outbreak and shows the use of syndromic surveillance in monitoring the extent of an outbreak using the HPA/NHS Direct and HPA/QSurveillance national surveillance systems. Cryptosporidiosis Cryptosporidium is a protozoan parasite that can cause an infection in people, cattle and sometimes other animals [22]. Cryptosporidiosis is most common in children aged between one and five years, but it can affect all ages. Those with impaired immune systems are likely to be most seriously affected. Symptoms usually appear between three and 12 days after initial exposure and include watery diarrhoea, stomach pains, dehydration and fever. In its transmissible form, called an oocyst, the parasite is protected by an outer shell, which allows it to survive in the environment for a long time. Transmission occurs most often via the faeco-oral route through person-to-person or animalto-person contact, but people may also be infected by consuming contaminated water or food or by swimming in contaminated water. Although uncommon, the largest outbreaks have occurred following contamination of drinking water [23,24]. Normal chlorine disinfection procedures do not kill the oocysts, so they are removed by filtration and water companies carry out routine monitoring of treated water. Description of the incident On 25 June 2008 the local Health Protection Unit was informed by Anglian Water of an exceedence in the level of Cryptosporidium oocysts found in water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, during 19 to 24 June 2008 [25]. The reservoir supplied a population of more than 250,000 in the Northampton area. A notice advising people in the affected areas to boil all drinking water was issued on 25 June 2008 and public health messages were circulated to local health services and to the general public via the media. Those members of the public who were concerned about health risks associated with the incident were asked to ring NHS Direct for clinical advice [26]. The HPA wrote to local GPs and hospitals asking them to monitor potential patients for signs and symptoms of Cryptosporidium infection and to submit faecal specimens to the local hospital diagnostic laboratory if patients presented with diarrhoea. Samples from 34 patients where Cryptosporidium infection was identified were sent to the UK Cryptosporidium reference unit for typing. On 30 June 2008, the Cryptosporidium oocysts found in the reservoir water were confirmed as being of the rabbit genotype Cryptosporidium cuniculus [27]. Subsequently, a dead rabbit was found in a treated water tank at the water treatment works. The genotype of Cryptosporidium oocysts in the rabbit’s large bowel was indistinguishable from that of the oocysts found in the water [27]. After remediation of the water supply and distribution, the ‘boil water notice’ was lifted on 4 July and the following day the first case of cryptosporidiosis linked to the incident was identified by the reference laboratory (this case was infected with C. cuniculus). During the course of the outbreak (24 June – 18 July 2008, the dates of symptom onset in the first and last case, respectively), 23 cases of cryptosporidiosis were confirmed as being infected with C. cuniculus; one of the 23 was a secondary case. The HPA Real-time Syndromic Surveillance Team provided data in order to aid the response to this incident and the first syndromic surveillance report was circulated to the incident management team and other relevant people in the HPA on 27 June 2008. Data from the HPA/NHS Direct and HPA/QSurveillance systems were provided in a series of regular reports, initially daily and eventually weekly, until the final report on 21 August 2008. Each report included a summary interpretation and more detailed data on diarrhoea, gastroenteritis and vomiting indicators. Methods Surveillance systems HPA/NHS Direct surveillance system NHS Direct is a 24-hour nurse-led telephone helpline that provides health information and advice to the general public. Nurses use a computerised clinical decision support system – the NHS Clinical Assessment System (NHS CAS) – to handle calls. This assessment system uses approximately 200 computerised symptom-based clinical algorithms. Nurses assign the call to the most appropriate algorithm and the patient’s symptoms determine the questions asked and the action to be taken following the call (call outcome), which could be guidance on self-care or they could be referred to their GP or advised to attend a hospital emergency department. No attempt is made to provide a formal diagnosis. Daily NHS Direct data are received by the Real-time Syndromic Surveillance Team, where the number and type of calls received during the previous day are analysed and interpreted. Call proportions are calculated by age group and algorithm against the total number of calls received. HPA/QSurveillance system The HPA/QSurveillance national surveillance system was set up by the University of Nottingham, United Kingdom, and Egton Medical Information Systems (EMIS), a supplier of general practice computer systems, in collaboration with the HPA. It comprises a network of more than 3,500 general practices throughout the UK, covering more than 22 million patients (about 38% of the population [28]). Aggregated data on GP consultations for a range of indicators are automatically uploaded daily from GP practice systems to a central database. Data are routinely reported on a weekly basis; however, daily reporting is possible for specific incidents. Reports are provided at national or regional level (strategic health authority, SHA) and by local health district (primary care trust, PCT). Analysis of surveillance data NHS Direct call proportions for gastrointestinal syndromes (diarrhoea and vomiting) for the East Midlands region in England, where Northampton is situated, were examined during the outbreak (24 June – 18 July 2008) and compared with those for England and Wales. A series of control charts for diarrhoea calls are routinely used to monitor significant rises in the numbers of calls received. Control charts are calculated by assuming that calls follow a Poisson distribution with the total number of calls as an offset: a model is fitted to each region and symptom separately [29]. The model takes into account call variation caused by weekends, public holidays and the time of year – variables that can affect the number of calls received by NHS Direct. A value above the upper limit of the 99.5% confidence interval would be considered to be unusual. The seven-day moving average for diarrhoea calls was also monitored. The number and percentage of calls for diarrhoea in the East Midlands region were presented by call outcome and the number of calls in the Northampton (NN) postcode districts and in particular the number of calls in the NN11 and NN12 postcode districts, which were most affected by the incident. QSurveillance national consultation rates per 100,000 population for diarrhoea (in the age groups under five years, five years and over, and all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with rates for the same period in 2007 (data not presented). Consultation rates by region for 2008 for diarrhoea (all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with those for the East Midlands region. The gastroenteritis indicator includes all cases of diarrhoea and/or vomiting. Consultation rates and standardised incidence ratios (SIRs) – calculated using the UK as the standard population – for diarrhoea, gastroenteritis and vomiting were compared for the UK, Yorkshire and Humberside, East Midlands, Leicestershire, Northamptonshire and Rutland SHA, and Daventry and South Northants PCT, Northamptonshire Heartlands PCT and Northampton PCT. Yorkshire and Humberside was not an affected region but was included as a control. The area supplied by the Pitsford Reservoir included the three PCTs, which were all within the Leicestershire, Northamptonshire and Rutland SHA. The consultation rates and SIRs were compared for the period from week 16 to week 35 of 2008 in order to compare the rates before and after the Cryptosporidium exceedance, which took place in week 26. Estimates of excess numbers of cases of diarrhoea occurring during and following the Cryptosporidium outbreak were made by calculating the mean consultation rate over a period of five weeks before and after the incident (weeks 20–24 and weeks 31–35, respectively). For each of the three PCTs, the calculated mean rate was applied to the PCT population to estimate the number of cases that would be expected each week. The actual consultation rates for diarrhoea for weeks 25 to 30 were used to estimate the number of cases for the PCT population each week. The expected number of cases was subtracted from the estimated number of cases in the PCT population to give the estimated number of excess cases. Results HPA/NHS Direct surveillance system A peak in the number of calls for diarrhoea in the East Midlands was recorded in 25–26 June 2008, the period that coincided with the contamination incident and the associated media coverage (Figure 1). The neighbouring areas of the West Midlands, Yorkshire and the Humber, and East of England showed no increase in the number of calls for diarrhoea. The peak produced a control chart exceedance for calls for diarrhoea on 25 June 2008 (Figure 2), when the proportion of calls exceeded the upper limit of the 99.5% confidence interval. There were further confidence interval exceedances on 26 and 28 June (which were not control chart exceedances). There was no peak in calls for vomiting or control chart exceedance for these calls in the East Midlands. HPA/QSurveillance national surveillance system The East Midlands region had significantly high consultation rates for diarrhoea and gastroenteritis in week 25 (16–22 June), week 26 (23–29 June 2008, when the contamination incident was reported) and in the following four weeks. Within this region. Leicestershire, Northamptonshire and Rutland SHA had slightly raised consultation rates and significant SIRs across weeks 25 to 30 that were not seen in the neighbouring Trent SHA. At PCT level, all three of the PCTs in the area affected by the incident showed increased consultation rates for diarrhoea (Table 1) and gastroenteritis (Table 2) with SIRs significantly above the UK rate in week 26. Daventry and South Northants PCT also had a raised SIR for both indicators in week 25, and although Northamptonshire Heartlands and Northampton PCTs did not have SIRs significantly above that of the UK in week 25, the rise in consultation rates for diarrhoea and gastroenteritis began during week 25. In Northampton PCT, consultations for both diarrhoea and gastroenteritis peaked in the week following the contamination incident, week 27, returning to normal levels by week 30 (Figure 3A and 3B). A similar effect can be seen in Northampton Heartlands PCT. Daventry and South Northants PCT also showed an increase, but appeared to have consistently higher rates for both indicators. This was the area with the smallest population so the rates were more variable than in the other PCTs and we therefore interpreted these results with caution. The consultation rates for vomiting during weeks 25 to 30 in the East Midlands were not unusual at SHA or PCT level (data not presented). Discussion We have demonstrated the sensitivity of syndromic surveillance in detecting this small Cryptosporidium outbreak and the value of the surveillance in being able to describe the extent of its spread. Both the HPA/NHS Direct and HPA/QSurveillance systems showed demonstrable increases in calls and consultations for diarrhoea that were linked to the outbreak. QSurveillance consultations appeared to increase across the PCTs immediately affected but not in the surrounding area. Both the HPA/NHS Direct and HPA/QSurveillance systems showed a clear signal at the time of the incident and we were able to describe the extent of the impact on pre-primary care and primary care services. The HPA/QSurveillance system showed a rise in consultation rates for gastrointestinal symptoms that began the week before the outbreak, consistent with the period when Cryptosporidium was present in the water leaving the Pitsford Reservoir (19–24 June 2008) and with the onset of symptoms in the first outbreak case on 24 June. Although only 33 cases were identified by the outbreak investigation team, of which 23 were confirmed as having the outbreak Cryptosporidium strain, our syndromic surveillance data detected this limited outbreak. Data also suggested a more widespread increase in general gastrointestinal symptoms around the time of the outbreak, with an estimated 422 excess diarrhoea cases; these excess cases represented an increase of about 25% above normally expected levels. It is highly probable that a proportion of these excess cases may have resulted from the increased publicity surrounding the incident – for example, it is likely that media reports contributed to the large peak in calls detected by the HPA/NHS Direct surveillance system on the day the boil water notice was issued, and could also have impacted on the GP consultation rate. It has been previously shown that reporting of mumps cases is sensitive to media coverage, with a rise in clinically reported cases following newspaper reports [30]. A similar mechanism could account for some of the excess GP consultations as cases experiencing gastrointestinal symptoms may have been more likely to consult their GP, whereas in normal circumstances they would have cared for themselves at home. It is also possible that the surveillance shows outbreak-associated cases that did not come to the attention of the outbreak team, perhaps because symptoms were not sufficiently severe to warrant further investigation, or stool samples were not provided for testing. It is interesting to note that there was no demonstrable impact on the number of calls for vomiting (which is not a prominent clinical feature of cryptosporidiosis). Other common community-based pathogens such as norovirus and rotavirus were at low levels, as is normal for that time of year [31]. In this instance, public health authorities had already been alerted to a potential problem by the water company, although the extent of the outbreak was detected by syndromic surveillance. In 2003 the syndromic surveillance systems in the city of New York, United States, were able to detect an increase in diarrhoeal illness following a power outage when there was no other indication of citywide illness [32]. The New York system covers a population of nine million, but does not regularly detect localised outbreaks [33]. It has been shown previously that the HPA/NHS Direct surveillance system would be unlikely to detect a Cryptosporidium outbreak unless call volumes are high (72% chance of detection if nine-tenths of cases called NHS Direct) [29], although the value of syndromic surveillance for such outbreaks has been recognised [34]. The system detected the East Midlands Cryptosporidium outbreak that affected a smaller population than that covered by the New York system. The three PCTs affected have a combined population of around 600,000, of which just over half use GP practices reporting to QSurveillance, yet this syndromic surveillance system was able to describe an increase in consultation rates for diarrhoea and gastroenteritis around the time of the outbreak. Limitations of the data There was extensive media reporting of the incident that may have affected both the HPA/NHS Direct and HPA/QSurveillance systems and contributed to the increase in reported gastrointestinal symptoms around the time of the contamination incident. However, the rise in consultation rates for diarrhoea began before the outbreak had been detected and therefore cannot be attributed to media coverage. The HPA/NHS Direct and HPA/QSurveillance systems monitor general symptoms and so could only monitor the relevant symptoms of diarrhoea and vomiting. They are not able to detect Cryptosporidium cases, as this would require laboratory confirmation of diagnosis, so some of the estimated excess cases could be unconnected with this incident. This outbreak was discovered by other means but both the HPA/NHS Direct and HPA/ QSurveillance systems were able to describe the extent of the disease in the general population and provide reassurance that there was no widespread impact. Compared with other populations, older people and ethnic minorities are less likely to call NHS Direct [29], and although this should not prevent detection of gastrointestinal symptoms as a result of drinking water contamination as this would affect the whole population, this may reduce the signal from the system [35]. With such large surveillance systems, there will be ‘background noise’ in the data, so procedures must be in place to correctly interpret the data and set appropriate thresholds for action. Conclusion To our knowledge, this is the first time that PCT-level data from a syndromic surveillance system, the HPA/ QSurveillance national surveillance system, have been able to show the extent of such a limited outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide PCT-level data and this Cryptosporidium contamination incident demonstrates the potential usefulness of this system.

In which street did this happen?

1501

Value of syndromic surveillance in monitoring a focal waterborne outbreak due to an unusual Cryptosporidium genotype in Northamptonshire

The United Kingdom (UK) has several national syndromic surveillance systems. The Health Protection Agency (HPA)/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data from a national telephone helpline, while the HPA/ QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems. Data from both of these systems were used to monitor a local outbreak of cryptosporidiosis that occurred following Cryptosporidium oocyst contamination of drinking water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, in June 2008. There was a peak in the number of calls to NHS Direct concerning diarrhoea that coincided with the incident. QSurveillance data for the local areas affected by the outbreak showed a significant increase in GP consultations for diarrhoea and gastroenteritis in the week of the incident but there was no increase in consultations for vomiting. A total of 33 clinical cases of cryptosporidiosis were identified in the outbreak investigation, of which 23 were confirmed as infected with the outbreak strain. However, QSurveillance data suggest that there were an estimated 422 excess diarrhoea cases during the outbreak, an increase of about 25% over baseline weekly levels. To our knowledge, this is the first time that data from a syndromic surveillance system, the HPA/QSurveillance national surveillance system, have been able to show the extent of such a small outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide data at local health district (primary care trust) level. The Cryptosporidium contamination incident described demonstrates the potential usefulness of this information, as it is unusual for syndromic surveillance systems to be able to help monitor such a small-scale outbreak. Introduction As syndromic surveillance systems usually capture data already collected for other purposes, and monitor generic symptoms and/or clinically diagnosed disease, they provide information at an earlier stage of illness (compared with laboratory-confirmed diagnoses), so that action can be taken in time to substantially reduce the impact of disease. Some systems, for example, the Royal College of General Practitioners Weekly Returns Service, are now well established, with many years of historical data that allow monitoring of longer-term disease trends [1]. They have the ability to provide early warning of, for example, seasonal rises in influenza and can trigger public health action, such as a recommendation to prescribe antiviral drugs in line with national guidance [2-4]. They can also provide reassurance to incident response teams and the general public that an incident has not caused adverse health effects – for example, following an explosion at the Buncefield oil storage depot in Hemel Hempstead, United Kingdom (UK), in 2005, syndromic surveillance confirmed that there were no unusual rises in community-based morbidity linked to the incident [5]; following the eruption of the Eyjafjallajökull volcano in Iceland in April 2010 similar assurance was given about lack of impact on community morbidity [6]. Health departments are increasingly expected to monitor health effects of natural events such as heat wave or flooding, or implement surveillance – of which syndromic surveillance plays a major role – for mass gatherings such as the Olympics or football World Cup [7-9]. Systems in France, Australia and Taiwan use data from emergency departments [10-12], a Canadian system uses over-the-counter pharmacy sales [13,14], and in the Netherlands data from both syndromic and surrogate data sources, such as employee absence records and prescription medications dispensed by pharmacies, are included in surveillance systems [15,16]. Currently systems based on Internet searches via search engines or on queries submitted to medical websites are being developed [17,18]. In the UK, the HPA/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data collected from the NHS Direct telephone helpline [19], while the HPA/QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems [20]. The HPA Real-time Syndromic Surveillance Team is a small team that coordinates a number of syndromic surveillance systems within the HPA and takes a lead for syndromic surveillance in England [21]. This paper describes the support provided by the team to the local incident management team during a local cryptosporidiosis outbreak and shows the use of syndromic surveillance in monitoring the extent of an outbreak using the HPA/NHS Direct and HPA/QSurveillance national surveillance systems. Cryptosporidiosis Cryptosporidium is a protozoan parasite that can cause an infection in people, cattle and sometimes other animals [22]. Cryptosporidiosis is most common in children aged between one and five years, but it can affect all ages. Those with impaired immune systems are likely to be most seriously affected. Symptoms usually appear between three and 12 days after initial exposure and include watery diarrhoea, stomach pains, dehydration and fever. In its transmissible form, called an oocyst, the parasite is protected by an outer shell, which allows it to survive in the environment for a long time. Transmission occurs most often via the faeco-oral route through person-to-person or animalto-person contact, but people may also be infected by consuming contaminated water or food or by swimming in contaminated water. Although uncommon, the largest outbreaks have occurred following contamination of drinking water [23,24]. Normal chlorine disinfection procedures do not kill the oocysts, so they are removed by filtration and water companies carry out routine monitoring of treated water. Description of the incident On 25 June 2008 the local Health Protection Unit was informed by Anglian Water of an exceedence in the level of Cryptosporidium oocysts found in water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, during 19 to 24 June 2008 [25]. The reservoir supplied a population of more than 250,000 in the Northampton area. A notice advising people in the affected areas to boil all drinking water was issued on 25 June 2008 and public health messages were circulated to local health services and to the general public via the media. Those members of the public who were concerned about health risks associated with the incident were asked to ring NHS Direct for clinical advice [26]. The HPA wrote to local GPs and hospitals asking them to monitor potential patients for signs and symptoms of Cryptosporidium infection and to submit faecal specimens to the local hospital diagnostic laboratory if patients presented with diarrhoea. Samples from 34 patients where Cryptosporidium infection was identified were sent to the UK Cryptosporidium reference unit for typing. On 30 June 2008, the Cryptosporidium oocysts found in the reservoir water were confirmed as being of the rabbit genotype Cryptosporidium cuniculus [27]. Subsequently, a dead rabbit was found in a treated water tank at the water treatment works. The genotype of Cryptosporidium oocysts in the rabbit’s large bowel was indistinguishable from that of the oocysts found in the water [27]. After remediation of the water supply and distribution, the ‘boil water notice’ was lifted on 4 July and the following day the first case of cryptosporidiosis linked to the incident was identified by the reference laboratory (this case was infected with C. cuniculus). During the course of the outbreak (24 June – 18 July 2008, the dates of symptom onset in the first and last case, respectively), 23 cases of cryptosporidiosis were confirmed as being infected with C. cuniculus; one of the 23 was a secondary case. The HPA Real-time Syndromic Surveillance Team provided data in order to aid the response to this incident and the first syndromic surveillance report was circulated to the incident management team and other relevant people in the HPA on 27 June 2008. Data from the HPA/NHS Direct and HPA/QSurveillance systems were provided in a series of regular reports, initially daily and eventually weekly, until the final report on 21 August 2008. Each report included a summary interpretation and more detailed data on diarrhoea, gastroenteritis and vomiting indicators. Methods Surveillance systems HPA/NHS Direct surveillance system NHS Direct is a 24-hour nurse-led telephone helpline that provides health information and advice to the general public. Nurses use a computerised clinical decision support system – the NHS Clinical Assessment System (NHS CAS) – to handle calls. This assessment system uses approximately 200 computerised symptom-based clinical algorithms. Nurses assign the call to the most appropriate algorithm and the patient’s symptoms determine the questions asked and the action to be taken following the call (call outcome), which could be guidance on self-care or they could be referred to their GP or advised to attend a hospital emergency department. No attempt is made to provide a formal diagnosis. Daily NHS Direct data are received by the Real-time Syndromic Surveillance Team, where the number and type of calls received during the previous day are analysed and interpreted. Call proportions are calculated by age group and algorithm against the total number of calls received. HPA/QSurveillance system The HPA/QSurveillance national surveillance system was set up by the University of Nottingham, United Kingdom, and Egton Medical Information Systems (EMIS), a supplier of general practice computer systems, in collaboration with the HPA. It comprises a network of more than 3,500 general practices throughout the UK, covering more than 22 million patients (about 38% of the population [28]). Aggregated data on GP consultations for a range of indicators are automatically uploaded daily from GP practice systems to a central database. Data are routinely reported on a weekly basis; however, daily reporting is possible for specific incidents. Reports are provided at national or regional level (strategic health authority, SHA) and by local health district (primary care trust, PCT). Analysis of surveillance data NHS Direct call proportions for gastrointestinal syndromes (diarrhoea and vomiting) for the East Midlands region in England, where Northampton is situated, were examined during the outbreak (24 June – 18 July 2008) and compared with those for England and Wales. A series of control charts for diarrhoea calls are routinely used to monitor significant rises in the numbers of calls received. Control charts are calculated by assuming that calls follow a Poisson distribution with the total number of calls as an offset: a model is fitted to each region and symptom separately [29]. The model takes into account call variation caused by weekends, public holidays and the time of year – variables that can affect the number of calls received by NHS Direct. A value above the upper limit of the 99.5% confidence interval would be considered to be unusual. The seven-day moving average for diarrhoea calls was also monitored. The number and percentage of calls for diarrhoea in the East Midlands region were presented by call outcome and the number of calls in the Northampton (NN) postcode districts and in particular the number of calls in the NN11 and NN12 postcode districts, which were most affected by the incident. QSurveillance national consultation rates per 100,000 population for diarrhoea (in the age groups under five years, five years and over, and all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with rates for the same period in 2007 (data not presented). Consultation rates by region for 2008 for diarrhoea (all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with those for the East Midlands region. The gastroenteritis indicator includes all cases of diarrhoea and/or vomiting. Consultation rates and standardised incidence ratios (SIRs) – calculated using the UK as the standard population – for diarrhoea, gastroenteritis and vomiting were compared for the UK, Yorkshire and Humberside, East Midlands, Leicestershire, Northamptonshire and Rutland SHA, and Daventry and South Northants PCT, Northamptonshire Heartlands PCT and Northampton PCT. Yorkshire and Humberside was not an affected region but was included as a control. The area supplied by the Pitsford Reservoir included the three PCTs, which were all within the Leicestershire, Northamptonshire and Rutland SHA. The consultation rates and SIRs were compared for the period from week 16 to week 35 of 2008 in order to compare the rates before and after the Cryptosporidium exceedance, which took place in week 26. Estimates of excess numbers of cases of diarrhoea occurring during and following the Cryptosporidium outbreak were made by calculating the mean consultation rate over a period of five weeks before and after the incident (weeks 20–24 and weeks 31–35, respectively). For each of the three PCTs, the calculated mean rate was applied to the PCT population to estimate the number of cases that would be expected each week. The actual consultation rates for diarrhoea for weeks 25 to 30 were used to estimate the number of cases for the PCT population each week. The expected number of cases was subtracted from the estimated number of cases in the PCT population to give the estimated number of excess cases. Results HPA/NHS Direct surveillance system A peak in the number of calls for diarrhoea in the East Midlands was recorded in 25–26 June 2008, the period that coincided with the contamination incident and the associated media coverage (Figure 1). The neighbouring areas of the West Midlands, Yorkshire and the Humber, and East of England showed no increase in the number of calls for diarrhoea. The peak produced a control chart exceedance for calls for diarrhoea on 25 June 2008 (Figure 2), when the proportion of calls exceeded the upper limit of the 99.5% confidence interval. There were further confidence interval exceedances on 26 and 28 June (which were not control chart exceedances). There was no peak in calls for vomiting or control chart exceedance for these calls in the East Midlands. HPA/QSurveillance national surveillance system The East Midlands region had significantly high consultation rates for diarrhoea and gastroenteritis in week 25 (16–22 June), week 26 (23–29 June 2008, when the contamination incident was reported) and in the following four weeks. Within this region. Leicestershire, Northamptonshire and Rutland SHA had slightly raised consultation rates and significant SIRs across weeks 25 to 30 that were not seen in the neighbouring Trent SHA. At PCT level, all three of the PCTs in the area affected by the incident showed increased consultation rates for diarrhoea (Table 1) and gastroenteritis (Table 2) with SIRs significantly above the UK rate in week 26. Daventry and South Northants PCT also had a raised SIR for both indicators in week 25, and although Northamptonshire Heartlands and Northampton PCTs did not have SIRs significantly above that of the UK in week 25, the rise in consultation rates for diarrhoea and gastroenteritis began during week 25. In Northampton PCT, consultations for both diarrhoea and gastroenteritis peaked in the week following the contamination incident, week 27, returning to normal levels by week 30 (Figure 3A and 3B). A similar effect can be seen in Northampton Heartlands PCT. Daventry and South Northants PCT also showed an increase, but appeared to have consistently higher rates for both indicators. This was the area with the smallest population so the rates were more variable than in the other PCTs and we therefore interpreted these results with caution. The consultation rates for vomiting during weeks 25 to 30 in the East Midlands were not unusual at SHA or PCT level (data not presented). Discussion We have demonstrated the sensitivity of syndromic surveillance in detecting this small Cryptosporidium outbreak and the value of the surveillance in being able to describe the extent of its spread. Both the HPA/NHS Direct and HPA/QSurveillance systems showed demonstrable increases in calls and consultations for diarrhoea that were linked to the outbreak. QSurveillance consultations appeared to increase across the PCTs immediately affected but not in the surrounding area. Both the HPA/NHS Direct and HPA/QSurveillance systems showed a clear signal at the time of the incident and we were able to describe the extent of the impact on pre-primary care and primary care services. The HPA/QSurveillance system showed a rise in consultation rates for gastrointestinal symptoms that began the week before the outbreak, consistent with the period when Cryptosporidium was present in the water leaving the Pitsford Reservoir (19–24 June 2008) and with the onset of symptoms in the first outbreak case on 24 June. Although only 33 cases were identified by the outbreak investigation team, of which 23 were confirmed as having the outbreak Cryptosporidium strain, our syndromic surveillance data detected this limited outbreak. Data also suggested a more widespread increase in general gastrointestinal symptoms around the time of the outbreak, with an estimated 422 excess diarrhoea cases; these excess cases represented an increase of about 25% above normally expected levels. It is highly probable that a proportion of these excess cases may have resulted from the increased publicity surrounding the incident – for example, it is likely that media reports contributed to the large peak in calls detected by the HPA/NHS Direct surveillance system on the day the boil water notice was issued, and could also have impacted on the GP consultation rate. It has been previously shown that reporting of mumps cases is sensitive to media coverage, with a rise in clinically reported cases following newspaper reports [30]. A similar mechanism could account for some of the excess GP consultations as cases experiencing gastrointestinal symptoms may have been more likely to consult their GP, whereas in normal circumstances they would have cared for themselves at home. It is also possible that the surveillance shows outbreak-associated cases that did not come to the attention of the outbreak team, perhaps because symptoms were not sufficiently severe to warrant further investigation, or stool samples were not provided for testing. It is interesting to note that there was no demonstrable impact on the number of calls for vomiting (which is not a prominent clinical feature of cryptosporidiosis). Other common community-based pathogens such as norovirus and rotavirus were at low levels, as is normal for that time of year [31]. In this instance, public health authorities had already been alerted to a potential problem by the water company, although the extent of the outbreak was detected by syndromic surveillance. In 2003 the syndromic surveillance systems in the city of New York, United States, were able to detect an increase in diarrhoeal illness following a power outage when there was no other indication of citywide illness [32]. The New York system covers a population of nine million, but does not regularly detect localised outbreaks [33]. It has been shown previously that the HPA/NHS Direct surveillance system would be unlikely to detect a Cryptosporidium outbreak unless call volumes are high (72% chance of detection if nine-tenths of cases called NHS Direct) [29], although the value of syndromic surveillance for such outbreaks has been recognised [34]. The system detected the East Midlands Cryptosporidium outbreak that affected a smaller population than that covered by the New York system. The three PCTs affected have a combined population of around 600,000, of which just over half use GP practices reporting to QSurveillance, yet this syndromic surveillance system was able to describe an increase in consultation rates for diarrhoea and gastroenteritis around the time of the outbreak. Limitations of the data There was extensive media reporting of the incident that may have affected both the HPA/NHS Direct and HPA/QSurveillance systems and contributed to the increase in reported gastrointestinal symptoms around the time of the contamination incident. However, the rise in consultation rates for diarrhoea began before the outbreak had been detected and therefore cannot be attributed to media coverage. The HPA/NHS Direct and HPA/QSurveillance systems monitor general symptoms and so could only monitor the relevant symptoms of diarrhoea and vomiting. They are not able to detect Cryptosporidium cases, as this would require laboratory confirmation of diagnosis, so some of the estimated excess cases could be unconnected with this incident. This outbreak was discovered by other means but both the HPA/NHS Direct and HPA/ QSurveillance systems were able to describe the extent of the disease in the general population and provide reassurance that there was no widespread impact. Compared with other populations, older people and ethnic minorities are less likely to call NHS Direct [29], and although this should not prevent detection of gastrointestinal symptoms as a result of drinking water contamination as this would affect the whole population, this may reduce the signal from the system [35]. With such large surveillance systems, there will be ‘background noise’ in the data, so procedures must be in place to correctly interpret the data and set appropriate thresholds for action. Conclusion To our knowledge, this is the first time that PCT-level data from a syndromic surveillance system, the HPA/ QSurveillance national surveillance system, have been able to show the extent of such a limited outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide PCT-level data and this Cryptosporidium contamination incident demonstrates the potential usefulness of this system.

In which city did this happen?

1502

Value of syndromic surveillance in monitoring a focal waterborne outbreak due to an unusual Cryptosporidium genotype in Northamptonshire

The United Kingdom (UK) has several national syndromic surveillance systems. The Health Protection Agency (HPA)/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data from a national telephone helpline, while the HPA/ QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems. Data from both of these systems were used to monitor a local outbreak of cryptosporidiosis that occurred following Cryptosporidium oocyst contamination of drinking water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, in June 2008. There was a peak in the number of calls to NHS Direct concerning diarrhoea that coincided with the incident. QSurveillance data for the local areas affected by the outbreak showed a significant increase in GP consultations for diarrhoea and gastroenteritis in the week of the incident but there was no increase in consultations for vomiting. A total of 33 clinical cases of cryptosporidiosis were identified in the outbreak investigation, of which 23 were confirmed as infected with the outbreak strain. However, QSurveillance data suggest that there were an estimated 422 excess diarrhoea cases during the outbreak, an increase of about 25% over baseline weekly levels. To our knowledge, this is the first time that data from a syndromic surveillance system, the HPA/QSurveillance national surveillance system, have been able to show the extent of such a small outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide data at local health district (primary care trust) level. The Cryptosporidium contamination incident described demonstrates the potential usefulness of this information, as it is unusual for syndromic surveillance systems to be able to help monitor such a small-scale outbreak. Introduction As syndromic surveillance systems usually capture data already collected for other purposes, and monitor generic symptoms and/or clinically diagnosed disease, they provide information at an earlier stage of illness (compared with laboratory-confirmed diagnoses), so that action can be taken in time to substantially reduce the impact of disease. Some systems, for example, the Royal College of General Practitioners Weekly Returns Service, are now well established, with many years of historical data that allow monitoring of longer-term disease trends [1]. They have the ability to provide early warning of, for example, seasonal rises in influenza and can trigger public health action, such as a recommendation to prescribe antiviral drugs in line with national guidance [2-4]. They can also provide reassurance to incident response teams and the general public that an incident has not caused adverse health effects – for example, following an explosion at the Buncefield oil storage depot in Hemel Hempstead, United Kingdom (UK), in 2005, syndromic surveillance confirmed that there were no unusual rises in community-based morbidity linked to the incident [5]; following the eruption of the Eyjafjallajökull volcano in Iceland in April 2010 similar assurance was given about lack of impact on community morbidity [6]. Health departments are increasingly expected to monitor health effects of natural events such as heat wave or flooding, or implement surveillance – of which syndromic surveillance plays a major role – for mass gatherings such as the Olympics or football World Cup [7-9]. Systems in France, Australia and Taiwan use data from emergency departments [10-12], a Canadian system uses over-the-counter pharmacy sales [13,14], and in the Netherlands data from both syndromic and surrogate data sources, such as employee absence records and prescription medications dispensed by pharmacies, are included in surveillance systems [15,16]. Currently systems based on Internet searches via search engines or on queries submitted to medical websites are being developed [17,18]. In the UK, the HPA/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data collected from the NHS Direct telephone helpline [19], while the HPA/QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems [20]. The HPA Real-time Syndromic Surveillance Team is a small team that coordinates a number of syndromic surveillance systems within the HPA and takes a lead for syndromic surveillance in England [21]. This paper describes the support provided by the team to the local incident management team during a local cryptosporidiosis outbreak and shows the use of syndromic surveillance in monitoring the extent of an outbreak using the HPA/NHS Direct and HPA/QSurveillance national surveillance systems. Cryptosporidiosis Cryptosporidium is a protozoan parasite that can cause an infection in people, cattle and sometimes other animals [22]. Cryptosporidiosis is most common in children aged between one and five years, but it can affect all ages. Those with impaired immune systems are likely to be most seriously affected. Symptoms usually appear between three and 12 days after initial exposure and include watery diarrhoea, stomach pains, dehydration and fever. In its transmissible form, called an oocyst, the parasite is protected by an outer shell, which allows it to survive in the environment for a long time. Transmission occurs most often via the faeco-oral route through person-to-person or animalto-person contact, but people may also be infected by consuming contaminated water or food or by swimming in contaminated water. Although uncommon, the largest outbreaks have occurred following contamination of drinking water [23,24]. Normal chlorine disinfection procedures do not kill the oocysts, so they are removed by filtration and water companies carry out routine monitoring of treated water. Description of the incident On 25 June 2008 the local Health Protection Unit was informed by Anglian Water of an exceedence in the level of Cryptosporidium oocysts found in water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, during 19 to 24 June 2008 [25]. The reservoir supplied a population of more than 250,000 in the Northampton area. A notice advising people in the affected areas to boil all drinking water was issued on 25 June 2008 and public health messages were circulated to local health services and to the general public via the media. Those members of the public who were concerned about health risks associated with the incident were asked to ring NHS Direct for clinical advice [26]. The HPA wrote to local GPs and hospitals asking them to monitor potential patients for signs and symptoms of Cryptosporidium infection and to submit faecal specimens to the local hospital diagnostic laboratory if patients presented with diarrhoea. Samples from 34 patients where Cryptosporidium infection was identified were sent to the UK Cryptosporidium reference unit for typing. On 30 June 2008, the Cryptosporidium oocysts found in the reservoir water were confirmed as being of the rabbit genotype Cryptosporidium cuniculus [27]. Subsequently, a dead rabbit was found in a treated water tank at the water treatment works. The genotype of Cryptosporidium oocysts in the rabbit’s large bowel was indistinguishable from that of the oocysts found in the water [27]. After remediation of the water supply and distribution, the ‘boil water notice’ was lifted on 4 July and the following day the first case of cryptosporidiosis linked to the incident was identified by the reference laboratory (this case was infected with C. cuniculus). During the course of the outbreak (24 June – 18 July 2008, the dates of symptom onset in the first and last case, respectively), 23 cases of cryptosporidiosis were confirmed as being infected with C. cuniculus; one of the 23 was a secondary case. The HPA Real-time Syndromic Surveillance Team provided data in order to aid the response to this incident and the first syndromic surveillance report was circulated to the incident management team and other relevant people in the HPA on 27 June 2008. Data from the HPA/NHS Direct and HPA/QSurveillance systems were provided in a series of regular reports, initially daily and eventually weekly, until the final report on 21 August 2008. Each report included a summary interpretation and more detailed data on diarrhoea, gastroenteritis and vomiting indicators. Methods Surveillance systems HPA/NHS Direct surveillance system NHS Direct is a 24-hour nurse-led telephone helpline that provides health information and advice to the general public. Nurses use a computerised clinical decision support system – the NHS Clinical Assessment System (NHS CAS) – to handle calls. This assessment system uses approximately 200 computerised symptom-based clinical algorithms. Nurses assign the call to the most appropriate algorithm and the patient’s symptoms determine the questions asked and the action to be taken following the call (call outcome), which could be guidance on self-care or they could be referred to their GP or advised to attend a hospital emergency department. No attempt is made to provide a formal diagnosis. Daily NHS Direct data are received by the Real-time Syndromic Surveillance Team, where the number and type of calls received during the previous day are analysed and interpreted. Call proportions are calculated by age group and algorithm against the total number of calls received. HPA/QSurveillance system The HPA/QSurveillance national surveillance system was set up by the University of Nottingham, United Kingdom, and Egton Medical Information Systems (EMIS), a supplier of general practice computer systems, in collaboration with the HPA. It comprises a network of more than 3,500 general practices throughout the UK, covering more than 22 million patients (about 38% of the population [28]). Aggregated data on GP consultations for a range of indicators are automatically uploaded daily from GP practice systems to a central database. Data are routinely reported on a weekly basis; however, daily reporting is possible for specific incidents. Reports are provided at national or regional level (strategic health authority, SHA) and by local health district (primary care trust, PCT). Analysis of surveillance data NHS Direct call proportions for gastrointestinal syndromes (diarrhoea and vomiting) for the East Midlands region in England, where Northampton is situated, were examined during the outbreak (24 June – 18 July 2008) and compared with those for England and Wales. A series of control charts for diarrhoea calls are routinely used to monitor significant rises in the numbers of calls received. Control charts are calculated by assuming that calls follow a Poisson distribution with the total number of calls as an offset: a model is fitted to each region and symptom separately [29]. The model takes into account call variation caused by weekends, public holidays and the time of year – variables that can affect the number of calls received by NHS Direct. A value above the upper limit of the 99.5% confidence interval would be considered to be unusual. The seven-day moving average for diarrhoea calls was also monitored. The number and percentage of calls for diarrhoea in the East Midlands region were presented by call outcome and the number of calls in the Northampton (NN) postcode districts and in particular the number of calls in the NN11 and NN12 postcode districts, which were most affected by the incident. QSurveillance national consultation rates per 100,000 population for diarrhoea (in the age groups under five years, five years and over, and all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with rates for the same period in 2007 (data not presented). Consultation rates by region for 2008 for diarrhoea (all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with those for the East Midlands region. The gastroenteritis indicator includes all cases of diarrhoea and/or vomiting. Consultation rates and standardised incidence ratios (SIRs) – calculated using the UK as the standard population – for diarrhoea, gastroenteritis and vomiting were compared for the UK, Yorkshire and Humberside, East Midlands, Leicestershire, Northamptonshire and Rutland SHA, and Daventry and South Northants PCT, Northamptonshire Heartlands PCT and Northampton PCT. Yorkshire and Humberside was not an affected region but was included as a control. The area supplied by the Pitsford Reservoir included the three PCTs, which were all within the Leicestershire, Northamptonshire and Rutland SHA. The consultation rates and SIRs were compared for the period from week 16 to week 35 of 2008 in order to compare the rates before and after the Cryptosporidium exceedance, which took place in week 26. Estimates of excess numbers of cases of diarrhoea occurring during and following the Cryptosporidium outbreak were made by calculating the mean consultation rate over a period of five weeks before and after the incident (weeks 20–24 and weeks 31–35, respectively). For each of the three PCTs, the calculated mean rate was applied to the PCT population to estimate the number of cases that would be expected each week. The actual consultation rates for diarrhoea for weeks 25 to 30 were used to estimate the number of cases for the PCT population each week. The expected number of cases was subtracted from the estimated number of cases in the PCT population to give the estimated number of excess cases. Results HPA/NHS Direct surveillance system A peak in the number of calls for diarrhoea in the East Midlands was recorded in 25–26 June 2008, the period that coincided with the contamination incident and the associated media coverage (Figure 1). The neighbouring areas of the West Midlands, Yorkshire and the Humber, and East of England showed no increase in the number of calls for diarrhoea. The peak produced a control chart exceedance for calls for diarrhoea on 25 June 2008 (Figure 2), when the proportion of calls exceeded the upper limit of the 99.5% confidence interval. There were further confidence interval exceedances on 26 and 28 June (which were not control chart exceedances). There was no peak in calls for vomiting or control chart exceedance for these calls in the East Midlands. HPA/QSurveillance national surveillance system The East Midlands region had significantly high consultation rates for diarrhoea and gastroenteritis in week 25 (16–22 June), week 26 (23–29 June 2008, when the contamination incident was reported) and in the following four weeks. Within this region. Leicestershire, Northamptonshire and Rutland SHA had slightly raised consultation rates and significant SIRs across weeks 25 to 30 that were not seen in the neighbouring Trent SHA. At PCT level, all three of the PCTs in the area affected by the incident showed increased consultation rates for diarrhoea (Table 1) and gastroenteritis (Table 2) with SIRs significantly above the UK rate in week 26. Daventry and South Northants PCT also had a raised SIR for both indicators in week 25, and although Northamptonshire Heartlands and Northampton PCTs did not have SIRs significantly above that of the UK in week 25, the rise in consultation rates for diarrhoea and gastroenteritis began during week 25. In Northampton PCT, consultations for both diarrhoea and gastroenteritis peaked in the week following the contamination incident, week 27, returning to normal levels by week 30 (Figure 3A and 3B). A similar effect can be seen in Northampton Heartlands PCT. Daventry and South Northants PCT also showed an increase, but appeared to have consistently higher rates for both indicators. This was the area with the smallest population so the rates were more variable than in the other PCTs and we therefore interpreted these results with caution. The consultation rates for vomiting during weeks 25 to 30 in the East Midlands were not unusual at SHA or PCT level (data not presented). Discussion We have demonstrated the sensitivity of syndromic surveillance in detecting this small Cryptosporidium outbreak and the value of the surveillance in being able to describe the extent of its spread. Both the HPA/NHS Direct and HPA/QSurveillance systems showed demonstrable increases in calls and consultations for diarrhoea that were linked to the outbreak. QSurveillance consultations appeared to increase across the PCTs immediately affected but not in the surrounding area. Both the HPA/NHS Direct and HPA/QSurveillance systems showed a clear signal at the time of the incident and we were able to describe the extent of the impact on pre-primary care and primary care services. The HPA/QSurveillance system showed a rise in consultation rates for gastrointestinal symptoms that began the week before the outbreak, consistent with the period when Cryptosporidium was present in the water leaving the Pitsford Reservoir (19–24 June 2008) and with the onset of symptoms in the first outbreak case on 24 June. Although only 33 cases were identified by the outbreak investigation team, of which 23 were confirmed as having the outbreak Cryptosporidium strain, our syndromic surveillance data detected this limited outbreak. Data also suggested a more widespread increase in general gastrointestinal symptoms around the time of the outbreak, with an estimated 422 excess diarrhoea cases; these excess cases represented an increase of about 25% above normally expected levels. It is highly probable that a proportion of these excess cases may have resulted from the increased publicity surrounding the incident – for example, it is likely that media reports contributed to the large peak in calls detected by the HPA/NHS Direct surveillance system on the day the boil water notice was issued, and could also have impacted on the GP consultation rate. It has been previously shown that reporting of mumps cases is sensitive to media coverage, with a rise in clinically reported cases following newspaper reports [30]. A similar mechanism could account for some of the excess GP consultations as cases experiencing gastrointestinal symptoms may have been more likely to consult their GP, whereas in normal circumstances they would have cared for themselves at home. It is also possible that the surveillance shows outbreak-associated cases that did not come to the attention of the outbreak team, perhaps because symptoms were not sufficiently severe to warrant further investigation, or stool samples were not provided for testing. It is interesting to note that there was no demonstrable impact on the number of calls for vomiting (which is not a prominent clinical feature of cryptosporidiosis). Other common community-based pathogens such as norovirus and rotavirus were at low levels, as is normal for that time of year [31]. In this instance, public health authorities had already been alerted to a potential problem by the water company, although the extent of the outbreak was detected by syndromic surveillance. In 2003 the syndromic surveillance systems in the city of New York, United States, were able to detect an increase in diarrhoeal illness following a power outage when there was no other indication of citywide illness [32]. The New York system covers a population of nine million, but does not regularly detect localised outbreaks [33]. It has been shown previously that the HPA/NHS Direct surveillance system would be unlikely to detect a Cryptosporidium outbreak unless call volumes are high (72% chance of detection if nine-tenths of cases called NHS Direct) [29], although the value of syndromic surveillance for such outbreaks has been recognised [34]. The system detected the East Midlands Cryptosporidium outbreak that affected a smaller population than that covered by the New York system. The three PCTs affected have a combined population of around 600,000, of which just over half use GP practices reporting to QSurveillance, yet this syndromic surveillance system was able to describe an increase in consultation rates for diarrhoea and gastroenteritis around the time of the outbreak. Limitations of the data There was extensive media reporting of the incident that may have affected both the HPA/NHS Direct and HPA/QSurveillance systems and contributed to the increase in reported gastrointestinal symptoms around the time of the contamination incident. However, the rise in consultation rates for diarrhoea began before the outbreak had been detected and therefore cannot be attributed to media coverage. The HPA/NHS Direct and HPA/QSurveillance systems monitor general symptoms and so could only monitor the relevant symptoms of diarrhoea and vomiting. They are not able to detect Cryptosporidium cases, as this would require laboratory confirmation of diagnosis, so some of the estimated excess cases could be unconnected with this incident. This outbreak was discovered by other means but both the HPA/NHS Direct and HPA/ QSurveillance systems were able to describe the extent of the disease in the general population and provide reassurance that there was no widespread impact. Compared with other populations, older people and ethnic minorities are less likely to call NHS Direct [29], and although this should not prevent detection of gastrointestinal symptoms as a result of drinking water contamination as this would affect the whole population, this may reduce the signal from the system [35]. With such large surveillance systems, there will be ‘background noise’ in the data, so procedures must be in place to correctly interpret the data and set appropriate thresholds for action. Conclusion To our knowledge, this is the first time that PCT-level data from a syndromic surveillance system, the HPA/ QSurveillance national surveillance system, have been able to show the extent of such a limited outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide PCT-level data and this Cryptosporidium contamination incident demonstrates the potential usefulness of this system.

In which region did this happen?

1503

Value of syndromic surveillance in monitoring a focal waterborne outbreak due to an unusual Cryptosporidium genotype in Northamptonshire

The United Kingdom (UK) has several national syndromic surveillance systems. The Health Protection Agency (HPA)/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data from a national telephone helpline, while the HPA/ QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems. Data from both of these systems were used to monitor a local outbreak of cryptosporidiosis that occurred following Cryptosporidium oocyst contamination of drinking water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, in June 2008. There was a peak in the number of calls to NHS Direct concerning diarrhoea that coincided with the incident. QSurveillance data for the local areas affected by the outbreak showed a significant increase in GP consultations for diarrhoea and gastroenteritis in the week of the incident but there was no increase in consultations for vomiting. A total of 33 clinical cases of cryptosporidiosis were identified in the outbreak investigation, of which 23 were confirmed as infected with the outbreak strain. However, QSurveillance data suggest that there were an estimated 422 excess diarrhoea cases during the outbreak, an increase of about 25% over baseline weekly levels. To our knowledge, this is the first time that data from a syndromic surveillance system, the HPA/QSurveillance national surveillance system, have been able to show the extent of such a small outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide data at local health district (primary care trust) level. The Cryptosporidium contamination incident described demonstrates the potential usefulness of this information, as it is unusual for syndromic surveillance systems to be able to help monitor such a small-scale outbreak. Introduction As syndromic surveillance systems usually capture data already collected for other purposes, and monitor generic symptoms and/or clinically diagnosed disease, they provide information at an earlier stage of illness (compared with laboratory-confirmed diagnoses), so that action can be taken in time to substantially reduce the impact of disease. Some systems, for example, the Royal College of General Practitioners Weekly Returns Service, are now well established, with many years of historical data that allow monitoring of longer-term disease trends [1]. They have the ability to provide early warning of, for example, seasonal rises in influenza and can trigger public health action, such as a recommendation to prescribe antiviral drugs in line with national guidance [2-4]. They can also provide reassurance to incident response teams and the general public that an incident has not caused adverse health effects – for example, following an explosion at the Buncefield oil storage depot in Hemel Hempstead, United Kingdom (UK), in 2005, syndromic surveillance confirmed that there were no unusual rises in community-based morbidity linked to the incident [5]; following the eruption of the Eyjafjallajökull volcano in Iceland in April 2010 similar assurance was given about lack of impact on community morbidity [6]. Health departments are increasingly expected to monitor health effects of natural events such as heat wave or flooding, or implement surveillance – of which syndromic surveillance plays a major role – for mass gatherings such as the Olympics or football World Cup [7-9]. Systems in France, Australia and Taiwan use data from emergency departments [10-12], a Canadian system uses over-the-counter pharmacy sales [13,14], and in the Netherlands data from both syndromic and surrogate data sources, such as employee absence records and prescription medications dispensed by pharmacies, are included in surveillance systems [15,16]. Currently systems based on Internet searches via search engines or on queries submitted to medical websites are being developed [17,18]. In the UK, the HPA/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data collected from the NHS Direct telephone helpline [19], while the HPA/QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems [20]. The HPA Real-time Syndromic Surveillance Team is a small team that coordinates a number of syndromic surveillance systems within the HPA and takes a lead for syndromic surveillance in England [21]. This paper describes the support provided by the team to the local incident management team during a local cryptosporidiosis outbreak and shows the use of syndromic surveillance in monitoring the extent of an outbreak using the HPA/NHS Direct and HPA/QSurveillance national surveillance systems. Cryptosporidiosis Cryptosporidium is a protozoan parasite that can cause an infection in people, cattle and sometimes other animals [22]. Cryptosporidiosis is most common in children aged between one and five years, but it can affect all ages. Those with impaired immune systems are likely to be most seriously affected. Symptoms usually appear between three and 12 days after initial exposure and include watery diarrhoea, stomach pains, dehydration and fever. In its transmissible form, called an oocyst, the parasite is protected by an outer shell, which allows it to survive in the environment for a long time. Transmission occurs most often via the faeco-oral route through person-to-person or animalto-person contact, but people may also be infected by consuming contaminated water or food or by swimming in contaminated water. Although uncommon, the largest outbreaks have occurred following contamination of drinking water [23,24]. Normal chlorine disinfection procedures do not kill the oocysts, so they are removed by filtration and water companies carry out routine monitoring of treated water. Description of the incident On 25 June 2008 the local Health Protection Unit was informed by Anglian Water of an exceedence in the level of Cryptosporidium oocysts found in water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, during 19 to 24 June 2008 [25]. The reservoir supplied a population of more than 250,000 in the Northampton area. A notice advising people in the affected areas to boil all drinking water was issued on 25 June 2008 and public health messages were circulated to local health services and to the general public via the media. Those members of the public who were concerned about health risks associated with the incident were asked to ring NHS Direct for clinical advice [26]. The HPA wrote to local GPs and hospitals asking them to monitor potential patients for signs and symptoms of Cryptosporidium infection and to submit faecal specimens to the local hospital diagnostic laboratory if patients presented with diarrhoea. Samples from 34 patients where Cryptosporidium infection was identified were sent to the UK Cryptosporidium reference unit for typing. On 30 June 2008, the Cryptosporidium oocysts found in the reservoir water were confirmed as being of the rabbit genotype Cryptosporidium cuniculus [27]. Subsequently, a dead rabbit was found in a treated water tank at the water treatment works. The genotype of Cryptosporidium oocysts in the rabbit’s large bowel was indistinguishable from that of the oocysts found in the water [27]. After remediation of the water supply and distribution, the ‘boil water notice’ was lifted on 4 July and the following day the first case of cryptosporidiosis linked to the incident was identified by the reference laboratory (this case was infected with C. cuniculus). During the course of the outbreak (24 June – 18 July 2008, the dates of symptom onset in the first and last case, respectively), 23 cases of cryptosporidiosis were confirmed as being infected with C. cuniculus; one of the 23 was a secondary case. The HPA Real-time Syndromic Surveillance Team provided data in order to aid the response to this incident and the first syndromic surveillance report was circulated to the incident management team and other relevant people in the HPA on 27 June 2008. Data from the HPA/NHS Direct and HPA/QSurveillance systems were provided in a series of regular reports, initially daily and eventually weekly, until the final report on 21 August 2008. Each report included a summary interpretation and more detailed data on diarrhoea, gastroenteritis and vomiting indicators. Methods Surveillance systems HPA/NHS Direct surveillance system NHS Direct is a 24-hour nurse-led telephone helpline that provides health information and advice to the general public. Nurses use a computerised clinical decision support system – the NHS Clinical Assessment System (NHS CAS) – to handle calls. This assessment system uses approximately 200 computerised symptom-based clinical algorithms. Nurses assign the call to the most appropriate algorithm and the patient’s symptoms determine the questions asked and the action to be taken following the call (call outcome), which could be guidance on self-care or they could be referred to their GP or advised to attend a hospital emergency department. No attempt is made to provide a formal diagnosis. Daily NHS Direct data are received by the Real-time Syndromic Surveillance Team, where the number and type of calls received during the previous day are analysed and interpreted. Call proportions are calculated by age group and algorithm against the total number of calls received. HPA/QSurveillance system The HPA/QSurveillance national surveillance system was set up by the University of Nottingham, United Kingdom, and Egton Medical Information Systems (EMIS), a supplier of general practice computer systems, in collaboration with the HPA. It comprises a network of more than 3,500 general practices throughout the UK, covering more than 22 million patients (about 38% of the population [28]). Aggregated data on GP consultations for a range of indicators are automatically uploaded daily from GP practice systems to a central database. Data are routinely reported on a weekly basis; however, daily reporting is possible for specific incidents. Reports are provided at national or regional level (strategic health authority, SHA) and by local health district (primary care trust, PCT). Analysis of surveillance data NHS Direct call proportions for gastrointestinal syndromes (diarrhoea and vomiting) for the East Midlands region in England, where Northampton is situated, were examined during the outbreak (24 June – 18 July 2008) and compared with those for England and Wales. A series of control charts for diarrhoea calls are routinely used to monitor significant rises in the numbers of calls received. Control charts are calculated by assuming that calls follow a Poisson distribution with the total number of calls as an offset: a model is fitted to each region and symptom separately [29]. The model takes into account call variation caused by weekends, public holidays and the time of year – variables that can affect the number of calls received by NHS Direct. A value above the upper limit of the 99.5% confidence interval would be considered to be unusual. The seven-day moving average for diarrhoea calls was also monitored. The number and percentage of calls for diarrhoea in the East Midlands region were presented by call outcome and the number of calls in the Northampton (NN) postcode districts and in particular the number of calls in the NN11 and NN12 postcode districts, which were most affected by the incident. QSurveillance national consultation rates per 100,000 population for diarrhoea (in the age groups under five years, five years and over, and all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with rates for the same period in 2007 (data not presented). Consultation rates by region for 2008 for diarrhoea (all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with those for the East Midlands region. The gastroenteritis indicator includes all cases of diarrhoea and/or vomiting. Consultation rates and standardised incidence ratios (SIRs) – calculated using the UK as the standard population – for diarrhoea, gastroenteritis and vomiting were compared for the UK, Yorkshire and Humberside, East Midlands, Leicestershire, Northamptonshire and Rutland SHA, and Daventry and South Northants PCT, Northamptonshire Heartlands PCT and Northampton PCT. Yorkshire and Humberside was not an affected region but was included as a control. The area supplied by the Pitsford Reservoir included the three PCTs, which were all within the Leicestershire, Northamptonshire and Rutland SHA. The consultation rates and SIRs were compared for the period from week 16 to week 35 of 2008 in order to compare the rates before and after the Cryptosporidium exceedance, which took place in week 26. Estimates of excess numbers of cases of diarrhoea occurring during and following the Cryptosporidium outbreak were made by calculating the mean consultation rate over a period of five weeks before and after the incident (weeks 20–24 and weeks 31–35, respectively). For each of the three PCTs, the calculated mean rate was applied to the PCT population to estimate the number of cases that would be expected each week. The actual consultation rates for diarrhoea for weeks 25 to 30 were used to estimate the number of cases for the PCT population each week. The expected number of cases was subtracted from the estimated number of cases in the PCT population to give the estimated number of excess cases. Results HPA/NHS Direct surveillance system A peak in the number of calls for diarrhoea in the East Midlands was recorded in 25–26 June 2008, the period that coincided with the contamination incident and the associated media coverage (Figure 1). The neighbouring areas of the West Midlands, Yorkshire and the Humber, and East of England showed no increase in the number of calls for diarrhoea. The peak produced a control chart exceedance for calls for diarrhoea on 25 June 2008 (Figure 2), when the proportion of calls exceeded the upper limit of the 99.5% confidence interval. There were further confidence interval exceedances on 26 and 28 June (which were not control chart exceedances). There was no peak in calls for vomiting or control chart exceedance for these calls in the East Midlands. HPA/QSurveillance national surveillance system The East Midlands region had significantly high consultation rates for diarrhoea and gastroenteritis in week 25 (16–22 June), week 26 (23–29 June 2008, when the contamination incident was reported) and in the following four weeks. Within this region. Leicestershire, Northamptonshire and Rutland SHA had slightly raised consultation rates and significant SIRs across weeks 25 to 30 that were not seen in the neighbouring Trent SHA. At PCT level, all three of the PCTs in the area affected by the incident showed increased consultation rates for diarrhoea (Table 1) and gastroenteritis (Table 2) with SIRs significantly above the UK rate in week 26. Daventry and South Northants PCT also had a raised SIR for both indicators in week 25, and although Northamptonshire Heartlands and Northampton PCTs did not have SIRs significantly above that of the UK in week 25, the rise in consultation rates for diarrhoea and gastroenteritis began during week 25. In Northampton PCT, consultations for both diarrhoea and gastroenteritis peaked in the week following the contamination incident, week 27, returning to normal levels by week 30 (Figure 3A and 3B). A similar effect can be seen in Northampton Heartlands PCT. Daventry and South Northants PCT also showed an increase, but appeared to have consistently higher rates for both indicators. This was the area with the smallest population so the rates were more variable than in the other PCTs and we therefore interpreted these results with caution. The consultation rates for vomiting during weeks 25 to 30 in the East Midlands were not unusual at SHA or PCT level (data not presented). Discussion We have demonstrated the sensitivity of syndromic surveillance in detecting this small Cryptosporidium outbreak and the value of the surveillance in being able to describe the extent of its spread. Both the HPA/NHS Direct and HPA/QSurveillance systems showed demonstrable increases in calls and consultations for diarrhoea that were linked to the outbreak. QSurveillance consultations appeared to increase across the PCTs immediately affected but not in the surrounding area. Both the HPA/NHS Direct and HPA/QSurveillance systems showed a clear signal at the time of the incident and we were able to describe the extent of the impact on pre-primary care and primary care services. The HPA/QSurveillance system showed a rise in consultation rates for gastrointestinal symptoms that began the week before the outbreak, consistent with the period when Cryptosporidium was present in the water leaving the Pitsford Reservoir (19–24 June 2008) and with the onset of symptoms in the first outbreak case on 24 June. Although only 33 cases were identified by the outbreak investigation team, of which 23 were confirmed as having the outbreak Cryptosporidium strain, our syndromic surveillance data detected this limited outbreak. Data also suggested a more widespread increase in general gastrointestinal symptoms around the time of the outbreak, with an estimated 422 excess diarrhoea cases; these excess cases represented an increase of about 25% above normally expected levels. It is highly probable that a proportion of these excess cases may have resulted from the increased publicity surrounding the incident – for example, it is likely that media reports contributed to the large peak in calls detected by the HPA/NHS Direct surveillance system on the day the boil water notice was issued, and could also have impacted on the GP consultation rate. It has been previously shown that reporting of mumps cases is sensitive to media coverage, with a rise in clinically reported cases following newspaper reports [30]. A similar mechanism could account for some of the excess GP consultations as cases experiencing gastrointestinal symptoms may have been more likely to consult their GP, whereas in normal circumstances they would have cared for themselves at home. It is also possible that the surveillance shows outbreak-associated cases that did not come to the attention of the outbreak team, perhaps because symptoms were not sufficiently severe to warrant further investigation, or stool samples were not provided for testing. It is interesting to note that there was no demonstrable impact on the number of calls for vomiting (which is not a prominent clinical feature of cryptosporidiosis). Other common community-based pathogens such as norovirus and rotavirus were at low levels, as is normal for that time of year [31]. In this instance, public health authorities had already been alerted to a potential problem by the water company, although the extent of the outbreak was detected by syndromic surveillance. In 2003 the syndromic surveillance systems in the city of New York, United States, were able to detect an increase in diarrhoeal illness following a power outage when there was no other indication of citywide illness [32]. The New York system covers a population of nine million, but does not regularly detect localised outbreaks [33]. It has been shown previously that the HPA/NHS Direct surveillance system would be unlikely to detect a Cryptosporidium outbreak unless call volumes are high (72% chance of detection if nine-tenths of cases called NHS Direct) [29], although the value of syndromic surveillance for such outbreaks has been recognised [34]. The system detected the East Midlands Cryptosporidium outbreak that affected a smaller population than that covered by the New York system. The three PCTs affected have a combined population of around 600,000, of which just over half use GP practices reporting to QSurveillance, yet this syndromic surveillance system was able to describe an increase in consultation rates for diarrhoea and gastroenteritis around the time of the outbreak. Limitations of the data There was extensive media reporting of the incident that may have affected both the HPA/NHS Direct and HPA/QSurveillance systems and contributed to the increase in reported gastrointestinal symptoms around the time of the contamination incident. However, the rise in consultation rates for diarrhoea began before the outbreak had been detected and therefore cannot be attributed to media coverage. The HPA/NHS Direct and HPA/QSurveillance systems monitor general symptoms and so could only monitor the relevant symptoms of diarrhoea and vomiting. They are not able to detect Cryptosporidium cases, as this would require laboratory confirmation of diagnosis, so some of the estimated excess cases could be unconnected with this incident. This outbreak was discovered by other means but both the HPA/NHS Direct and HPA/ QSurveillance systems were able to describe the extent of the disease in the general population and provide reassurance that there was no widespread impact. Compared with other populations, older people and ethnic minorities are less likely to call NHS Direct [29], and although this should not prevent detection of gastrointestinal symptoms as a result of drinking water contamination as this would affect the whole population, this may reduce the signal from the system [35]. With such large surveillance systems, there will be ‘background noise’ in the data, so procedures must be in place to correctly interpret the data and set appropriate thresholds for action. Conclusion To our knowledge, this is the first time that PCT-level data from a syndromic surveillance system, the HPA/ QSurveillance national surveillance system, have been able to show the extent of such a limited outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide PCT-level data and this Cryptosporidium contamination incident demonstrates the potential usefulness of this system.

In which country did this happen?

1504

Value of syndromic surveillance in monitoring a focal waterborne outbreak due to an unusual Cryptosporidium genotype in Northamptonshire

The United Kingdom (UK) has several national syndromic surveillance systems. The Health Protection Agency (HPA)/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data from a national telephone helpline, while the HPA/ QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems. Data from both of these systems were used to monitor a local outbreak of cryptosporidiosis that occurred following Cryptosporidium oocyst contamination of drinking water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, in June 2008. There was a peak in the number of calls to NHS Direct concerning diarrhoea that coincided with the incident. QSurveillance data for the local areas affected by the outbreak showed a significant increase in GP consultations for diarrhoea and gastroenteritis in the week of the incident but there was no increase in consultations for vomiting. A total of 33 clinical cases of cryptosporidiosis were identified in the outbreak investigation, of which 23 were confirmed as infected with the outbreak strain. However, QSurveillance data suggest that there were an estimated 422 excess diarrhoea cases during the outbreak, an increase of about 25% over baseline weekly levels. To our knowledge, this is the first time that data from a syndromic surveillance system, the HPA/QSurveillance national surveillance system, have been able to show the extent of such a small outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide data at local health district (primary care trust) level. The Cryptosporidium contamination incident described demonstrates the potential usefulness of this information, as it is unusual for syndromic surveillance systems to be able to help monitor such a small-scale outbreak. Introduction As syndromic surveillance systems usually capture data already collected for other purposes, and monitor generic symptoms and/or clinically diagnosed disease, they provide information at an earlier stage of illness (compared with laboratory-confirmed diagnoses), so that action can be taken in time to substantially reduce the impact of disease. Some systems, for example, the Royal College of General Practitioners Weekly Returns Service, are now well established, with many years of historical data that allow monitoring of longer-term disease trends [1]. They have the ability to provide early warning of, for example, seasonal rises in influenza and can trigger public health action, such as a recommendation to prescribe antiviral drugs in line with national guidance [2-4]. They can also provide reassurance to incident response teams and the general public that an incident has not caused adverse health effects – for example, following an explosion at the Buncefield oil storage depot in Hemel Hempstead, United Kingdom (UK), in 2005, syndromic surveillance confirmed that there were no unusual rises in community-based morbidity linked to the incident [5]; following the eruption of the Eyjafjallajökull volcano in Iceland in April 2010 similar assurance was given about lack of impact on community morbidity [6]. Health departments are increasingly expected to monitor health effects of natural events such as heat wave or flooding, or implement surveillance – of which syndromic surveillance plays a major role – for mass gatherings such as the Olympics or football World Cup [7-9]. Systems in France, Australia and Taiwan use data from emergency departments [10-12], a Canadian system uses over-the-counter pharmacy sales [13,14], and in the Netherlands data from both syndromic and surrogate data sources, such as employee absence records and prescription medications dispensed by pharmacies, are included in surveillance systems [15,16]. Currently systems based on Internet searches via search engines or on queries submitted to medical websites are being developed [17,18]. In the UK, the HPA/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data collected from the NHS Direct telephone helpline [19], while the HPA/QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems [20]. The HPA Real-time Syndromic Surveillance Team is a small team that coordinates a number of syndromic surveillance systems within the HPA and takes a lead for syndromic surveillance in England [21]. This paper describes the support provided by the team to the local incident management team during a local cryptosporidiosis outbreak and shows the use of syndromic surveillance in monitoring the extent of an outbreak using the HPA/NHS Direct and HPA/QSurveillance national surveillance systems. Cryptosporidiosis Cryptosporidium is a protozoan parasite that can cause an infection in people, cattle and sometimes other animals [22]. Cryptosporidiosis is most common in children aged between one and five years, but it can affect all ages. Those with impaired immune systems are likely to be most seriously affected. Symptoms usually appear between three and 12 days after initial exposure and include watery diarrhoea, stomach pains, dehydration and fever. In its transmissible form, called an oocyst, the parasite is protected by an outer shell, which allows it to survive in the environment for a long time. Transmission occurs most often via the faeco-oral route through person-to-person or animalto-person contact, but people may also be infected by consuming contaminated water or food or by swimming in contaminated water. Although uncommon, the largest outbreaks have occurred following contamination of drinking water [23,24]. Normal chlorine disinfection procedures do not kill the oocysts, so they are removed by filtration and water companies carry out routine monitoring of treated water. Description of the incident On 25 June 2008 the local Health Protection Unit was informed by Anglian Water of an exceedence in the level of Cryptosporidium oocysts found in water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, during 19 to 24 June 2008 [25]. The reservoir supplied a population of more than 250,000 in the Northampton area. A notice advising people in the affected areas to boil all drinking water was issued on 25 June 2008 and public health messages were circulated to local health services and to the general public via the media. Those members of the public who were concerned about health risks associated with the incident were asked to ring NHS Direct for clinical advice [26]. The HPA wrote to local GPs and hospitals asking them to monitor potential patients for signs and symptoms of Cryptosporidium infection and to submit faecal specimens to the local hospital diagnostic laboratory if patients presented with diarrhoea. Samples from 34 patients where Cryptosporidium infection was identified were sent to the UK Cryptosporidium reference unit for typing. On 30 June 2008, the Cryptosporidium oocysts found in the reservoir water were confirmed as being of the rabbit genotype Cryptosporidium cuniculus [27]. Subsequently, a dead rabbit was found in a treated water tank at the water treatment works. The genotype of Cryptosporidium oocysts in the rabbit’s large bowel was indistinguishable from that of the oocysts found in the water [27]. After remediation of the water supply and distribution, the ‘boil water notice’ was lifted on 4 July and the following day the first case of cryptosporidiosis linked to the incident was identified by the reference laboratory (this case was infected with C. cuniculus). During the course of the outbreak (24 June – 18 July 2008, the dates of symptom onset in the first and last case, respectively), 23 cases of cryptosporidiosis were confirmed as being infected with C. cuniculus; one of the 23 was a secondary case. The HPA Real-time Syndromic Surveillance Team provided data in order to aid the response to this incident and the first syndromic surveillance report was circulated to the incident management team and other relevant people in the HPA on 27 June 2008. Data from the HPA/NHS Direct and HPA/QSurveillance systems were provided in a series of regular reports, initially daily and eventually weekly, until the final report on 21 August 2008. Each report included a summary interpretation and more detailed data on diarrhoea, gastroenteritis and vomiting indicators. Methods Surveillance systems HPA/NHS Direct surveillance system NHS Direct is a 24-hour nurse-led telephone helpline that provides health information and advice to the general public. Nurses use a computerised clinical decision support system – the NHS Clinical Assessment System (NHS CAS) – to handle calls. This assessment system uses approximately 200 computerised symptom-based clinical algorithms. Nurses assign the call to the most appropriate algorithm and the patient’s symptoms determine the questions asked and the action to be taken following the call (call outcome), which could be guidance on self-care or they could be referred to their GP or advised to attend a hospital emergency department. No attempt is made to provide a formal diagnosis. Daily NHS Direct data are received by the Real-time Syndromic Surveillance Team, where the number and type of calls received during the previous day are analysed and interpreted. Call proportions are calculated by age group and algorithm against the total number of calls received. HPA/QSurveillance system The HPA/QSurveillance national surveillance system was set up by the University of Nottingham, United Kingdom, and Egton Medical Information Systems (EMIS), a supplier of general practice computer systems, in collaboration with the HPA. It comprises a network of more than 3,500 general practices throughout the UK, covering more than 22 million patients (about 38% of the population [28]). Aggregated data on GP consultations for a range of indicators are automatically uploaded daily from GP practice systems to a central database. Data are routinely reported on a weekly basis; however, daily reporting is possible for specific incidents. Reports are provided at national or regional level (strategic health authority, SHA) and by local health district (primary care trust, PCT). Analysis of surveillance data NHS Direct call proportions for gastrointestinal syndromes (diarrhoea and vomiting) for the East Midlands region in England, where Northampton is situated, were examined during the outbreak (24 June – 18 July 2008) and compared with those for England and Wales. A series of control charts for diarrhoea calls are routinely used to monitor significant rises in the numbers of calls received. Control charts are calculated by assuming that calls follow a Poisson distribution with the total number of calls as an offset: a model is fitted to each region and symptom separately [29]. The model takes into account call variation caused by weekends, public holidays and the time of year – variables that can affect the number of calls received by NHS Direct. A value above the upper limit of the 99.5% confidence interval would be considered to be unusual. The seven-day moving average for diarrhoea calls was also monitored. The number and percentage of calls for diarrhoea in the East Midlands region were presented by call outcome and the number of calls in the Northampton (NN) postcode districts and in particular the number of calls in the NN11 and NN12 postcode districts, which were most affected by the incident. QSurveillance national consultation rates per 100,000 population for diarrhoea (in the age groups under five years, five years and over, and all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with rates for the same period in 2007 (data not presented). Consultation rates by region for 2008 for diarrhoea (all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with those for the East Midlands region. The gastroenteritis indicator includes all cases of diarrhoea and/or vomiting. Consultation rates and standardised incidence ratios (SIRs) – calculated using the UK as the standard population – for diarrhoea, gastroenteritis and vomiting were compared for the UK, Yorkshire and Humberside, East Midlands, Leicestershire, Northamptonshire and Rutland SHA, and Daventry and South Northants PCT, Northamptonshire Heartlands PCT and Northampton PCT. Yorkshire and Humberside was not an affected region but was included as a control. The area supplied by the Pitsford Reservoir included the three PCTs, which were all within the Leicestershire, Northamptonshire and Rutland SHA. The consultation rates and SIRs were compared for the period from week 16 to week 35 of 2008 in order to compare the rates before and after the Cryptosporidium exceedance, which took place in week 26. Estimates of excess numbers of cases of diarrhoea occurring during and following the Cryptosporidium outbreak were made by calculating the mean consultation rate over a period of five weeks before and after the incident (weeks 20–24 and weeks 31–35, respectively). For each of the three PCTs, the calculated mean rate was applied to the PCT population to estimate the number of cases that would be expected each week. The actual consultation rates for diarrhoea for weeks 25 to 30 were used to estimate the number of cases for the PCT population each week. The expected number of cases was subtracted from the estimated number of cases in the PCT population to give the estimated number of excess cases. Results HPA/NHS Direct surveillance system A peak in the number of calls for diarrhoea in the East Midlands was recorded in 25–26 June 2008, the period that coincided with the contamination incident and the associated media coverage (Figure 1). The neighbouring areas of the West Midlands, Yorkshire and the Humber, and East of England showed no increase in the number of calls for diarrhoea. The peak produced a control chart exceedance for calls for diarrhoea on 25 June 2008 (Figure 2), when the proportion of calls exceeded the upper limit of the 99.5% confidence interval. There were further confidence interval exceedances on 26 and 28 June (which were not control chart exceedances). There was no peak in calls for vomiting or control chart exceedance for these calls in the East Midlands. HPA/QSurveillance national surveillance system The East Midlands region had significantly high consultation rates for diarrhoea and gastroenteritis in week 25 (16–22 June), week 26 (23–29 June 2008, when the contamination incident was reported) and in the following four weeks. Within this region. Leicestershire, Northamptonshire and Rutland SHA had slightly raised consultation rates and significant SIRs across weeks 25 to 30 that were not seen in the neighbouring Trent SHA. At PCT level, all three of the PCTs in the area affected by the incident showed increased consultation rates for diarrhoea (Table 1) and gastroenteritis (Table 2) with SIRs significantly above the UK rate in week 26. Daventry and South Northants PCT also had a raised SIR for both indicators in week 25, and although Northamptonshire Heartlands and Northampton PCTs did not have SIRs significantly above that of the UK in week 25, the rise in consultation rates for diarrhoea and gastroenteritis began during week 25. In Northampton PCT, consultations for both diarrhoea and gastroenteritis peaked in the week following the contamination incident, week 27, returning to normal levels by week 30 (Figure 3A and 3B). A similar effect can be seen in Northampton Heartlands PCT. Daventry and South Northants PCT also showed an increase, but appeared to have consistently higher rates for both indicators. This was the area with the smallest population so the rates were more variable than in the other PCTs and we therefore interpreted these results with caution. The consultation rates for vomiting during weeks 25 to 30 in the East Midlands were not unusual at SHA or PCT level (data not presented). Discussion We have demonstrated the sensitivity of syndromic surveillance in detecting this small Cryptosporidium outbreak and the value of the surveillance in being able to describe the extent of its spread. Both the HPA/NHS Direct and HPA/QSurveillance systems showed demonstrable increases in calls and consultations for diarrhoea that were linked to the outbreak. QSurveillance consultations appeared to increase across the PCTs immediately affected but not in the surrounding area. Both the HPA/NHS Direct and HPA/QSurveillance systems showed a clear signal at the time of the incident and we were able to describe the extent of the impact on pre-primary care and primary care services. The HPA/QSurveillance system showed a rise in consultation rates for gastrointestinal symptoms that began the week before the outbreak, consistent with the period when Cryptosporidium was present in the water leaving the Pitsford Reservoir (19–24 June 2008) and with the onset of symptoms in the first outbreak case on 24 June. Although only 33 cases were identified by the outbreak investigation team, of which 23 were confirmed as having the outbreak Cryptosporidium strain, our syndromic surveillance data detected this limited outbreak. Data also suggested a more widespread increase in general gastrointestinal symptoms around the time of the outbreak, with an estimated 422 excess diarrhoea cases; these excess cases represented an increase of about 25% above normally expected levels. It is highly probable that a proportion of these excess cases may have resulted from the increased publicity surrounding the incident – for example, it is likely that media reports contributed to the large peak in calls detected by the HPA/NHS Direct surveillance system on the day the boil water notice was issued, and could also have impacted on the GP consultation rate. It has been previously shown that reporting of mumps cases is sensitive to media coverage, with a rise in clinically reported cases following newspaper reports [30]. A similar mechanism could account for some of the excess GP consultations as cases experiencing gastrointestinal symptoms may have been more likely to consult their GP, whereas in normal circumstances they would have cared for themselves at home. It is also possible that the surveillance shows outbreak-associated cases that did not come to the attention of the outbreak team, perhaps because symptoms were not sufficiently severe to warrant further investigation, or stool samples were not provided for testing. It is interesting to note that there was no demonstrable impact on the number of calls for vomiting (which is not a prominent clinical feature of cryptosporidiosis). Other common community-based pathogens such as norovirus and rotavirus were at low levels, as is normal for that time of year [31]. In this instance, public health authorities had already been alerted to a potential problem by the water company, although the extent of the outbreak was detected by syndromic surveillance. In 2003 the syndromic surveillance systems in the city of New York, United States, were able to detect an increase in diarrhoeal illness following a power outage when there was no other indication of citywide illness [32]. The New York system covers a population of nine million, but does not regularly detect localised outbreaks [33]. It has been shown previously that the HPA/NHS Direct surveillance system would be unlikely to detect a Cryptosporidium outbreak unless call volumes are high (72% chance of detection if nine-tenths of cases called NHS Direct) [29], although the value of syndromic surveillance for such outbreaks has been recognised [34]. The system detected the East Midlands Cryptosporidium outbreak that affected a smaller population than that covered by the New York system. The three PCTs affected have a combined population of around 600,000, of which just over half use GP practices reporting to QSurveillance, yet this syndromic surveillance system was able to describe an increase in consultation rates for diarrhoea and gastroenteritis around the time of the outbreak. Limitations of the data There was extensive media reporting of the incident that may have affected both the HPA/NHS Direct and HPA/QSurveillance systems and contributed to the increase in reported gastrointestinal symptoms around the time of the contamination incident. However, the rise in consultation rates for diarrhoea began before the outbreak had been detected and therefore cannot be attributed to media coverage. The HPA/NHS Direct and HPA/QSurveillance systems monitor general symptoms and so could only monitor the relevant symptoms of diarrhoea and vomiting. They are not able to detect Cryptosporidium cases, as this would require laboratory confirmation of diagnosis, so some of the estimated excess cases could be unconnected with this incident. This outbreak was discovered by other means but both the HPA/NHS Direct and HPA/ QSurveillance systems were able to describe the extent of the disease in the general population and provide reassurance that there was no widespread impact. Compared with other populations, older people and ethnic minorities are less likely to call NHS Direct [29], and although this should not prevent detection of gastrointestinal symptoms as a result of drinking water contamination as this would affect the whole population, this may reduce the signal from the system [35]. With such large surveillance systems, there will be ‘background noise’ in the data, so procedures must be in place to correctly interpret the data and set appropriate thresholds for action. Conclusion To our knowledge, this is the first time that PCT-level data from a syndromic surveillance system, the HPA/ QSurveillance national surveillance system, have been able to show the extent of such a limited outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide PCT-level data and this Cryptosporidium contamination incident demonstrates the potential usefulness of this system.

Where did this happen?

1505

Value of syndromic surveillance in monitoring a focal waterborne outbreak due to an unusual Cryptosporidium genotype in Northamptonshire

The United Kingdom (UK) has several national syndromic surveillance systems. The Health Protection Agency (HPA)/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data from a national telephone helpline, while the HPA/ QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems. Data from both of these systems were used to monitor a local outbreak of cryptosporidiosis that occurred following Cryptosporidium oocyst contamination of drinking water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, in June 2008. There was a peak in the number of calls to NHS Direct concerning diarrhoea that coincided with the incident. QSurveillance data for the local areas affected by the outbreak showed a significant increase in GP consultations for diarrhoea and gastroenteritis in the week of the incident but there was no increase in consultations for vomiting. A total of 33 clinical cases of cryptosporidiosis were identified in the outbreak investigation, of which 23 were confirmed as infected with the outbreak strain. However, QSurveillance data suggest that there were an estimated 422 excess diarrhoea cases during the outbreak, an increase of about 25% over baseline weekly levels. To our knowledge, this is the first time that data from a syndromic surveillance system, the HPA/QSurveillance national surveillance system, have been able to show the extent of such a small outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide data at local health district (primary care trust) level. The Cryptosporidium contamination incident described demonstrates the potential usefulness of this information, as it is unusual for syndromic surveillance systems to be able to help monitor such a small-scale outbreak. Introduction As syndromic surveillance systems usually capture data already collected for other purposes, and monitor generic symptoms and/or clinically diagnosed disease, they provide information at an earlier stage of illness (compared with laboratory-confirmed diagnoses), so that action can be taken in time to substantially reduce the impact of disease. Some systems, for example, the Royal College of General Practitioners Weekly Returns Service, are now well established, with many years of historical data that allow monitoring of longer-term disease trends [1]. They have the ability to provide early warning of, for example, seasonal rises in influenza and can trigger public health action, such as a recommendation to prescribe antiviral drugs in line with national guidance [2-4]. They can also provide reassurance to incident response teams and the general public that an incident has not caused adverse health effects – for example, following an explosion at the Buncefield oil storage depot in Hemel Hempstead, United Kingdom (UK), in 2005, syndromic surveillance confirmed that there were no unusual rises in community-based morbidity linked to the incident [5]; following the eruption of the Eyjafjallajökull volcano in Iceland in April 2010 similar assurance was given about lack of impact on community morbidity [6]. Health departments are increasingly expected to monitor health effects of natural events such as heat wave or flooding, or implement surveillance – of which syndromic surveillance plays a major role – for mass gatherings such as the Olympics or football World Cup [7-9]. Systems in France, Australia and Taiwan use data from emergency departments [10-12], a Canadian system uses over-the-counter pharmacy sales [13,14], and in the Netherlands data from both syndromic and surrogate data sources, such as employee absence records and prescription medications dispensed by pharmacies, are included in surveillance systems [15,16]. Currently systems based on Internet searches via search engines or on queries submitted to medical websites are being developed [17,18]. In the UK, the HPA/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data collected from the NHS Direct telephone helpline [19], while the HPA/QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems [20]. The HPA Real-time Syndromic Surveillance Team is a small team that coordinates a number of syndromic surveillance systems within the HPA and takes a lead for syndromic surveillance in England [21]. This paper describes the support provided by the team to the local incident management team during a local cryptosporidiosis outbreak and shows the use of syndromic surveillance in monitoring the extent of an outbreak using the HPA/NHS Direct and HPA/QSurveillance national surveillance systems. Cryptosporidiosis Cryptosporidium is a protozoan parasite that can cause an infection in people, cattle and sometimes other animals [22]. Cryptosporidiosis is most common in children aged between one and five years, but it can affect all ages. Those with impaired immune systems are likely to be most seriously affected. Symptoms usually appear between three and 12 days after initial exposure and include watery diarrhoea, stomach pains, dehydration and fever. In its transmissible form, called an oocyst, the parasite is protected by an outer shell, which allows it to survive in the environment for a long time. Transmission occurs most often via the faeco-oral route through person-to-person or animalto-person contact, but people may also be infected by consuming contaminated water or food or by swimming in contaminated water. Although uncommon, the largest outbreaks have occurred following contamination of drinking water [23,24]. Normal chlorine disinfection procedures do not kill the oocysts, so they are removed by filtration and water companies carry out routine monitoring of treated water. Description of the incident On 25 June 2008 the local Health Protection Unit was informed by Anglian Water of an exceedence in the level of Cryptosporidium oocysts found in water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, during 19 to 24 June 2008 [25]. The reservoir supplied a population of more than 250,000 in the Northampton area. A notice advising people in the affected areas to boil all drinking water was issued on 25 June 2008 and public health messages were circulated to local health services and to the general public via the media. Those members of the public who were concerned about health risks associated with the incident were asked to ring NHS Direct for clinical advice [26]. The HPA wrote to local GPs and hospitals asking them to monitor potential patients for signs and symptoms of Cryptosporidium infection and to submit faecal specimens to the local hospital diagnostic laboratory if patients presented with diarrhoea. Samples from 34 patients where Cryptosporidium infection was identified were sent to the UK Cryptosporidium reference unit for typing. On 30 June 2008, the Cryptosporidium oocysts found in the reservoir water were confirmed as being of the rabbit genotype Cryptosporidium cuniculus [27]. Subsequently, a dead rabbit was found in a treated water tank at the water treatment works. The genotype of Cryptosporidium oocysts in the rabbit’s large bowel was indistinguishable from that of the oocysts found in the water [27]. After remediation of the water supply and distribution, the ‘boil water notice’ was lifted on 4 July and the following day the first case of cryptosporidiosis linked to the incident was identified by the reference laboratory (this case was infected with C. cuniculus). During the course of the outbreak (24 June – 18 July 2008, the dates of symptom onset in the first and last case, respectively), 23 cases of cryptosporidiosis were confirmed as being infected with C. cuniculus; one of the 23 was a secondary case. The HPA Real-time Syndromic Surveillance Team provided data in order to aid the response to this incident and the first syndromic surveillance report was circulated to the incident management team and other relevant people in the HPA on 27 June 2008. Data from the HPA/NHS Direct and HPA/QSurveillance systems were provided in a series of regular reports, initially daily and eventually weekly, until the final report on 21 August 2008. Each report included a summary interpretation and more detailed data on diarrhoea, gastroenteritis and vomiting indicators. Methods Surveillance systems HPA/NHS Direct surveillance system NHS Direct is a 24-hour nurse-led telephone helpline that provides health information and advice to the general public. Nurses use a computerised clinical decision support system – the NHS Clinical Assessment System (NHS CAS) – to handle calls. This assessment system uses approximately 200 computerised symptom-based clinical algorithms. Nurses assign the call to the most appropriate algorithm and the patient’s symptoms determine the questions asked and the action to be taken following the call (call outcome), which could be guidance on self-care or they could be referred to their GP or advised to attend a hospital emergency department. No attempt is made to provide a formal diagnosis. Daily NHS Direct data are received by the Real-time Syndromic Surveillance Team, where the number and type of calls received during the previous day are analysed and interpreted. Call proportions are calculated by age group and algorithm against the total number of calls received. HPA/QSurveillance system The HPA/QSurveillance national surveillance system was set up by the University of Nottingham, United Kingdom, and Egton Medical Information Systems (EMIS), a supplier of general practice computer systems, in collaboration with the HPA. It comprises a network of more than 3,500 general practices throughout the UK, covering more than 22 million patients (about 38% of the population [28]). Aggregated data on GP consultations for a range of indicators are automatically uploaded daily from GP practice systems to a central database. Data are routinely reported on a weekly basis; however, daily reporting is possible for specific incidents. Reports are provided at national or regional level (strategic health authority, SHA) and by local health district (primary care trust, PCT). Analysis of surveillance data NHS Direct call proportions for gastrointestinal syndromes (diarrhoea and vomiting) for the East Midlands region in England, where Northampton is situated, were examined during the outbreak (24 June – 18 July 2008) and compared with those for England and Wales. A series of control charts for diarrhoea calls are routinely used to monitor significant rises in the numbers of calls received. Control charts are calculated by assuming that calls follow a Poisson distribution with the total number of calls as an offset: a model is fitted to each region and symptom separately [29]. The model takes into account call variation caused by weekends, public holidays and the time of year – variables that can affect the number of calls received by NHS Direct. A value above the upper limit of the 99.5% confidence interval would be considered to be unusual. The seven-day moving average for diarrhoea calls was also monitored. The number and percentage of calls for diarrhoea in the East Midlands region were presented by call outcome and the number of calls in the Northampton (NN) postcode districts and in particular the number of calls in the NN11 and NN12 postcode districts, which were most affected by the incident. QSurveillance national consultation rates per 100,000 population for diarrhoea (in the age groups under five years, five years and over, and all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with rates for the same period in 2007 (data not presented). Consultation rates by region for 2008 for diarrhoea (all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with those for the East Midlands region. The gastroenteritis indicator includes all cases of diarrhoea and/or vomiting. Consultation rates and standardised incidence ratios (SIRs) – calculated using the UK as the standard population – for diarrhoea, gastroenteritis and vomiting were compared for the UK, Yorkshire and Humberside, East Midlands, Leicestershire, Northamptonshire and Rutland SHA, and Daventry and South Northants PCT, Northamptonshire Heartlands PCT and Northampton PCT. Yorkshire and Humberside was not an affected region but was included as a control. The area supplied by the Pitsford Reservoir included the three PCTs, which were all within the Leicestershire, Northamptonshire and Rutland SHA. The consultation rates and SIRs were compared for the period from week 16 to week 35 of 2008 in order to compare the rates before and after the Cryptosporidium exceedance, which took place in week 26. Estimates of excess numbers of cases of diarrhoea occurring during and following the Cryptosporidium outbreak were made by calculating the mean consultation rate over a period of five weeks before and after the incident (weeks 20–24 and weeks 31–35, respectively). For each of the three PCTs, the calculated mean rate was applied to the PCT population to estimate the number of cases that would be expected each week. The actual consultation rates for diarrhoea for weeks 25 to 30 were used to estimate the number of cases for the PCT population each week. The expected number of cases was subtracted from the estimated number of cases in the PCT population to give the estimated number of excess cases. Results HPA/NHS Direct surveillance system A peak in the number of calls for diarrhoea in the East Midlands was recorded in 25–26 June 2008, the period that coincided with the contamination incident and the associated media coverage (Figure 1). The neighbouring areas of the West Midlands, Yorkshire and the Humber, and East of England showed no increase in the number of calls for diarrhoea. The peak produced a control chart exceedance for calls for diarrhoea on 25 June 2008 (Figure 2), when the proportion of calls exceeded the upper limit of the 99.5% confidence interval. There were further confidence interval exceedances on 26 and 28 June (which were not control chart exceedances). There was no peak in calls for vomiting or control chart exceedance for these calls in the East Midlands. HPA/QSurveillance national surveillance system The East Midlands region had significantly high consultation rates for diarrhoea and gastroenteritis in week 25 (16–22 June), week 26 (23–29 June 2008, when the contamination incident was reported) and in the following four weeks. Within this region. Leicestershire, Northamptonshire and Rutland SHA had slightly raised consultation rates and significant SIRs across weeks 25 to 30 that were not seen in the neighbouring Trent SHA. At PCT level, all three of the PCTs in the area affected by the incident showed increased consultation rates for diarrhoea (Table 1) and gastroenteritis (Table 2) with SIRs significantly above the UK rate in week 26. Daventry and South Northants PCT also had a raised SIR for both indicators in week 25, and although Northamptonshire Heartlands and Northampton PCTs did not have SIRs significantly above that of the UK in week 25, the rise in consultation rates for diarrhoea and gastroenteritis began during week 25. In Northampton PCT, consultations for both diarrhoea and gastroenteritis peaked in the week following the contamination incident, week 27, returning to normal levels by week 30 (Figure 3A and 3B). A similar effect can be seen in Northampton Heartlands PCT. Daventry and South Northants PCT also showed an increase, but appeared to have consistently higher rates for both indicators. This was the area with the smallest population so the rates were more variable than in the other PCTs and we therefore interpreted these results with caution. The consultation rates for vomiting during weeks 25 to 30 in the East Midlands were not unusual at SHA or PCT level (data not presented). Discussion We have demonstrated the sensitivity of syndromic surveillance in detecting this small Cryptosporidium outbreak and the value of the surveillance in being able to describe the extent of its spread. Both the HPA/NHS Direct and HPA/QSurveillance systems showed demonstrable increases in calls and consultations for diarrhoea that were linked to the outbreak. QSurveillance consultations appeared to increase across the PCTs immediately affected but not in the surrounding area. Both the HPA/NHS Direct and HPA/QSurveillance systems showed a clear signal at the time of the incident and we were able to describe the extent of the impact on pre-primary care and primary care services. The HPA/QSurveillance system showed a rise in consultation rates for gastrointestinal symptoms that began the week before the outbreak, consistent with the period when Cryptosporidium was present in the water leaving the Pitsford Reservoir (19–24 June 2008) and with the onset of symptoms in the first outbreak case on 24 June. Although only 33 cases were identified by the outbreak investigation team, of which 23 were confirmed as having the outbreak Cryptosporidium strain, our syndromic surveillance data detected this limited outbreak. Data also suggested a more widespread increase in general gastrointestinal symptoms around the time of the outbreak, with an estimated 422 excess diarrhoea cases; these excess cases represented an increase of about 25% above normally expected levels. It is highly probable that a proportion of these excess cases may have resulted from the increased publicity surrounding the incident – for example, it is likely that media reports contributed to the large peak in calls detected by the HPA/NHS Direct surveillance system on the day the boil water notice was issued, and could also have impacted on the GP consultation rate. It has been previously shown that reporting of mumps cases is sensitive to media coverage, with a rise in clinically reported cases following newspaper reports [30]. A similar mechanism could account for some of the excess GP consultations as cases experiencing gastrointestinal symptoms may have been more likely to consult their GP, whereas in normal circumstances they would have cared for themselves at home. It is also possible that the surveillance shows outbreak-associated cases that did not come to the attention of the outbreak team, perhaps because symptoms were not sufficiently severe to warrant further investigation, or stool samples were not provided for testing. It is interesting to note that there was no demonstrable impact on the number of calls for vomiting (which is not a prominent clinical feature of cryptosporidiosis). Other common community-based pathogens such as norovirus and rotavirus were at low levels, as is normal for that time of year [31]. In this instance, public health authorities had already been alerted to a potential problem by the water company, although the extent of the outbreak was detected by syndromic surveillance. In 2003 the syndromic surveillance systems in the city of New York, United States, were able to detect an increase in diarrhoeal illness following a power outage when there was no other indication of citywide illness [32]. The New York system covers a population of nine million, but does not regularly detect localised outbreaks [33]. It has been shown previously that the HPA/NHS Direct surveillance system would be unlikely to detect a Cryptosporidium outbreak unless call volumes are high (72% chance of detection if nine-tenths of cases called NHS Direct) [29], although the value of syndromic surveillance for such outbreaks has been recognised [34]. The system detected the East Midlands Cryptosporidium outbreak that affected a smaller population than that covered by the New York system. The three PCTs affected have a combined population of around 600,000, of which just over half use GP practices reporting to QSurveillance, yet this syndromic surveillance system was able to describe an increase in consultation rates for diarrhoea and gastroenteritis around the time of the outbreak. Limitations of the data There was extensive media reporting of the incident that may have affected both the HPA/NHS Direct and HPA/QSurveillance systems and contributed to the increase in reported gastrointestinal symptoms around the time of the contamination incident. However, the rise in consultation rates for diarrhoea began before the outbreak had been detected and therefore cannot be attributed to media coverage. The HPA/NHS Direct and HPA/QSurveillance systems monitor general symptoms and so could only monitor the relevant symptoms of diarrhoea and vomiting. They are not able to detect Cryptosporidium cases, as this would require laboratory confirmation of diagnosis, so some of the estimated excess cases could be unconnected with this incident. This outbreak was discovered by other means but both the HPA/NHS Direct and HPA/ QSurveillance systems were able to describe the extent of the disease in the general population and provide reassurance that there was no widespread impact. Compared with other populations, older people and ethnic minorities are less likely to call NHS Direct [29], and although this should not prevent detection of gastrointestinal symptoms as a result of drinking water contamination as this would affect the whole population, this may reduce the signal from the system [35]. With such large surveillance systems, there will be ‘background noise’ in the data, so procedures must be in place to correctly interpret the data and set appropriate thresholds for action. Conclusion To our knowledge, this is the first time that PCT-level data from a syndromic surveillance system, the HPA/ QSurveillance national surveillance system, have been able to show the extent of such a limited outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide PCT-level data and this Cryptosporidium contamination incident demonstrates the potential usefulness of this system.

What caused the event?

1506

Value of syndromic surveillance in monitoring a focal waterborne outbreak due to an unusual Cryptosporidium genotype in Northamptonshire

The United Kingdom (UK) has several national syndromic surveillance systems. The Health Protection Agency (HPA)/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data from a national telephone helpline, while the HPA/ QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems. Data from both of these systems were used to monitor a local outbreak of cryptosporidiosis that occurred following Cryptosporidium oocyst contamination of drinking water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, in June 2008. There was a peak in the number of calls to NHS Direct concerning diarrhoea that coincided with the incident. QSurveillance data for the local areas affected by the outbreak showed a significant increase in GP consultations for diarrhoea and gastroenteritis in the week of the incident but there was no increase in consultations for vomiting. A total of 33 clinical cases of cryptosporidiosis were identified in the outbreak investigation, of which 23 were confirmed as infected with the outbreak strain. However, QSurveillance data suggest that there were an estimated 422 excess diarrhoea cases during the outbreak, an increase of about 25% over baseline weekly levels. To our knowledge, this is the first time that data from a syndromic surveillance system, the HPA/QSurveillance national surveillance system, have been able to show the extent of such a small outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide data at local health district (primary care trust) level. The Cryptosporidium contamination incident described demonstrates the potential usefulness of this information, as it is unusual for syndromic surveillance systems to be able to help monitor such a small-scale outbreak. Introduction As syndromic surveillance systems usually capture data already collected for other purposes, and monitor generic symptoms and/or clinically diagnosed disease, they provide information at an earlier stage of illness (compared with laboratory-confirmed diagnoses), so that action can be taken in time to substantially reduce the impact of disease. Some systems, for example, the Royal College of General Practitioners Weekly Returns Service, are now well established, with many years of historical data that allow monitoring of longer-term disease trends [1]. They have the ability to provide early warning of, for example, seasonal rises in influenza and can trigger public health action, such as a recommendation to prescribe antiviral drugs in line with national guidance [2-4]. They can also provide reassurance to incident response teams and the general public that an incident has not caused adverse health effects – for example, following an explosion at the Buncefield oil storage depot in Hemel Hempstead, United Kingdom (UK), in 2005, syndromic surveillance confirmed that there were no unusual rises in community-based morbidity linked to the incident [5]; following the eruption of the Eyjafjallajökull volcano in Iceland in April 2010 similar assurance was given about lack of impact on community morbidity [6]. Health departments are increasingly expected to monitor health effects of natural events such as heat wave or flooding, or implement surveillance – of which syndromic surveillance plays a major role – for mass gatherings such as the Olympics or football World Cup [7-9]. Systems in France, Australia and Taiwan use data from emergency departments [10-12], a Canadian system uses over-the-counter pharmacy sales [13,14], and in the Netherlands data from both syndromic and surrogate data sources, such as employee absence records and prescription medications dispensed by pharmacies, are included in surveillance systems [15,16]. Currently systems based on Internet searches via search engines or on queries submitted to medical websites are being developed [17,18]. In the UK, the HPA/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data collected from the NHS Direct telephone helpline [19], while the HPA/QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems [20]. The HPA Real-time Syndromic Surveillance Team is a small team that coordinates a number of syndromic surveillance systems within the HPA and takes a lead for syndromic surveillance in England [21]. This paper describes the support provided by the team to the local incident management team during a local cryptosporidiosis outbreak and shows the use of syndromic surveillance in monitoring the extent of an outbreak using the HPA/NHS Direct and HPA/QSurveillance national surveillance systems. Cryptosporidiosis Cryptosporidium is a protozoan parasite that can cause an infection in people, cattle and sometimes other animals [22]. Cryptosporidiosis is most common in children aged between one and five years, but it can affect all ages. Those with impaired immune systems are likely to be most seriously affected. Symptoms usually appear between three and 12 days after initial exposure and include watery diarrhoea, stomach pains, dehydration and fever. In its transmissible form, called an oocyst, the parasite is protected by an outer shell, which allows it to survive in the environment for a long time. Transmission occurs most often via the faeco-oral route through person-to-person or animalto-person contact, but people may also be infected by consuming contaminated water or food or by swimming in contaminated water. Although uncommon, the largest outbreaks have occurred following contamination of drinking water [23,24]. Normal chlorine disinfection procedures do not kill the oocysts, so they are removed by filtration and water companies carry out routine monitoring of treated water. Description of the incident On 25 June 2008 the local Health Protection Unit was informed by Anglian Water of an exceedence in the level of Cryptosporidium oocysts found in water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, during 19 to 24 June 2008 [25]. The reservoir supplied a population of more than 250,000 in the Northampton area. A notice advising people in the affected areas to boil all drinking water was issued on 25 June 2008 and public health messages were circulated to local health services and to the general public via the media. Those members of the public who were concerned about health risks associated with the incident were asked to ring NHS Direct for clinical advice [26]. The HPA wrote to local GPs and hospitals asking them to monitor potential patients for signs and symptoms of Cryptosporidium infection and to submit faecal specimens to the local hospital diagnostic laboratory if patients presented with diarrhoea. Samples from 34 patients where Cryptosporidium infection was identified were sent to the UK Cryptosporidium reference unit for typing. On 30 June 2008, the Cryptosporidium oocysts found in the reservoir water were confirmed as being of the rabbit genotype Cryptosporidium cuniculus [27]. Subsequently, a dead rabbit was found in a treated water tank at the water treatment works. The genotype of Cryptosporidium oocysts in the rabbit’s large bowel was indistinguishable from that of the oocysts found in the water [27]. After remediation of the water supply and distribution, the ‘boil water notice’ was lifted on 4 July and the following day the first case of cryptosporidiosis linked to the incident was identified by the reference laboratory (this case was infected with C. cuniculus). During the course of the outbreak (24 June – 18 July 2008, the dates of symptom onset in the first and last case, respectively), 23 cases of cryptosporidiosis were confirmed as being infected with C. cuniculus; one of the 23 was a secondary case. The HPA Real-time Syndromic Surveillance Team provided data in order to aid the response to this incident and the first syndromic surveillance report was circulated to the incident management team and other relevant people in the HPA on 27 June 2008. Data from the HPA/NHS Direct and HPA/QSurveillance systems were provided in a series of regular reports, initially daily and eventually weekly, until the final report on 21 August 2008. Each report included a summary interpretation and more detailed data on diarrhoea, gastroenteritis and vomiting indicators. Methods Surveillance systems HPA/NHS Direct surveillance system NHS Direct is a 24-hour nurse-led telephone helpline that provides health information and advice to the general public. Nurses use a computerised clinical decision support system – the NHS Clinical Assessment System (NHS CAS) – to handle calls. This assessment system uses approximately 200 computerised symptom-based clinical algorithms. Nurses assign the call to the most appropriate algorithm and the patient’s symptoms determine the questions asked and the action to be taken following the call (call outcome), which could be guidance on self-care or they could be referred to their GP or advised to attend a hospital emergency department. No attempt is made to provide a formal diagnosis. Daily NHS Direct data are received by the Real-time Syndromic Surveillance Team, where the number and type of calls received during the previous day are analysed and interpreted. Call proportions are calculated by age group and algorithm against the total number of calls received. HPA/QSurveillance system The HPA/QSurveillance national surveillance system was set up by the University of Nottingham, United Kingdom, and Egton Medical Information Systems (EMIS), a supplier of general practice computer systems, in collaboration with the HPA. It comprises a network of more than 3,500 general practices throughout the UK, covering more than 22 million patients (about 38% of the population [28]). Aggregated data on GP consultations for a range of indicators are automatically uploaded daily from GP practice systems to a central database. Data are routinely reported on a weekly basis; however, daily reporting is possible for specific incidents. Reports are provided at national or regional level (strategic health authority, SHA) and by local health district (primary care trust, PCT). Analysis of surveillance data NHS Direct call proportions for gastrointestinal syndromes (diarrhoea and vomiting) for the East Midlands region in England, where Northampton is situated, were examined during the outbreak (24 June – 18 July 2008) and compared with those for England and Wales. A series of control charts for diarrhoea calls are routinely used to monitor significant rises in the numbers of calls received. Control charts are calculated by assuming that calls follow a Poisson distribution with the total number of calls as an offset: a model is fitted to each region and symptom separately [29]. The model takes into account call variation caused by weekends, public holidays and the time of year – variables that can affect the number of calls received by NHS Direct. A value above the upper limit of the 99.5% confidence interval would be considered to be unusual. The seven-day moving average for diarrhoea calls was also monitored. The number and percentage of calls for diarrhoea in the East Midlands region were presented by call outcome and the number of calls in the Northampton (NN) postcode districts and in particular the number of calls in the NN11 and NN12 postcode districts, which were most affected by the incident. QSurveillance national consultation rates per 100,000 population for diarrhoea (in the age groups under five years, five years and over, and all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with rates for the same period in 2007 (data not presented). Consultation rates by region for 2008 for diarrhoea (all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with those for the East Midlands region. The gastroenteritis indicator includes all cases of diarrhoea and/or vomiting. Consultation rates and standardised incidence ratios (SIRs) – calculated using the UK as the standard population – for diarrhoea, gastroenteritis and vomiting were compared for the UK, Yorkshire and Humberside, East Midlands, Leicestershire, Northamptonshire and Rutland SHA, and Daventry and South Northants PCT, Northamptonshire Heartlands PCT and Northampton PCT. Yorkshire and Humberside was not an affected region but was included as a control. The area supplied by the Pitsford Reservoir included the three PCTs, which were all within the Leicestershire, Northamptonshire and Rutland SHA. The consultation rates and SIRs were compared for the period from week 16 to week 35 of 2008 in order to compare the rates before and after the Cryptosporidium exceedance, which took place in week 26. Estimates of excess numbers of cases of diarrhoea occurring during and following the Cryptosporidium outbreak were made by calculating the mean consultation rate over a period of five weeks before and after the incident (weeks 20–24 and weeks 31–35, respectively). For each of the three PCTs, the calculated mean rate was applied to the PCT population to estimate the number of cases that would be expected each week. The actual consultation rates for diarrhoea for weeks 25 to 30 were used to estimate the number of cases for the PCT population each week. The expected number of cases was subtracted from the estimated number of cases in the PCT population to give the estimated number of excess cases. Results HPA/NHS Direct surveillance system A peak in the number of calls for diarrhoea in the East Midlands was recorded in 25–26 June 2008, the period that coincided with the contamination incident and the associated media coverage (Figure 1). The neighbouring areas of the West Midlands, Yorkshire and the Humber, and East of England showed no increase in the number of calls for diarrhoea. The peak produced a control chart exceedance for calls for diarrhoea on 25 June 2008 (Figure 2), when the proportion of calls exceeded the upper limit of the 99.5% confidence interval. There were further confidence interval exceedances on 26 and 28 June (which were not control chart exceedances). There was no peak in calls for vomiting or control chart exceedance for these calls in the East Midlands. HPA/QSurveillance national surveillance system The East Midlands region had significantly high consultation rates for diarrhoea and gastroenteritis in week 25 (16–22 June), week 26 (23–29 June 2008, when the contamination incident was reported) and in the following four weeks. Within this region. Leicestershire, Northamptonshire and Rutland SHA had slightly raised consultation rates and significant SIRs across weeks 25 to 30 that were not seen in the neighbouring Trent SHA. At PCT level, all three of the PCTs in the area affected by the incident showed increased consultation rates for diarrhoea (Table 1) and gastroenteritis (Table 2) with SIRs significantly above the UK rate in week 26. Daventry and South Northants PCT also had a raised SIR for both indicators in week 25, and although Northamptonshire Heartlands and Northampton PCTs did not have SIRs significantly above that of the UK in week 25, the rise in consultation rates for diarrhoea and gastroenteritis began during week 25. In Northampton PCT, consultations for both diarrhoea and gastroenteritis peaked in the week following the contamination incident, week 27, returning to normal levels by week 30 (Figure 3A and 3B). A similar effect can be seen in Northampton Heartlands PCT. Daventry and South Northants PCT also showed an increase, but appeared to have consistently higher rates for both indicators. This was the area with the smallest population so the rates were more variable than in the other PCTs and we therefore interpreted these results with caution. The consultation rates for vomiting during weeks 25 to 30 in the East Midlands were not unusual at SHA or PCT level (data not presented). Discussion We have demonstrated the sensitivity of syndromic surveillance in detecting this small Cryptosporidium outbreak and the value of the surveillance in being able to describe the extent of its spread. Both the HPA/NHS Direct and HPA/QSurveillance systems showed demonstrable increases in calls and consultations for diarrhoea that were linked to the outbreak. QSurveillance consultations appeared to increase across the PCTs immediately affected but not in the surrounding area. Both the HPA/NHS Direct and HPA/QSurveillance systems showed a clear signal at the time of the incident and we were able to describe the extent of the impact on pre-primary care and primary care services. The HPA/QSurveillance system showed a rise in consultation rates for gastrointestinal symptoms that began the week before the outbreak, consistent with the period when Cryptosporidium was present in the water leaving the Pitsford Reservoir (19–24 June 2008) and with the onset of symptoms in the first outbreak case on 24 June. Although only 33 cases were identified by the outbreak investigation team, of which 23 were confirmed as having the outbreak Cryptosporidium strain, our syndromic surveillance data detected this limited outbreak. Data also suggested a more widespread increase in general gastrointestinal symptoms around the time of the outbreak, with an estimated 422 excess diarrhoea cases; these excess cases represented an increase of about 25% above normally expected levels. It is highly probable that a proportion of these excess cases may have resulted from the increased publicity surrounding the incident – for example, it is likely that media reports contributed to the large peak in calls detected by the HPA/NHS Direct surveillance system on the day the boil water notice was issued, and could also have impacted on the GP consultation rate. It has been previously shown that reporting of mumps cases is sensitive to media coverage, with a rise in clinically reported cases following newspaper reports [30]. A similar mechanism could account for some of the excess GP consultations as cases experiencing gastrointestinal symptoms may have been more likely to consult their GP, whereas in normal circumstances they would have cared for themselves at home. It is also possible that the surveillance shows outbreak-associated cases that did not come to the attention of the outbreak team, perhaps because symptoms were not sufficiently severe to warrant further investigation, or stool samples were not provided for testing. It is interesting to note that there was no demonstrable impact on the number of calls for vomiting (which is not a prominent clinical feature of cryptosporidiosis). Other common community-based pathogens such as norovirus and rotavirus were at low levels, as is normal for that time of year [31]. In this instance, public health authorities had already been alerted to a potential problem by the water company, although the extent of the outbreak was detected by syndromic surveillance. In 2003 the syndromic surveillance systems in the city of New York, United States, were able to detect an increase in diarrhoeal illness following a power outage when there was no other indication of citywide illness [32]. The New York system covers a population of nine million, but does not regularly detect localised outbreaks [33]. It has been shown previously that the HPA/NHS Direct surveillance system would be unlikely to detect a Cryptosporidium outbreak unless call volumes are high (72% chance of detection if nine-tenths of cases called NHS Direct) [29], although the value of syndromic surveillance for such outbreaks has been recognised [34]. The system detected the East Midlands Cryptosporidium outbreak that affected a smaller population than that covered by the New York system. The three PCTs affected have a combined population of around 600,000, of which just over half use GP practices reporting to QSurveillance, yet this syndromic surveillance system was able to describe an increase in consultation rates for diarrhoea and gastroenteritis around the time of the outbreak. Limitations of the data There was extensive media reporting of the incident that may have affected both the HPA/NHS Direct and HPA/QSurveillance systems and contributed to the increase in reported gastrointestinal symptoms around the time of the contamination incident. However, the rise in consultation rates for diarrhoea began before the outbreak had been detected and therefore cannot be attributed to media coverage. The HPA/NHS Direct and HPA/QSurveillance systems monitor general symptoms and so could only monitor the relevant symptoms of diarrhoea and vomiting. They are not able to detect Cryptosporidium cases, as this would require laboratory confirmation of diagnosis, so some of the estimated excess cases could be unconnected with this incident. This outbreak was discovered by other means but both the HPA/NHS Direct and HPA/ QSurveillance systems were able to describe the extent of the disease in the general population and provide reassurance that there was no widespread impact. Compared with other populations, older people and ethnic minorities are less likely to call NHS Direct [29], and although this should not prevent detection of gastrointestinal symptoms as a result of drinking water contamination as this would affect the whole population, this may reduce the signal from the system [35]. With such large surveillance systems, there will be ‘background noise’ in the data, so procedures must be in place to correctly interpret the data and set appropriate thresholds for action. Conclusion To our knowledge, this is the first time that PCT-level data from a syndromic surveillance system, the HPA/ QSurveillance national surveillance system, have been able to show the extent of such a limited outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide PCT-level data and this Cryptosporidium contamination incident demonstrates the potential usefulness of this system.

What was the cause of the event?

1507

Value of syndromic surveillance in monitoring a focal waterborne outbreak due to an unusual Cryptosporidium genotype in Northamptonshire

The United Kingdom (UK) has several national syndromic surveillance systems. The Health Protection Agency (HPA)/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data from a national telephone helpline, while the HPA/ QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems. Data from both of these systems were used to monitor a local outbreak of cryptosporidiosis that occurred following Cryptosporidium oocyst contamination of drinking water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, in June 2008. There was a peak in the number of calls to NHS Direct concerning diarrhoea that coincided with the incident. QSurveillance data for the local areas affected by the outbreak showed a significant increase in GP consultations for diarrhoea and gastroenteritis in the week of the incident but there was no increase in consultations for vomiting. A total of 33 clinical cases of cryptosporidiosis were identified in the outbreak investigation, of which 23 were confirmed as infected with the outbreak strain. However, QSurveillance data suggest that there were an estimated 422 excess diarrhoea cases during the outbreak, an increase of about 25% over baseline weekly levels. To our knowledge, this is the first time that data from a syndromic surveillance system, the HPA/QSurveillance national surveillance system, have been able to show the extent of such a small outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide data at local health district (primary care trust) level. The Cryptosporidium contamination incident described demonstrates the potential usefulness of this information, as it is unusual for syndromic surveillance systems to be able to help monitor such a small-scale outbreak. Introduction As syndromic surveillance systems usually capture data already collected for other purposes, and monitor generic symptoms and/or clinically diagnosed disease, they provide information at an earlier stage of illness (compared with laboratory-confirmed diagnoses), so that action can be taken in time to substantially reduce the impact of disease. Some systems, for example, the Royal College of General Practitioners Weekly Returns Service, are now well established, with many years of historical data that allow monitoring of longer-term disease trends [1]. They have the ability to provide early warning of, for example, seasonal rises in influenza and can trigger public health action, such as a recommendation to prescribe antiviral drugs in line with national guidance [2-4]. They can also provide reassurance to incident response teams and the general public that an incident has not caused adverse health effects – for example, following an explosion at the Buncefield oil storage depot in Hemel Hempstead, United Kingdom (UK), in 2005, syndromic surveillance confirmed that there were no unusual rises in community-based morbidity linked to the incident [5]; following the eruption of the Eyjafjallajökull volcano in Iceland in April 2010 similar assurance was given about lack of impact on community morbidity [6]. Health departments are increasingly expected to monitor health effects of natural events such as heat wave or flooding, or implement surveillance – of which syndromic surveillance plays a major role – for mass gatherings such as the Olympics or football World Cup [7-9]. Systems in France, Australia and Taiwan use data from emergency departments [10-12], a Canadian system uses over-the-counter pharmacy sales [13,14], and in the Netherlands data from both syndromic and surrogate data sources, such as employee absence records and prescription medications dispensed by pharmacies, are included in surveillance systems [15,16]. Currently systems based on Internet searches via search engines or on queries submitted to medical websites are being developed [17,18]. In the UK, the HPA/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data collected from the NHS Direct telephone helpline [19], while the HPA/QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems [20]. The HPA Real-time Syndromic Surveillance Team is a small team that coordinates a number of syndromic surveillance systems within the HPA and takes a lead for syndromic surveillance in England [21]. This paper describes the support provided by the team to the local incident management team during a local cryptosporidiosis outbreak and shows the use of syndromic surveillance in monitoring the extent of an outbreak using the HPA/NHS Direct and HPA/QSurveillance national surveillance systems. Cryptosporidiosis Cryptosporidium is a protozoan parasite that can cause an infection in people, cattle and sometimes other animals [22]. Cryptosporidiosis is most common in children aged between one and five years, but it can affect all ages. Those with impaired immune systems are likely to be most seriously affected. Symptoms usually appear between three and 12 days after initial exposure and include watery diarrhoea, stomach pains, dehydration and fever. In its transmissible form, called an oocyst, the parasite is protected by an outer shell, which allows it to survive in the environment for a long time. Transmission occurs most often via the faeco-oral route through person-to-person or animalto-person contact, but people may also be infected by consuming contaminated water or food or by swimming in contaminated water. Although uncommon, the largest outbreaks have occurred following contamination of drinking water [23,24]. Normal chlorine disinfection procedures do not kill the oocysts, so they are removed by filtration and water companies carry out routine monitoring of treated water. Description of the incident On 25 June 2008 the local Health Protection Unit was informed by Anglian Water of an exceedence in the level of Cryptosporidium oocysts found in water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, during 19 to 24 June 2008 [25]. The reservoir supplied a population of more than 250,000 in the Northampton area. A notice advising people in the affected areas to boil all drinking water was issued on 25 June 2008 and public health messages were circulated to local health services and to the general public via the media. Those members of the public who were concerned about health risks associated with the incident were asked to ring NHS Direct for clinical advice [26]. The HPA wrote to local GPs and hospitals asking them to monitor potential patients for signs and symptoms of Cryptosporidium infection and to submit faecal specimens to the local hospital diagnostic laboratory if patients presented with diarrhoea. Samples from 34 patients where Cryptosporidium infection was identified were sent to the UK Cryptosporidium reference unit for typing. On 30 June 2008, the Cryptosporidium oocysts found in the reservoir water were confirmed as being of the rabbit genotype Cryptosporidium cuniculus [27]. Subsequently, a dead rabbit was found in a treated water tank at the water treatment works. The genotype of Cryptosporidium oocysts in the rabbit’s large bowel was indistinguishable from that of the oocysts found in the water [27]. After remediation of the water supply and distribution, the ‘boil water notice’ was lifted on 4 July and the following day the first case of cryptosporidiosis linked to the incident was identified by the reference laboratory (this case was infected with C. cuniculus). During the course of the outbreak (24 June – 18 July 2008, the dates of symptom onset in the first and last case, respectively), 23 cases of cryptosporidiosis were confirmed as being infected with C. cuniculus; one of the 23 was a secondary case. The HPA Real-time Syndromic Surveillance Team provided data in order to aid the response to this incident and the first syndromic surveillance report was circulated to the incident management team and other relevant people in the HPA on 27 June 2008. Data from the HPA/NHS Direct and HPA/QSurveillance systems were provided in a series of regular reports, initially daily and eventually weekly, until the final report on 21 August 2008. Each report included a summary interpretation and more detailed data on diarrhoea, gastroenteritis and vomiting indicators. Methods Surveillance systems HPA/NHS Direct surveillance system NHS Direct is a 24-hour nurse-led telephone helpline that provides health information and advice to the general public. Nurses use a computerised clinical decision support system – the NHS Clinical Assessment System (NHS CAS) – to handle calls. This assessment system uses approximately 200 computerised symptom-based clinical algorithms. Nurses assign the call to the most appropriate algorithm and the patient’s symptoms determine the questions asked and the action to be taken following the call (call outcome), which could be guidance on self-care or they could be referred to their GP or advised to attend a hospital emergency department. No attempt is made to provide a formal diagnosis. Daily NHS Direct data are received by the Real-time Syndromic Surveillance Team, where the number and type of calls received during the previous day are analysed and interpreted. Call proportions are calculated by age group and algorithm against the total number of calls received. HPA/QSurveillance system The HPA/QSurveillance national surveillance system was set up by the University of Nottingham, United Kingdom, and Egton Medical Information Systems (EMIS), a supplier of general practice computer systems, in collaboration with the HPA. It comprises a network of more than 3,500 general practices throughout the UK, covering more than 22 million patients (about 38% of the population [28]). Aggregated data on GP consultations for a range of indicators are automatically uploaded daily from GP practice systems to a central database. Data are routinely reported on a weekly basis; however, daily reporting is possible for specific incidents. Reports are provided at national or regional level (strategic health authority, SHA) and by local health district (primary care trust, PCT). Analysis of surveillance data NHS Direct call proportions for gastrointestinal syndromes (diarrhoea and vomiting) for the East Midlands region in England, where Northampton is situated, were examined during the outbreak (24 June – 18 July 2008) and compared with those for England and Wales. A series of control charts for diarrhoea calls are routinely used to monitor significant rises in the numbers of calls received. Control charts are calculated by assuming that calls follow a Poisson distribution with the total number of calls as an offset: a model is fitted to each region and symptom separately [29]. The model takes into account call variation caused by weekends, public holidays and the time of year – variables that can affect the number of calls received by NHS Direct. A value above the upper limit of the 99.5% confidence interval would be considered to be unusual. The seven-day moving average for diarrhoea calls was also monitored. The number and percentage of calls for diarrhoea in the East Midlands region were presented by call outcome and the number of calls in the Northampton (NN) postcode districts and in particular the number of calls in the NN11 and NN12 postcode districts, which were most affected by the incident. QSurveillance national consultation rates per 100,000 population for diarrhoea (in the age groups under five years, five years and over, and all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with rates for the same period in 2007 (data not presented). Consultation rates by region for 2008 for diarrhoea (all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with those for the East Midlands region. The gastroenteritis indicator includes all cases of diarrhoea and/or vomiting. Consultation rates and standardised incidence ratios (SIRs) – calculated using the UK as the standard population – for diarrhoea, gastroenteritis and vomiting were compared for the UK, Yorkshire and Humberside, East Midlands, Leicestershire, Northamptonshire and Rutland SHA, and Daventry and South Northants PCT, Northamptonshire Heartlands PCT and Northampton PCT. Yorkshire and Humberside was not an affected region but was included as a control. The area supplied by the Pitsford Reservoir included the three PCTs, which were all within the Leicestershire, Northamptonshire and Rutland SHA. The consultation rates and SIRs were compared for the period from week 16 to week 35 of 2008 in order to compare the rates before and after the Cryptosporidium exceedance, which took place in week 26. Estimates of excess numbers of cases of diarrhoea occurring during and following the Cryptosporidium outbreak were made by calculating the mean consultation rate over a period of five weeks before and after the incident (weeks 20–24 and weeks 31–35, respectively). For each of the three PCTs, the calculated mean rate was applied to the PCT population to estimate the number of cases that would be expected each week. The actual consultation rates for diarrhoea for weeks 25 to 30 were used to estimate the number of cases for the PCT population each week. The expected number of cases was subtracted from the estimated number of cases in the PCT population to give the estimated number of excess cases. Results HPA/NHS Direct surveillance system A peak in the number of calls for diarrhoea in the East Midlands was recorded in 25–26 June 2008, the period that coincided with the contamination incident and the associated media coverage (Figure 1). The neighbouring areas of the West Midlands, Yorkshire and the Humber, and East of England showed no increase in the number of calls for diarrhoea. The peak produced a control chart exceedance for calls for diarrhoea on 25 June 2008 (Figure 2), when the proportion of calls exceeded the upper limit of the 99.5% confidence interval. There were further confidence interval exceedances on 26 and 28 June (which were not control chart exceedances). There was no peak in calls for vomiting or control chart exceedance for these calls in the East Midlands. HPA/QSurveillance national surveillance system The East Midlands region had significantly high consultation rates for diarrhoea and gastroenteritis in week 25 (16–22 June), week 26 (23–29 June 2008, when the contamination incident was reported) and in the following four weeks. Within this region. Leicestershire, Northamptonshire and Rutland SHA had slightly raised consultation rates and significant SIRs across weeks 25 to 30 that were not seen in the neighbouring Trent SHA. At PCT level, all three of the PCTs in the area affected by the incident showed increased consultation rates for diarrhoea (Table 1) and gastroenteritis (Table 2) with SIRs significantly above the UK rate in week 26. Daventry and South Northants PCT also had a raised SIR for both indicators in week 25, and although Northamptonshire Heartlands and Northampton PCTs did not have SIRs significantly above that of the UK in week 25, the rise in consultation rates for diarrhoea and gastroenteritis began during week 25. In Northampton PCT, consultations for both diarrhoea and gastroenteritis peaked in the week following the contamination incident, week 27, returning to normal levels by week 30 (Figure 3A and 3B). A similar effect can be seen in Northampton Heartlands PCT. Daventry and South Northants PCT also showed an increase, but appeared to have consistently higher rates for both indicators. This was the area with the smallest population so the rates were more variable than in the other PCTs and we therefore interpreted these results with caution. The consultation rates for vomiting during weeks 25 to 30 in the East Midlands were not unusual at SHA or PCT level (data not presented). Discussion We have demonstrated the sensitivity of syndromic surveillance in detecting this small Cryptosporidium outbreak and the value of the surveillance in being able to describe the extent of its spread. Both the HPA/NHS Direct and HPA/QSurveillance systems showed demonstrable increases in calls and consultations for diarrhoea that were linked to the outbreak. QSurveillance consultations appeared to increase across the PCTs immediately affected but not in the surrounding area. Both the HPA/NHS Direct and HPA/QSurveillance systems showed a clear signal at the time of the incident and we were able to describe the extent of the impact on pre-primary care and primary care services. The HPA/QSurveillance system showed a rise in consultation rates for gastrointestinal symptoms that began the week before the outbreak, consistent with the period when Cryptosporidium was present in the water leaving the Pitsford Reservoir (19–24 June 2008) and with the onset of symptoms in the first outbreak case on 24 June. Although only 33 cases were identified by the outbreak investigation team, of which 23 were confirmed as having the outbreak Cryptosporidium strain, our syndromic surveillance data detected this limited outbreak. Data also suggested a more widespread increase in general gastrointestinal symptoms around the time of the outbreak, with an estimated 422 excess diarrhoea cases; these excess cases represented an increase of about 25% above normally expected levels. It is highly probable that a proportion of these excess cases may have resulted from the increased publicity surrounding the incident – for example, it is likely that media reports contributed to the large peak in calls detected by the HPA/NHS Direct surveillance system on the day the boil water notice was issued, and could also have impacted on the GP consultation rate. It has been previously shown that reporting of mumps cases is sensitive to media coverage, with a rise in clinically reported cases following newspaper reports [30]. A similar mechanism could account for some of the excess GP consultations as cases experiencing gastrointestinal symptoms may have been more likely to consult their GP, whereas in normal circumstances they would have cared for themselves at home. It is also possible that the surveillance shows outbreak-associated cases that did not come to the attention of the outbreak team, perhaps because symptoms were not sufficiently severe to warrant further investigation, or stool samples were not provided for testing. It is interesting to note that there was no demonstrable impact on the number of calls for vomiting (which is not a prominent clinical feature of cryptosporidiosis). Other common community-based pathogens such as norovirus and rotavirus were at low levels, as is normal for that time of year [31]. In this instance, public health authorities had already been alerted to a potential problem by the water company, although the extent of the outbreak was detected by syndromic surveillance. In 2003 the syndromic surveillance systems in the city of New York, United States, were able to detect an increase in diarrhoeal illness following a power outage when there was no other indication of citywide illness [32]. The New York system covers a population of nine million, but does not regularly detect localised outbreaks [33]. It has been shown previously that the HPA/NHS Direct surveillance system would be unlikely to detect a Cryptosporidium outbreak unless call volumes are high (72% chance of detection if nine-tenths of cases called NHS Direct) [29], although the value of syndromic surveillance for such outbreaks has been recognised [34]. The system detected the East Midlands Cryptosporidium outbreak that affected a smaller population than that covered by the New York system. The three PCTs affected have a combined population of around 600,000, of which just over half use GP practices reporting to QSurveillance, yet this syndromic surveillance system was able to describe an increase in consultation rates for diarrhoea and gastroenteritis around the time of the outbreak. Limitations of the data There was extensive media reporting of the incident that may have affected both the HPA/NHS Direct and HPA/QSurveillance systems and contributed to the increase in reported gastrointestinal symptoms around the time of the contamination incident. However, the rise in consultation rates for diarrhoea began before the outbreak had been detected and therefore cannot be attributed to media coverage. The HPA/NHS Direct and HPA/QSurveillance systems monitor general symptoms and so could only monitor the relevant symptoms of diarrhoea and vomiting. They are not able to detect Cryptosporidium cases, as this would require laboratory confirmation of diagnosis, so some of the estimated excess cases could be unconnected with this incident. This outbreak was discovered by other means but both the HPA/NHS Direct and HPA/ QSurveillance systems were able to describe the extent of the disease in the general population and provide reassurance that there was no widespread impact. Compared with other populations, older people and ethnic minorities are less likely to call NHS Direct [29], and although this should not prevent detection of gastrointestinal symptoms as a result of drinking water contamination as this would affect the whole population, this may reduce the signal from the system [35]. With such large surveillance systems, there will be ‘background noise’ in the data, so procedures must be in place to correctly interpret the data and set appropriate thresholds for action. Conclusion To our knowledge, this is the first time that PCT-level data from a syndromic surveillance system, the HPA/ QSurveillance national surveillance system, have been able to show the extent of such a limited outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide PCT-level data and this Cryptosporidium contamination incident demonstrates the potential usefulness of this system.

What source started the event?

1508

Value of syndromic surveillance in monitoring a focal waterborne outbreak due to an unusual Cryptosporidium genotype in Northamptonshire

The United Kingdom (UK) has several national syndromic surveillance systems. The Health Protection Agency (HPA)/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data from a national telephone helpline, while the HPA/ QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems. Data from both of these systems were used to monitor a local outbreak of cryptosporidiosis that occurred following Cryptosporidium oocyst contamination of drinking water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, in June 2008. There was a peak in the number of calls to NHS Direct concerning diarrhoea that coincided with the incident. QSurveillance data for the local areas affected by the outbreak showed a significant increase in GP consultations for diarrhoea and gastroenteritis in the week of the incident but there was no increase in consultations for vomiting. A total of 33 clinical cases of cryptosporidiosis were identified in the outbreak investigation, of which 23 were confirmed as infected with the outbreak strain. However, QSurveillance data suggest that there were an estimated 422 excess diarrhoea cases during the outbreak, an increase of about 25% over baseline weekly levels. To our knowledge, this is the first time that data from a syndromic surveillance system, the HPA/QSurveillance national surveillance system, have been able to show the extent of such a small outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide data at local health district (primary care trust) level. The Cryptosporidium contamination incident described demonstrates the potential usefulness of this information, as it is unusual for syndromic surveillance systems to be able to help monitor such a small-scale outbreak. Introduction As syndromic surveillance systems usually capture data already collected for other purposes, and monitor generic symptoms and/or clinically diagnosed disease, they provide information at an earlier stage of illness (compared with laboratory-confirmed diagnoses), so that action can be taken in time to substantially reduce the impact of disease. Some systems, for example, the Royal College of General Practitioners Weekly Returns Service, are now well established, with many years of historical data that allow monitoring of longer-term disease trends [1]. They have the ability to provide early warning of, for example, seasonal rises in influenza and can trigger public health action, such as a recommendation to prescribe antiviral drugs in line with national guidance [2-4]. They can also provide reassurance to incident response teams and the general public that an incident has not caused adverse health effects – for example, following an explosion at the Buncefield oil storage depot in Hemel Hempstead, United Kingdom (UK), in 2005, syndromic surveillance confirmed that there were no unusual rises in community-based morbidity linked to the incident [5]; following the eruption of the Eyjafjallajökull volcano in Iceland in April 2010 similar assurance was given about lack of impact on community morbidity [6]. Health departments are increasingly expected to monitor health effects of natural events such as heat wave or flooding, or implement surveillance – of which syndromic surveillance plays a major role – for mass gatherings such as the Olympics or football World Cup [7-9]. Systems in France, Australia and Taiwan use data from emergency departments [10-12], a Canadian system uses over-the-counter pharmacy sales [13,14], and in the Netherlands data from both syndromic and surrogate data sources, such as employee absence records and prescription medications dispensed by pharmacies, are included in surveillance systems [15,16]. Currently systems based on Internet searches via search engines or on queries submitted to medical websites are being developed [17,18]. In the UK, the HPA/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data collected from the NHS Direct telephone helpline [19], while the HPA/QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems [20]. The HPA Real-time Syndromic Surveillance Team is a small team that coordinates a number of syndromic surveillance systems within the HPA and takes a lead for syndromic surveillance in England [21]. This paper describes the support provided by the team to the local incident management team during a local cryptosporidiosis outbreak and shows the use of syndromic surveillance in monitoring the extent of an outbreak using the HPA/NHS Direct and HPA/QSurveillance national surveillance systems. Cryptosporidiosis Cryptosporidium is a protozoan parasite that can cause an infection in people, cattle and sometimes other animals [22]. Cryptosporidiosis is most common in children aged between one and five years, but it can affect all ages. Those with impaired immune systems are likely to be most seriously affected. Symptoms usually appear between three and 12 days after initial exposure and include watery diarrhoea, stomach pains, dehydration and fever. In its transmissible form, called an oocyst, the parasite is protected by an outer shell, which allows it to survive in the environment for a long time. Transmission occurs most often via the faeco-oral route through person-to-person or animalto-person contact, but people may also be infected by consuming contaminated water or food or by swimming in contaminated water. Although uncommon, the largest outbreaks have occurred following contamination of drinking water [23,24]. Normal chlorine disinfection procedures do not kill the oocysts, so they are removed by filtration and water companies carry out routine monitoring of treated water. Description of the incident On 25 June 2008 the local Health Protection Unit was informed by Anglian Water of an exceedence in the level of Cryptosporidium oocysts found in water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, during 19 to 24 June 2008 [25]. The reservoir supplied a population of more than 250,000 in the Northampton area. A notice advising people in the affected areas to boil all drinking water was issued on 25 June 2008 and public health messages were circulated to local health services and to the general public via the media. Those members of the public who were concerned about health risks associated with the incident were asked to ring NHS Direct for clinical advice [26]. The HPA wrote to local GPs and hospitals asking them to monitor potential patients for signs and symptoms of Cryptosporidium infection and to submit faecal specimens to the local hospital diagnostic laboratory if patients presented with diarrhoea. Samples from 34 patients where Cryptosporidium infection was identified were sent to the UK Cryptosporidium reference unit for typing. On 30 June 2008, the Cryptosporidium oocysts found in the reservoir water were confirmed as being of the rabbit genotype Cryptosporidium cuniculus [27]. Subsequently, a dead rabbit was found in a treated water tank at the water treatment works. The genotype of Cryptosporidium oocysts in the rabbit’s large bowel was indistinguishable from that of the oocysts found in the water [27]. After remediation of the water supply and distribution, the ‘boil water notice’ was lifted on 4 July and the following day the first case of cryptosporidiosis linked to the incident was identified by the reference laboratory (this case was infected with C. cuniculus). During the course of the outbreak (24 June – 18 July 2008, the dates of symptom onset in the first and last case, respectively), 23 cases of cryptosporidiosis were confirmed as being infected with C. cuniculus; one of the 23 was a secondary case. The HPA Real-time Syndromic Surveillance Team provided data in order to aid the response to this incident and the first syndromic surveillance report was circulated to the incident management team and other relevant people in the HPA on 27 June 2008. Data from the HPA/NHS Direct and HPA/QSurveillance systems were provided in a series of regular reports, initially daily and eventually weekly, until the final report on 21 August 2008. Each report included a summary interpretation and more detailed data on diarrhoea, gastroenteritis and vomiting indicators. Methods Surveillance systems HPA/NHS Direct surveillance system NHS Direct is a 24-hour nurse-led telephone helpline that provides health information and advice to the general public. Nurses use a computerised clinical decision support system – the NHS Clinical Assessment System (NHS CAS) – to handle calls. This assessment system uses approximately 200 computerised symptom-based clinical algorithms. Nurses assign the call to the most appropriate algorithm and the patient’s symptoms determine the questions asked and the action to be taken following the call (call outcome), which could be guidance on self-care or they could be referred to their GP or advised to attend a hospital emergency department. No attempt is made to provide a formal diagnosis. Daily NHS Direct data are received by the Real-time Syndromic Surveillance Team, where the number and type of calls received during the previous day are analysed and interpreted. Call proportions are calculated by age group and algorithm against the total number of calls received. HPA/QSurveillance system The HPA/QSurveillance national surveillance system was set up by the University of Nottingham, United Kingdom, and Egton Medical Information Systems (EMIS), a supplier of general practice computer systems, in collaboration with the HPA. It comprises a network of more than 3,500 general practices throughout the UK, covering more than 22 million patients (about 38% of the population [28]). Aggregated data on GP consultations for a range of indicators are automatically uploaded daily from GP practice systems to a central database. Data are routinely reported on a weekly basis; however, daily reporting is possible for specific incidents. Reports are provided at national or regional level (strategic health authority, SHA) and by local health district (primary care trust, PCT). Analysis of surveillance data NHS Direct call proportions for gastrointestinal syndromes (diarrhoea and vomiting) for the East Midlands region in England, where Northampton is situated, were examined during the outbreak (24 June – 18 July 2008) and compared with those for England and Wales. A series of control charts for diarrhoea calls are routinely used to monitor significant rises in the numbers of calls received. Control charts are calculated by assuming that calls follow a Poisson distribution with the total number of calls as an offset: a model is fitted to each region and symptom separately [29]. The model takes into account call variation caused by weekends, public holidays and the time of year – variables that can affect the number of calls received by NHS Direct. A value above the upper limit of the 99.5% confidence interval would be considered to be unusual. The seven-day moving average for diarrhoea calls was also monitored. The number and percentage of calls for diarrhoea in the East Midlands region were presented by call outcome and the number of calls in the Northampton (NN) postcode districts and in particular the number of calls in the NN11 and NN12 postcode districts, which were most affected by the incident. QSurveillance national consultation rates per 100,000 population for diarrhoea (in the age groups under five years, five years and over, and all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with rates for the same period in 2007 (data not presented). Consultation rates by region for 2008 for diarrhoea (all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with those for the East Midlands region. The gastroenteritis indicator includes all cases of diarrhoea and/or vomiting. Consultation rates and standardised incidence ratios (SIRs) – calculated using the UK as the standard population – for diarrhoea, gastroenteritis and vomiting were compared for the UK, Yorkshire and Humberside, East Midlands, Leicestershire, Northamptonshire and Rutland SHA, and Daventry and South Northants PCT, Northamptonshire Heartlands PCT and Northampton PCT. Yorkshire and Humberside was not an affected region but was included as a control. The area supplied by the Pitsford Reservoir included the three PCTs, which were all within the Leicestershire, Northamptonshire and Rutland SHA. The consultation rates and SIRs were compared for the period from week 16 to week 35 of 2008 in order to compare the rates before and after the Cryptosporidium exceedance, which took place in week 26. Estimates of excess numbers of cases of diarrhoea occurring during and following the Cryptosporidium outbreak were made by calculating the mean consultation rate over a period of five weeks before and after the incident (weeks 20–24 and weeks 31–35, respectively). For each of the three PCTs, the calculated mean rate was applied to the PCT population to estimate the number of cases that would be expected each week. The actual consultation rates for diarrhoea for weeks 25 to 30 were used to estimate the number of cases for the PCT population each week. The expected number of cases was subtracted from the estimated number of cases in the PCT population to give the estimated number of excess cases. Results HPA/NHS Direct surveillance system A peak in the number of calls for diarrhoea in the East Midlands was recorded in 25–26 June 2008, the period that coincided with the contamination incident and the associated media coverage (Figure 1). The neighbouring areas of the West Midlands, Yorkshire and the Humber, and East of England showed no increase in the number of calls for diarrhoea. The peak produced a control chart exceedance for calls for diarrhoea on 25 June 2008 (Figure 2), when the proportion of calls exceeded the upper limit of the 99.5% confidence interval. There were further confidence interval exceedances on 26 and 28 June (which were not control chart exceedances). There was no peak in calls for vomiting or control chart exceedance for these calls in the East Midlands. HPA/QSurveillance national surveillance system The East Midlands region had significantly high consultation rates for diarrhoea and gastroenteritis in week 25 (16–22 June), week 26 (23–29 June 2008, when the contamination incident was reported) and in the following four weeks. Within this region. Leicestershire, Northamptonshire and Rutland SHA had slightly raised consultation rates and significant SIRs across weeks 25 to 30 that were not seen in the neighbouring Trent SHA. At PCT level, all three of the PCTs in the area affected by the incident showed increased consultation rates for diarrhoea (Table 1) and gastroenteritis (Table 2) with SIRs significantly above the UK rate in week 26. Daventry and South Northants PCT also had a raised SIR for both indicators in week 25, and although Northamptonshire Heartlands and Northampton PCTs did not have SIRs significantly above that of the UK in week 25, the rise in consultation rates for diarrhoea and gastroenteritis began during week 25. In Northampton PCT, consultations for both diarrhoea and gastroenteritis peaked in the week following the contamination incident, week 27, returning to normal levels by week 30 (Figure 3A and 3B). A similar effect can be seen in Northampton Heartlands PCT. Daventry and South Northants PCT also showed an increase, but appeared to have consistently higher rates for both indicators. This was the area with the smallest population so the rates were more variable than in the other PCTs and we therefore interpreted these results with caution. The consultation rates for vomiting during weeks 25 to 30 in the East Midlands were not unusual at SHA or PCT level (data not presented). Discussion We have demonstrated the sensitivity of syndromic surveillance in detecting this small Cryptosporidium outbreak and the value of the surveillance in being able to describe the extent of its spread. Both the HPA/NHS Direct and HPA/QSurveillance systems showed demonstrable increases in calls and consultations for diarrhoea that were linked to the outbreak. QSurveillance consultations appeared to increase across the PCTs immediately affected but not in the surrounding area. Both the HPA/NHS Direct and HPA/QSurveillance systems showed a clear signal at the time of the incident and we were able to describe the extent of the impact on pre-primary care and primary care services. The HPA/QSurveillance system showed a rise in consultation rates for gastrointestinal symptoms that began the week before the outbreak, consistent with the period when Cryptosporidium was present in the water leaving the Pitsford Reservoir (19–24 June 2008) and with the onset of symptoms in the first outbreak case on 24 June. Although only 33 cases were identified by the outbreak investigation team, of which 23 were confirmed as having the outbreak Cryptosporidium strain, our syndromic surveillance data detected this limited outbreak. Data also suggested a more widespread increase in general gastrointestinal symptoms around the time of the outbreak, with an estimated 422 excess diarrhoea cases; these excess cases represented an increase of about 25% above normally expected levels. It is highly probable that a proportion of these excess cases may have resulted from the increased publicity surrounding the incident – for example, it is likely that media reports contributed to the large peak in calls detected by the HPA/NHS Direct surveillance system on the day the boil water notice was issued, and could also have impacted on the GP consultation rate. It has been previously shown that reporting of mumps cases is sensitive to media coverage, with a rise in clinically reported cases following newspaper reports [30]. A similar mechanism could account for some of the excess GP consultations as cases experiencing gastrointestinal symptoms may have been more likely to consult their GP, whereas in normal circumstances they would have cared for themselves at home. It is also possible that the surveillance shows outbreak-associated cases that did not come to the attention of the outbreak team, perhaps because symptoms were not sufficiently severe to warrant further investigation, or stool samples were not provided for testing. It is interesting to note that there was no demonstrable impact on the number of calls for vomiting (which is not a prominent clinical feature of cryptosporidiosis). Other common community-based pathogens such as norovirus and rotavirus were at low levels, as is normal for that time of year [31]. In this instance, public health authorities had already been alerted to a potential problem by the water company, although the extent of the outbreak was detected by syndromic surveillance. In 2003 the syndromic surveillance systems in the city of New York, United States, were able to detect an increase in diarrhoeal illness following a power outage when there was no other indication of citywide illness [32]. The New York system covers a population of nine million, but does not regularly detect localised outbreaks [33]. It has been shown previously that the HPA/NHS Direct surveillance system would be unlikely to detect a Cryptosporidium outbreak unless call volumes are high (72% chance of detection if nine-tenths of cases called NHS Direct) [29], although the value of syndromic surveillance for such outbreaks has been recognised [34]. The system detected the East Midlands Cryptosporidium outbreak that affected a smaller population than that covered by the New York system. The three PCTs affected have a combined population of around 600,000, of which just over half use GP practices reporting to QSurveillance, yet this syndromic surveillance system was able to describe an increase in consultation rates for diarrhoea and gastroenteritis around the time of the outbreak. Limitations of the data There was extensive media reporting of the incident that may have affected both the HPA/NHS Direct and HPA/QSurveillance systems and contributed to the increase in reported gastrointestinal symptoms around the time of the contamination incident. However, the rise in consultation rates for diarrhoea began before the outbreak had been detected and therefore cannot be attributed to media coverage. The HPA/NHS Direct and HPA/QSurveillance systems monitor general symptoms and so could only monitor the relevant symptoms of diarrhoea and vomiting. They are not able to detect Cryptosporidium cases, as this would require laboratory confirmation of diagnosis, so some of the estimated excess cases could be unconnected with this incident. This outbreak was discovered by other means but both the HPA/NHS Direct and HPA/ QSurveillance systems were able to describe the extent of the disease in the general population and provide reassurance that there was no widespread impact. Compared with other populations, older people and ethnic minorities are less likely to call NHS Direct [29], and although this should not prevent detection of gastrointestinal symptoms as a result of drinking water contamination as this would affect the whole population, this may reduce the signal from the system [35]. With such large surveillance systems, there will be ‘background noise’ in the data, so procedures must be in place to correctly interpret the data and set appropriate thresholds for action. Conclusion To our knowledge, this is the first time that PCT-level data from a syndromic surveillance system, the HPA/ QSurveillance national surveillance system, have been able to show the extent of such a limited outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide PCT-level data and this Cryptosporidium contamination incident demonstrates the potential usefulness of this system.

How was the event first detected?

1509

Value of syndromic surveillance in monitoring a focal waterborne outbreak due to an unusual Cryptosporidium genotype in Northamptonshire

The United Kingdom (UK) has several national syndromic surveillance systems. The Health Protection Agency (HPA)/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data from a national telephone helpline, while the HPA/ QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems. Data from both of these systems were used to monitor a local outbreak of cryptosporidiosis that occurred following Cryptosporidium oocyst contamination of drinking water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, in June 2008. There was a peak in the number of calls to NHS Direct concerning diarrhoea that coincided with the incident. QSurveillance data for the local areas affected by the outbreak showed a significant increase in GP consultations for diarrhoea and gastroenteritis in the week of the incident but there was no increase in consultations for vomiting. A total of 33 clinical cases of cryptosporidiosis were identified in the outbreak investigation, of which 23 were confirmed as infected with the outbreak strain. However, QSurveillance data suggest that there were an estimated 422 excess diarrhoea cases during the outbreak, an increase of about 25% over baseline weekly levels. To our knowledge, this is the first time that data from a syndromic surveillance system, the HPA/QSurveillance national surveillance system, have been able to show the extent of such a small outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide data at local health district (primary care trust) level. The Cryptosporidium contamination incident described demonstrates the potential usefulness of this information, as it is unusual for syndromic surveillance systems to be able to help monitor such a small-scale outbreak. Introduction As syndromic surveillance systems usually capture data already collected for other purposes, and monitor generic symptoms and/or clinically diagnosed disease, they provide information at an earlier stage of illness (compared with laboratory-confirmed diagnoses), so that action can be taken in time to substantially reduce the impact of disease. Some systems, for example, the Royal College of General Practitioners Weekly Returns Service, are now well established, with many years of historical data that allow monitoring of longer-term disease trends [1]. They have the ability to provide early warning of, for example, seasonal rises in influenza and can trigger public health action, such as a recommendation to prescribe antiviral drugs in line with national guidance [2-4]. They can also provide reassurance to incident response teams and the general public that an incident has not caused adverse health effects – for example, following an explosion at the Buncefield oil storage depot in Hemel Hempstead, United Kingdom (UK), in 2005, syndromic surveillance confirmed that there were no unusual rises in community-based morbidity linked to the incident [5]; following the eruption of the Eyjafjallajökull volcano in Iceland in April 2010 similar assurance was given about lack of impact on community morbidity [6]. Health departments are increasingly expected to monitor health effects of natural events such as heat wave or flooding, or implement surveillance – of which syndromic surveillance plays a major role – for mass gatherings such as the Olympics or football World Cup [7-9]. Systems in France, Australia and Taiwan use data from emergency departments [10-12], a Canadian system uses over-the-counter pharmacy sales [13,14], and in the Netherlands data from both syndromic and surrogate data sources, such as employee absence records and prescription medications dispensed by pharmacies, are included in surveillance systems [15,16]. Currently systems based on Internet searches via search engines or on queries submitted to medical websites are being developed [17,18]. In the UK, the HPA/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data collected from the NHS Direct telephone helpline [19], while the HPA/QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems [20]. The HPA Real-time Syndromic Surveillance Team is a small team that coordinates a number of syndromic surveillance systems within the HPA and takes a lead for syndromic surveillance in England [21]. This paper describes the support provided by the team to the local incident management team during a local cryptosporidiosis outbreak and shows the use of syndromic surveillance in monitoring the extent of an outbreak using the HPA/NHS Direct and HPA/QSurveillance national surveillance systems. Cryptosporidiosis Cryptosporidium is a protozoan parasite that can cause an infection in people, cattle and sometimes other animals [22]. Cryptosporidiosis is most common in children aged between one and five years, but it can affect all ages. Those with impaired immune systems are likely to be most seriously affected. Symptoms usually appear between three and 12 days after initial exposure and include watery diarrhoea, stomach pains, dehydration and fever. In its transmissible form, called an oocyst, the parasite is protected by an outer shell, which allows it to survive in the environment for a long time. Transmission occurs most often via the faeco-oral route through person-to-person or animalto-person contact, but people may also be infected by consuming contaminated water or food or by swimming in contaminated water. Although uncommon, the largest outbreaks have occurred following contamination of drinking water [23,24]. Normal chlorine disinfection procedures do not kill the oocysts, so they are removed by filtration and water companies carry out routine monitoring of treated water. Description of the incident On 25 June 2008 the local Health Protection Unit was informed by Anglian Water of an exceedence in the level of Cryptosporidium oocysts found in water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, during 19 to 24 June 2008 [25]. The reservoir supplied a population of more than 250,000 in the Northampton area. A notice advising people in the affected areas to boil all drinking water was issued on 25 June 2008 and public health messages were circulated to local health services and to the general public via the media. Those members of the public who were concerned about health risks associated with the incident were asked to ring NHS Direct for clinical advice [26]. The HPA wrote to local GPs and hospitals asking them to monitor potential patients for signs and symptoms of Cryptosporidium infection and to submit faecal specimens to the local hospital diagnostic laboratory if patients presented with diarrhoea. Samples from 34 patients where Cryptosporidium infection was identified were sent to the UK Cryptosporidium reference unit for typing. On 30 June 2008, the Cryptosporidium oocysts found in the reservoir water were confirmed as being of the rabbit genotype Cryptosporidium cuniculus [27]. Subsequently, a dead rabbit was found in a treated water tank at the water treatment works. The genotype of Cryptosporidium oocysts in the rabbit’s large bowel was indistinguishable from that of the oocysts found in the water [27]. After remediation of the water supply and distribution, the ‘boil water notice’ was lifted on 4 July and the following day the first case of cryptosporidiosis linked to the incident was identified by the reference laboratory (this case was infected with C. cuniculus). During the course of the outbreak (24 June – 18 July 2008, the dates of symptom onset in the first and last case, respectively), 23 cases of cryptosporidiosis were confirmed as being infected with C. cuniculus; one of the 23 was a secondary case. The HPA Real-time Syndromic Surveillance Team provided data in order to aid the response to this incident and the first syndromic surveillance report was circulated to the incident management team and other relevant people in the HPA on 27 June 2008. Data from the HPA/NHS Direct and HPA/QSurveillance systems were provided in a series of regular reports, initially daily and eventually weekly, until the final report on 21 August 2008. Each report included a summary interpretation and more detailed data on diarrhoea, gastroenteritis and vomiting indicators. Methods Surveillance systems HPA/NHS Direct surveillance system NHS Direct is a 24-hour nurse-led telephone helpline that provides health information and advice to the general public. Nurses use a computerised clinical decision support system – the NHS Clinical Assessment System (NHS CAS) – to handle calls. This assessment system uses approximately 200 computerised symptom-based clinical algorithms. Nurses assign the call to the most appropriate algorithm and the patient’s symptoms determine the questions asked and the action to be taken following the call (call outcome), which could be guidance on self-care or they could be referred to their GP or advised to attend a hospital emergency department. No attempt is made to provide a formal diagnosis. Daily NHS Direct data are received by the Real-time Syndromic Surveillance Team, where the number and type of calls received during the previous day are analysed and interpreted. Call proportions are calculated by age group and algorithm against the total number of calls received. HPA/QSurveillance system The HPA/QSurveillance national surveillance system was set up by the University of Nottingham, United Kingdom, and Egton Medical Information Systems (EMIS), a supplier of general practice computer systems, in collaboration with the HPA. It comprises a network of more than 3,500 general practices throughout the UK, covering more than 22 million patients (about 38% of the population [28]). Aggregated data on GP consultations for a range of indicators are automatically uploaded daily from GP practice systems to a central database. Data are routinely reported on a weekly basis; however, daily reporting is possible for specific incidents. Reports are provided at national or regional level (strategic health authority, SHA) and by local health district (primary care trust, PCT). Analysis of surveillance data NHS Direct call proportions for gastrointestinal syndromes (diarrhoea and vomiting) for the East Midlands region in England, where Northampton is situated, were examined during the outbreak (24 June – 18 July 2008) and compared with those for England and Wales. A series of control charts for diarrhoea calls are routinely used to monitor significant rises in the numbers of calls received. Control charts are calculated by assuming that calls follow a Poisson distribution with the total number of calls as an offset: a model is fitted to each region and symptom separately [29]. The model takes into account call variation caused by weekends, public holidays and the time of year – variables that can affect the number of calls received by NHS Direct. A value above the upper limit of the 99.5% confidence interval would be considered to be unusual. The seven-day moving average for diarrhoea calls was also monitored. The number and percentage of calls for diarrhoea in the East Midlands region were presented by call outcome and the number of calls in the Northampton (NN) postcode districts and in particular the number of calls in the NN11 and NN12 postcode districts, which were most affected by the incident. QSurveillance national consultation rates per 100,000 population for diarrhoea (in the age groups under five years, five years and over, and all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with rates for the same period in 2007 (data not presented). Consultation rates by region for 2008 for diarrhoea (all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with those for the East Midlands region. The gastroenteritis indicator includes all cases of diarrhoea and/or vomiting. Consultation rates and standardised incidence ratios (SIRs) – calculated using the UK as the standard population – for diarrhoea, gastroenteritis and vomiting were compared for the UK, Yorkshire and Humberside, East Midlands, Leicestershire, Northamptonshire and Rutland SHA, and Daventry and South Northants PCT, Northamptonshire Heartlands PCT and Northampton PCT. Yorkshire and Humberside was not an affected region but was included as a control. The area supplied by the Pitsford Reservoir included the three PCTs, which were all within the Leicestershire, Northamptonshire and Rutland SHA. The consultation rates and SIRs were compared for the period from week 16 to week 35 of 2008 in order to compare the rates before and after the Cryptosporidium exceedance, which took place in week 26. Estimates of excess numbers of cases of diarrhoea occurring during and following the Cryptosporidium outbreak were made by calculating the mean consultation rate over a period of five weeks before and after the incident (weeks 20–24 and weeks 31–35, respectively). For each of the three PCTs, the calculated mean rate was applied to the PCT population to estimate the number of cases that would be expected each week. The actual consultation rates for diarrhoea for weeks 25 to 30 were used to estimate the number of cases for the PCT population each week. The expected number of cases was subtracted from the estimated number of cases in the PCT population to give the estimated number of excess cases. Results HPA/NHS Direct surveillance system A peak in the number of calls for diarrhoea in the East Midlands was recorded in 25–26 June 2008, the period that coincided with the contamination incident and the associated media coverage (Figure 1). The neighbouring areas of the West Midlands, Yorkshire and the Humber, and East of England showed no increase in the number of calls for diarrhoea. The peak produced a control chart exceedance for calls for diarrhoea on 25 June 2008 (Figure 2), when the proportion of calls exceeded the upper limit of the 99.5% confidence interval. There were further confidence interval exceedances on 26 and 28 June (which were not control chart exceedances). There was no peak in calls for vomiting or control chart exceedance for these calls in the East Midlands. HPA/QSurveillance national surveillance system The East Midlands region had significantly high consultation rates for diarrhoea and gastroenteritis in week 25 (16–22 June), week 26 (23–29 June 2008, when the contamination incident was reported) and in the following four weeks. Within this region. Leicestershire, Northamptonshire and Rutland SHA had slightly raised consultation rates and significant SIRs across weeks 25 to 30 that were not seen in the neighbouring Trent SHA. At PCT level, all three of the PCTs in the area affected by the incident showed increased consultation rates for diarrhoea (Table 1) and gastroenteritis (Table 2) with SIRs significantly above the UK rate in week 26. Daventry and South Northants PCT also had a raised SIR for both indicators in week 25, and although Northamptonshire Heartlands and Northampton PCTs did not have SIRs significantly above that of the UK in week 25, the rise in consultation rates for diarrhoea and gastroenteritis began during week 25. In Northampton PCT, consultations for both diarrhoea and gastroenteritis peaked in the week following the contamination incident, week 27, returning to normal levels by week 30 (Figure 3A and 3B). A similar effect can be seen in Northampton Heartlands PCT. Daventry and South Northants PCT also showed an increase, but appeared to have consistently higher rates for both indicators. This was the area with the smallest population so the rates were more variable than in the other PCTs and we therefore interpreted these results with caution. The consultation rates for vomiting during weeks 25 to 30 in the East Midlands were not unusual at SHA or PCT level (data not presented). Discussion We have demonstrated the sensitivity of syndromic surveillance in detecting this small Cryptosporidium outbreak and the value of the surveillance in being able to describe the extent of its spread. Both the HPA/NHS Direct and HPA/QSurveillance systems showed demonstrable increases in calls and consultations for diarrhoea that were linked to the outbreak. QSurveillance consultations appeared to increase across the PCTs immediately affected but not in the surrounding area. Both the HPA/NHS Direct and HPA/QSurveillance systems showed a clear signal at the time of the incident and we were able to describe the extent of the impact on pre-primary care and primary care services. The HPA/QSurveillance system showed a rise in consultation rates for gastrointestinal symptoms that began the week before the outbreak, consistent with the period when Cryptosporidium was present in the water leaving the Pitsford Reservoir (19–24 June 2008) and with the onset of symptoms in the first outbreak case on 24 June. Although only 33 cases were identified by the outbreak investigation team, of which 23 were confirmed as having the outbreak Cryptosporidium strain, our syndromic surveillance data detected this limited outbreak. Data also suggested a more widespread increase in general gastrointestinal symptoms around the time of the outbreak, with an estimated 422 excess diarrhoea cases; these excess cases represented an increase of about 25% above normally expected levels. It is highly probable that a proportion of these excess cases may have resulted from the increased publicity surrounding the incident – for example, it is likely that media reports contributed to the large peak in calls detected by the HPA/NHS Direct surveillance system on the day the boil water notice was issued, and could also have impacted on the GP consultation rate. It has been previously shown that reporting of mumps cases is sensitive to media coverage, with a rise in clinically reported cases following newspaper reports [30]. A similar mechanism could account for some of the excess GP consultations as cases experiencing gastrointestinal symptoms may have been more likely to consult their GP, whereas in normal circumstances they would have cared for themselves at home. It is also possible that the surveillance shows outbreak-associated cases that did not come to the attention of the outbreak team, perhaps because symptoms were not sufficiently severe to warrant further investigation, or stool samples were not provided for testing. It is interesting to note that there was no demonstrable impact on the number of calls for vomiting (which is not a prominent clinical feature of cryptosporidiosis). Other common community-based pathogens such as norovirus and rotavirus were at low levels, as is normal for that time of year [31]. In this instance, public health authorities had already been alerted to a potential problem by the water company, although the extent of the outbreak was detected by syndromic surveillance. In 2003 the syndromic surveillance systems in the city of New York, United States, were able to detect an increase in diarrhoeal illness following a power outage when there was no other indication of citywide illness [32]. The New York system covers a population of nine million, but does not regularly detect localised outbreaks [33]. It has been shown previously that the HPA/NHS Direct surveillance system would be unlikely to detect a Cryptosporidium outbreak unless call volumes are high (72% chance of detection if nine-tenths of cases called NHS Direct) [29], although the value of syndromic surveillance for such outbreaks has been recognised [34]. The system detected the East Midlands Cryptosporidium outbreak that affected a smaller population than that covered by the New York system. The three PCTs affected have a combined population of around 600,000, of which just over half use GP practices reporting to QSurveillance, yet this syndromic surveillance system was able to describe an increase in consultation rates for diarrhoea and gastroenteritis around the time of the outbreak. Limitations of the data There was extensive media reporting of the incident that may have affected both the HPA/NHS Direct and HPA/QSurveillance systems and contributed to the increase in reported gastrointestinal symptoms around the time of the contamination incident. However, the rise in consultation rates for diarrhoea began before the outbreak had been detected and therefore cannot be attributed to media coverage. The HPA/NHS Direct and HPA/QSurveillance systems monitor general symptoms and so could only monitor the relevant symptoms of diarrhoea and vomiting. They are not able to detect Cryptosporidium cases, as this would require laboratory confirmation of diagnosis, so some of the estimated excess cases could be unconnected with this incident. This outbreak was discovered by other means but both the HPA/NHS Direct and HPA/ QSurveillance systems were able to describe the extent of the disease in the general population and provide reassurance that there was no widespread impact. Compared with other populations, older people and ethnic minorities are less likely to call NHS Direct [29], and although this should not prevent detection of gastrointestinal symptoms as a result of drinking water contamination as this would affect the whole population, this may reduce the signal from the system [35]. With such large surveillance systems, there will be ‘background noise’ in the data, so procedures must be in place to correctly interpret the data and set appropriate thresholds for action. Conclusion To our knowledge, this is the first time that PCT-level data from a syndromic surveillance system, the HPA/ QSurveillance national surveillance system, have been able to show the extent of such a limited outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide PCT-level data and this Cryptosporidium contamination incident demonstrates the potential usefulness of this system.

How many people were ill?

1510

Value of syndromic surveillance in monitoring a focal waterborne outbreak due to an unusual Cryptosporidium genotype in Northamptonshire

The United Kingdom (UK) has several national syndromic surveillance systems. The Health Protection Agency (HPA)/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data from a national telephone helpline, while the HPA/ QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems. Data from both of these systems were used to monitor a local outbreak of cryptosporidiosis that occurred following Cryptosporidium oocyst contamination of drinking water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, in June 2008. There was a peak in the number of calls to NHS Direct concerning diarrhoea that coincided with the incident. QSurveillance data for the local areas affected by the outbreak showed a significant increase in GP consultations for diarrhoea and gastroenteritis in the week of the incident but there was no increase in consultations for vomiting. A total of 33 clinical cases of cryptosporidiosis were identified in the outbreak investigation, of which 23 were confirmed as infected with the outbreak strain. However, QSurveillance data suggest that there were an estimated 422 excess diarrhoea cases during the outbreak, an increase of about 25% over baseline weekly levels. To our knowledge, this is the first time that data from a syndromic surveillance system, the HPA/QSurveillance national surveillance system, have been able to show the extent of such a small outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide data at local health district (primary care trust) level. The Cryptosporidium contamination incident described demonstrates the potential usefulness of this information, as it is unusual for syndromic surveillance systems to be able to help monitor such a small-scale outbreak. Introduction As syndromic surveillance systems usually capture data already collected for other purposes, and monitor generic symptoms and/or clinically diagnosed disease, they provide information at an earlier stage of illness (compared with laboratory-confirmed diagnoses), so that action can be taken in time to substantially reduce the impact of disease. Some systems, for example, the Royal College of General Practitioners Weekly Returns Service, are now well established, with many years of historical data that allow monitoring of longer-term disease trends [1]. They have the ability to provide early warning of, for example, seasonal rises in influenza and can trigger public health action, such as a recommendation to prescribe antiviral drugs in line with national guidance [2-4]. They can also provide reassurance to incident response teams and the general public that an incident has not caused adverse health effects – for example, following an explosion at the Buncefield oil storage depot in Hemel Hempstead, United Kingdom (UK), in 2005, syndromic surveillance confirmed that there were no unusual rises in community-based morbidity linked to the incident [5]; following the eruption of the Eyjafjallajökull volcano in Iceland in April 2010 similar assurance was given about lack of impact on community morbidity [6]. Health departments are increasingly expected to monitor health effects of natural events such as heat wave or flooding, or implement surveillance – of which syndromic surveillance plays a major role – for mass gatherings such as the Olympics or football World Cup [7-9]. Systems in France, Australia and Taiwan use data from emergency departments [10-12], a Canadian system uses over-the-counter pharmacy sales [13,14], and in the Netherlands data from both syndromic and surrogate data sources, such as employee absence records and prescription medications dispensed by pharmacies, are included in surveillance systems [15,16]. Currently systems based on Internet searches via search engines or on queries submitted to medical websites are being developed [17,18]. In the UK, the HPA/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data collected from the NHS Direct telephone helpline [19], while the HPA/QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems [20]. The HPA Real-time Syndromic Surveillance Team is a small team that coordinates a number of syndromic surveillance systems within the HPA and takes a lead for syndromic surveillance in England [21]. This paper describes the support provided by the team to the local incident management team during a local cryptosporidiosis outbreak and shows the use of syndromic surveillance in monitoring the extent of an outbreak using the HPA/NHS Direct and HPA/QSurveillance national surveillance systems. Cryptosporidiosis Cryptosporidium is a protozoan parasite that can cause an infection in people, cattle and sometimes other animals [22]. Cryptosporidiosis is most common in children aged between one and five years, but it can affect all ages. Those with impaired immune systems are likely to be most seriously affected. Symptoms usually appear between three and 12 days after initial exposure and include watery diarrhoea, stomach pains, dehydration and fever. In its transmissible form, called an oocyst, the parasite is protected by an outer shell, which allows it to survive in the environment for a long time. Transmission occurs most often via the faeco-oral route through person-to-person or animalto-person contact, but people may also be infected by consuming contaminated water or food or by swimming in contaminated water. Although uncommon, the largest outbreaks have occurred following contamination of drinking water [23,24]. Normal chlorine disinfection procedures do not kill the oocysts, so they are removed by filtration and water companies carry out routine monitoring of treated water. Description of the incident On 25 June 2008 the local Health Protection Unit was informed by Anglian Water of an exceedence in the level of Cryptosporidium oocysts found in water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, during 19 to 24 June 2008 [25]. The reservoir supplied a population of more than 250,000 in the Northampton area. A notice advising people in the affected areas to boil all drinking water was issued on 25 June 2008 and public health messages were circulated to local health services and to the general public via the media. Those members of the public who were concerned about health risks associated with the incident were asked to ring NHS Direct for clinical advice [26]. The HPA wrote to local GPs and hospitals asking them to monitor potential patients for signs and symptoms of Cryptosporidium infection and to submit faecal specimens to the local hospital diagnostic laboratory if patients presented with diarrhoea. Samples from 34 patients where Cryptosporidium infection was identified were sent to the UK Cryptosporidium reference unit for typing. On 30 June 2008, the Cryptosporidium oocysts found in the reservoir water were confirmed as being of the rabbit genotype Cryptosporidium cuniculus [27]. Subsequently, a dead rabbit was found in a treated water tank at the water treatment works. The genotype of Cryptosporidium oocysts in the rabbit’s large bowel was indistinguishable from that of the oocysts found in the water [27]. After remediation of the water supply and distribution, the ‘boil water notice’ was lifted on 4 July and the following day the first case of cryptosporidiosis linked to the incident was identified by the reference laboratory (this case was infected with C. cuniculus). During the course of the outbreak (24 June – 18 July 2008, the dates of symptom onset in the first and last case, respectively), 23 cases of cryptosporidiosis were confirmed as being infected with C. cuniculus; one of the 23 was a secondary case. The HPA Real-time Syndromic Surveillance Team provided data in order to aid the response to this incident and the first syndromic surveillance report was circulated to the incident management team and other relevant people in the HPA on 27 June 2008. Data from the HPA/NHS Direct and HPA/QSurveillance systems were provided in a series of regular reports, initially daily and eventually weekly, until the final report on 21 August 2008. Each report included a summary interpretation and more detailed data on diarrhoea, gastroenteritis and vomiting indicators. Methods Surveillance systems HPA/NHS Direct surveillance system NHS Direct is a 24-hour nurse-led telephone helpline that provides health information and advice to the general public. Nurses use a computerised clinical decision support system – the NHS Clinical Assessment System (NHS CAS) – to handle calls. This assessment system uses approximately 200 computerised symptom-based clinical algorithms. Nurses assign the call to the most appropriate algorithm and the patient’s symptoms determine the questions asked and the action to be taken following the call (call outcome), which could be guidance on self-care or they could be referred to their GP or advised to attend a hospital emergency department. No attempt is made to provide a formal diagnosis. Daily NHS Direct data are received by the Real-time Syndromic Surveillance Team, where the number and type of calls received during the previous day are analysed and interpreted. Call proportions are calculated by age group and algorithm against the total number of calls received. HPA/QSurveillance system The HPA/QSurveillance national surveillance system was set up by the University of Nottingham, United Kingdom, and Egton Medical Information Systems (EMIS), a supplier of general practice computer systems, in collaboration with the HPA. It comprises a network of more than 3,500 general practices throughout the UK, covering more than 22 million patients (about 38% of the population [28]). Aggregated data on GP consultations for a range of indicators are automatically uploaded daily from GP practice systems to a central database. Data are routinely reported on a weekly basis; however, daily reporting is possible for specific incidents. Reports are provided at national or regional level (strategic health authority, SHA) and by local health district (primary care trust, PCT). Analysis of surveillance data NHS Direct call proportions for gastrointestinal syndromes (diarrhoea and vomiting) for the East Midlands region in England, where Northampton is situated, were examined during the outbreak (24 June – 18 July 2008) and compared with those for England and Wales. A series of control charts for diarrhoea calls are routinely used to monitor significant rises in the numbers of calls received. Control charts are calculated by assuming that calls follow a Poisson distribution with the total number of calls as an offset: a model is fitted to each region and symptom separately [29]. The model takes into account call variation caused by weekends, public holidays and the time of year – variables that can affect the number of calls received by NHS Direct. A value above the upper limit of the 99.5% confidence interval would be considered to be unusual. The seven-day moving average for diarrhoea calls was also monitored. The number and percentage of calls for diarrhoea in the East Midlands region were presented by call outcome and the number of calls in the Northampton (NN) postcode districts and in particular the number of calls in the NN11 and NN12 postcode districts, which were most affected by the incident. QSurveillance national consultation rates per 100,000 population for diarrhoea (in the age groups under five years, five years and over, and all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with rates for the same period in 2007 (data not presented). Consultation rates by region for 2008 for diarrhoea (all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with those for the East Midlands region. The gastroenteritis indicator includes all cases of diarrhoea and/or vomiting. Consultation rates and standardised incidence ratios (SIRs) – calculated using the UK as the standard population – for diarrhoea, gastroenteritis and vomiting were compared for the UK, Yorkshire and Humberside, East Midlands, Leicestershire, Northamptonshire and Rutland SHA, and Daventry and South Northants PCT, Northamptonshire Heartlands PCT and Northampton PCT. Yorkshire and Humberside was not an affected region but was included as a control. The area supplied by the Pitsford Reservoir included the three PCTs, which were all within the Leicestershire, Northamptonshire and Rutland SHA. The consultation rates and SIRs were compared for the period from week 16 to week 35 of 2008 in order to compare the rates before and after the Cryptosporidium exceedance, which took place in week 26. Estimates of excess numbers of cases of diarrhoea occurring during and following the Cryptosporidium outbreak were made by calculating the mean consultation rate over a period of five weeks before and after the incident (weeks 20–24 and weeks 31–35, respectively). For each of the three PCTs, the calculated mean rate was applied to the PCT population to estimate the number of cases that would be expected each week. The actual consultation rates for diarrhoea for weeks 25 to 30 were used to estimate the number of cases for the PCT population each week. The expected number of cases was subtracted from the estimated number of cases in the PCT population to give the estimated number of excess cases. Results HPA/NHS Direct surveillance system A peak in the number of calls for diarrhoea in the East Midlands was recorded in 25–26 June 2008, the period that coincided with the contamination incident and the associated media coverage (Figure 1). The neighbouring areas of the West Midlands, Yorkshire and the Humber, and East of England showed no increase in the number of calls for diarrhoea. The peak produced a control chart exceedance for calls for diarrhoea on 25 June 2008 (Figure 2), when the proportion of calls exceeded the upper limit of the 99.5% confidence interval. There were further confidence interval exceedances on 26 and 28 June (which were not control chart exceedances). There was no peak in calls for vomiting or control chart exceedance for these calls in the East Midlands. HPA/QSurveillance national surveillance system The East Midlands region had significantly high consultation rates for diarrhoea and gastroenteritis in week 25 (16–22 June), week 26 (23–29 June 2008, when the contamination incident was reported) and in the following four weeks. Within this region. Leicestershire, Northamptonshire and Rutland SHA had slightly raised consultation rates and significant SIRs across weeks 25 to 30 that were not seen in the neighbouring Trent SHA. At PCT level, all three of the PCTs in the area affected by the incident showed increased consultation rates for diarrhoea (Table 1) and gastroenteritis (Table 2) with SIRs significantly above the UK rate in week 26. Daventry and South Northants PCT also had a raised SIR for both indicators in week 25, and although Northamptonshire Heartlands and Northampton PCTs did not have SIRs significantly above that of the UK in week 25, the rise in consultation rates for diarrhoea and gastroenteritis began during week 25. In Northampton PCT, consultations for both diarrhoea and gastroenteritis peaked in the week following the contamination incident, week 27, returning to normal levels by week 30 (Figure 3A and 3B). A similar effect can be seen in Northampton Heartlands PCT. Daventry and South Northants PCT also showed an increase, but appeared to have consistently higher rates for both indicators. This was the area with the smallest population so the rates were more variable than in the other PCTs and we therefore interpreted these results with caution. The consultation rates for vomiting during weeks 25 to 30 in the East Midlands were not unusual at SHA or PCT level (data not presented). Discussion We have demonstrated the sensitivity of syndromic surveillance in detecting this small Cryptosporidium outbreak and the value of the surveillance in being able to describe the extent of its spread. Both the HPA/NHS Direct and HPA/QSurveillance systems showed demonstrable increases in calls and consultations for diarrhoea that were linked to the outbreak. QSurveillance consultations appeared to increase across the PCTs immediately affected but not in the surrounding area. Both the HPA/NHS Direct and HPA/QSurveillance systems showed a clear signal at the time of the incident and we were able to describe the extent of the impact on pre-primary care and primary care services. The HPA/QSurveillance system showed a rise in consultation rates for gastrointestinal symptoms that began the week before the outbreak, consistent with the period when Cryptosporidium was present in the water leaving the Pitsford Reservoir (19–24 June 2008) and with the onset of symptoms in the first outbreak case on 24 June. Although only 33 cases were identified by the outbreak investigation team, of which 23 were confirmed as having the outbreak Cryptosporidium strain, our syndromic surveillance data detected this limited outbreak. Data also suggested a more widespread increase in general gastrointestinal symptoms around the time of the outbreak, with an estimated 422 excess diarrhoea cases; these excess cases represented an increase of about 25% above normally expected levels. It is highly probable that a proportion of these excess cases may have resulted from the increased publicity surrounding the incident – for example, it is likely that media reports contributed to the large peak in calls detected by the HPA/NHS Direct surveillance system on the day the boil water notice was issued, and could also have impacted on the GP consultation rate. It has been previously shown that reporting of mumps cases is sensitive to media coverage, with a rise in clinically reported cases following newspaper reports [30]. A similar mechanism could account for some of the excess GP consultations as cases experiencing gastrointestinal symptoms may have been more likely to consult their GP, whereas in normal circumstances they would have cared for themselves at home. It is also possible that the surveillance shows outbreak-associated cases that did not come to the attention of the outbreak team, perhaps because symptoms were not sufficiently severe to warrant further investigation, or stool samples were not provided for testing. It is interesting to note that there was no demonstrable impact on the number of calls for vomiting (which is not a prominent clinical feature of cryptosporidiosis). Other common community-based pathogens such as norovirus and rotavirus were at low levels, as is normal for that time of year [31]. In this instance, public health authorities had already been alerted to a potential problem by the water company, although the extent of the outbreak was detected by syndromic surveillance. In 2003 the syndromic surveillance systems in the city of New York, United States, were able to detect an increase in diarrhoeal illness following a power outage when there was no other indication of citywide illness [32]. The New York system covers a population of nine million, but does not regularly detect localised outbreaks [33]. It has been shown previously that the HPA/NHS Direct surveillance system would be unlikely to detect a Cryptosporidium outbreak unless call volumes are high (72% chance of detection if nine-tenths of cases called NHS Direct) [29], although the value of syndromic surveillance for such outbreaks has been recognised [34]. The system detected the East Midlands Cryptosporidium outbreak that affected a smaller population than that covered by the New York system. The three PCTs affected have a combined population of around 600,000, of which just over half use GP practices reporting to QSurveillance, yet this syndromic surveillance system was able to describe an increase in consultation rates for diarrhoea and gastroenteritis around the time of the outbreak. Limitations of the data There was extensive media reporting of the incident that may have affected both the HPA/NHS Direct and HPA/QSurveillance systems and contributed to the increase in reported gastrointestinal symptoms around the time of the contamination incident. However, the rise in consultation rates for diarrhoea began before the outbreak had been detected and therefore cannot be attributed to media coverage. The HPA/NHS Direct and HPA/QSurveillance systems monitor general symptoms and so could only monitor the relevant symptoms of diarrhoea and vomiting. They are not able to detect Cryptosporidium cases, as this would require laboratory confirmation of diagnosis, so some of the estimated excess cases could be unconnected with this incident. This outbreak was discovered by other means but both the HPA/NHS Direct and HPA/ QSurveillance systems were able to describe the extent of the disease in the general population and provide reassurance that there was no widespread impact. Compared with other populations, older people and ethnic minorities are less likely to call NHS Direct [29], and although this should not prevent detection of gastrointestinal symptoms as a result of drinking water contamination as this would affect the whole population, this may reduce the signal from the system [35]. With such large surveillance systems, there will be ‘background noise’ in the data, so procedures must be in place to correctly interpret the data and set appropriate thresholds for action. Conclusion To our knowledge, this is the first time that PCT-level data from a syndromic surveillance system, the HPA/ QSurveillance national surveillance system, have been able to show the extent of such a limited outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide PCT-level data and this Cryptosporidium contamination incident demonstrates the potential usefulness of this system.

How many people were hospitalized?

1511

Value of syndromic surveillance in monitoring a focal waterborne outbreak due to an unusual Cryptosporidium genotype in Northamptonshire

The United Kingdom (UK) has several national syndromic surveillance systems. The Health Protection Agency (HPA)/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data from a national telephone helpline, while the HPA/ QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems. Data from both of these systems were used to monitor a local outbreak of cryptosporidiosis that occurred following Cryptosporidium oocyst contamination of drinking water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, in June 2008. There was a peak in the number of calls to NHS Direct concerning diarrhoea that coincided with the incident. QSurveillance data for the local areas affected by the outbreak showed a significant increase in GP consultations for diarrhoea and gastroenteritis in the week of the incident but there was no increase in consultations for vomiting. A total of 33 clinical cases of cryptosporidiosis were identified in the outbreak investigation, of which 23 were confirmed as infected with the outbreak strain. However, QSurveillance data suggest that there were an estimated 422 excess diarrhoea cases during the outbreak, an increase of about 25% over baseline weekly levels. To our knowledge, this is the first time that data from a syndromic surveillance system, the HPA/QSurveillance national surveillance system, have been able to show the extent of such a small outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide data at local health district (primary care trust) level. The Cryptosporidium contamination incident described demonstrates the potential usefulness of this information, as it is unusual for syndromic surveillance systems to be able to help monitor such a small-scale outbreak. Introduction As syndromic surveillance systems usually capture data already collected for other purposes, and monitor generic symptoms and/or clinically diagnosed disease, they provide information at an earlier stage of illness (compared with laboratory-confirmed diagnoses), so that action can be taken in time to substantially reduce the impact of disease. Some systems, for example, the Royal College of General Practitioners Weekly Returns Service, are now well established, with many years of historical data that allow monitoring of longer-term disease trends [1]. They have the ability to provide early warning of, for example, seasonal rises in influenza and can trigger public health action, such as a recommendation to prescribe antiviral drugs in line with national guidance [2-4]. They can also provide reassurance to incident response teams and the general public that an incident has not caused adverse health effects – for example, following an explosion at the Buncefield oil storage depot in Hemel Hempstead, United Kingdom (UK), in 2005, syndromic surveillance confirmed that there were no unusual rises in community-based morbidity linked to the incident [5]; following the eruption of the Eyjafjallajökull volcano in Iceland in April 2010 similar assurance was given about lack of impact on community morbidity [6]. Health departments are increasingly expected to monitor health effects of natural events such as heat wave or flooding, or implement surveillance – of which syndromic surveillance plays a major role – for mass gatherings such as the Olympics or football World Cup [7-9]. Systems in France, Australia and Taiwan use data from emergency departments [10-12], a Canadian system uses over-the-counter pharmacy sales [13,14], and in the Netherlands data from both syndromic and surrogate data sources, such as employee absence records and prescription medications dispensed by pharmacies, are included in surveillance systems [15,16]. Currently systems based on Internet searches via search engines or on queries submitted to medical websites are being developed [17,18]. In the UK, the HPA/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data collected from the NHS Direct telephone helpline [19], while the HPA/QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems [20]. The HPA Real-time Syndromic Surveillance Team is a small team that coordinates a number of syndromic surveillance systems within the HPA and takes a lead for syndromic surveillance in England [21]. This paper describes the support provided by the team to the local incident management team during a local cryptosporidiosis outbreak and shows the use of syndromic surveillance in monitoring the extent of an outbreak using the HPA/NHS Direct and HPA/QSurveillance national surveillance systems. Cryptosporidiosis Cryptosporidium is a protozoan parasite that can cause an infection in people, cattle and sometimes other animals [22]. Cryptosporidiosis is most common in children aged between one and five years, but it can affect all ages. Those with impaired immune systems are likely to be most seriously affected. Symptoms usually appear between three and 12 days after initial exposure and include watery diarrhoea, stomach pains, dehydration and fever. In its transmissible form, called an oocyst, the parasite is protected by an outer shell, which allows it to survive in the environment for a long time. Transmission occurs most often via the faeco-oral route through person-to-person or animalto-person contact, but people may also be infected by consuming contaminated water or food or by swimming in contaminated water. Although uncommon, the largest outbreaks have occurred following contamination of drinking water [23,24]. Normal chlorine disinfection procedures do not kill the oocysts, so they are removed by filtration and water companies carry out routine monitoring of treated water. Description of the incident On 25 June 2008 the local Health Protection Unit was informed by Anglian Water of an exceedence in the level of Cryptosporidium oocysts found in water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, during 19 to 24 June 2008 [25]. The reservoir supplied a population of more than 250,000 in the Northampton area. A notice advising people in the affected areas to boil all drinking water was issued on 25 June 2008 and public health messages were circulated to local health services and to the general public via the media. Those members of the public who were concerned about health risks associated with the incident were asked to ring NHS Direct for clinical advice [26]. The HPA wrote to local GPs and hospitals asking them to monitor potential patients for signs and symptoms of Cryptosporidium infection and to submit faecal specimens to the local hospital diagnostic laboratory if patients presented with diarrhoea. Samples from 34 patients where Cryptosporidium infection was identified were sent to the UK Cryptosporidium reference unit for typing. On 30 June 2008, the Cryptosporidium oocysts found in the reservoir water were confirmed as being of the rabbit genotype Cryptosporidium cuniculus [27]. Subsequently, a dead rabbit was found in a treated water tank at the water treatment works. The genotype of Cryptosporidium oocysts in the rabbit’s large bowel was indistinguishable from that of the oocysts found in the water [27]. After remediation of the water supply and distribution, the ‘boil water notice’ was lifted on 4 July and the following day the first case of cryptosporidiosis linked to the incident was identified by the reference laboratory (this case was infected with C. cuniculus). During the course of the outbreak (24 June – 18 July 2008, the dates of symptom onset in the first and last case, respectively), 23 cases of cryptosporidiosis were confirmed as being infected with C. cuniculus; one of the 23 was a secondary case. The HPA Real-time Syndromic Surveillance Team provided data in order to aid the response to this incident and the first syndromic surveillance report was circulated to the incident management team and other relevant people in the HPA on 27 June 2008. Data from the HPA/NHS Direct and HPA/QSurveillance systems were provided in a series of regular reports, initially daily and eventually weekly, until the final report on 21 August 2008. Each report included a summary interpretation and more detailed data on diarrhoea, gastroenteritis and vomiting indicators. Methods Surveillance systems HPA/NHS Direct surveillance system NHS Direct is a 24-hour nurse-led telephone helpline that provides health information and advice to the general public. Nurses use a computerised clinical decision support system – the NHS Clinical Assessment System (NHS CAS) – to handle calls. This assessment system uses approximately 200 computerised symptom-based clinical algorithms. Nurses assign the call to the most appropriate algorithm and the patient’s symptoms determine the questions asked and the action to be taken following the call (call outcome), which could be guidance on self-care or they could be referred to their GP or advised to attend a hospital emergency department. No attempt is made to provide a formal diagnosis. Daily NHS Direct data are received by the Real-time Syndromic Surveillance Team, where the number and type of calls received during the previous day are analysed and interpreted. Call proportions are calculated by age group and algorithm against the total number of calls received. HPA/QSurveillance system The HPA/QSurveillance national surveillance system was set up by the University of Nottingham, United Kingdom, and Egton Medical Information Systems (EMIS), a supplier of general practice computer systems, in collaboration with the HPA. It comprises a network of more than 3,500 general practices throughout the UK, covering more than 22 million patients (about 38% of the population [28]). Aggregated data on GP consultations for a range of indicators are automatically uploaded daily from GP practice systems to a central database. Data are routinely reported on a weekly basis; however, daily reporting is possible for specific incidents. Reports are provided at national or regional level (strategic health authority, SHA) and by local health district (primary care trust, PCT). Analysis of surveillance data NHS Direct call proportions for gastrointestinal syndromes (diarrhoea and vomiting) for the East Midlands region in England, where Northampton is situated, were examined during the outbreak (24 June – 18 July 2008) and compared with those for England and Wales. A series of control charts for diarrhoea calls are routinely used to monitor significant rises in the numbers of calls received. Control charts are calculated by assuming that calls follow a Poisson distribution with the total number of calls as an offset: a model is fitted to each region and symptom separately [29]. The model takes into account call variation caused by weekends, public holidays and the time of year – variables that can affect the number of calls received by NHS Direct. A value above the upper limit of the 99.5% confidence interval would be considered to be unusual. The seven-day moving average for diarrhoea calls was also monitored. The number and percentage of calls for diarrhoea in the East Midlands region were presented by call outcome and the number of calls in the Northampton (NN) postcode districts and in particular the number of calls in the NN11 and NN12 postcode districts, which were most affected by the incident. QSurveillance national consultation rates per 100,000 population for diarrhoea (in the age groups under five years, five years and over, and all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with rates for the same period in 2007 (data not presented). Consultation rates by region for 2008 for diarrhoea (all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with those for the East Midlands region. The gastroenteritis indicator includes all cases of diarrhoea and/or vomiting. Consultation rates and standardised incidence ratios (SIRs) – calculated using the UK as the standard population – for diarrhoea, gastroenteritis and vomiting were compared for the UK, Yorkshire and Humberside, East Midlands, Leicestershire, Northamptonshire and Rutland SHA, and Daventry and South Northants PCT, Northamptonshire Heartlands PCT and Northampton PCT. Yorkshire and Humberside was not an affected region but was included as a control. The area supplied by the Pitsford Reservoir included the three PCTs, which were all within the Leicestershire, Northamptonshire and Rutland SHA. The consultation rates and SIRs were compared for the period from week 16 to week 35 of 2008 in order to compare the rates before and after the Cryptosporidium exceedance, which took place in week 26. Estimates of excess numbers of cases of diarrhoea occurring during and following the Cryptosporidium outbreak were made by calculating the mean consultation rate over a period of five weeks before and after the incident (weeks 20–24 and weeks 31–35, respectively). For each of the three PCTs, the calculated mean rate was applied to the PCT population to estimate the number of cases that would be expected each week. The actual consultation rates for diarrhoea for weeks 25 to 30 were used to estimate the number of cases for the PCT population each week. The expected number of cases was subtracted from the estimated number of cases in the PCT population to give the estimated number of excess cases. Results HPA/NHS Direct surveillance system A peak in the number of calls for diarrhoea in the East Midlands was recorded in 25–26 June 2008, the period that coincided with the contamination incident and the associated media coverage (Figure 1). The neighbouring areas of the West Midlands, Yorkshire and the Humber, and East of England showed no increase in the number of calls for diarrhoea. The peak produced a control chart exceedance for calls for diarrhoea on 25 June 2008 (Figure 2), when the proportion of calls exceeded the upper limit of the 99.5% confidence interval. There were further confidence interval exceedances on 26 and 28 June (which were not control chart exceedances). There was no peak in calls for vomiting or control chart exceedance for these calls in the East Midlands. HPA/QSurveillance national surveillance system The East Midlands region had significantly high consultation rates for diarrhoea and gastroenteritis in week 25 (16–22 June), week 26 (23–29 June 2008, when the contamination incident was reported) and in the following four weeks. Within this region. Leicestershire, Northamptonshire and Rutland SHA had slightly raised consultation rates and significant SIRs across weeks 25 to 30 that were not seen in the neighbouring Trent SHA. At PCT level, all three of the PCTs in the area affected by the incident showed increased consultation rates for diarrhoea (Table 1) and gastroenteritis (Table 2) with SIRs significantly above the UK rate in week 26. Daventry and South Northants PCT also had a raised SIR for both indicators in week 25, and although Northamptonshire Heartlands and Northampton PCTs did not have SIRs significantly above that of the UK in week 25, the rise in consultation rates for diarrhoea and gastroenteritis began during week 25. In Northampton PCT, consultations for both diarrhoea and gastroenteritis peaked in the week following the contamination incident, week 27, returning to normal levels by week 30 (Figure 3A and 3B). A similar effect can be seen in Northampton Heartlands PCT. Daventry and South Northants PCT also showed an increase, but appeared to have consistently higher rates for both indicators. This was the area with the smallest population so the rates were more variable than in the other PCTs and we therefore interpreted these results with caution. The consultation rates for vomiting during weeks 25 to 30 in the East Midlands were not unusual at SHA or PCT level (data not presented). Discussion We have demonstrated the sensitivity of syndromic surveillance in detecting this small Cryptosporidium outbreak and the value of the surveillance in being able to describe the extent of its spread. Both the HPA/NHS Direct and HPA/QSurveillance systems showed demonstrable increases in calls and consultations for diarrhoea that were linked to the outbreak. QSurveillance consultations appeared to increase across the PCTs immediately affected but not in the surrounding area. Both the HPA/NHS Direct and HPA/QSurveillance systems showed a clear signal at the time of the incident and we were able to describe the extent of the impact on pre-primary care and primary care services. The HPA/QSurveillance system showed a rise in consultation rates for gastrointestinal symptoms that began the week before the outbreak, consistent with the period when Cryptosporidium was present in the water leaving the Pitsford Reservoir (19–24 June 2008) and with the onset of symptoms in the first outbreak case on 24 June. Although only 33 cases were identified by the outbreak investigation team, of which 23 were confirmed as having the outbreak Cryptosporidium strain, our syndromic surveillance data detected this limited outbreak. Data also suggested a more widespread increase in general gastrointestinal symptoms around the time of the outbreak, with an estimated 422 excess diarrhoea cases; these excess cases represented an increase of about 25% above normally expected levels. It is highly probable that a proportion of these excess cases may have resulted from the increased publicity surrounding the incident – for example, it is likely that media reports contributed to the large peak in calls detected by the HPA/NHS Direct surveillance system on the day the boil water notice was issued, and could also have impacted on the GP consultation rate. It has been previously shown that reporting of mumps cases is sensitive to media coverage, with a rise in clinically reported cases following newspaper reports [30]. A similar mechanism could account for some of the excess GP consultations as cases experiencing gastrointestinal symptoms may have been more likely to consult their GP, whereas in normal circumstances they would have cared for themselves at home. It is also possible that the surveillance shows outbreak-associated cases that did not come to the attention of the outbreak team, perhaps because symptoms were not sufficiently severe to warrant further investigation, or stool samples were not provided for testing. It is interesting to note that there was no demonstrable impact on the number of calls for vomiting (which is not a prominent clinical feature of cryptosporidiosis). Other common community-based pathogens such as norovirus and rotavirus were at low levels, as is normal for that time of year [31]. In this instance, public health authorities had already been alerted to a potential problem by the water company, although the extent of the outbreak was detected by syndromic surveillance. In 2003 the syndromic surveillance systems in the city of New York, United States, were able to detect an increase in diarrhoeal illness following a power outage when there was no other indication of citywide illness [32]. The New York system covers a population of nine million, but does not regularly detect localised outbreaks [33]. It has been shown previously that the HPA/NHS Direct surveillance system would be unlikely to detect a Cryptosporidium outbreak unless call volumes are high (72% chance of detection if nine-tenths of cases called NHS Direct) [29], although the value of syndromic surveillance for such outbreaks has been recognised [34]. The system detected the East Midlands Cryptosporidium outbreak that affected a smaller population than that covered by the New York system. The three PCTs affected have a combined population of around 600,000, of which just over half use GP practices reporting to QSurveillance, yet this syndromic surveillance system was able to describe an increase in consultation rates for diarrhoea and gastroenteritis around the time of the outbreak. Limitations of the data There was extensive media reporting of the incident that may have affected both the HPA/NHS Direct and HPA/QSurveillance systems and contributed to the increase in reported gastrointestinal symptoms around the time of the contamination incident. However, the rise in consultation rates for diarrhoea began before the outbreak had been detected and therefore cannot be attributed to media coverage. The HPA/NHS Direct and HPA/QSurveillance systems monitor general symptoms and so could only monitor the relevant symptoms of diarrhoea and vomiting. They are not able to detect Cryptosporidium cases, as this would require laboratory confirmation of diagnosis, so some of the estimated excess cases could be unconnected with this incident. This outbreak was discovered by other means but both the HPA/NHS Direct and HPA/ QSurveillance systems were able to describe the extent of the disease in the general population and provide reassurance that there was no widespread impact. Compared with other populations, older people and ethnic minorities are less likely to call NHS Direct [29], and although this should not prevent detection of gastrointestinal symptoms as a result of drinking water contamination as this would affect the whole population, this may reduce the signal from the system [35]. With such large surveillance systems, there will be ‘background noise’ in the data, so procedures must be in place to correctly interpret the data and set appropriate thresholds for action. Conclusion To our knowledge, this is the first time that PCT-level data from a syndromic surveillance system, the HPA/ QSurveillance national surveillance system, have been able to show the extent of such a limited outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide PCT-level data and this Cryptosporidium contamination incident demonstrates the potential usefulness of this system.

How many people were dead?

1512

Value of syndromic surveillance in monitoring a focal waterborne outbreak due to an unusual Cryptosporidium genotype in Northamptonshire

The United Kingdom (UK) has several national syndromic surveillance systems. The Health Protection Agency (HPA)/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data from a national telephone helpline, while the HPA/ QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems. Data from both of these systems were used to monitor a local outbreak of cryptosporidiosis that occurred following Cryptosporidium oocyst contamination of drinking water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, in June 2008. There was a peak in the number of calls to NHS Direct concerning diarrhoea that coincided with the incident. QSurveillance data for the local areas affected by the outbreak showed a significant increase in GP consultations for diarrhoea and gastroenteritis in the week of the incident but there was no increase in consultations for vomiting. A total of 33 clinical cases of cryptosporidiosis were identified in the outbreak investigation, of which 23 were confirmed as infected with the outbreak strain. However, QSurveillance data suggest that there were an estimated 422 excess diarrhoea cases during the outbreak, an increase of about 25% over baseline weekly levels. To our knowledge, this is the first time that data from a syndromic surveillance system, the HPA/QSurveillance national surveillance system, have been able to show the extent of such a small outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide data at local health district (primary care trust) level. The Cryptosporidium contamination incident described demonstrates the potential usefulness of this information, as it is unusual for syndromic surveillance systems to be able to help monitor such a small-scale outbreak. Introduction As syndromic surveillance systems usually capture data already collected for other purposes, and monitor generic symptoms and/or clinically diagnosed disease, they provide information at an earlier stage of illness (compared with laboratory-confirmed diagnoses), so that action can be taken in time to substantially reduce the impact of disease. Some systems, for example, the Royal College of General Practitioners Weekly Returns Service, are now well established, with many years of historical data that allow monitoring of longer-term disease trends [1]. They have the ability to provide early warning of, for example, seasonal rises in influenza and can trigger public health action, such as a recommendation to prescribe antiviral drugs in line with national guidance [2-4]. They can also provide reassurance to incident response teams and the general public that an incident has not caused adverse health effects – for example, following an explosion at the Buncefield oil storage depot in Hemel Hempstead, United Kingdom (UK), in 2005, syndromic surveillance confirmed that there were no unusual rises in community-based morbidity linked to the incident [5]; following the eruption of the Eyjafjallajökull volcano in Iceland in April 2010 similar assurance was given about lack of impact on community morbidity [6]. Health departments are increasingly expected to monitor health effects of natural events such as heat wave or flooding, or implement surveillance – of which syndromic surveillance plays a major role – for mass gatherings such as the Olympics or football World Cup [7-9]. Systems in France, Australia and Taiwan use data from emergency departments [10-12], a Canadian system uses over-the-counter pharmacy sales [13,14], and in the Netherlands data from both syndromic and surrogate data sources, such as employee absence records and prescription medications dispensed by pharmacies, are included in surveillance systems [15,16]. Currently systems based on Internet searches via search engines or on queries submitted to medical websites are being developed [17,18]. In the UK, the HPA/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data collected from the NHS Direct telephone helpline [19], while the HPA/QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems [20]. The HPA Real-time Syndromic Surveillance Team is a small team that coordinates a number of syndromic surveillance systems within the HPA and takes a lead for syndromic surveillance in England [21]. This paper describes the support provided by the team to the local incident management team during a local cryptosporidiosis outbreak and shows the use of syndromic surveillance in monitoring the extent of an outbreak using the HPA/NHS Direct and HPA/QSurveillance national surveillance systems. Cryptosporidiosis Cryptosporidium is a protozoan parasite that can cause an infection in people, cattle and sometimes other animals [22]. Cryptosporidiosis is most common in children aged between one and five years, but it can affect all ages. Those with impaired immune systems are likely to be most seriously affected. Symptoms usually appear between three and 12 days after initial exposure and include watery diarrhoea, stomach pains, dehydration and fever. In its transmissible form, called an oocyst, the parasite is protected by an outer shell, which allows it to survive in the environment for a long time. Transmission occurs most often via the faeco-oral route through person-to-person or animalto-person contact, but people may also be infected by consuming contaminated water or food or by swimming in contaminated water. Although uncommon, the largest outbreaks have occurred following contamination of drinking water [23,24]. Normal chlorine disinfection procedures do not kill the oocysts, so they are removed by filtration and water companies carry out routine monitoring of treated water. Description of the incident On 25 June 2008 the local Health Protection Unit was informed by Anglian Water of an exceedence in the level of Cryptosporidium oocysts found in water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, during 19 to 24 June 2008 [25]. The reservoir supplied a population of more than 250,000 in the Northampton area. A notice advising people in the affected areas to boil all drinking water was issued on 25 June 2008 and public health messages were circulated to local health services and to the general public via the media. Those members of the public who were concerned about health risks associated with the incident were asked to ring NHS Direct for clinical advice [26]. The HPA wrote to local GPs and hospitals asking them to monitor potential patients for signs and symptoms of Cryptosporidium infection and to submit faecal specimens to the local hospital diagnostic laboratory if patients presented with diarrhoea. Samples from 34 patients where Cryptosporidium infection was identified were sent to the UK Cryptosporidium reference unit for typing. On 30 June 2008, the Cryptosporidium oocysts found in the reservoir water were confirmed as being of the rabbit genotype Cryptosporidium cuniculus [27]. Subsequently, a dead rabbit was found in a treated water tank at the water treatment works. The genotype of Cryptosporidium oocysts in the rabbit’s large bowel was indistinguishable from that of the oocysts found in the water [27]. After remediation of the water supply and distribution, the ‘boil water notice’ was lifted on 4 July and the following day the first case of cryptosporidiosis linked to the incident was identified by the reference laboratory (this case was infected with C. cuniculus). During the course of the outbreak (24 June – 18 July 2008, the dates of symptom onset in the first and last case, respectively), 23 cases of cryptosporidiosis were confirmed as being infected with C. cuniculus; one of the 23 was a secondary case. The HPA Real-time Syndromic Surveillance Team provided data in order to aid the response to this incident and the first syndromic surveillance report was circulated to the incident management team and other relevant people in the HPA on 27 June 2008. Data from the HPA/NHS Direct and HPA/QSurveillance systems were provided in a series of regular reports, initially daily and eventually weekly, until the final report on 21 August 2008. Each report included a summary interpretation and more detailed data on diarrhoea, gastroenteritis and vomiting indicators. Methods Surveillance systems HPA/NHS Direct surveillance system NHS Direct is a 24-hour nurse-led telephone helpline that provides health information and advice to the general public. Nurses use a computerised clinical decision support system – the NHS Clinical Assessment System (NHS CAS) – to handle calls. This assessment system uses approximately 200 computerised symptom-based clinical algorithms. Nurses assign the call to the most appropriate algorithm and the patient’s symptoms determine the questions asked and the action to be taken following the call (call outcome), which could be guidance on self-care or they could be referred to their GP or advised to attend a hospital emergency department. No attempt is made to provide a formal diagnosis. Daily NHS Direct data are received by the Real-time Syndromic Surveillance Team, where the number and type of calls received during the previous day are analysed and interpreted. Call proportions are calculated by age group and algorithm against the total number of calls received. HPA/QSurveillance system The HPA/QSurveillance national surveillance system was set up by the University of Nottingham, United Kingdom, and Egton Medical Information Systems (EMIS), a supplier of general practice computer systems, in collaboration with the HPA. It comprises a network of more than 3,500 general practices throughout the UK, covering more than 22 million patients (about 38% of the population [28]). Aggregated data on GP consultations for a range of indicators are automatically uploaded daily from GP practice systems to a central database. Data are routinely reported on a weekly basis; however, daily reporting is possible for specific incidents. Reports are provided at national or regional level (strategic health authority, SHA) and by local health district (primary care trust, PCT). Analysis of surveillance data NHS Direct call proportions for gastrointestinal syndromes (diarrhoea and vomiting) for the East Midlands region in England, where Northampton is situated, were examined during the outbreak (24 June – 18 July 2008) and compared with those for England and Wales. A series of control charts for diarrhoea calls are routinely used to monitor significant rises in the numbers of calls received. Control charts are calculated by assuming that calls follow a Poisson distribution with the total number of calls as an offset: a model is fitted to each region and symptom separately [29]. The model takes into account call variation caused by weekends, public holidays and the time of year – variables that can affect the number of calls received by NHS Direct. A value above the upper limit of the 99.5% confidence interval would be considered to be unusual. The seven-day moving average for diarrhoea calls was also monitored. The number and percentage of calls for diarrhoea in the East Midlands region were presented by call outcome and the number of calls in the Northampton (NN) postcode districts and in particular the number of calls in the NN11 and NN12 postcode districts, which were most affected by the incident. QSurveillance national consultation rates per 100,000 population for diarrhoea (in the age groups under five years, five years and over, and all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with rates for the same period in 2007 (data not presented). Consultation rates by region for 2008 for diarrhoea (all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with those for the East Midlands region. The gastroenteritis indicator includes all cases of diarrhoea and/or vomiting. Consultation rates and standardised incidence ratios (SIRs) – calculated using the UK as the standard population – for diarrhoea, gastroenteritis and vomiting were compared for the UK, Yorkshire and Humberside, East Midlands, Leicestershire, Northamptonshire and Rutland SHA, and Daventry and South Northants PCT, Northamptonshire Heartlands PCT and Northampton PCT. Yorkshire and Humberside was not an affected region but was included as a control. The area supplied by the Pitsford Reservoir included the three PCTs, which were all within the Leicestershire, Northamptonshire and Rutland SHA. The consultation rates and SIRs were compared for the period from week 16 to week 35 of 2008 in order to compare the rates before and after the Cryptosporidium exceedance, which took place in week 26. Estimates of excess numbers of cases of diarrhoea occurring during and following the Cryptosporidium outbreak were made by calculating the mean consultation rate over a period of five weeks before and after the incident (weeks 20–24 and weeks 31–35, respectively). For each of the three PCTs, the calculated mean rate was applied to the PCT population to estimate the number of cases that would be expected each week. The actual consultation rates for diarrhoea for weeks 25 to 30 were used to estimate the number of cases for the PCT population each week. The expected number of cases was subtracted from the estimated number of cases in the PCT population to give the estimated number of excess cases. Results HPA/NHS Direct surveillance system A peak in the number of calls for diarrhoea in the East Midlands was recorded in 25–26 June 2008, the period that coincided with the contamination incident and the associated media coverage (Figure 1). The neighbouring areas of the West Midlands, Yorkshire and the Humber, and East of England showed no increase in the number of calls for diarrhoea. The peak produced a control chart exceedance for calls for diarrhoea on 25 June 2008 (Figure 2), when the proportion of calls exceeded the upper limit of the 99.5% confidence interval. There were further confidence interval exceedances on 26 and 28 June (which were not control chart exceedances). There was no peak in calls for vomiting or control chart exceedance for these calls in the East Midlands. HPA/QSurveillance national surveillance system The East Midlands region had significantly high consultation rates for diarrhoea and gastroenteritis in week 25 (16–22 June), week 26 (23–29 June 2008, when the contamination incident was reported) and in the following four weeks. Within this region. Leicestershire, Northamptonshire and Rutland SHA had slightly raised consultation rates and significant SIRs across weeks 25 to 30 that were not seen in the neighbouring Trent SHA. At PCT level, all three of the PCTs in the area affected by the incident showed increased consultation rates for diarrhoea (Table 1) and gastroenteritis (Table 2) with SIRs significantly above the UK rate in week 26. Daventry and South Northants PCT also had a raised SIR for both indicators in week 25, and although Northamptonshire Heartlands and Northampton PCTs did not have SIRs significantly above that of the UK in week 25, the rise in consultation rates for diarrhoea and gastroenteritis began during week 25. In Northampton PCT, consultations for both diarrhoea and gastroenteritis peaked in the week following the contamination incident, week 27, returning to normal levels by week 30 (Figure 3A and 3B). A similar effect can be seen in Northampton Heartlands PCT. Daventry and South Northants PCT also showed an increase, but appeared to have consistently higher rates for both indicators. This was the area with the smallest population so the rates were more variable than in the other PCTs and we therefore interpreted these results with caution. The consultation rates for vomiting during weeks 25 to 30 in the East Midlands were not unusual at SHA or PCT level (data not presented). Discussion We have demonstrated the sensitivity of syndromic surveillance in detecting this small Cryptosporidium outbreak and the value of the surveillance in being able to describe the extent of its spread. Both the HPA/NHS Direct and HPA/QSurveillance systems showed demonstrable increases in calls and consultations for diarrhoea that were linked to the outbreak. QSurveillance consultations appeared to increase across the PCTs immediately affected but not in the surrounding area. Both the HPA/NHS Direct and HPA/QSurveillance systems showed a clear signal at the time of the incident and we were able to describe the extent of the impact on pre-primary care and primary care services. The HPA/QSurveillance system showed a rise in consultation rates for gastrointestinal symptoms that began the week before the outbreak, consistent with the period when Cryptosporidium was present in the water leaving the Pitsford Reservoir (19–24 June 2008) and with the onset of symptoms in the first outbreak case on 24 June. Although only 33 cases were identified by the outbreak investigation team, of which 23 were confirmed as having the outbreak Cryptosporidium strain, our syndromic surveillance data detected this limited outbreak. Data also suggested a more widespread increase in general gastrointestinal symptoms around the time of the outbreak, with an estimated 422 excess diarrhoea cases; these excess cases represented an increase of about 25% above normally expected levels. It is highly probable that a proportion of these excess cases may have resulted from the increased publicity surrounding the incident – for example, it is likely that media reports contributed to the large peak in calls detected by the HPA/NHS Direct surveillance system on the day the boil water notice was issued, and could also have impacted on the GP consultation rate. It has been previously shown that reporting of mumps cases is sensitive to media coverage, with a rise in clinically reported cases following newspaper reports [30]. A similar mechanism could account for some of the excess GP consultations as cases experiencing gastrointestinal symptoms may have been more likely to consult their GP, whereas in normal circumstances they would have cared for themselves at home. It is also possible that the surveillance shows outbreak-associated cases that did not come to the attention of the outbreak team, perhaps because symptoms were not sufficiently severe to warrant further investigation, or stool samples were not provided for testing. It is interesting to note that there was no demonstrable impact on the number of calls for vomiting (which is not a prominent clinical feature of cryptosporidiosis). Other common community-based pathogens such as norovirus and rotavirus were at low levels, as is normal for that time of year [31]. In this instance, public health authorities had already been alerted to a potential problem by the water company, although the extent of the outbreak was detected by syndromic surveillance. In 2003 the syndromic surveillance systems in the city of New York, United States, were able to detect an increase in diarrhoeal illness following a power outage when there was no other indication of citywide illness [32]. The New York system covers a population of nine million, but does not regularly detect localised outbreaks [33]. It has been shown previously that the HPA/NHS Direct surveillance system would be unlikely to detect a Cryptosporidium outbreak unless call volumes are high (72% chance of detection if nine-tenths of cases called NHS Direct) [29], although the value of syndromic surveillance for such outbreaks has been recognised [34]. The system detected the East Midlands Cryptosporidium outbreak that affected a smaller population than that covered by the New York system. The three PCTs affected have a combined population of around 600,000, of which just over half use GP practices reporting to QSurveillance, yet this syndromic surveillance system was able to describe an increase in consultation rates for diarrhoea and gastroenteritis around the time of the outbreak. Limitations of the data There was extensive media reporting of the incident that may have affected both the HPA/NHS Direct and HPA/QSurveillance systems and contributed to the increase in reported gastrointestinal symptoms around the time of the contamination incident. However, the rise in consultation rates for diarrhoea began before the outbreak had been detected and therefore cannot be attributed to media coverage. The HPA/NHS Direct and HPA/QSurveillance systems monitor general symptoms and so could only monitor the relevant symptoms of diarrhoea and vomiting. They are not able to detect Cryptosporidium cases, as this would require laboratory confirmation of diagnosis, so some of the estimated excess cases could be unconnected with this incident. This outbreak was discovered by other means but both the HPA/NHS Direct and HPA/ QSurveillance systems were able to describe the extent of the disease in the general population and provide reassurance that there was no widespread impact. Compared with other populations, older people and ethnic minorities are less likely to call NHS Direct [29], and although this should not prevent detection of gastrointestinal symptoms as a result of drinking water contamination as this would affect the whole population, this may reduce the signal from the system [35]. With such large surveillance systems, there will be ‘background noise’ in the data, so procedures must be in place to correctly interpret the data and set appropriate thresholds for action. Conclusion To our knowledge, this is the first time that PCT-level data from a syndromic surveillance system, the HPA/ QSurveillance national surveillance system, have been able to show the extent of such a limited outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide PCT-level data and this Cryptosporidium contamination incident demonstrates the potential usefulness of this system.

Which contaminants or viruses or bacteria were found?

1513

Value of syndromic surveillance in monitoring a focal waterborne outbreak due to an unusual Cryptosporidium genotype in Northamptonshire

The United Kingdom (UK) has several national syndromic surveillance systems. The Health Protection Agency (HPA)/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data from a national telephone helpline, while the HPA/ QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems. Data from both of these systems were used to monitor a local outbreak of cryptosporidiosis that occurred following Cryptosporidium oocyst contamination of drinking water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, in June 2008. There was a peak in the number of calls to NHS Direct concerning diarrhoea that coincided with the incident. QSurveillance data for the local areas affected by the outbreak showed a significant increase in GP consultations for diarrhoea and gastroenteritis in the week of the incident but there was no increase in consultations for vomiting. A total of 33 clinical cases of cryptosporidiosis were identified in the outbreak investigation, of which 23 were confirmed as infected with the outbreak strain. However, QSurveillance data suggest that there were an estimated 422 excess diarrhoea cases during the outbreak, an increase of about 25% over baseline weekly levels. To our knowledge, this is the first time that data from a syndromic surveillance system, the HPA/QSurveillance national surveillance system, have been able to show the extent of such a small outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide data at local health district (primary care trust) level. The Cryptosporidium contamination incident described demonstrates the potential usefulness of this information, as it is unusual for syndromic surveillance systems to be able to help monitor such a small-scale outbreak. Introduction As syndromic surveillance systems usually capture data already collected for other purposes, and monitor generic symptoms and/or clinically diagnosed disease, they provide information at an earlier stage of illness (compared with laboratory-confirmed diagnoses), so that action can be taken in time to substantially reduce the impact of disease. Some systems, for example, the Royal College of General Practitioners Weekly Returns Service, are now well established, with many years of historical data that allow monitoring of longer-term disease trends [1]. They have the ability to provide early warning of, for example, seasonal rises in influenza and can trigger public health action, such as a recommendation to prescribe antiviral drugs in line with national guidance [2-4]. They can also provide reassurance to incident response teams and the general public that an incident has not caused adverse health effects – for example, following an explosion at the Buncefield oil storage depot in Hemel Hempstead, United Kingdom (UK), in 2005, syndromic surveillance confirmed that there were no unusual rises in community-based morbidity linked to the incident [5]; following the eruption of the Eyjafjallajökull volcano in Iceland in April 2010 similar assurance was given about lack of impact on community morbidity [6]. Health departments are increasingly expected to monitor health effects of natural events such as heat wave or flooding, or implement surveillance – of which syndromic surveillance plays a major role – for mass gatherings such as the Olympics or football World Cup [7-9]. Systems in France, Australia and Taiwan use data from emergency departments [10-12], a Canadian system uses over-the-counter pharmacy sales [13,14], and in the Netherlands data from both syndromic and surrogate data sources, such as employee absence records and prescription medications dispensed by pharmacies, are included in surveillance systems [15,16]. Currently systems based on Internet searches via search engines or on queries submitted to medical websites are being developed [17,18]. In the UK, the HPA/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data collected from the NHS Direct telephone helpline [19], while the HPA/QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems [20]. The HPA Real-time Syndromic Surveillance Team is a small team that coordinates a number of syndromic surveillance systems within the HPA and takes a lead for syndromic surveillance in England [21]. This paper describes the support provided by the team to the local incident management team during a local cryptosporidiosis outbreak and shows the use of syndromic surveillance in monitoring the extent of an outbreak using the HPA/NHS Direct and HPA/QSurveillance national surveillance systems. Cryptosporidiosis Cryptosporidium is a protozoan parasite that can cause an infection in people, cattle and sometimes other animals [22]. Cryptosporidiosis is most common in children aged between one and five years, but it can affect all ages. Those with impaired immune systems are likely to be most seriously affected. Symptoms usually appear between three and 12 days after initial exposure and include watery diarrhoea, stomach pains, dehydration and fever. In its transmissible form, called an oocyst, the parasite is protected by an outer shell, which allows it to survive in the environment for a long time. Transmission occurs most often via the faeco-oral route through person-to-person or animalto-person contact, but people may also be infected by consuming contaminated water or food or by swimming in contaminated water. Although uncommon, the largest outbreaks have occurred following contamination of drinking water [23,24]. Normal chlorine disinfection procedures do not kill the oocysts, so they are removed by filtration and water companies carry out routine monitoring of treated water. Description of the incident On 25 June 2008 the local Health Protection Unit was informed by Anglian Water of an exceedence in the level of Cryptosporidium oocysts found in water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, during 19 to 24 June 2008 [25]. The reservoir supplied a population of more than 250,000 in the Northampton area. A notice advising people in the affected areas to boil all drinking water was issued on 25 June 2008 and public health messages were circulated to local health services and to the general public via the media. Those members of the public who were concerned about health risks associated with the incident were asked to ring NHS Direct for clinical advice [26]. The HPA wrote to local GPs and hospitals asking them to monitor potential patients for signs and symptoms of Cryptosporidium infection and to submit faecal specimens to the local hospital diagnostic laboratory if patients presented with diarrhoea. Samples from 34 patients where Cryptosporidium infection was identified were sent to the UK Cryptosporidium reference unit for typing. On 30 June 2008, the Cryptosporidium oocysts found in the reservoir water were confirmed as being of the rabbit genotype Cryptosporidium cuniculus [27]. Subsequently, a dead rabbit was found in a treated water tank at the water treatment works. The genotype of Cryptosporidium oocysts in the rabbit’s large bowel was indistinguishable from that of the oocysts found in the water [27]. After remediation of the water supply and distribution, the ‘boil water notice’ was lifted on 4 July and the following day the first case of cryptosporidiosis linked to the incident was identified by the reference laboratory (this case was infected with C. cuniculus). During the course of the outbreak (24 June – 18 July 2008, the dates of symptom onset in the first and last case, respectively), 23 cases of cryptosporidiosis were confirmed as being infected with C. cuniculus; one of the 23 was a secondary case. The HPA Real-time Syndromic Surveillance Team provided data in order to aid the response to this incident and the first syndromic surveillance report was circulated to the incident management team and other relevant people in the HPA on 27 June 2008. Data from the HPA/NHS Direct and HPA/QSurveillance systems were provided in a series of regular reports, initially daily and eventually weekly, until the final report on 21 August 2008. Each report included a summary interpretation and more detailed data on diarrhoea, gastroenteritis and vomiting indicators. Methods Surveillance systems HPA/NHS Direct surveillance system NHS Direct is a 24-hour nurse-led telephone helpline that provides health information and advice to the general public. Nurses use a computerised clinical decision support system – the NHS Clinical Assessment System (NHS CAS) – to handle calls. This assessment system uses approximately 200 computerised symptom-based clinical algorithms. Nurses assign the call to the most appropriate algorithm and the patient’s symptoms determine the questions asked and the action to be taken following the call (call outcome), which could be guidance on self-care or they could be referred to their GP or advised to attend a hospital emergency department. No attempt is made to provide a formal diagnosis. Daily NHS Direct data are received by the Real-time Syndromic Surveillance Team, where the number and type of calls received during the previous day are analysed and interpreted. Call proportions are calculated by age group and algorithm against the total number of calls received. HPA/QSurveillance system The HPA/QSurveillance national surveillance system was set up by the University of Nottingham, United Kingdom, and Egton Medical Information Systems (EMIS), a supplier of general practice computer systems, in collaboration with the HPA. It comprises a network of more than 3,500 general practices throughout the UK, covering more than 22 million patients (about 38% of the population [28]). Aggregated data on GP consultations for a range of indicators are automatically uploaded daily from GP practice systems to a central database. Data are routinely reported on a weekly basis; however, daily reporting is possible for specific incidents. Reports are provided at national or regional level (strategic health authority, SHA) and by local health district (primary care trust, PCT). Analysis of surveillance data NHS Direct call proportions for gastrointestinal syndromes (diarrhoea and vomiting) for the East Midlands region in England, where Northampton is situated, were examined during the outbreak (24 June – 18 July 2008) and compared with those for England and Wales. A series of control charts for diarrhoea calls are routinely used to monitor significant rises in the numbers of calls received. Control charts are calculated by assuming that calls follow a Poisson distribution with the total number of calls as an offset: a model is fitted to each region and symptom separately [29]. The model takes into account call variation caused by weekends, public holidays and the time of year – variables that can affect the number of calls received by NHS Direct. A value above the upper limit of the 99.5% confidence interval would be considered to be unusual. The seven-day moving average for diarrhoea calls was also monitored. The number and percentage of calls for diarrhoea in the East Midlands region were presented by call outcome and the number of calls in the Northampton (NN) postcode districts and in particular the number of calls in the NN11 and NN12 postcode districts, which were most affected by the incident. QSurveillance national consultation rates per 100,000 population for diarrhoea (in the age groups under five years, five years and over, and all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with rates for the same period in 2007 (data not presented). Consultation rates by region for 2008 for diarrhoea (all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with those for the East Midlands region. The gastroenteritis indicator includes all cases of diarrhoea and/or vomiting. Consultation rates and standardised incidence ratios (SIRs) – calculated using the UK as the standard population – for diarrhoea, gastroenteritis and vomiting were compared for the UK, Yorkshire and Humberside, East Midlands, Leicestershire, Northamptonshire and Rutland SHA, and Daventry and South Northants PCT, Northamptonshire Heartlands PCT and Northampton PCT. Yorkshire and Humberside was not an affected region but was included as a control. The area supplied by the Pitsford Reservoir included the three PCTs, which were all within the Leicestershire, Northamptonshire and Rutland SHA. The consultation rates and SIRs were compared for the period from week 16 to week 35 of 2008 in order to compare the rates before and after the Cryptosporidium exceedance, which took place in week 26. Estimates of excess numbers of cases of diarrhoea occurring during and following the Cryptosporidium outbreak were made by calculating the mean consultation rate over a period of five weeks before and after the incident (weeks 20–24 and weeks 31–35, respectively). For each of the three PCTs, the calculated mean rate was applied to the PCT population to estimate the number of cases that would be expected each week. The actual consultation rates for diarrhoea for weeks 25 to 30 were used to estimate the number of cases for the PCT population each week. The expected number of cases was subtracted from the estimated number of cases in the PCT population to give the estimated number of excess cases. Results HPA/NHS Direct surveillance system A peak in the number of calls for diarrhoea in the East Midlands was recorded in 25–26 June 2008, the period that coincided with the contamination incident and the associated media coverage (Figure 1). The neighbouring areas of the West Midlands, Yorkshire and the Humber, and East of England showed no increase in the number of calls for diarrhoea. The peak produced a control chart exceedance for calls for diarrhoea on 25 June 2008 (Figure 2), when the proportion of calls exceeded the upper limit of the 99.5% confidence interval. There were further confidence interval exceedances on 26 and 28 June (which were not control chart exceedances). There was no peak in calls for vomiting or control chart exceedance for these calls in the East Midlands. HPA/QSurveillance national surveillance system The East Midlands region had significantly high consultation rates for diarrhoea and gastroenteritis in week 25 (16–22 June), week 26 (23–29 June 2008, when the contamination incident was reported) and in the following four weeks. Within this region. Leicestershire, Northamptonshire and Rutland SHA had slightly raised consultation rates and significant SIRs across weeks 25 to 30 that were not seen in the neighbouring Trent SHA. At PCT level, all three of the PCTs in the area affected by the incident showed increased consultation rates for diarrhoea (Table 1) and gastroenteritis (Table 2) with SIRs significantly above the UK rate in week 26. Daventry and South Northants PCT also had a raised SIR for both indicators in week 25, and although Northamptonshire Heartlands and Northampton PCTs did not have SIRs significantly above that of the UK in week 25, the rise in consultation rates for diarrhoea and gastroenteritis began during week 25. In Northampton PCT, consultations for both diarrhoea and gastroenteritis peaked in the week following the contamination incident, week 27, returning to normal levels by week 30 (Figure 3A and 3B). A similar effect can be seen in Northampton Heartlands PCT. Daventry and South Northants PCT also showed an increase, but appeared to have consistently higher rates for both indicators. This was the area with the smallest population so the rates were more variable than in the other PCTs and we therefore interpreted these results with caution. The consultation rates for vomiting during weeks 25 to 30 in the East Midlands were not unusual at SHA or PCT level (data not presented). Discussion We have demonstrated the sensitivity of syndromic surveillance in detecting this small Cryptosporidium outbreak and the value of the surveillance in being able to describe the extent of its spread. Both the HPA/NHS Direct and HPA/QSurveillance systems showed demonstrable increases in calls and consultations for diarrhoea that were linked to the outbreak. QSurveillance consultations appeared to increase across the PCTs immediately affected but not in the surrounding area. Both the HPA/NHS Direct and HPA/QSurveillance systems showed a clear signal at the time of the incident and we were able to describe the extent of the impact on pre-primary care and primary care services. The HPA/QSurveillance system showed a rise in consultation rates for gastrointestinal symptoms that began the week before the outbreak, consistent with the period when Cryptosporidium was present in the water leaving the Pitsford Reservoir (19–24 June 2008) and with the onset of symptoms in the first outbreak case on 24 June. Although only 33 cases were identified by the outbreak investigation team, of which 23 were confirmed as having the outbreak Cryptosporidium strain, our syndromic surveillance data detected this limited outbreak. Data also suggested a more widespread increase in general gastrointestinal symptoms around the time of the outbreak, with an estimated 422 excess diarrhoea cases; these excess cases represented an increase of about 25% above normally expected levels. It is highly probable that a proportion of these excess cases may have resulted from the increased publicity surrounding the incident – for example, it is likely that media reports contributed to the large peak in calls detected by the HPA/NHS Direct surveillance system on the day the boil water notice was issued, and could also have impacted on the GP consultation rate. It has been previously shown that reporting of mumps cases is sensitive to media coverage, with a rise in clinically reported cases following newspaper reports [30]. A similar mechanism could account for some of the excess GP consultations as cases experiencing gastrointestinal symptoms may have been more likely to consult their GP, whereas in normal circumstances they would have cared for themselves at home. It is also possible that the surveillance shows outbreak-associated cases that did not come to the attention of the outbreak team, perhaps because symptoms were not sufficiently severe to warrant further investigation, or stool samples were not provided for testing. It is interesting to note that there was no demonstrable impact on the number of calls for vomiting (which is not a prominent clinical feature of cryptosporidiosis). Other common community-based pathogens such as norovirus and rotavirus were at low levels, as is normal for that time of year [31]. In this instance, public health authorities had already been alerted to a potential problem by the water company, although the extent of the outbreak was detected by syndromic surveillance. In 2003 the syndromic surveillance systems in the city of New York, United States, were able to detect an increase in diarrhoeal illness following a power outage when there was no other indication of citywide illness [32]. The New York system covers a population of nine million, but does not regularly detect localised outbreaks [33]. It has been shown previously that the HPA/NHS Direct surveillance system would be unlikely to detect a Cryptosporidium outbreak unless call volumes are high (72% chance of detection if nine-tenths of cases called NHS Direct) [29], although the value of syndromic surveillance for such outbreaks has been recognised [34]. The system detected the East Midlands Cryptosporidium outbreak that affected a smaller population than that covered by the New York system. The three PCTs affected have a combined population of around 600,000, of which just over half use GP practices reporting to QSurveillance, yet this syndromic surveillance system was able to describe an increase in consultation rates for diarrhoea and gastroenteritis around the time of the outbreak. Limitations of the data There was extensive media reporting of the incident that may have affected both the HPA/NHS Direct and HPA/QSurveillance systems and contributed to the increase in reported gastrointestinal symptoms around the time of the contamination incident. However, the rise in consultation rates for diarrhoea began before the outbreak had been detected and therefore cannot be attributed to media coverage. The HPA/NHS Direct and HPA/QSurveillance systems monitor general symptoms and so could only monitor the relevant symptoms of diarrhoea and vomiting. They are not able to detect Cryptosporidium cases, as this would require laboratory confirmation of diagnosis, so some of the estimated excess cases could be unconnected with this incident. This outbreak was discovered by other means but both the HPA/NHS Direct and HPA/ QSurveillance systems were able to describe the extent of the disease in the general population and provide reassurance that there was no widespread impact. Compared with other populations, older people and ethnic minorities are less likely to call NHS Direct [29], and although this should not prevent detection of gastrointestinal symptoms as a result of drinking water contamination as this would affect the whole population, this may reduce the signal from the system [35]. With such large surveillance systems, there will be ‘background noise’ in the data, so procedures must be in place to correctly interpret the data and set appropriate thresholds for action. Conclusion To our knowledge, this is the first time that PCT-level data from a syndromic surveillance system, the HPA/ QSurveillance national surveillance system, have been able to show the extent of such a limited outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide PCT-level data and this Cryptosporidium contamination incident demonstrates the potential usefulness of this system.

Which were the symptoms?

1514

Value of syndromic surveillance in monitoring a focal waterborne outbreak due to an unusual Cryptosporidium genotype in Northamptonshire

The United Kingdom (UK) has several national syndromic surveillance systems. The Health Protection Agency (HPA)/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data from a national telephone helpline, while the HPA/ QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems. Data from both of these systems were used to monitor a local outbreak of cryptosporidiosis that occurred following Cryptosporidium oocyst contamination of drinking water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, in June 2008. There was a peak in the number of calls to NHS Direct concerning diarrhoea that coincided with the incident. QSurveillance data for the local areas affected by the outbreak showed a significant increase in GP consultations for diarrhoea and gastroenteritis in the week of the incident but there was no increase in consultations for vomiting. A total of 33 clinical cases of cryptosporidiosis were identified in the outbreak investigation, of which 23 were confirmed as infected with the outbreak strain. However, QSurveillance data suggest that there were an estimated 422 excess diarrhoea cases during the outbreak, an increase of about 25% over baseline weekly levels. To our knowledge, this is the first time that data from a syndromic surveillance system, the HPA/QSurveillance national surveillance system, have been able to show the extent of such a small outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide data at local health district (primary care trust) level. The Cryptosporidium contamination incident described demonstrates the potential usefulness of this information, as it is unusual for syndromic surveillance systems to be able to help monitor such a small-scale outbreak. Introduction As syndromic surveillance systems usually capture data already collected for other purposes, and monitor generic symptoms and/or clinically diagnosed disease, they provide information at an earlier stage of illness (compared with laboratory-confirmed diagnoses), so that action can be taken in time to substantially reduce the impact of disease. Some systems, for example, the Royal College of General Practitioners Weekly Returns Service, are now well established, with many years of historical data that allow monitoring of longer-term disease trends [1]. They have the ability to provide early warning of, for example, seasonal rises in influenza and can trigger public health action, such as a recommendation to prescribe antiviral drugs in line with national guidance [2-4]. They can also provide reassurance to incident response teams and the general public that an incident has not caused adverse health effects – for example, following an explosion at the Buncefield oil storage depot in Hemel Hempstead, United Kingdom (UK), in 2005, syndromic surveillance confirmed that there were no unusual rises in community-based morbidity linked to the incident [5]; following the eruption of the Eyjafjallajökull volcano in Iceland in April 2010 similar assurance was given about lack of impact on community morbidity [6]. Health departments are increasingly expected to monitor health effects of natural events such as heat wave or flooding, or implement surveillance – of which syndromic surveillance plays a major role – for mass gatherings such as the Olympics or football World Cup [7-9]. Systems in France, Australia and Taiwan use data from emergency departments [10-12], a Canadian system uses over-the-counter pharmacy sales [13,14], and in the Netherlands data from both syndromic and surrogate data sources, such as employee absence records and prescription medications dispensed by pharmacies, are included in surveillance systems [15,16]. Currently systems based on Internet searches via search engines or on queries submitted to medical websites are being developed [17,18]. In the UK, the HPA/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data collected from the NHS Direct telephone helpline [19], while the HPA/QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems [20]. The HPA Real-time Syndromic Surveillance Team is a small team that coordinates a number of syndromic surveillance systems within the HPA and takes a lead for syndromic surveillance in England [21]. This paper describes the support provided by the team to the local incident management team during a local cryptosporidiosis outbreak and shows the use of syndromic surveillance in monitoring the extent of an outbreak using the HPA/NHS Direct and HPA/QSurveillance national surveillance systems. Cryptosporidiosis Cryptosporidium is a protozoan parasite that can cause an infection in people, cattle and sometimes other animals [22]. Cryptosporidiosis is most common in children aged between one and five years, but it can affect all ages. Those with impaired immune systems are likely to be most seriously affected. Symptoms usually appear between three and 12 days after initial exposure and include watery diarrhoea, stomach pains, dehydration and fever. In its transmissible form, called an oocyst, the parasite is protected by an outer shell, which allows it to survive in the environment for a long time. Transmission occurs most often via the faeco-oral route through person-to-person or animalto-person contact, but people may also be infected by consuming contaminated water or food or by swimming in contaminated water. Although uncommon, the largest outbreaks have occurred following contamination of drinking water [23,24]. Normal chlorine disinfection procedures do not kill the oocysts, so they are removed by filtration and water companies carry out routine monitoring of treated water. Description of the incident On 25 June 2008 the local Health Protection Unit was informed by Anglian Water of an exceedence in the level of Cryptosporidium oocysts found in water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, during 19 to 24 June 2008 [25]. The reservoir supplied a population of more than 250,000 in the Northampton area. A notice advising people in the affected areas to boil all drinking water was issued on 25 June 2008 and public health messages were circulated to local health services and to the general public via the media. Those members of the public who were concerned about health risks associated with the incident were asked to ring NHS Direct for clinical advice [26]. The HPA wrote to local GPs and hospitals asking them to monitor potential patients for signs and symptoms of Cryptosporidium infection and to submit faecal specimens to the local hospital diagnostic laboratory if patients presented with diarrhoea. Samples from 34 patients where Cryptosporidium infection was identified were sent to the UK Cryptosporidium reference unit for typing. On 30 June 2008, the Cryptosporidium oocysts found in the reservoir water were confirmed as being of the rabbit genotype Cryptosporidium cuniculus [27]. Subsequently, a dead rabbit was found in a treated water tank at the water treatment works. The genotype of Cryptosporidium oocysts in the rabbit’s large bowel was indistinguishable from that of the oocysts found in the water [27]. After remediation of the water supply and distribution, the ‘boil water notice’ was lifted on 4 July and the following day the first case of cryptosporidiosis linked to the incident was identified by the reference laboratory (this case was infected with C. cuniculus). During the course of the outbreak (24 June – 18 July 2008, the dates of symptom onset in the first and last case, respectively), 23 cases of cryptosporidiosis were confirmed as being infected with C. cuniculus; one of the 23 was a secondary case. The HPA Real-time Syndromic Surveillance Team provided data in order to aid the response to this incident and the first syndromic surveillance report was circulated to the incident management team and other relevant people in the HPA on 27 June 2008. Data from the HPA/NHS Direct and HPA/QSurveillance systems were provided in a series of regular reports, initially daily and eventually weekly, until the final report on 21 August 2008. Each report included a summary interpretation and more detailed data on diarrhoea, gastroenteritis and vomiting indicators. Methods Surveillance systems HPA/NHS Direct surveillance system NHS Direct is a 24-hour nurse-led telephone helpline that provides health information and advice to the general public. Nurses use a computerised clinical decision support system – the NHS Clinical Assessment System (NHS CAS) – to handle calls. This assessment system uses approximately 200 computerised symptom-based clinical algorithms. Nurses assign the call to the most appropriate algorithm and the patient’s symptoms determine the questions asked and the action to be taken following the call (call outcome), which could be guidance on self-care or they could be referred to their GP or advised to attend a hospital emergency department. No attempt is made to provide a formal diagnosis. Daily NHS Direct data are received by the Real-time Syndromic Surveillance Team, where the number and type of calls received during the previous day are analysed and interpreted. Call proportions are calculated by age group and algorithm against the total number of calls received. HPA/QSurveillance system The HPA/QSurveillance national surveillance system was set up by the University of Nottingham, United Kingdom, and Egton Medical Information Systems (EMIS), a supplier of general practice computer systems, in collaboration with the HPA. It comprises a network of more than 3,500 general practices throughout the UK, covering more than 22 million patients (about 38% of the population [28]). Aggregated data on GP consultations for a range of indicators are automatically uploaded daily from GP practice systems to a central database. Data are routinely reported on a weekly basis; however, daily reporting is possible for specific incidents. Reports are provided at national or regional level (strategic health authority, SHA) and by local health district (primary care trust, PCT). Analysis of surveillance data NHS Direct call proportions for gastrointestinal syndromes (diarrhoea and vomiting) for the East Midlands region in England, where Northampton is situated, were examined during the outbreak (24 June – 18 July 2008) and compared with those for England and Wales. A series of control charts for diarrhoea calls are routinely used to monitor significant rises in the numbers of calls received. Control charts are calculated by assuming that calls follow a Poisson distribution with the total number of calls as an offset: a model is fitted to each region and symptom separately [29]. The model takes into account call variation caused by weekends, public holidays and the time of year – variables that can affect the number of calls received by NHS Direct. A value above the upper limit of the 99.5% confidence interval would be considered to be unusual. The seven-day moving average for diarrhoea calls was also monitored. The number and percentage of calls for diarrhoea in the East Midlands region were presented by call outcome and the number of calls in the Northampton (NN) postcode districts and in particular the number of calls in the NN11 and NN12 postcode districts, which were most affected by the incident. QSurveillance national consultation rates per 100,000 population for diarrhoea (in the age groups under five years, five years and over, and all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with rates for the same period in 2007 (data not presented). Consultation rates by region for 2008 for diarrhoea (all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with those for the East Midlands region. The gastroenteritis indicator includes all cases of diarrhoea and/or vomiting. Consultation rates and standardised incidence ratios (SIRs) – calculated using the UK as the standard population – for diarrhoea, gastroenteritis and vomiting were compared for the UK, Yorkshire and Humberside, East Midlands, Leicestershire, Northamptonshire and Rutland SHA, and Daventry and South Northants PCT, Northamptonshire Heartlands PCT and Northampton PCT. Yorkshire and Humberside was not an affected region but was included as a control. The area supplied by the Pitsford Reservoir included the three PCTs, which were all within the Leicestershire, Northamptonshire and Rutland SHA. The consultation rates and SIRs were compared for the period from week 16 to week 35 of 2008 in order to compare the rates before and after the Cryptosporidium exceedance, which took place in week 26. Estimates of excess numbers of cases of diarrhoea occurring during and following the Cryptosporidium outbreak were made by calculating the mean consultation rate over a period of five weeks before and after the incident (weeks 20–24 and weeks 31–35, respectively). For each of the three PCTs, the calculated mean rate was applied to the PCT population to estimate the number of cases that would be expected each week. The actual consultation rates for diarrhoea for weeks 25 to 30 were used to estimate the number of cases for the PCT population each week. The expected number of cases was subtracted from the estimated number of cases in the PCT population to give the estimated number of excess cases. Results HPA/NHS Direct surveillance system A peak in the number of calls for diarrhoea in the East Midlands was recorded in 25–26 June 2008, the period that coincided with the contamination incident and the associated media coverage (Figure 1). The neighbouring areas of the West Midlands, Yorkshire and the Humber, and East of England showed no increase in the number of calls for diarrhoea. The peak produced a control chart exceedance for calls for diarrhoea on 25 June 2008 (Figure 2), when the proportion of calls exceeded the upper limit of the 99.5% confidence interval. There were further confidence interval exceedances on 26 and 28 June (which were not control chart exceedances). There was no peak in calls for vomiting or control chart exceedance for these calls in the East Midlands. HPA/QSurveillance national surveillance system The East Midlands region had significantly high consultation rates for diarrhoea and gastroenteritis in week 25 (16–22 June), week 26 (23–29 June 2008, when the contamination incident was reported) and in the following four weeks. Within this region. Leicestershire, Northamptonshire and Rutland SHA had slightly raised consultation rates and significant SIRs across weeks 25 to 30 that were not seen in the neighbouring Trent SHA. At PCT level, all three of the PCTs in the area affected by the incident showed increased consultation rates for diarrhoea (Table 1) and gastroenteritis (Table 2) with SIRs significantly above the UK rate in week 26. Daventry and South Northants PCT also had a raised SIR for both indicators in week 25, and although Northamptonshire Heartlands and Northampton PCTs did not have SIRs significantly above that of the UK in week 25, the rise in consultation rates for diarrhoea and gastroenteritis began during week 25. In Northampton PCT, consultations for both diarrhoea and gastroenteritis peaked in the week following the contamination incident, week 27, returning to normal levels by week 30 (Figure 3A and 3B). A similar effect can be seen in Northampton Heartlands PCT. Daventry and South Northants PCT also showed an increase, but appeared to have consistently higher rates for both indicators. This was the area with the smallest population so the rates were more variable than in the other PCTs and we therefore interpreted these results with caution. The consultation rates for vomiting during weeks 25 to 30 in the East Midlands were not unusual at SHA or PCT level (data not presented). Discussion We have demonstrated the sensitivity of syndromic surveillance in detecting this small Cryptosporidium outbreak and the value of the surveillance in being able to describe the extent of its spread. Both the HPA/NHS Direct and HPA/QSurveillance systems showed demonstrable increases in calls and consultations for diarrhoea that were linked to the outbreak. QSurveillance consultations appeared to increase across the PCTs immediately affected but not in the surrounding area. Both the HPA/NHS Direct and HPA/QSurveillance systems showed a clear signal at the time of the incident and we were able to describe the extent of the impact on pre-primary care and primary care services. The HPA/QSurveillance system showed a rise in consultation rates for gastrointestinal symptoms that began the week before the outbreak, consistent with the period when Cryptosporidium was present in the water leaving the Pitsford Reservoir (19–24 June 2008) and with the onset of symptoms in the first outbreak case on 24 June. Although only 33 cases were identified by the outbreak investigation team, of which 23 were confirmed as having the outbreak Cryptosporidium strain, our syndromic surveillance data detected this limited outbreak. Data also suggested a more widespread increase in general gastrointestinal symptoms around the time of the outbreak, with an estimated 422 excess diarrhoea cases; these excess cases represented an increase of about 25% above normally expected levels. It is highly probable that a proportion of these excess cases may have resulted from the increased publicity surrounding the incident – for example, it is likely that media reports contributed to the large peak in calls detected by the HPA/NHS Direct surveillance system on the day the boil water notice was issued, and could also have impacted on the GP consultation rate. It has been previously shown that reporting of mumps cases is sensitive to media coverage, with a rise in clinically reported cases following newspaper reports [30]. A similar mechanism could account for some of the excess GP consultations as cases experiencing gastrointestinal symptoms may have been more likely to consult their GP, whereas in normal circumstances they would have cared for themselves at home. It is also possible that the surveillance shows outbreak-associated cases that did not come to the attention of the outbreak team, perhaps because symptoms were not sufficiently severe to warrant further investigation, or stool samples were not provided for testing. It is interesting to note that there was no demonstrable impact on the number of calls for vomiting (which is not a prominent clinical feature of cryptosporidiosis). Other common community-based pathogens such as norovirus and rotavirus were at low levels, as is normal for that time of year [31]. In this instance, public health authorities had already been alerted to a potential problem by the water company, although the extent of the outbreak was detected by syndromic surveillance. In 2003 the syndromic surveillance systems in the city of New York, United States, were able to detect an increase in diarrhoeal illness following a power outage when there was no other indication of citywide illness [32]. The New York system covers a population of nine million, but does not regularly detect localised outbreaks [33]. It has been shown previously that the HPA/NHS Direct surveillance system would be unlikely to detect a Cryptosporidium outbreak unless call volumes are high (72% chance of detection if nine-tenths of cases called NHS Direct) [29], although the value of syndromic surveillance for such outbreaks has been recognised [34]. The system detected the East Midlands Cryptosporidium outbreak that affected a smaller population than that covered by the New York system. The three PCTs affected have a combined population of around 600,000, of which just over half use GP practices reporting to QSurveillance, yet this syndromic surveillance system was able to describe an increase in consultation rates for diarrhoea and gastroenteritis around the time of the outbreak. Limitations of the data There was extensive media reporting of the incident that may have affected both the HPA/NHS Direct and HPA/QSurveillance systems and contributed to the increase in reported gastrointestinal symptoms around the time of the contamination incident. However, the rise in consultation rates for diarrhoea began before the outbreak had been detected and therefore cannot be attributed to media coverage. The HPA/NHS Direct and HPA/QSurveillance systems monitor general symptoms and so could only monitor the relevant symptoms of diarrhoea and vomiting. They are not able to detect Cryptosporidium cases, as this would require laboratory confirmation of diagnosis, so some of the estimated excess cases could be unconnected with this incident. This outbreak was discovered by other means but both the HPA/NHS Direct and HPA/ QSurveillance systems were able to describe the extent of the disease in the general population and provide reassurance that there was no widespread impact. Compared with other populations, older people and ethnic minorities are less likely to call NHS Direct [29], and although this should not prevent detection of gastrointestinal symptoms as a result of drinking water contamination as this would affect the whole population, this may reduce the signal from the system [35]. With such large surveillance systems, there will be ‘background noise’ in the data, so procedures must be in place to correctly interpret the data and set appropriate thresholds for action. Conclusion To our knowledge, this is the first time that PCT-level data from a syndromic surveillance system, the HPA/ QSurveillance national surveillance system, have been able to show the extent of such a limited outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide PCT-level data and this Cryptosporidium contamination incident demonstrates the potential usefulness of this system.

What did the patients have?

1515

Value of syndromic surveillance in monitoring a focal waterborne outbreak due to an unusual Cryptosporidium genotype in Northamptonshire

The United Kingdom (UK) has several national syndromic surveillance systems. The Health Protection Agency (HPA)/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data from a national telephone helpline, while the HPA/ QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems. Data from both of these systems were used to monitor a local outbreak of cryptosporidiosis that occurred following Cryptosporidium oocyst contamination of drinking water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, in June 2008. There was a peak in the number of calls to NHS Direct concerning diarrhoea that coincided with the incident. QSurveillance data for the local areas affected by the outbreak showed a significant increase in GP consultations for diarrhoea and gastroenteritis in the week of the incident but there was no increase in consultations for vomiting. A total of 33 clinical cases of cryptosporidiosis were identified in the outbreak investigation, of which 23 were confirmed as infected with the outbreak strain. However, QSurveillance data suggest that there were an estimated 422 excess diarrhoea cases during the outbreak, an increase of about 25% over baseline weekly levels. To our knowledge, this is the first time that data from a syndromic surveillance system, the HPA/QSurveillance national surveillance system, have been able to show the extent of such a small outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide data at local health district (primary care trust) level. The Cryptosporidium contamination incident described demonstrates the potential usefulness of this information, as it is unusual for syndromic surveillance systems to be able to help monitor such a small-scale outbreak. Introduction As syndromic surveillance systems usually capture data already collected for other purposes, and monitor generic symptoms and/or clinically diagnosed disease, they provide information at an earlier stage of illness (compared with laboratory-confirmed diagnoses), so that action can be taken in time to substantially reduce the impact of disease. Some systems, for example, the Royal College of General Practitioners Weekly Returns Service, are now well established, with many years of historical data that allow monitoring of longer-term disease trends [1]. They have the ability to provide early warning of, for example, seasonal rises in influenza and can trigger public health action, such as a recommendation to prescribe antiviral drugs in line with national guidance [2-4]. They can also provide reassurance to incident response teams and the general public that an incident has not caused adverse health effects – for example, following an explosion at the Buncefield oil storage depot in Hemel Hempstead, United Kingdom (UK), in 2005, syndromic surveillance confirmed that there were no unusual rises in community-based morbidity linked to the incident [5]; following the eruption of the Eyjafjallajökull volcano in Iceland in April 2010 similar assurance was given about lack of impact on community morbidity [6]. Health departments are increasingly expected to monitor health effects of natural events such as heat wave or flooding, or implement surveillance – of which syndromic surveillance plays a major role – for mass gatherings such as the Olympics or football World Cup [7-9]. Systems in France, Australia and Taiwan use data from emergency departments [10-12], a Canadian system uses over-the-counter pharmacy sales [13,14], and in the Netherlands data from both syndromic and surrogate data sources, such as employee absence records and prescription medications dispensed by pharmacies, are included in surveillance systems [15,16]. Currently systems based on Internet searches via search engines or on queries submitted to medical websites are being developed [17,18]. In the UK, the HPA/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data collected from the NHS Direct telephone helpline [19], while the HPA/QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems [20]. The HPA Real-time Syndromic Surveillance Team is a small team that coordinates a number of syndromic surveillance systems within the HPA and takes a lead for syndromic surveillance in England [21]. This paper describes the support provided by the team to the local incident management team during a local cryptosporidiosis outbreak and shows the use of syndromic surveillance in monitoring the extent of an outbreak using the HPA/NHS Direct and HPA/QSurveillance national surveillance systems. Cryptosporidiosis Cryptosporidium is a protozoan parasite that can cause an infection in people, cattle and sometimes other animals [22]. Cryptosporidiosis is most common in children aged between one and five years, but it can affect all ages. Those with impaired immune systems are likely to be most seriously affected. Symptoms usually appear between three and 12 days after initial exposure and include watery diarrhoea, stomach pains, dehydration and fever. In its transmissible form, called an oocyst, the parasite is protected by an outer shell, which allows it to survive in the environment for a long time. Transmission occurs most often via the faeco-oral route through person-to-person or animalto-person contact, but people may also be infected by consuming contaminated water or food or by swimming in contaminated water. Although uncommon, the largest outbreaks have occurred following contamination of drinking water [23,24]. Normal chlorine disinfection procedures do not kill the oocysts, so they are removed by filtration and water companies carry out routine monitoring of treated water. Description of the incident On 25 June 2008 the local Health Protection Unit was informed by Anglian Water of an exceedence in the level of Cryptosporidium oocysts found in water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, during 19 to 24 June 2008 [25]. The reservoir supplied a population of more than 250,000 in the Northampton area. A notice advising people in the affected areas to boil all drinking water was issued on 25 June 2008 and public health messages were circulated to local health services and to the general public via the media. Those members of the public who were concerned about health risks associated with the incident were asked to ring NHS Direct for clinical advice [26]. The HPA wrote to local GPs and hospitals asking them to monitor potential patients for signs and symptoms of Cryptosporidium infection and to submit faecal specimens to the local hospital diagnostic laboratory if patients presented with diarrhoea. Samples from 34 patients where Cryptosporidium infection was identified were sent to the UK Cryptosporidium reference unit for typing. On 30 June 2008, the Cryptosporidium oocysts found in the reservoir water were confirmed as being of the rabbit genotype Cryptosporidium cuniculus [27]. Subsequently, a dead rabbit was found in a treated water tank at the water treatment works. The genotype of Cryptosporidium oocysts in the rabbit’s large bowel was indistinguishable from that of the oocysts found in the water [27]. After remediation of the water supply and distribution, the ‘boil water notice’ was lifted on 4 July and the following day the first case of cryptosporidiosis linked to the incident was identified by the reference laboratory (this case was infected with C. cuniculus). During the course of the outbreak (24 June – 18 July 2008, the dates of symptom onset in the first and last case, respectively), 23 cases of cryptosporidiosis were confirmed as being infected with C. cuniculus; one of the 23 was a secondary case. The HPA Real-time Syndromic Surveillance Team provided data in order to aid the response to this incident and the first syndromic surveillance report was circulated to the incident management team and other relevant people in the HPA on 27 June 2008. Data from the HPA/NHS Direct and HPA/QSurveillance systems were provided in a series of regular reports, initially daily and eventually weekly, until the final report on 21 August 2008. Each report included a summary interpretation and more detailed data on diarrhoea, gastroenteritis and vomiting indicators. Methods Surveillance systems HPA/NHS Direct surveillance system NHS Direct is a 24-hour nurse-led telephone helpline that provides health information and advice to the general public. Nurses use a computerised clinical decision support system – the NHS Clinical Assessment System (NHS CAS) – to handle calls. This assessment system uses approximately 200 computerised symptom-based clinical algorithms. Nurses assign the call to the most appropriate algorithm and the patient’s symptoms determine the questions asked and the action to be taken following the call (call outcome), which could be guidance on self-care or they could be referred to their GP or advised to attend a hospital emergency department. No attempt is made to provide a formal diagnosis. Daily NHS Direct data are received by the Real-time Syndromic Surveillance Team, where the number and type of calls received during the previous day are analysed and interpreted. Call proportions are calculated by age group and algorithm against the total number of calls received. HPA/QSurveillance system The HPA/QSurveillance national surveillance system was set up by the University of Nottingham, United Kingdom, and Egton Medical Information Systems (EMIS), a supplier of general practice computer systems, in collaboration with the HPA. It comprises a network of more than 3,500 general practices throughout the UK, covering more than 22 million patients (about 38% of the population [28]). Aggregated data on GP consultations for a range of indicators are automatically uploaded daily from GP practice systems to a central database. Data are routinely reported on a weekly basis; however, daily reporting is possible for specific incidents. Reports are provided at national or regional level (strategic health authority, SHA) and by local health district (primary care trust, PCT). Analysis of surveillance data NHS Direct call proportions for gastrointestinal syndromes (diarrhoea and vomiting) for the East Midlands region in England, where Northampton is situated, were examined during the outbreak (24 June – 18 July 2008) and compared with those for England and Wales. A series of control charts for diarrhoea calls are routinely used to monitor significant rises in the numbers of calls received. Control charts are calculated by assuming that calls follow a Poisson distribution with the total number of calls as an offset: a model is fitted to each region and symptom separately [29]. The model takes into account call variation caused by weekends, public holidays and the time of year – variables that can affect the number of calls received by NHS Direct. A value above the upper limit of the 99.5% confidence interval would be considered to be unusual. The seven-day moving average for diarrhoea calls was also monitored. The number and percentage of calls for diarrhoea in the East Midlands region were presented by call outcome and the number of calls in the Northampton (NN) postcode districts and in particular the number of calls in the NN11 and NN12 postcode districts, which were most affected by the incident. QSurveillance national consultation rates per 100,000 population for diarrhoea (in the age groups under five years, five years and over, and all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with rates for the same period in 2007 (data not presented). Consultation rates by region for 2008 for diarrhoea (all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with those for the East Midlands region. The gastroenteritis indicator includes all cases of diarrhoea and/or vomiting. Consultation rates and standardised incidence ratios (SIRs) – calculated using the UK as the standard population – for diarrhoea, gastroenteritis and vomiting were compared for the UK, Yorkshire and Humberside, East Midlands, Leicestershire, Northamptonshire and Rutland SHA, and Daventry and South Northants PCT, Northamptonshire Heartlands PCT and Northampton PCT. Yorkshire and Humberside was not an affected region but was included as a control. The area supplied by the Pitsford Reservoir included the three PCTs, which were all within the Leicestershire, Northamptonshire and Rutland SHA. The consultation rates and SIRs were compared for the period from week 16 to week 35 of 2008 in order to compare the rates before and after the Cryptosporidium exceedance, which took place in week 26. Estimates of excess numbers of cases of diarrhoea occurring during and following the Cryptosporidium outbreak were made by calculating the mean consultation rate over a period of five weeks before and after the incident (weeks 20–24 and weeks 31–35, respectively). For each of the three PCTs, the calculated mean rate was applied to the PCT population to estimate the number of cases that would be expected each week. The actual consultation rates for diarrhoea for weeks 25 to 30 were used to estimate the number of cases for the PCT population each week. The expected number of cases was subtracted from the estimated number of cases in the PCT population to give the estimated number of excess cases. Results HPA/NHS Direct surveillance system A peak in the number of calls for diarrhoea in the East Midlands was recorded in 25–26 June 2008, the period that coincided with the contamination incident and the associated media coverage (Figure 1). The neighbouring areas of the West Midlands, Yorkshire and the Humber, and East of England showed no increase in the number of calls for diarrhoea. The peak produced a control chart exceedance for calls for diarrhoea on 25 June 2008 (Figure 2), when the proportion of calls exceeded the upper limit of the 99.5% confidence interval. There were further confidence interval exceedances on 26 and 28 June (which were not control chart exceedances). There was no peak in calls for vomiting or control chart exceedance for these calls in the East Midlands. HPA/QSurveillance national surveillance system The East Midlands region had significantly high consultation rates for diarrhoea and gastroenteritis in week 25 (16–22 June), week 26 (23–29 June 2008, when the contamination incident was reported) and in the following four weeks. Within this region. Leicestershire, Northamptonshire and Rutland SHA had slightly raised consultation rates and significant SIRs across weeks 25 to 30 that were not seen in the neighbouring Trent SHA. At PCT level, all three of the PCTs in the area affected by the incident showed increased consultation rates for diarrhoea (Table 1) and gastroenteritis (Table 2) with SIRs significantly above the UK rate in week 26. Daventry and South Northants PCT also had a raised SIR for both indicators in week 25, and although Northamptonshire Heartlands and Northampton PCTs did not have SIRs significantly above that of the UK in week 25, the rise in consultation rates for diarrhoea and gastroenteritis began during week 25. In Northampton PCT, consultations for both diarrhoea and gastroenteritis peaked in the week following the contamination incident, week 27, returning to normal levels by week 30 (Figure 3A and 3B). A similar effect can be seen in Northampton Heartlands PCT. Daventry and South Northants PCT also showed an increase, but appeared to have consistently higher rates for both indicators. This was the area with the smallest population so the rates were more variable than in the other PCTs and we therefore interpreted these results with caution. The consultation rates for vomiting during weeks 25 to 30 in the East Midlands were not unusual at SHA or PCT level (data not presented). Discussion We have demonstrated the sensitivity of syndromic surveillance in detecting this small Cryptosporidium outbreak and the value of the surveillance in being able to describe the extent of its spread. Both the HPA/NHS Direct and HPA/QSurveillance systems showed demonstrable increases in calls and consultations for diarrhoea that were linked to the outbreak. QSurveillance consultations appeared to increase across the PCTs immediately affected but not in the surrounding area. Both the HPA/NHS Direct and HPA/QSurveillance systems showed a clear signal at the time of the incident and we were able to describe the extent of the impact on pre-primary care and primary care services. The HPA/QSurveillance system showed a rise in consultation rates for gastrointestinal symptoms that began the week before the outbreak, consistent with the period when Cryptosporidium was present in the water leaving the Pitsford Reservoir (19–24 June 2008) and with the onset of symptoms in the first outbreak case on 24 June. Although only 33 cases were identified by the outbreak investigation team, of which 23 were confirmed as having the outbreak Cryptosporidium strain, our syndromic surveillance data detected this limited outbreak. Data also suggested a more widespread increase in general gastrointestinal symptoms around the time of the outbreak, with an estimated 422 excess diarrhoea cases; these excess cases represented an increase of about 25% above normally expected levels. It is highly probable that a proportion of these excess cases may have resulted from the increased publicity surrounding the incident – for example, it is likely that media reports contributed to the large peak in calls detected by the HPA/NHS Direct surveillance system on the day the boil water notice was issued, and could also have impacted on the GP consultation rate. It has been previously shown that reporting of mumps cases is sensitive to media coverage, with a rise in clinically reported cases following newspaper reports [30]. A similar mechanism could account for some of the excess GP consultations as cases experiencing gastrointestinal symptoms may have been more likely to consult their GP, whereas in normal circumstances they would have cared for themselves at home. It is also possible that the surveillance shows outbreak-associated cases that did not come to the attention of the outbreak team, perhaps because symptoms were not sufficiently severe to warrant further investigation, or stool samples were not provided for testing. It is interesting to note that there was no demonstrable impact on the number of calls for vomiting (which is not a prominent clinical feature of cryptosporidiosis). Other common community-based pathogens such as norovirus and rotavirus were at low levels, as is normal for that time of year [31]. In this instance, public health authorities had already been alerted to a potential problem by the water company, although the extent of the outbreak was detected by syndromic surveillance. In 2003 the syndromic surveillance systems in the city of New York, United States, were able to detect an increase in diarrhoeal illness following a power outage when there was no other indication of citywide illness [32]. The New York system covers a population of nine million, but does not regularly detect localised outbreaks [33]. It has been shown previously that the HPA/NHS Direct surveillance system would be unlikely to detect a Cryptosporidium outbreak unless call volumes are high (72% chance of detection if nine-tenths of cases called NHS Direct) [29], although the value of syndromic surveillance for such outbreaks has been recognised [34]. The system detected the East Midlands Cryptosporidium outbreak that affected a smaller population than that covered by the New York system. The three PCTs affected have a combined population of around 600,000, of which just over half use GP practices reporting to QSurveillance, yet this syndromic surveillance system was able to describe an increase in consultation rates for diarrhoea and gastroenteritis around the time of the outbreak. Limitations of the data There was extensive media reporting of the incident that may have affected both the HPA/NHS Direct and HPA/QSurveillance systems and contributed to the increase in reported gastrointestinal symptoms around the time of the contamination incident. However, the rise in consultation rates for diarrhoea began before the outbreak had been detected and therefore cannot be attributed to media coverage. The HPA/NHS Direct and HPA/QSurveillance systems monitor general symptoms and so could only monitor the relevant symptoms of diarrhoea and vomiting. They are not able to detect Cryptosporidium cases, as this would require laboratory confirmation of diagnosis, so some of the estimated excess cases could be unconnected with this incident. This outbreak was discovered by other means but both the HPA/NHS Direct and HPA/ QSurveillance systems were able to describe the extent of the disease in the general population and provide reassurance that there was no widespread impact. Compared with other populations, older people and ethnic minorities are less likely to call NHS Direct [29], and although this should not prevent detection of gastrointestinal symptoms as a result of drinking water contamination as this would affect the whole population, this may reduce the signal from the system [35]. With such large surveillance systems, there will be ‘background noise’ in the data, so procedures must be in place to correctly interpret the data and set appropriate thresholds for action. Conclusion To our knowledge, this is the first time that PCT-level data from a syndromic surveillance system, the HPA/ QSurveillance national surveillance system, have been able to show the extent of such a limited outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide PCT-level data and this Cryptosporidium contamination incident demonstrates the potential usefulness of this system.

What were the first steps?

1516

Value of syndromic surveillance in monitoring a focal waterborne outbreak due to an unusual Cryptosporidium genotype in Northamptonshire

The United Kingdom (UK) has several national syndromic surveillance systems. The Health Protection Agency (HPA)/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data from a national telephone helpline, while the HPA/ QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems. Data from both of these systems were used to monitor a local outbreak of cryptosporidiosis that occurred following Cryptosporidium oocyst contamination of drinking water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, in June 2008. There was a peak in the number of calls to NHS Direct concerning diarrhoea that coincided with the incident. QSurveillance data for the local areas affected by the outbreak showed a significant increase in GP consultations for diarrhoea and gastroenteritis in the week of the incident but there was no increase in consultations for vomiting. A total of 33 clinical cases of cryptosporidiosis were identified in the outbreak investigation, of which 23 were confirmed as infected with the outbreak strain. However, QSurveillance data suggest that there were an estimated 422 excess diarrhoea cases during the outbreak, an increase of about 25% over baseline weekly levels. To our knowledge, this is the first time that data from a syndromic surveillance system, the HPA/QSurveillance national surveillance system, have been able to show the extent of such a small outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide data at local health district (primary care trust) level. The Cryptosporidium contamination incident described demonstrates the potential usefulness of this information, as it is unusual for syndromic surveillance systems to be able to help monitor such a small-scale outbreak. Introduction As syndromic surveillance systems usually capture data already collected for other purposes, and monitor generic symptoms and/or clinically diagnosed disease, they provide information at an earlier stage of illness (compared with laboratory-confirmed diagnoses), so that action can be taken in time to substantially reduce the impact of disease. Some systems, for example, the Royal College of General Practitioners Weekly Returns Service, are now well established, with many years of historical data that allow monitoring of longer-term disease trends [1]. They have the ability to provide early warning of, for example, seasonal rises in influenza and can trigger public health action, such as a recommendation to prescribe antiviral drugs in line with national guidance [2-4]. They can also provide reassurance to incident response teams and the general public that an incident has not caused adverse health effects – for example, following an explosion at the Buncefield oil storage depot in Hemel Hempstead, United Kingdom (UK), in 2005, syndromic surveillance confirmed that there were no unusual rises in community-based morbidity linked to the incident [5]; following the eruption of the Eyjafjallajökull volcano in Iceland in April 2010 similar assurance was given about lack of impact on community morbidity [6]. Health departments are increasingly expected to monitor health effects of natural events such as heat wave or flooding, or implement surveillance – of which syndromic surveillance plays a major role – for mass gatherings such as the Olympics or football World Cup [7-9]. Systems in France, Australia and Taiwan use data from emergency departments [10-12], a Canadian system uses over-the-counter pharmacy sales [13,14], and in the Netherlands data from both syndromic and surrogate data sources, such as employee absence records and prescription medications dispensed by pharmacies, are included in surveillance systems [15,16]. Currently systems based on Internet searches via search engines or on queries submitted to medical websites are being developed [17,18]. In the UK, the HPA/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data collected from the NHS Direct telephone helpline [19], while the HPA/QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems [20]. The HPA Real-time Syndromic Surveillance Team is a small team that coordinates a number of syndromic surveillance systems within the HPA and takes a lead for syndromic surveillance in England [21]. This paper describes the support provided by the team to the local incident management team during a local cryptosporidiosis outbreak and shows the use of syndromic surveillance in monitoring the extent of an outbreak using the HPA/NHS Direct and HPA/QSurveillance national surveillance systems. Cryptosporidiosis Cryptosporidium is a protozoan parasite that can cause an infection in people, cattle and sometimes other animals [22]. Cryptosporidiosis is most common in children aged between one and five years, but it can affect all ages. Those with impaired immune systems are likely to be most seriously affected. Symptoms usually appear between three and 12 days after initial exposure and include watery diarrhoea, stomach pains, dehydration and fever. In its transmissible form, called an oocyst, the parasite is protected by an outer shell, which allows it to survive in the environment for a long time. Transmission occurs most often via the faeco-oral route through person-to-person or animalto-person contact, but people may also be infected by consuming contaminated water or food or by swimming in contaminated water. Although uncommon, the largest outbreaks have occurred following contamination of drinking water [23,24]. Normal chlorine disinfection procedures do not kill the oocysts, so they are removed by filtration and water companies carry out routine monitoring of treated water. Description of the incident On 25 June 2008 the local Health Protection Unit was informed by Anglian Water of an exceedence in the level of Cryptosporidium oocysts found in water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, during 19 to 24 June 2008 [25]. The reservoir supplied a population of more than 250,000 in the Northampton area. A notice advising people in the affected areas to boil all drinking water was issued on 25 June 2008 and public health messages were circulated to local health services and to the general public via the media. Those members of the public who were concerned about health risks associated with the incident were asked to ring NHS Direct for clinical advice [26]. The HPA wrote to local GPs and hospitals asking them to monitor potential patients for signs and symptoms of Cryptosporidium infection and to submit faecal specimens to the local hospital diagnostic laboratory if patients presented with diarrhoea. Samples from 34 patients where Cryptosporidium infection was identified were sent to the UK Cryptosporidium reference unit for typing. On 30 June 2008, the Cryptosporidium oocysts found in the reservoir water were confirmed as being of the rabbit genotype Cryptosporidium cuniculus [27]. Subsequently, a dead rabbit was found in a treated water tank at the water treatment works. The genotype of Cryptosporidium oocysts in the rabbit’s large bowel was indistinguishable from that of the oocysts found in the water [27]. After remediation of the water supply and distribution, the ‘boil water notice’ was lifted on 4 July and the following day the first case of cryptosporidiosis linked to the incident was identified by the reference laboratory (this case was infected with C. cuniculus). During the course of the outbreak (24 June – 18 July 2008, the dates of symptom onset in the first and last case, respectively), 23 cases of cryptosporidiosis were confirmed as being infected with C. cuniculus; one of the 23 was a secondary case. The HPA Real-time Syndromic Surveillance Team provided data in order to aid the response to this incident and the first syndromic surveillance report was circulated to the incident management team and other relevant people in the HPA on 27 June 2008. Data from the HPA/NHS Direct and HPA/QSurveillance systems were provided in a series of regular reports, initially daily and eventually weekly, until the final report on 21 August 2008. Each report included a summary interpretation and more detailed data on diarrhoea, gastroenteritis and vomiting indicators. Methods Surveillance systems HPA/NHS Direct surveillance system NHS Direct is a 24-hour nurse-led telephone helpline that provides health information and advice to the general public. Nurses use a computerised clinical decision support system – the NHS Clinical Assessment System (NHS CAS) – to handle calls. This assessment system uses approximately 200 computerised symptom-based clinical algorithms. Nurses assign the call to the most appropriate algorithm and the patient’s symptoms determine the questions asked and the action to be taken following the call (call outcome), which could be guidance on self-care or they could be referred to their GP or advised to attend a hospital emergency department. No attempt is made to provide a formal diagnosis. Daily NHS Direct data are received by the Real-time Syndromic Surveillance Team, where the number and type of calls received during the previous day are analysed and interpreted. Call proportions are calculated by age group and algorithm against the total number of calls received. HPA/QSurveillance system The HPA/QSurveillance national surveillance system was set up by the University of Nottingham, United Kingdom, and Egton Medical Information Systems (EMIS), a supplier of general practice computer systems, in collaboration with the HPA. It comprises a network of more than 3,500 general practices throughout the UK, covering more than 22 million patients (about 38% of the population [28]). Aggregated data on GP consultations for a range of indicators are automatically uploaded daily from GP practice systems to a central database. Data are routinely reported on a weekly basis; however, daily reporting is possible for specific incidents. Reports are provided at national or regional level (strategic health authority, SHA) and by local health district (primary care trust, PCT). Analysis of surveillance data NHS Direct call proportions for gastrointestinal syndromes (diarrhoea and vomiting) for the East Midlands region in England, where Northampton is situated, were examined during the outbreak (24 June – 18 July 2008) and compared with those for England and Wales. A series of control charts for diarrhoea calls are routinely used to monitor significant rises in the numbers of calls received. Control charts are calculated by assuming that calls follow a Poisson distribution with the total number of calls as an offset: a model is fitted to each region and symptom separately [29]. The model takes into account call variation caused by weekends, public holidays and the time of year – variables that can affect the number of calls received by NHS Direct. A value above the upper limit of the 99.5% confidence interval would be considered to be unusual. The seven-day moving average for diarrhoea calls was also monitored. The number and percentage of calls for diarrhoea in the East Midlands region were presented by call outcome and the number of calls in the Northampton (NN) postcode districts and in particular the number of calls in the NN11 and NN12 postcode districts, which were most affected by the incident. QSurveillance national consultation rates per 100,000 population for diarrhoea (in the age groups under five years, five years and over, and all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with rates for the same period in 2007 (data not presented). Consultation rates by region for 2008 for diarrhoea (all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with those for the East Midlands region. The gastroenteritis indicator includes all cases of diarrhoea and/or vomiting. Consultation rates and standardised incidence ratios (SIRs) – calculated using the UK as the standard population – for diarrhoea, gastroenteritis and vomiting were compared for the UK, Yorkshire and Humberside, East Midlands, Leicestershire, Northamptonshire and Rutland SHA, and Daventry and South Northants PCT, Northamptonshire Heartlands PCT and Northampton PCT. Yorkshire and Humberside was not an affected region but was included as a control. The area supplied by the Pitsford Reservoir included the three PCTs, which were all within the Leicestershire, Northamptonshire and Rutland SHA. The consultation rates and SIRs were compared for the period from week 16 to week 35 of 2008 in order to compare the rates before and after the Cryptosporidium exceedance, which took place in week 26. Estimates of excess numbers of cases of diarrhoea occurring during and following the Cryptosporidium outbreak were made by calculating the mean consultation rate over a period of five weeks before and after the incident (weeks 20–24 and weeks 31–35, respectively). For each of the three PCTs, the calculated mean rate was applied to the PCT population to estimate the number of cases that would be expected each week. The actual consultation rates for diarrhoea for weeks 25 to 30 were used to estimate the number of cases for the PCT population each week. The expected number of cases was subtracted from the estimated number of cases in the PCT population to give the estimated number of excess cases. Results HPA/NHS Direct surveillance system A peak in the number of calls for diarrhoea in the East Midlands was recorded in 25–26 June 2008, the period that coincided with the contamination incident and the associated media coverage (Figure 1). The neighbouring areas of the West Midlands, Yorkshire and the Humber, and East of England showed no increase in the number of calls for diarrhoea. The peak produced a control chart exceedance for calls for diarrhoea on 25 June 2008 (Figure 2), when the proportion of calls exceeded the upper limit of the 99.5% confidence interval. There were further confidence interval exceedances on 26 and 28 June (which were not control chart exceedances). There was no peak in calls for vomiting or control chart exceedance for these calls in the East Midlands. HPA/QSurveillance national surveillance system The East Midlands region had significantly high consultation rates for diarrhoea and gastroenteritis in week 25 (16–22 June), week 26 (23–29 June 2008, when the contamination incident was reported) and in the following four weeks. Within this region. Leicestershire, Northamptonshire and Rutland SHA had slightly raised consultation rates and significant SIRs across weeks 25 to 30 that were not seen in the neighbouring Trent SHA. At PCT level, all three of the PCTs in the area affected by the incident showed increased consultation rates for diarrhoea (Table 1) and gastroenteritis (Table 2) with SIRs significantly above the UK rate in week 26. Daventry and South Northants PCT also had a raised SIR for both indicators in week 25, and although Northamptonshire Heartlands and Northampton PCTs did not have SIRs significantly above that of the UK in week 25, the rise in consultation rates for diarrhoea and gastroenteritis began during week 25. In Northampton PCT, consultations for both diarrhoea and gastroenteritis peaked in the week following the contamination incident, week 27, returning to normal levels by week 30 (Figure 3A and 3B). A similar effect can be seen in Northampton Heartlands PCT. Daventry and South Northants PCT also showed an increase, but appeared to have consistently higher rates for both indicators. This was the area with the smallest population so the rates were more variable than in the other PCTs and we therefore interpreted these results with caution. The consultation rates for vomiting during weeks 25 to 30 in the East Midlands were not unusual at SHA or PCT level (data not presented). Discussion We have demonstrated the sensitivity of syndromic surveillance in detecting this small Cryptosporidium outbreak and the value of the surveillance in being able to describe the extent of its spread. Both the HPA/NHS Direct and HPA/QSurveillance systems showed demonstrable increases in calls and consultations for diarrhoea that were linked to the outbreak. QSurveillance consultations appeared to increase across the PCTs immediately affected but not in the surrounding area. Both the HPA/NHS Direct and HPA/QSurveillance systems showed a clear signal at the time of the incident and we were able to describe the extent of the impact on pre-primary care and primary care services. The HPA/QSurveillance system showed a rise in consultation rates for gastrointestinal symptoms that began the week before the outbreak, consistent with the period when Cryptosporidium was present in the water leaving the Pitsford Reservoir (19–24 June 2008) and with the onset of symptoms in the first outbreak case on 24 June. Although only 33 cases were identified by the outbreak investigation team, of which 23 were confirmed as having the outbreak Cryptosporidium strain, our syndromic surveillance data detected this limited outbreak. Data also suggested a more widespread increase in general gastrointestinal symptoms around the time of the outbreak, with an estimated 422 excess diarrhoea cases; these excess cases represented an increase of about 25% above normally expected levels. It is highly probable that a proportion of these excess cases may have resulted from the increased publicity surrounding the incident – for example, it is likely that media reports contributed to the large peak in calls detected by the HPA/NHS Direct surveillance system on the day the boil water notice was issued, and could also have impacted on the GP consultation rate. It has been previously shown that reporting of mumps cases is sensitive to media coverage, with a rise in clinically reported cases following newspaper reports [30]. A similar mechanism could account for some of the excess GP consultations as cases experiencing gastrointestinal symptoms may have been more likely to consult their GP, whereas in normal circumstances they would have cared for themselves at home. It is also possible that the surveillance shows outbreak-associated cases that did not come to the attention of the outbreak team, perhaps because symptoms were not sufficiently severe to warrant further investigation, or stool samples were not provided for testing. It is interesting to note that there was no demonstrable impact on the number of calls for vomiting (which is not a prominent clinical feature of cryptosporidiosis). Other common community-based pathogens such as norovirus and rotavirus were at low levels, as is normal for that time of year [31]. In this instance, public health authorities had already been alerted to a potential problem by the water company, although the extent of the outbreak was detected by syndromic surveillance. In 2003 the syndromic surveillance systems in the city of New York, United States, were able to detect an increase in diarrhoeal illness following a power outage when there was no other indication of citywide illness [32]. The New York system covers a population of nine million, but does not regularly detect localised outbreaks [33]. It has been shown previously that the HPA/NHS Direct surveillance system would be unlikely to detect a Cryptosporidium outbreak unless call volumes are high (72% chance of detection if nine-tenths of cases called NHS Direct) [29], although the value of syndromic surveillance for such outbreaks has been recognised [34]. The system detected the East Midlands Cryptosporidium outbreak that affected a smaller population than that covered by the New York system. The three PCTs affected have a combined population of around 600,000, of which just over half use GP practices reporting to QSurveillance, yet this syndromic surveillance system was able to describe an increase in consultation rates for diarrhoea and gastroenteritis around the time of the outbreak. Limitations of the data There was extensive media reporting of the incident that may have affected both the HPA/NHS Direct and HPA/QSurveillance systems and contributed to the increase in reported gastrointestinal symptoms around the time of the contamination incident. However, the rise in consultation rates for diarrhoea began before the outbreak had been detected and therefore cannot be attributed to media coverage. The HPA/NHS Direct and HPA/QSurveillance systems monitor general symptoms and so could only monitor the relevant symptoms of diarrhoea and vomiting. They are not able to detect Cryptosporidium cases, as this would require laboratory confirmation of diagnosis, so some of the estimated excess cases could be unconnected with this incident. This outbreak was discovered by other means but both the HPA/NHS Direct and HPA/ QSurveillance systems were able to describe the extent of the disease in the general population and provide reassurance that there was no widespread impact. Compared with other populations, older people and ethnic minorities are less likely to call NHS Direct [29], and although this should not prevent detection of gastrointestinal symptoms as a result of drinking water contamination as this would affect the whole population, this may reduce the signal from the system [35]. With such large surveillance systems, there will be ‘background noise’ in the data, so procedures must be in place to correctly interpret the data and set appropriate thresholds for action. Conclusion To our knowledge, this is the first time that PCT-level data from a syndromic surveillance system, the HPA/ QSurveillance national surveillance system, have been able to show the extent of such a limited outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide PCT-level data and this Cryptosporidium contamination incident demonstrates the potential usefulness of this system.

What did they do to control the problem?

1517

Value of syndromic surveillance in monitoring a focal waterborne outbreak due to an unusual Cryptosporidium genotype in Northamptonshire

The United Kingdom (UK) has several national syndromic surveillance systems. The Health Protection Agency (HPA)/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data from a national telephone helpline, while the HPA/ QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems. Data from both of these systems were used to monitor a local outbreak of cryptosporidiosis that occurred following Cryptosporidium oocyst contamination of drinking water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, in June 2008. There was a peak in the number of calls to NHS Direct concerning diarrhoea that coincided with the incident. QSurveillance data for the local areas affected by the outbreak showed a significant increase in GP consultations for diarrhoea and gastroenteritis in the week of the incident but there was no increase in consultations for vomiting. A total of 33 clinical cases of cryptosporidiosis were identified in the outbreak investigation, of which 23 were confirmed as infected with the outbreak strain. However, QSurveillance data suggest that there were an estimated 422 excess diarrhoea cases during the outbreak, an increase of about 25% over baseline weekly levels. To our knowledge, this is the first time that data from a syndromic surveillance system, the HPA/QSurveillance national surveillance system, have been able to show the extent of such a small outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide data at local health district (primary care trust) level. The Cryptosporidium contamination incident described demonstrates the potential usefulness of this information, as it is unusual for syndromic surveillance systems to be able to help monitor such a small-scale outbreak. Introduction As syndromic surveillance systems usually capture data already collected for other purposes, and monitor generic symptoms and/or clinically diagnosed disease, they provide information at an earlier stage of illness (compared with laboratory-confirmed diagnoses), so that action can be taken in time to substantially reduce the impact of disease. Some systems, for example, the Royal College of General Practitioners Weekly Returns Service, are now well established, with many years of historical data that allow monitoring of longer-term disease trends [1]. They have the ability to provide early warning of, for example, seasonal rises in influenza and can trigger public health action, such as a recommendation to prescribe antiviral drugs in line with national guidance [2-4]. They can also provide reassurance to incident response teams and the general public that an incident has not caused adverse health effects – for example, following an explosion at the Buncefield oil storage depot in Hemel Hempstead, United Kingdom (UK), in 2005, syndromic surveillance confirmed that there were no unusual rises in community-based morbidity linked to the incident [5]; following the eruption of the Eyjafjallajökull volcano in Iceland in April 2010 similar assurance was given about lack of impact on community morbidity [6]. Health departments are increasingly expected to monitor health effects of natural events such as heat wave or flooding, or implement surveillance – of which syndromic surveillance plays a major role – for mass gatherings such as the Olympics or football World Cup [7-9]. Systems in France, Australia and Taiwan use data from emergency departments [10-12], a Canadian system uses over-the-counter pharmacy sales [13,14], and in the Netherlands data from both syndromic and surrogate data sources, such as employee absence records and prescription medications dispensed by pharmacies, are included in surveillance systems [15,16]. Currently systems based on Internet searches via search engines or on queries submitted to medical websites are being developed [17,18]. In the UK, the HPA/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data collected from the NHS Direct telephone helpline [19], while the HPA/QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems [20]. The HPA Real-time Syndromic Surveillance Team is a small team that coordinates a number of syndromic surveillance systems within the HPA and takes a lead for syndromic surveillance in England [21]. This paper describes the support provided by the team to the local incident management team during a local cryptosporidiosis outbreak and shows the use of syndromic surveillance in monitoring the extent of an outbreak using the HPA/NHS Direct and HPA/QSurveillance national surveillance systems. Cryptosporidiosis Cryptosporidium is a protozoan parasite that can cause an infection in people, cattle and sometimes other animals [22]. Cryptosporidiosis is most common in children aged between one and five years, but it can affect all ages. Those with impaired immune systems are likely to be most seriously affected. Symptoms usually appear between three and 12 days after initial exposure and include watery diarrhoea, stomach pains, dehydration and fever. In its transmissible form, called an oocyst, the parasite is protected by an outer shell, which allows it to survive in the environment for a long time. Transmission occurs most often via the faeco-oral route through person-to-person or animalto-person contact, but people may also be infected by consuming contaminated water or food or by swimming in contaminated water. Although uncommon, the largest outbreaks have occurred following contamination of drinking water [23,24]. Normal chlorine disinfection procedures do not kill the oocysts, so they are removed by filtration and water companies carry out routine monitoring of treated water. Description of the incident On 25 June 2008 the local Health Protection Unit was informed by Anglian Water of an exceedence in the level of Cryptosporidium oocysts found in water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, during 19 to 24 June 2008 [25]. The reservoir supplied a population of more than 250,000 in the Northampton area. A notice advising people in the affected areas to boil all drinking water was issued on 25 June 2008 and public health messages were circulated to local health services and to the general public via the media. Those members of the public who were concerned about health risks associated with the incident were asked to ring NHS Direct for clinical advice [26]. The HPA wrote to local GPs and hospitals asking them to monitor potential patients for signs and symptoms of Cryptosporidium infection and to submit faecal specimens to the local hospital diagnostic laboratory if patients presented with diarrhoea. Samples from 34 patients where Cryptosporidium infection was identified were sent to the UK Cryptosporidium reference unit for typing. On 30 June 2008, the Cryptosporidium oocysts found in the reservoir water were confirmed as being of the rabbit genotype Cryptosporidium cuniculus [27]. Subsequently, a dead rabbit was found in a treated water tank at the water treatment works. The genotype of Cryptosporidium oocysts in the rabbit’s large bowel was indistinguishable from that of the oocysts found in the water [27]. After remediation of the water supply and distribution, the ‘boil water notice’ was lifted on 4 July and the following day the first case of cryptosporidiosis linked to the incident was identified by the reference laboratory (this case was infected with C. cuniculus). During the course of the outbreak (24 June – 18 July 2008, the dates of symptom onset in the first and last case, respectively), 23 cases of cryptosporidiosis were confirmed as being infected with C. cuniculus; one of the 23 was a secondary case. The HPA Real-time Syndromic Surveillance Team provided data in order to aid the response to this incident and the first syndromic surveillance report was circulated to the incident management team and other relevant people in the HPA on 27 June 2008. Data from the HPA/NHS Direct and HPA/QSurveillance systems were provided in a series of regular reports, initially daily and eventually weekly, until the final report on 21 August 2008. Each report included a summary interpretation and more detailed data on diarrhoea, gastroenteritis and vomiting indicators. Methods Surveillance systems HPA/NHS Direct surveillance system NHS Direct is a 24-hour nurse-led telephone helpline that provides health information and advice to the general public. Nurses use a computerised clinical decision support system – the NHS Clinical Assessment System (NHS CAS) – to handle calls. This assessment system uses approximately 200 computerised symptom-based clinical algorithms. Nurses assign the call to the most appropriate algorithm and the patient’s symptoms determine the questions asked and the action to be taken following the call (call outcome), which could be guidance on self-care or they could be referred to their GP or advised to attend a hospital emergency department. No attempt is made to provide a formal diagnosis. Daily NHS Direct data are received by the Real-time Syndromic Surveillance Team, where the number and type of calls received during the previous day are analysed and interpreted. Call proportions are calculated by age group and algorithm against the total number of calls received. HPA/QSurveillance system The HPA/QSurveillance national surveillance system was set up by the University of Nottingham, United Kingdom, and Egton Medical Information Systems (EMIS), a supplier of general practice computer systems, in collaboration with the HPA. It comprises a network of more than 3,500 general practices throughout the UK, covering more than 22 million patients (about 38% of the population [28]). Aggregated data on GP consultations for a range of indicators are automatically uploaded daily from GP practice systems to a central database. Data are routinely reported on a weekly basis; however, daily reporting is possible for specific incidents. Reports are provided at national or regional level (strategic health authority, SHA) and by local health district (primary care trust, PCT). Analysis of surveillance data NHS Direct call proportions for gastrointestinal syndromes (diarrhoea and vomiting) for the East Midlands region in England, where Northampton is situated, were examined during the outbreak (24 June – 18 July 2008) and compared with those for England and Wales. A series of control charts for diarrhoea calls are routinely used to monitor significant rises in the numbers of calls received. Control charts are calculated by assuming that calls follow a Poisson distribution with the total number of calls as an offset: a model is fitted to each region and symptom separately [29]. The model takes into account call variation caused by weekends, public holidays and the time of year – variables that can affect the number of calls received by NHS Direct. A value above the upper limit of the 99.5% confidence interval would be considered to be unusual. The seven-day moving average for diarrhoea calls was also monitored. The number and percentage of calls for diarrhoea in the East Midlands region were presented by call outcome and the number of calls in the Northampton (NN) postcode districts and in particular the number of calls in the NN11 and NN12 postcode districts, which were most affected by the incident. QSurveillance national consultation rates per 100,000 population for diarrhoea (in the age groups under five years, five years and over, and all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with rates for the same period in 2007 (data not presented). Consultation rates by region for 2008 for diarrhoea (all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with those for the East Midlands region. The gastroenteritis indicator includes all cases of diarrhoea and/or vomiting. Consultation rates and standardised incidence ratios (SIRs) – calculated using the UK as the standard population – for diarrhoea, gastroenteritis and vomiting were compared for the UK, Yorkshire and Humberside, East Midlands, Leicestershire, Northamptonshire and Rutland SHA, and Daventry and South Northants PCT, Northamptonshire Heartlands PCT and Northampton PCT. Yorkshire and Humberside was not an affected region but was included as a control. The area supplied by the Pitsford Reservoir included the three PCTs, which were all within the Leicestershire, Northamptonshire and Rutland SHA. The consultation rates and SIRs were compared for the period from week 16 to week 35 of 2008 in order to compare the rates before and after the Cryptosporidium exceedance, which took place in week 26. Estimates of excess numbers of cases of diarrhoea occurring during and following the Cryptosporidium outbreak were made by calculating the mean consultation rate over a period of five weeks before and after the incident (weeks 20–24 and weeks 31–35, respectively). For each of the three PCTs, the calculated mean rate was applied to the PCT population to estimate the number of cases that would be expected each week. The actual consultation rates for diarrhoea for weeks 25 to 30 were used to estimate the number of cases for the PCT population each week. The expected number of cases was subtracted from the estimated number of cases in the PCT population to give the estimated number of excess cases. Results HPA/NHS Direct surveillance system A peak in the number of calls for diarrhoea in the East Midlands was recorded in 25–26 June 2008, the period that coincided with the contamination incident and the associated media coverage (Figure 1). The neighbouring areas of the West Midlands, Yorkshire and the Humber, and East of England showed no increase in the number of calls for diarrhoea. The peak produced a control chart exceedance for calls for diarrhoea on 25 June 2008 (Figure 2), when the proportion of calls exceeded the upper limit of the 99.5% confidence interval. There were further confidence interval exceedances on 26 and 28 June (which were not control chart exceedances). There was no peak in calls for vomiting or control chart exceedance for these calls in the East Midlands. HPA/QSurveillance national surveillance system The East Midlands region had significantly high consultation rates for diarrhoea and gastroenteritis in week 25 (16–22 June), week 26 (23–29 June 2008, when the contamination incident was reported) and in the following four weeks. Within this region. Leicestershire, Northamptonshire and Rutland SHA had slightly raised consultation rates and significant SIRs across weeks 25 to 30 that were not seen in the neighbouring Trent SHA. At PCT level, all three of the PCTs in the area affected by the incident showed increased consultation rates for diarrhoea (Table 1) and gastroenteritis (Table 2) with SIRs significantly above the UK rate in week 26. Daventry and South Northants PCT also had a raised SIR for both indicators in week 25, and although Northamptonshire Heartlands and Northampton PCTs did not have SIRs significantly above that of the UK in week 25, the rise in consultation rates for diarrhoea and gastroenteritis began during week 25. In Northampton PCT, consultations for both diarrhoea and gastroenteritis peaked in the week following the contamination incident, week 27, returning to normal levels by week 30 (Figure 3A and 3B). A similar effect can be seen in Northampton Heartlands PCT. Daventry and South Northants PCT also showed an increase, but appeared to have consistently higher rates for both indicators. This was the area with the smallest population so the rates were more variable than in the other PCTs and we therefore interpreted these results with caution. The consultation rates for vomiting during weeks 25 to 30 in the East Midlands were not unusual at SHA or PCT level (data not presented). Discussion We have demonstrated the sensitivity of syndromic surveillance in detecting this small Cryptosporidium outbreak and the value of the surveillance in being able to describe the extent of its spread. Both the HPA/NHS Direct and HPA/QSurveillance systems showed demonstrable increases in calls and consultations for diarrhoea that were linked to the outbreak. QSurveillance consultations appeared to increase across the PCTs immediately affected but not in the surrounding area. Both the HPA/NHS Direct and HPA/QSurveillance systems showed a clear signal at the time of the incident and we were able to describe the extent of the impact on pre-primary care and primary care services. The HPA/QSurveillance system showed a rise in consultation rates for gastrointestinal symptoms that began the week before the outbreak, consistent with the period when Cryptosporidium was present in the water leaving the Pitsford Reservoir (19–24 June 2008) and with the onset of symptoms in the first outbreak case on 24 June. Although only 33 cases were identified by the outbreak investigation team, of which 23 were confirmed as having the outbreak Cryptosporidium strain, our syndromic surveillance data detected this limited outbreak. Data also suggested a more widespread increase in general gastrointestinal symptoms around the time of the outbreak, with an estimated 422 excess diarrhoea cases; these excess cases represented an increase of about 25% above normally expected levels. It is highly probable that a proportion of these excess cases may have resulted from the increased publicity surrounding the incident – for example, it is likely that media reports contributed to the large peak in calls detected by the HPA/NHS Direct surveillance system on the day the boil water notice was issued, and could also have impacted on the GP consultation rate. It has been previously shown that reporting of mumps cases is sensitive to media coverage, with a rise in clinically reported cases following newspaper reports [30]. A similar mechanism could account for some of the excess GP consultations as cases experiencing gastrointestinal symptoms may have been more likely to consult their GP, whereas in normal circumstances they would have cared for themselves at home. It is also possible that the surveillance shows outbreak-associated cases that did not come to the attention of the outbreak team, perhaps because symptoms were not sufficiently severe to warrant further investigation, or stool samples were not provided for testing. It is interesting to note that there was no demonstrable impact on the number of calls for vomiting (which is not a prominent clinical feature of cryptosporidiosis). Other common community-based pathogens such as norovirus and rotavirus were at low levels, as is normal for that time of year [31]. In this instance, public health authorities had already been alerted to a potential problem by the water company, although the extent of the outbreak was detected by syndromic surveillance. In 2003 the syndromic surveillance systems in the city of New York, United States, were able to detect an increase in diarrhoeal illness following a power outage when there was no other indication of citywide illness [32]. The New York system covers a population of nine million, but does not regularly detect localised outbreaks [33]. It has been shown previously that the HPA/NHS Direct surveillance system would be unlikely to detect a Cryptosporidium outbreak unless call volumes are high (72% chance of detection if nine-tenths of cases called NHS Direct) [29], although the value of syndromic surveillance for such outbreaks has been recognised [34]. The system detected the East Midlands Cryptosporidium outbreak that affected a smaller population than that covered by the New York system. The three PCTs affected have a combined population of around 600,000, of which just over half use GP practices reporting to QSurveillance, yet this syndromic surveillance system was able to describe an increase in consultation rates for diarrhoea and gastroenteritis around the time of the outbreak. Limitations of the data There was extensive media reporting of the incident that may have affected both the HPA/NHS Direct and HPA/QSurveillance systems and contributed to the increase in reported gastrointestinal symptoms around the time of the contamination incident. However, the rise in consultation rates for diarrhoea began before the outbreak had been detected and therefore cannot be attributed to media coverage. The HPA/NHS Direct and HPA/QSurveillance systems monitor general symptoms and so could only monitor the relevant symptoms of diarrhoea and vomiting. They are not able to detect Cryptosporidium cases, as this would require laboratory confirmation of diagnosis, so some of the estimated excess cases could be unconnected with this incident. This outbreak was discovered by other means but both the HPA/NHS Direct and HPA/ QSurveillance systems were able to describe the extent of the disease in the general population and provide reassurance that there was no widespread impact. Compared with other populations, older people and ethnic minorities are less likely to call NHS Direct [29], and although this should not prevent detection of gastrointestinal symptoms as a result of drinking water contamination as this would affect the whole population, this may reduce the signal from the system [35]. With such large surveillance systems, there will be ‘background noise’ in the data, so procedures must be in place to correctly interpret the data and set appropriate thresholds for action. Conclusion To our knowledge, this is the first time that PCT-level data from a syndromic surveillance system, the HPA/ QSurveillance national surveillance system, have been able to show the extent of such a limited outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide PCT-level data and this Cryptosporidium contamination incident demonstrates the potential usefulness of this system.

What did the local authorities advise?

1518

Value of syndromic surveillance in monitoring a focal waterborne outbreak due to an unusual Cryptosporidium genotype in Northamptonshire

The United Kingdom (UK) has several national syndromic surveillance systems. The Health Protection Agency (HPA)/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data from a national telephone helpline, while the HPA/ QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems. Data from both of these systems were used to monitor a local outbreak of cryptosporidiosis that occurred following Cryptosporidium oocyst contamination of drinking water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, in June 2008. There was a peak in the number of calls to NHS Direct concerning diarrhoea that coincided with the incident. QSurveillance data for the local areas affected by the outbreak showed a significant increase in GP consultations for diarrhoea and gastroenteritis in the week of the incident but there was no increase in consultations for vomiting. A total of 33 clinical cases of cryptosporidiosis were identified in the outbreak investigation, of which 23 were confirmed as infected with the outbreak strain. However, QSurveillance data suggest that there were an estimated 422 excess diarrhoea cases during the outbreak, an increase of about 25% over baseline weekly levels. To our knowledge, this is the first time that data from a syndromic surveillance system, the HPA/QSurveillance national surveillance system, have been able to show the extent of such a small outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide data at local health district (primary care trust) level. The Cryptosporidium contamination incident described demonstrates the potential usefulness of this information, as it is unusual for syndromic surveillance systems to be able to help monitor such a small-scale outbreak. Introduction As syndromic surveillance systems usually capture data already collected for other purposes, and monitor generic symptoms and/or clinically diagnosed disease, they provide information at an earlier stage of illness (compared with laboratory-confirmed diagnoses), so that action can be taken in time to substantially reduce the impact of disease. Some systems, for example, the Royal College of General Practitioners Weekly Returns Service, are now well established, with many years of historical data that allow monitoring of longer-term disease trends [1]. They have the ability to provide early warning of, for example, seasonal rises in influenza and can trigger public health action, such as a recommendation to prescribe antiviral drugs in line with national guidance [2-4]. They can also provide reassurance to incident response teams and the general public that an incident has not caused adverse health effects – for example, following an explosion at the Buncefield oil storage depot in Hemel Hempstead, United Kingdom (UK), in 2005, syndromic surveillance confirmed that there were no unusual rises in community-based morbidity linked to the incident [5]; following the eruption of the Eyjafjallajökull volcano in Iceland in April 2010 similar assurance was given about lack of impact on community morbidity [6]. Health departments are increasingly expected to monitor health effects of natural events such as heat wave or flooding, or implement surveillance – of which syndromic surveillance plays a major role – for mass gatherings such as the Olympics or football World Cup [7-9]. Systems in France, Australia and Taiwan use data from emergency departments [10-12], a Canadian system uses over-the-counter pharmacy sales [13,14], and in the Netherlands data from both syndromic and surrogate data sources, such as employee absence records and prescription medications dispensed by pharmacies, are included in surveillance systems [15,16]. Currently systems based on Internet searches via search engines or on queries submitted to medical websites are being developed [17,18]. In the UK, the HPA/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data collected from the NHS Direct telephone helpline [19], while the HPA/QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems [20]. The HPA Real-time Syndromic Surveillance Team is a small team that coordinates a number of syndromic surveillance systems within the HPA and takes a lead for syndromic surveillance in England [21]. This paper describes the support provided by the team to the local incident management team during a local cryptosporidiosis outbreak and shows the use of syndromic surveillance in monitoring the extent of an outbreak using the HPA/NHS Direct and HPA/QSurveillance national surveillance systems. Cryptosporidiosis Cryptosporidium is a protozoan parasite that can cause an infection in people, cattle and sometimes other animals [22]. Cryptosporidiosis is most common in children aged between one and five years, but it can affect all ages. Those with impaired immune systems are likely to be most seriously affected. Symptoms usually appear between three and 12 days after initial exposure and include watery diarrhoea, stomach pains, dehydration and fever. In its transmissible form, called an oocyst, the parasite is protected by an outer shell, which allows it to survive in the environment for a long time. Transmission occurs most often via the faeco-oral route through person-to-person or animalto-person contact, but people may also be infected by consuming contaminated water or food or by swimming in contaminated water. Although uncommon, the largest outbreaks have occurred following contamination of drinking water [23,24]. Normal chlorine disinfection procedures do not kill the oocysts, so they are removed by filtration and water companies carry out routine monitoring of treated water. Description of the incident On 25 June 2008 the local Health Protection Unit was informed by Anglian Water of an exceedence in the level of Cryptosporidium oocysts found in water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, during 19 to 24 June 2008 [25]. The reservoir supplied a population of more than 250,000 in the Northampton area. A notice advising people in the affected areas to boil all drinking water was issued on 25 June 2008 and public health messages were circulated to local health services and to the general public via the media. Those members of the public who were concerned about health risks associated with the incident were asked to ring NHS Direct for clinical advice [26]. The HPA wrote to local GPs and hospitals asking them to monitor potential patients for signs and symptoms of Cryptosporidium infection and to submit faecal specimens to the local hospital diagnostic laboratory if patients presented with diarrhoea. Samples from 34 patients where Cryptosporidium infection was identified were sent to the UK Cryptosporidium reference unit for typing. On 30 June 2008, the Cryptosporidium oocysts found in the reservoir water were confirmed as being of the rabbit genotype Cryptosporidium cuniculus [27]. Subsequently, a dead rabbit was found in a treated water tank at the water treatment works. The genotype of Cryptosporidium oocysts in the rabbit’s large bowel was indistinguishable from that of the oocysts found in the water [27]. After remediation of the water supply and distribution, the ‘boil water notice’ was lifted on 4 July and the following day the first case of cryptosporidiosis linked to the incident was identified by the reference laboratory (this case was infected with C. cuniculus). During the course of the outbreak (24 June – 18 July 2008, the dates of symptom onset in the first and last case, respectively), 23 cases of cryptosporidiosis were confirmed as being infected with C. cuniculus; one of the 23 was a secondary case. The HPA Real-time Syndromic Surveillance Team provided data in order to aid the response to this incident and the first syndromic surveillance report was circulated to the incident management team and other relevant people in the HPA on 27 June 2008. Data from the HPA/NHS Direct and HPA/QSurveillance systems were provided in a series of regular reports, initially daily and eventually weekly, until the final report on 21 August 2008. Each report included a summary interpretation and more detailed data on diarrhoea, gastroenteritis and vomiting indicators. Methods Surveillance systems HPA/NHS Direct surveillance system NHS Direct is a 24-hour nurse-led telephone helpline that provides health information and advice to the general public. Nurses use a computerised clinical decision support system – the NHS Clinical Assessment System (NHS CAS) – to handle calls. This assessment system uses approximately 200 computerised symptom-based clinical algorithms. Nurses assign the call to the most appropriate algorithm and the patient’s symptoms determine the questions asked and the action to be taken following the call (call outcome), which could be guidance on self-care or they could be referred to their GP or advised to attend a hospital emergency department. No attempt is made to provide a formal diagnosis. Daily NHS Direct data are received by the Real-time Syndromic Surveillance Team, where the number and type of calls received during the previous day are analysed and interpreted. Call proportions are calculated by age group and algorithm against the total number of calls received. HPA/QSurveillance system The HPA/QSurveillance national surveillance system was set up by the University of Nottingham, United Kingdom, and Egton Medical Information Systems (EMIS), a supplier of general practice computer systems, in collaboration with the HPA. It comprises a network of more than 3,500 general practices throughout the UK, covering more than 22 million patients (about 38% of the population [28]). Aggregated data on GP consultations for a range of indicators are automatically uploaded daily from GP practice systems to a central database. Data are routinely reported on a weekly basis; however, daily reporting is possible for specific incidents. Reports are provided at national or regional level (strategic health authority, SHA) and by local health district (primary care trust, PCT). Analysis of surveillance data NHS Direct call proportions for gastrointestinal syndromes (diarrhoea and vomiting) for the East Midlands region in England, where Northampton is situated, were examined during the outbreak (24 June – 18 July 2008) and compared with those for England and Wales. A series of control charts for diarrhoea calls are routinely used to monitor significant rises in the numbers of calls received. Control charts are calculated by assuming that calls follow a Poisson distribution with the total number of calls as an offset: a model is fitted to each region and symptom separately [29]. The model takes into account call variation caused by weekends, public holidays and the time of year – variables that can affect the number of calls received by NHS Direct. A value above the upper limit of the 99.5% confidence interval would be considered to be unusual. The seven-day moving average for diarrhoea calls was also monitored. The number and percentage of calls for diarrhoea in the East Midlands region were presented by call outcome and the number of calls in the Northampton (NN) postcode districts and in particular the number of calls in the NN11 and NN12 postcode districts, which were most affected by the incident. QSurveillance national consultation rates per 100,000 population for diarrhoea (in the age groups under five years, five years and over, and all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with rates for the same period in 2007 (data not presented). Consultation rates by region for 2008 for diarrhoea (all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with those for the East Midlands region. The gastroenteritis indicator includes all cases of diarrhoea and/or vomiting. Consultation rates and standardised incidence ratios (SIRs) – calculated using the UK as the standard population – for diarrhoea, gastroenteritis and vomiting were compared for the UK, Yorkshire and Humberside, East Midlands, Leicestershire, Northamptonshire and Rutland SHA, and Daventry and South Northants PCT, Northamptonshire Heartlands PCT and Northampton PCT. Yorkshire and Humberside was not an affected region but was included as a control. The area supplied by the Pitsford Reservoir included the three PCTs, which were all within the Leicestershire, Northamptonshire and Rutland SHA. The consultation rates and SIRs were compared for the period from week 16 to week 35 of 2008 in order to compare the rates before and after the Cryptosporidium exceedance, which took place in week 26. Estimates of excess numbers of cases of diarrhoea occurring during and following the Cryptosporidium outbreak were made by calculating the mean consultation rate over a period of five weeks before and after the incident (weeks 20–24 and weeks 31–35, respectively). For each of the three PCTs, the calculated mean rate was applied to the PCT population to estimate the number of cases that would be expected each week. The actual consultation rates for diarrhoea for weeks 25 to 30 were used to estimate the number of cases for the PCT population each week. The expected number of cases was subtracted from the estimated number of cases in the PCT population to give the estimated number of excess cases. Results HPA/NHS Direct surveillance system A peak in the number of calls for diarrhoea in the East Midlands was recorded in 25–26 June 2008, the period that coincided with the contamination incident and the associated media coverage (Figure 1). The neighbouring areas of the West Midlands, Yorkshire and the Humber, and East of England showed no increase in the number of calls for diarrhoea. The peak produced a control chart exceedance for calls for diarrhoea on 25 June 2008 (Figure 2), when the proportion of calls exceeded the upper limit of the 99.5% confidence interval. There were further confidence interval exceedances on 26 and 28 June (which were not control chart exceedances). There was no peak in calls for vomiting or control chart exceedance for these calls in the East Midlands. HPA/QSurveillance national surveillance system The East Midlands region had significantly high consultation rates for diarrhoea and gastroenteritis in week 25 (16–22 June), week 26 (23–29 June 2008, when the contamination incident was reported) and in the following four weeks. Within this region. Leicestershire, Northamptonshire and Rutland SHA had slightly raised consultation rates and significant SIRs across weeks 25 to 30 that were not seen in the neighbouring Trent SHA. At PCT level, all three of the PCTs in the area affected by the incident showed increased consultation rates for diarrhoea (Table 1) and gastroenteritis (Table 2) with SIRs significantly above the UK rate in week 26. Daventry and South Northants PCT also had a raised SIR for both indicators in week 25, and although Northamptonshire Heartlands and Northampton PCTs did not have SIRs significantly above that of the UK in week 25, the rise in consultation rates for diarrhoea and gastroenteritis began during week 25. In Northampton PCT, consultations for both diarrhoea and gastroenteritis peaked in the week following the contamination incident, week 27, returning to normal levels by week 30 (Figure 3A and 3B). A similar effect can be seen in Northampton Heartlands PCT. Daventry and South Northants PCT also showed an increase, but appeared to have consistently higher rates for both indicators. This was the area with the smallest population so the rates were more variable than in the other PCTs and we therefore interpreted these results with caution. The consultation rates for vomiting during weeks 25 to 30 in the East Midlands were not unusual at SHA or PCT level (data not presented). Discussion We have demonstrated the sensitivity of syndromic surveillance in detecting this small Cryptosporidium outbreak and the value of the surveillance in being able to describe the extent of its spread. Both the HPA/NHS Direct and HPA/QSurveillance systems showed demonstrable increases in calls and consultations for diarrhoea that were linked to the outbreak. QSurveillance consultations appeared to increase across the PCTs immediately affected but not in the surrounding area. Both the HPA/NHS Direct and HPA/QSurveillance systems showed a clear signal at the time of the incident and we were able to describe the extent of the impact on pre-primary care and primary care services. The HPA/QSurveillance system showed a rise in consultation rates for gastrointestinal symptoms that began the week before the outbreak, consistent with the period when Cryptosporidium was present in the water leaving the Pitsford Reservoir (19–24 June 2008) and with the onset of symptoms in the first outbreak case on 24 June. Although only 33 cases were identified by the outbreak investigation team, of which 23 were confirmed as having the outbreak Cryptosporidium strain, our syndromic surveillance data detected this limited outbreak. Data also suggested a more widespread increase in general gastrointestinal symptoms around the time of the outbreak, with an estimated 422 excess diarrhoea cases; these excess cases represented an increase of about 25% above normally expected levels. It is highly probable that a proportion of these excess cases may have resulted from the increased publicity surrounding the incident – for example, it is likely that media reports contributed to the large peak in calls detected by the HPA/NHS Direct surveillance system on the day the boil water notice was issued, and could also have impacted on the GP consultation rate. It has been previously shown that reporting of mumps cases is sensitive to media coverage, with a rise in clinically reported cases following newspaper reports [30]. A similar mechanism could account for some of the excess GP consultations as cases experiencing gastrointestinal symptoms may have been more likely to consult their GP, whereas in normal circumstances they would have cared for themselves at home. It is also possible that the surveillance shows outbreak-associated cases that did not come to the attention of the outbreak team, perhaps because symptoms were not sufficiently severe to warrant further investigation, or stool samples were not provided for testing. It is interesting to note that there was no demonstrable impact on the number of calls for vomiting (which is not a prominent clinical feature of cryptosporidiosis). Other common community-based pathogens such as norovirus and rotavirus were at low levels, as is normal for that time of year [31]. In this instance, public health authorities had already been alerted to a potential problem by the water company, although the extent of the outbreak was detected by syndromic surveillance. In 2003 the syndromic surveillance systems in the city of New York, United States, were able to detect an increase in diarrhoeal illness following a power outage when there was no other indication of citywide illness [32]. The New York system covers a population of nine million, but does not regularly detect localised outbreaks [33]. It has been shown previously that the HPA/NHS Direct surveillance system would be unlikely to detect a Cryptosporidium outbreak unless call volumes are high (72% chance of detection if nine-tenths of cases called NHS Direct) [29], although the value of syndromic surveillance for such outbreaks has been recognised [34]. The system detected the East Midlands Cryptosporidium outbreak that affected a smaller population than that covered by the New York system. The three PCTs affected have a combined population of around 600,000, of which just over half use GP practices reporting to QSurveillance, yet this syndromic surveillance system was able to describe an increase in consultation rates for diarrhoea and gastroenteritis around the time of the outbreak. Limitations of the data There was extensive media reporting of the incident that may have affected both the HPA/NHS Direct and HPA/QSurveillance systems and contributed to the increase in reported gastrointestinal symptoms around the time of the contamination incident. However, the rise in consultation rates for diarrhoea began before the outbreak had been detected and therefore cannot be attributed to media coverage. The HPA/NHS Direct and HPA/QSurveillance systems monitor general symptoms and so could only monitor the relevant symptoms of diarrhoea and vomiting. They are not able to detect Cryptosporidium cases, as this would require laboratory confirmation of diagnosis, so some of the estimated excess cases could be unconnected with this incident. This outbreak was discovered by other means but both the HPA/NHS Direct and HPA/ QSurveillance systems were able to describe the extent of the disease in the general population and provide reassurance that there was no widespread impact. Compared with other populations, older people and ethnic minorities are less likely to call NHS Direct [29], and although this should not prevent detection of gastrointestinal symptoms as a result of drinking water contamination as this would affect the whole population, this may reduce the signal from the system [35]. With such large surveillance systems, there will be ‘background noise’ in the data, so procedures must be in place to correctly interpret the data and set appropriate thresholds for action. Conclusion To our knowledge, this is the first time that PCT-level data from a syndromic surveillance system, the HPA/ QSurveillance national surveillance system, have been able to show the extent of such a limited outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide PCT-level data and this Cryptosporidium contamination incident demonstrates the potential usefulness of this system.

What were the control measures?

1519

Value of syndromic surveillance in monitoring a focal waterborne outbreak due to an unusual Cryptosporidium genotype in Northamptonshire

The United Kingdom (UK) has several national syndromic surveillance systems. The Health Protection Agency (HPA)/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data from a national telephone helpline, while the HPA/ QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems. Data from both of these systems were used to monitor a local outbreak of cryptosporidiosis that occurred following Cryptosporidium oocyst contamination of drinking water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, in June 2008. There was a peak in the number of calls to NHS Direct concerning diarrhoea that coincided with the incident. QSurveillance data for the local areas affected by the outbreak showed a significant increase in GP consultations for diarrhoea and gastroenteritis in the week of the incident but there was no increase in consultations for vomiting. A total of 33 clinical cases of cryptosporidiosis were identified in the outbreak investigation, of which 23 were confirmed as infected with the outbreak strain. However, QSurveillance data suggest that there were an estimated 422 excess diarrhoea cases during the outbreak, an increase of about 25% over baseline weekly levels. To our knowledge, this is the first time that data from a syndromic surveillance system, the HPA/QSurveillance national surveillance system, have been able to show the extent of such a small outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide data at local health district (primary care trust) level. The Cryptosporidium contamination incident described demonstrates the potential usefulness of this information, as it is unusual for syndromic surveillance systems to be able to help monitor such a small-scale outbreak. Introduction As syndromic surveillance systems usually capture data already collected for other purposes, and monitor generic symptoms and/or clinically diagnosed disease, they provide information at an earlier stage of illness (compared with laboratory-confirmed diagnoses), so that action can be taken in time to substantially reduce the impact of disease. Some systems, for example, the Royal College of General Practitioners Weekly Returns Service, are now well established, with many years of historical data that allow monitoring of longer-term disease trends [1]. They have the ability to provide early warning of, for example, seasonal rises in influenza and can trigger public health action, such as a recommendation to prescribe antiviral drugs in line with national guidance [2-4]. They can also provide reassurance to incident response teams and the general public that an incident has not caused adverse health effects – for example, following an explosion at the Buncefield oil storage depot in Hemel Hempstead, United Kingdom (UK), in 2005, syndromic surveillance confirmed that there were no unusual rises in community-based morbidity linked to the incident [5]; following the eruption of the Eyjafjallajökull volcano in Iceland in April 2010 similar assurance was given about lack of impact on community morbidity [6]. Health departments are increasingly expected to monitor health effects of natural events such as heat wave or flooding, or implement surveillance – of which syndromic surveillance plays a major role – for mass gatherings such as the Olympics or football World Cup [7-9]. Systems in France, Australia and Taiwan use data from emergency departments [10-12], a Canadian system uses over-the-counter pharmacy sales [13,14], and in the Netherlands data from both syndromic and surrogate data sources, such as employee absence records and prescription medications dispensed by pharmacies, are included in surveillance systems [15,16]. Currently systems based on Internet searches via search engines or on queries submitted to medical websites are being developed [17,18]. In the UK, the HPA/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data collected from the NHS Direct telephone helpline [19], while the HPA/QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems [20]. The HPA Real-time Syndromic Surveillance Team is a small team that coordinates a number of syndromic surveillance systems within the HPA and takes a lead for syndromic surveillance in England [21]. This paper describes the support provided by the team to the local incident management team during a local cryptosporidiosis outbreak and shows the use of syndromic surveillance in monitoring the extent of an outbreak using the HPA/NHS Direct and HPA/QSurveillance national surveillance systems. Cryptosporidiosis Cryptosporidium is a protozoan parasite that can cause an infection in people, cattle and sometimes other animals [22]. Cryptosporidiosis is most common in children aged between one and five years, but it can affect all ages. Those with impaired immune systems are likely to be most seriously affected. Symptoms usually appear between three and 12 days after initial exposure and include watery diarrhoea, stomach pains, dehydration and fever. In its transmissible form, called an oocyst, the parasite is protected by an outer shell, which allows it to survive in the environment for a long time. Transmission occurs most often via the faeco-oral route through person-to-person or animalto-person contact, but people may also be infected by consuming contaminated water or food or by swimming in contaminated water. Although uncommon, the largest outbreaks have occurred following contamination of drinking water [23,24]. Normal chlorine disinfection procedures do not kill the oocysts, so they are removed by filtration and water companies carry out routine monitoring of treated water. Description of the incident On 25 June 2008 the local Health Protection Unit was informed by Anglian Water of an exceedence in the level of Cryptosporidium oocysts found in water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, during 19 to 24 June 2008 [25]. The reservoir supplied a population of more than 250,000 in the Northampton area. A notice advising people in the affected areas to boil all drinking water was issued on 25 June 2008 and public health messages were circulated to local health services and to the general public via the media. Those members of the public who were concerned about health risks associated with the incident were asked to ring NHS Direct for clinical advice [26]. The HPA wrote to local GPs and hospitals asking them to monitor potential patients for signs and symptoms of Cryptosporidium infection and to submit faecal specimens to the local hospital diagnostic laboratory if patients presented with diarrhoea. Samples from 34 patients where Cryptosporidium infection was identified were sent to the UK Cryptosporidium reference unit for typing. On 30 June 2008, the Cryptosporidium oocysts found in the reservoir water were confirmed as being of the rabbit genotype Cryptosporidium cuniculus [27]. Subsequently, a dead rabbit was found in a treated water tank at the water treatment works. The genotype of Cryptosporidium oocysts in the rabbit’s large bowel was indistinguishable from that of the oocysts found in the water [27]. After remediation of the water supply and distribution, the ‘boil water notice’ was lifted on 4 July and the following day the first case of cryptosporidiosis linked to the incident was identified by the reference laboratory (this case was infected with C. cuniculus). During the course of the outbreak (24 June – 18 July 2008, the dates of symptom onset in the first and last case, respectively), 23 cases of cryptosporidiosis were confirmed as being infected with C. cuniculus; one of the 23 was a secondary case. The HPA Real-time Syndromic Surveillance Team provided data in order to aid the response to this incident and the first syndromic surveillance report was circulated to the incident management team and other relevant people in the HPA on 27 June 2008. Data from the HPA/NHS Direct and HPA/QSurveillance systems were provided in a series of regular reports, initially daily and eventually weekly, until the final report on 21 August 2008. Each report included a summary interpretation and more detailed data on diarrhoea, gastroenteritis and vomiting indicators. Methods Surveillance systems HPA/NHS Direct surveillance system NHS Direct is a 24-hour nurse-led telephone helpline that provides health information and advice to the general public. Nurses use a computerised clinical decision support system – the NHS Clinical Assessment System (NHS CAS) – to handle calls. This assessment system uses approximately 200 computerised symptom-based clinical algorithms. Nurses assign the call to the most appropriate algorithm and the patient’s symptoms determine the questions asked and the action to be taken following the call (call outcome), which could be guidance on self-care or they could be referred to their GP or advised to attend a hospital emergency department. No attempt is made to provide a formal diagnosis. Daily NHS Direct data are received by the Real-time Syndromic Surveillance Team, where the number and type of calls received during the previous day are analysed and interpreted. Call proportions are calculated by age group and algorithm against the total number of calls received. HPA/QSurveillance system The HPA/QSurveillance national surveillance system was set up by the University of Nottingham, United Kingdom, and Egton Medical Information Systems (EMIS), a supplier of general practice computer systems, in collaboration with the HPA. It comprises a network of more than 3,500 general practices throughout the UK, covering more than 22 million patients (about 38% of the population [28]). Aggregated data on GP consultations for a range of indicators are automatically uploaded daily from GP practice systems to a central database. Data are routinely reported on a weekly basis; however, daily reporting is possible for specific incidents. Reports are provided at national or regional level (strategic health authority, SHA) and by local health district (primary care trust, PCT). Analysis of surveillance data NHS Direct call proportions for gastrointestinal syndromes (diarrhoea and vomiting) for the East Midlands region in England, where Northampton is situated, were examined during the outbreak (24 June – 18 July 2008) and compared with those for England and Wales. A series of control charts for diarrhoea calls are routinely used to monitor significant rises in the numbers of calls received. Control charts are calculated by assuming that calls follow a Poisson distribution with the total number of calls as an offset: a model is fitted to each region and symptom separately [29]. The model takes into account call variation caused by weekends, public holidays and the time of year – variables that can affect the number of calls received by NHS Direct. A value above the upper limit of the 99.5% confidence interval would be considered to be unusual. The seven-day moving average for diarrhoea calls was also monitored. The number and percentage of calls for diarrhoea in the East Midlands region were presented by call outcome and the number of calls in the Northampton (NN) postcode districts and in particular the number of calls in the NN11 and NN12 postcode districts, which were most affected by the incident. QSurveillance national consultation rates per 100,000 population for diarrhoea (in the age groups under five years, five years and over, and all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with rates for the same period in 2007 (data not presented). Consultation rates by region for 2008 for diarrhoea (all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with those for the East Midlands region. The gastroenteritis indicator includes all cases of diarrhoea and/or vomiting. Consultation rates and standardised incidence ratios (SIRs) – calculated using the UK as the standard population – for diarrhoea, gastroenteritis and vomiting were compared for the UK, Yorkshire and Humberside, East Midlands, Leicestershire, Northamptonshire and Rutland SHA, and Daventry and South Northants PCT, Northamptonshire Heartlands PCT and Northampton PCT. Yorkshire and Humberside was not an affected region but was included as a control. The area supplied by the Pitsford Reservoir included the three PCTs, which were all within the Leicestershire, Northamptonshire and Rutland SHA. The consultation rates and SIRs were compared for the period from week 16 to week 35 of 2008 in order to compare the rates before and after the Cryptosporidium exceedance, which took place in week 26. Estimates of excess numbers of cases of diarrhoea occurring during and following the Cryptosporidium outbreak were made by calculating the mean consultation rate over a period of five weeks before and after the incident (weeks 20–24 and weeks 31–35, respectively). For each of the three PCTs, the calculated mean rate was applied to the PCT population to estimate the number of cases that would be expected each week. The actual consultation rates for diarrhoea for weeks 25 to 30 were used to estimate the number of cases for the PCT population each week. The expected number of cases was subtracted from the estimated number of cases in the PCT population to give the estimated number of excess cases. Results HPA/NHS Direct surveillance system A peak in the number of calls for diarrhoea in the East Midlands was recorded in 25–26 June 2008, the period that coincided with the contamination incident and the associated media coverage (Figure 1). The neighbouring areas of the West Midlands, Yorkshire and the Humber, and East of England showed no increase in the number of calls for diarrhoea. The peak produced a control chart exceedance for calls for diarrhoea on 25 June 2008 (Figure 2), when the proportion of calls exceeded the upper limit of the 99.5% confidence interval. There were further confidence interval exceedances on 26 and 28 June (which were not control chart exceedances). There was no peak in calls for vomiting or control chart exceedance for these calls in the East Midlands. HPA/QSurveillance national surveillance system The East Midlands region had significantly high consultation rates for diarrhoea and gastroenteritis in week 25 (16–22 June), week 26 (23–29 June 2008, when the contamination incident was reported) and in the following four weeks. Within this region. Leicestershire, Northamptonshire and Rutland SHA had slightly raised consultation rates and significant SIRs across weeks 25 to 30 that were not seen in the neighbouring Trent SHA. At PCT level, all three of the PCTs in the area affected by the incident showed increased consultation rates for diarrhoea (Table 1) and gastroenteritis (Table 2) with SIRs significantly above the UK rate in week 26. Daventry and South Northants PCT also had a raised SIR for both indicators in week 25, and although Northamptonshire Heartlands and Northampton PCTs did not have SIRs significantly above that of the UK in week 25, the rise in consultation rates for diarrhoea and gastroenteritis began during week 25. In Northampton PCT, consultations for both diarrhoea and gastroenteritis peaked in the week following the contamination incident, week 27, returning to normal levels by week 30 (Figure 3A and 3B). A similar effect can be seen in Northampton Heartlands PCT. Daventry and South Northants PCT also showed an increase, but appeared to have consistently higher rates for both indicators. This was the area with the smallest population so the rates were more variable than in the other PCTs and we therefore interpreted these results with caution. The consultation rates for vomiting during weeks 25 to 30 in the East Midlands were not unusual at SHA or PCT level (data not presented). Discussion We have demonstrated the sensitivity of syndromic surveillance in detecting this small Cryptosporidium outbreak and the value of the surveillance in being able to describe the extent of its spread. Both the HPA/NHS Direct and HPA/QSurveillance systems showed demonstrable increases in calls and consultations for diarrhoea that were linked to the outbreak. QSurveillance consultations appeared to increase across the PCTs immediately affected but not in the surrounding area. Both the HPA/NHS Direct and HPA/QSurveillance systems showed a clear signal at the time of the incident and we were able to describe the extent of the impact on pre-primary care and primary care services. The HPA/QSurveillance system showed a rise in consultation rates for gastrointestinal symptoms that began the week before the outbreak, consistent with the period when Cryptosporidium was present in the water leaving the Pitsford Reservoir (19–24 June 2008) and with the onset of symptoms in the first outbreak case on 24 June. Although only 33 cases were identified by the outbreak investigation team, of which 23 were confirmed as having the outbreak Cryptosporidium strain, our syndromic surveillance data detected this limited outbreak. Data also suggested a more widespread increase in general gastrointestinal symptoms around the time of the outbreak, with an estimated 422 excess diarrhoea cases; these excess cases represented an increase of about 25% above normally expected levels. It is highly probable that a proportion of these excess cases may have resulted from the increased publicity surrounding the incident – for example, it is likely that media reports contributed to the large peak in calls detected by the HPA/NHS Direct surveillance system on the day the boil water notice was issued, and could also have impacted on the GP consultation rate. It has been previously shown that reporting of mumps cases is sensitive to media coverage, with a rise in clinically reported cases following newspaper reports [30]. A similar mechanism could account for some of the excess GP consultations as cases experiencing gastrointestinal symptoms may have been more likely to consult their GP, whereas in normal circumstances they would have cared for themselves at home. It is also possible that the surveillance shows outbreak-associated cases that did not come to the attention of the outbreak team, perhaps because symptoms were not sufficiently severe to warrant further investigation, or stool samples were not provided for testing. It is interesting to note that there was no demonstrable impact on the number of calls for vomiting (which is not a prominent clinical feature of cryptosporidiosis). Other common community-based pathogens such as norovirus and rotavirus were at low levels, as is normal for that time of year [31]. In this instance, public health authorities had already been alerted to a potential problem by the water company, although the extent of the outbreak was detected by syndromic surveillance. In 2003 the syndromic surveillance systems in the city of New York, United States, were able to detect an increase in diarrhoeal illness following a power outage when there was no other indication of citywide illness [32]. The New York system covers a population of nine million, but does not regularly detect localised outbreaks [33]. It has been shown previously that the HPA/NHS Direct surveillance system would be unlikely to detect a Cryptosporidium outbreak unless call volumes are high (72% chance of detection if nine-tenths of cases called NHS Direct) [29], although the value of syndromic surveillance for such outbreaks has been recognised [34]. The system detected the East Midlands Cryptosporidium outbreak that affected a smaller population than that covered by the New York system. The three PCTs affected have a combined population of around 600,000, of which just over half use GP practices reporting to QSurveillance, yet this syndromic surveillance system was able to describe an increase in consultation rates for diarrhoea and gastroenteritis around the time of the outbreak. Limitations of the data There was extensive media reporting of the incident that may have affected both the HPA/NHS Direct and HPA/QSurveillance systems and contributed to the increase in reported gastrointestinal symptoms around the time of the contamination incident. However, the rise in consultation rates for diarrhoea began before the outbreak had been detected and therefore cannot be attributed to media coverage. The HPA/NHS Direct and HPA/QSurveillance systems monitor general symptoms and so could only monitor the relevant symptoms of diarrhoea and vomiting. They are not able to detect Cryptosporidium cases, as this would require laboratory confirmation of diagnosis, so some of the estimated excess cases could be unconnected with this incident. This outbreak was discovered by other means but both the HPA/NHS Direct and HPA/ QSurveillance systems were able to describe the extent of the disease in the general population and provide reassurance that there was no widespread impact. Compared with other populations, older people and ethnic minorities are less likely to call NHS Direct [29], and although this should not prevent detection of gastrointestinal symptoms as a result of drinking water contamination as this would affect the whole population, this may reduce the signal from the system [35]. With such large surveillance systems, there will be ‘background noise’ in the data, so procedures must be in place to correctly interpret the data and set appropriate thresholds for action. Conclusion To our knowledge, this is the first time that PCT-level data from a syndromic surveillance system, the HPA/ QSurveillance national surveillance system, have been able to show the extent of such a limited outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide PCT-level data and this Cryptosporidium contamination incident demonstrates the potential usefulness of this system.

What type of samples were examined?

1520

Value of syndromic surveillance in monitoring a focal waterborne outbreak due to an unusual Cryptosporidium genotype in Northamptonshire

The United Kingdom (UK) has several national syndromic surveillance systems. The Health Protection Agency (HPA)/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data from a national telephone helpline, while the HPA/ QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems. Data from both of these systems were used to monitor a local outbreak of cryptosporidiosis that occurred following Cryptosporidium oocyst contamination of drinking water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, in June 2008. There was a peak in the number of calls to NHS Direct concerning diarrhoea that coincided with the incident. QSurveillance data for the local areas affected by the outbreak showed a significant increase in GP consultations for diarrhoea and gastroenteritis in the week of the incident but there was no increase in consultations for vomiting. A total of 33 clinical cases of cryptosporidiosis were identified in the outbreak investigation, of which 23 were confirmed as infected with the outbreak strain. However, QSurveillance data suggest that there were an estimated 422 excess diarrhoea cases during the outbreak, an increase of about 25% over baseline weekly levels. To our knowledge, this is the first time that data from a syndromic surveillance system, the HPA/QSurveillance national surveillance system, have been able to show the extent of such a small outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide data at local health district (primary care trust) level. The Cryptosporidium contamination incident described demonstrates the potential usefulness of this information, as it is unusual for syndromic surveillance systems to be able to help monitor such a small-scale outbreak. Introduction As syndromic surveillance systems usually capture data already collected for other purposes, and monitor generic symptoms and/or clinically diagnosed disease, they provide information at an earlier stage of illness (compared with laboratory-confirmed diagnoses), so that action can be taken in time to substantially reduce the impact of disease. Some systems, for example, the Royal College of General Practitioners Weekly Returns Service, are now well established, with many years of historical data that allow monitoring of longer-term disease trends [1]. They have the ability to provide early warning of, for example, seasonal rises in influenza and can trigger public health action, such as a recommendation to prescribe antiviral drugs in line with national guidance [2-4]. They can also provide reassurance to incident response teams and the general public that an incident has not caused adverse health effects – for example, following an explosion at the Buncefield oil storage depot in Hemel Hempstead, United Kingdom (UK), in 2005, syndromic surveillance confirmed that there were no unusual rises in community-based morbidity linked to the incident [5]; following the eruption of the Eyjafjallajökull volcano in Iceland in April 2010 similar assurance was given about lack of impact on community morbidity [6]. Health departments are increasingly expected to monitor health effects of natural events such as heat wave or flooding, or implement surveillance – of which syndromic surveillance plays a major role – for mass gatherings such as the Olympics or football World Cup [7-9]. Systems in France, Australia and Taiwan use data from emergency departments [10-12], a Canadian system uses over-the-counter pharmacy sales [13,14], and in the Netherlands data from both syndromic and surrogate data sources, such as employee absence records and prescription medications dispensed by pharmacies, are included in surveillance systems [15,16]. Currently systems based on Internet searches via search engines or on queries submitted to medical websites are being developed [17,18]. In the UK, the HPA/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data collected from the NHS Direct telephone helpline [19], while the HPA/QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems [20]. The HPA Real-time Syndromic Surveillance Team is a small team that coordinates a number of syndromic surveillance systems within the HPA and takes a lead for syndromic surveillance in England [21]. This paper describes the support provided by the team to the local incident management team during a local cryptosporidiosis outbreak and shows the use of syndromic surveillance in monitoring the extent of an outbreak using the HPA/NHS Direct and HPA/QSurveillance national surveillance systems. Cryptosporidiosis Cryptosporidium is a protozoan parasite that can cause an infection in people, cattle and sometimes other animals [22]. Cryptosporidiosis is most common in children aged between one and five years, but it can affect all ages. Those with impaired immune systems are likely to be most seriously affected. Symptoms usually appear between three and 12 days after initial exposure and include watery diarrhoea, stomach pains, dehydration and fever. In its transmissible form, called an oocyst, the parasite is protected by an outer shell, which allows it to survive in the environment for a long time. Transmission occurs most often via the faeco-oral route through person-to-person or animalto-person contact, but people may also be infected by consuming contaminated water or food or by swimming in contaminated water. Although uncommon, the largest outbreaks have occurred following contamination of drinking water [23,24]. Normal chlorine disinfection procedures do not kill the oocysts, so they are removed by filtration and water companies carry out routine monitoring of treated water. Description of the incident On 25 June 2008 the local Health Protection Unit was informed by Anglian Water of an exceedence in the level of Cryptosporidium oocysts found in water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, during 19 to 24 June 2008 [25]. The reservoir supplied a population of more than 250,000 in the Northampton area. A notice advising people in the affected areas to boil all drinking water was issued on 25 June 2008 and public health messages were circulated to local health services and to the general public via the media. Those members of the public who were concerned about health risks associated with the incident were asked to ring NHS Direct for clinical advice [26]. The HPA wrote to local GPs and hospitals asking them to monitor potential patients for signs and symptoms of Cryptosporidium infection and to submit faecal specimens to the local hospital diagnostic laboratory if patients presented with diarrhoea. Samples from 34 patients where Cryptosporidium infection was identified were sent to the UK Cryptosporidium reference unit for typing. On 30 June 2008, the Cryptosporidium oocysts found in the reservoir water were confirmed as being of the rabbit genotype Cryptosporidium cuniculus [27]. Subsequently, a dead rabbit was found in a treated water tank at the water treatment works. The genotype of Cryptosporidium oocysts in the rabbit’s large bowel was indistinguishable from that of the oocysts found in the water [27]. After remediation of the water supply and distribution, the ‘boil water notice’ was lifted on 4 July and the following day the first case of cryptosporidiosis linked to the incident was identified by the reference laboratory (this case was infected with C. cuniculus). During the course of the outbreak (24 June – 18 July 2008, the dates of symptom onset in the first and last case, respectively), 23 cases of cryptosporidiosis were confirmed as being infected with C. cuniculus; one of the 23 was a secondary case. The HPA Real-time Syndromic Surveillance Team provided data in order to aid the response to this incident and the first syndromic surveillance report was circulated to the incident management team and other relevant people in the HPA on 27 June 2008. Data from the HPA/NHS Direct and HPA/QSurveillance systems were provided in a series of regular reports, initially daily and eventually weekly, until the final report on 21 August 2008. Each report included a summary interpretation and more detailed data on diarrhoea, gastroenteritis and vomiting indicators. Methods Surveillance systems HPA/NHS Direct surveillance system NHS Direct is a 24-hour nurse-led telephone helpline that provides health information and advice to the general public. Nurses use a computerised clinical decision support system – the NHS Clinical Assessment System (NHS CAS) – to handle calls. This assessment system uses approximately 200 computerised symptom-based clinical algorithms. Nurses assign the call to the most appropriate algorithm and the patient’s symptoms determine the questions asked and the action to be taken following the call (call outcome), which could be guidance on self-care or they could be referred to their GP or advised to attend a hospital emergency department. No attempt is made to provide a formal diagnosis. Daily NHS Direct data are received by the Real-time Syndromic Surveillance Team, where the number and type of calls received during the previous day are analysed and interpreted. Call proportions are calculated by age group and algorithm against the total number of calls received. HPA/QSurveillance system The HPA/QSurveillance national surveillance system was set up by the University of Nottingham, United Kingdom, and Egton Medical Information Systems (EMIS), a supplier of general practice computer systems, in collaboration with the HPA. It comprises a network of more than 3,500 general practices throughout the UK, covering more than 22 million patients (about 38% of the population [28]). Aggregated data on GP consultations for a range of indicators are automatically uploaded daily from GP practice systems to a central database. Data are routinely reported on a weekly basis; however, daily reporting is possible for specific incidents. Reports are provided at national or regional level (strategic health authority, SHA) and by local health district (primary care trust, PCT). Analysis of surveillance data NHS Direct call proportions for gastrointestinal syndromes (diarrhoea and vomiting) for the East Midlands region in England, where Northampton is situated, were examined during the outbreak (24 June – 18 July 2008) and compared with those for England and Wales. A series of control charts for diarrhoea calls are routinely used to monitor significant rises in the numbers of calls received. Control charts are calculated by assuming that calls follow a Poisson distribution with the total number of calls as an offset: a model is fitted to each region and symptom separately [29]. The model takes into account call variation caused by weekends, public holidays and the time of year – variables that can affect the number of calls received by NHS Direct. A value above the upper limit of the 99.5% confidence interval would be considered to be unusual. The seven-day moving average for diarrhoea calls was also monitored. The number and percentage of calls for diarrhoea in the East Midlands region were presented by call outcome and the number of calls in the Northampton (NN) postcode districts and in particular the number of calls in the NN11 and NN12 postcode districts, which were most affected by the incident. QSurveillance national consultation rates per 100,000 population for diarrhoea (in the age groups under five years, five years and over, and all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with rates for the same period in 2007 (data not presented). Consultation rates by region for 2008 for diarrhoea (all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with those for the East Midlands region. The gastroenteritis indicator includes all cases of diarrhoea and/or vomiting. Consultation rates and standardised incidence ratios (SIRs) – calculated using the UK as the standard population – for diarrhoea, gastroenteritis and vomiting were compared for the UK, Yorkshire and Humberside, East Midlands, Leicestershire, Northamptonshire and Rutland SHA, and Daventry and South Northants PCT, Northamptonshire Heartlands PCT and Northampton PCT. Yorkshire and Humberside was not an affected region but was included as a control. The area supplied by the Pitsford Reservoir included the three PCTs, which were all within the Leicestershire, Northamptonshire and Rutland SHA. The consultation rates and SIRs were compared for the period from week 16 to week 35 of 2008 in order to compare the rates before and after the Cryptosporidium exceedance, which took place in week 26. Estimates of excess numbers of cases of diarrhoea occurring during and following the Cryptosporidium outbreak were made by calculating the mean consultation rate over a period of five weeks before and after the incident (weeks 20–24 and weeks 31–35, respectively). For each of the three PCTs, the calculated mean rate was applied to the PCT population to estimate the number of cases that would be expected each week. The actual consultation rates for diarrhoea for weeks 25 to 30 were used to estimate the number of cases for the PCT population each week. The expected number of cases was subtracted from the estimated number of cases in the PCT population to give the estimated number of excess cases. Results HPA/NHS Direct surveillance system A peak in the number of calls for diarrhoea in the East Midlands was recorded in 25–26 June 2008, the period that coincided with the contamination incident and the associated media coverage (Figure 1). The neighbouring areas of the West Midlands, Yorkshire and the Humber, and East of England showed no increase in the number of calls for diarrhoea. The peak produced a control chart exceedance for calls for diarrhoea on 25 June 2008 (Figure 2), when the proportion of calls exceeded the upper limit of the 99.5% confidence interval. There were further confidence interval exceedances on 26 and 28 June (which were not control chart exceedances). There was no peak in calls for vomiting or control chart exceedance for these calls in the East Midlands. HPA/QSurveillance national surveillance system The East Midlands region had significantly high consultation rates for diarrhoea and gastroenteritis in week 25 (16–22 June), week 26 (23–29 June 2008, when the contamination incident was reported) and in the following four weeks. Within this region. Leicestershire, Northamptonshire and Rutland SHA had slightly raised consultation rates and significant SIRs across weeks 25 to 30 that were not seen in the neighbouring Trent SHA. At PCT level, all three of the PCTs in the area affected by the incident showed increased consultation rates for diarrhoea (Table 1) and gastroenteritis (Table 2) with SIRs significantly above the UK rate in week 26. Daventry and South Northants PCT also had a raised SIR for both indicators in week 25, and although Northamptonshire Heartlands and Northampton PCTs did not have SIRs significantly above that of the UK in week 25, the rise in consultation rates for diarrhoea and gastroenteritis began during week 25. In Northampton PCT, consultations for both diarrhoea and gastroenteritis peaked in the week following the contamination incident, week 27, returning to normal levels by week 30 (Figure 3A and 3B). A similar effect can be seen in Northampton Heartlands PCT. Daventry and South Northants PCT also showed an increase, but appeared to have consistently higher rates for both indicators. This was the area with the smallest population so the rates were more variable than in the other PCTs and we therefore interpreted these results with caution. The consultation rates for vomiting during weeks 25 to 30 in the East Midlands were not unusual at SHA or PCT level (data not presented). Discussion We have demonstrated the sensitivity of syndromic surveillance in detecting this small Cryptosporidium outbreak and the value of the surveillance in being able to describe the extent of its spread. Both the HPA/NHS Direct and HPA/QSurveillance systems showed demonstrable increases in calls and consultations for diarrhoea that were linked to the outbreak. QSurveillance consultations appeared to increase across the PCTs immediately affected but not in the surrounding area. Both the HPA/NHS Direct and HPA/QSurveillance systems showed a clear signal at the time of the incident and we were able to describe the extent of the impact on pre-primary care and primary care services. The HPA/QSurveillance system showed a rise in consultation rates for gastrointestinal symptoms that began the week before the outbreak, consistent with the period when Cryptosporidium was present in the water leaving the Pitsford Reservoir (19–24 June 2008) and with the onset of symptoms in the first outbreak case on 24 June. Although only 33 cases were identified by the outbreak investigation team, of which 23 were confirmed as having the outbreak Cryptosporidium strain, our syndromic surveillance data detected this limited outbreak. Data also suggested a more widespread increase in general gastrointestinal symptoms around the time of the outbreak, with an estimated 422 excess diarrhoea cases; these excess cases represented an increase of about 25% above normally expected levels. It is highly probable that a proportion of these excess cases may have resulted from the increased publicity surrounding the incident – for example, it is likely that media reports contributed to the large peak in calls detected by the HPA/NHS Direct surveillance system on the day the boil water notice was issued, and could also have impacted on the GP consultation rate. It has been previously shown that reporting of mumps cases is sensitive to media coverage, with a rise in clinically reported cases following newspaper reports [30]. A similar mechanism could account for some of the excess GP consultations as cases experiencing gastrointestinal symptoms may have been more likely to consult their GP, whereas in normal circumstances they would have cared for themselves at home. It is also possible that the surveillance shows outbreak-associated cases that did not come to the attention of the outbreak team, perhaps because symptoms were not sufficiently severe to warrant further investigation, or stool samples were not provided for testing. It is interesting to note that there was no demonstrable impact on the number of calls for vomiting (which is not a prominent clinical feature of cryptosporidiosis). Other common community-based pathogens such as norovirus and rotavirus were at low levels, as is normal for that time of year [31]. In this instance, public health authorities had already been alerted to a potential problem by the water company, although the extent of the outbreak was detected by syndromic surveillance. In 2003 the syndromic surveillance systems in the city of New York, United States, were able to detect an increase in diarrhoeal illness following a power outage when there was no other indication of citywide illness [32]. The New York system covers a population of nine million, but does not regularly detect localised outbreaks [33]. It has been shown previously that the HPA/NHS Direct surveillance system would be unlikely to detect a Cryptosporidium outbreak unless call volumes are high (72% chance of detection if nine-tenths of cases called NHS Direct) [29], although the value of syndromic surveillance for such outbreaks has been recognised [34]. The system detected the East Midlands Cryptosporidium outbreak that affected a smaller population than that covered by the New York system. The three PCTs affected have a combined population of around 600,000, of which just over half use GP practices reporting to QSurveillance, yet this syndromic surveillance system was able to describe an increase in consultation rates for diarrhoea and gastroenteritis around the time of the outbreak. Limitations of the data There was extensive media reporting of the incident that may have affected both the HPA/NHS Direct and HPA/QSurveillance systems and contributed to the increase in reported gastrointestinal symptoms around the time of the contamination incident. However, the rise in consultation rates for diarrhoea began before the outbreak had been detected and therefore cannot be attributed to media coverage. The HPA/NHS Direct and HPA/QSurveillance systems monitor general symptoms and so could only monitor the relevant symptoms of diarrhoea and vomiting. They are not able to detect Cryptosporidium cases, as this would require laboratory confirmation of diagnosis, so some of the estimated excess cases could be unconnected with this incident. This outbreak was discovered by other means but both the HPA/NHS Direct and HPA/ QSurveillance systems were able to describe the extent of the disease in the general population and provide reassurance that there was no widespread impact. Compared with other populations, older people and ethnic minorities are less likely to call NHS Direct [29], and although this should not prevent detection of gastrointestinal symptoms as a result of drinking water contamination as this would affect the whole population, this may reduce the signal from the system [35]. With such large surveillance systems, there will be ‘background noise’ in the data, so procedures must be in place to correctly interpret the data and set appropriate thresholds for action. Conclusion To our knowledge, this is the first time that PCT-level data from a syndromic surveillance system, the HPA/ QSurveillance national surveillance system, have been able to show the extent of such a limited outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide PCT-level data and this Cryptosporidium contamination incident demonstrates the potential usefulness of this system.

What did they test for in the samples?

1521

Value of syndromic surveillance in monitoring a focal waterborne outbreak due to an unusual Cryptosporidium genotype in Northamptonshire

The United Kingdom (UK) has several national syndromic surveillance systems. The Health Protection Agency (HPA)/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data from a national telephone helpline, while the HPA/ QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems. Data from both of these systems were used to monitor a local outbreak of cryptosporidiosis that occurred following Cryptosporidium oocyst contamination of drinking water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, in June 2008. There was a peak in the number of calls to NHS Direct concerning diarrhoea that coincided with the incident. QSurveillance data for the local areas affected by the outbreak showed a significant increase in GP consultations for diarrhoea and gastroenteritis in the week of the incident but there was no increase in consultations for vomiting. A total of 33 clinical cases of cryptosporidiosis were identified in the outbreak investigation, of which 23 were confirmed as infected with the outbreak strain. However, QSurveillance data suggest that there were an estimated 422 excess diarrhoea cases during the outbreak, an increase of about 25% over baseline weekly levels. To our knowledge, this is the first time that data from a syndromic surveillance system, the HPA/QSurveillance national surveillance system, have been able to show the extent of such a small outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide data at local health district (primary care trust) level. The Cryptosporidium contamination incident described demonstrates the potential usefulness of this information, as it is unusual for syndromic surveillance systems to be able to help monitor such a small-scale outbreak. Introduction As syndromic surveillance systems usually capture data already collected for other purposes, and monitor generic symptoms and/or clinically diagnosed disease, they provide information at an earlier stage of illness (compared with laboratory-confirmed diagnoses), so that action can be taken in time to substantially reduce the impact of disease. Some systems, for example, the Royal College of General Practitioners Weekly Returns Service, are now well established, with many years of historical data that allow monitoring of longer-term disease trends [1]. They have the ability to provide early warning of, for example, seasonal rises in influenza and can trigger public health action, such as a recommendation to prescribe antiviral drugs in line with national guidance [2-4]. They can also provide reassurance to incident response teams and the general public that an incident has not caused adverse health effects – for example, following an explosion at the Buncefield oil storage depot in Hemel Hempstead, United Kingdom (UK), in 2005, syndromic surveillance confirmed that there were no unusual rises in community-based morbidity linked to the incident [5]; following the eruption of the Eyjafjallajökull volcano in Iceland in April 2010 similar assurance was given about lack of impact on community morbidity [6]. Health departments are increasingly expected to monitor health effects of natural events such as heat wave or flooding, or implement surveillance – of which syndromic surveillance plays a major role – for mass gatherings such as the Olympics or football World Cup [7-9]. Systems in France, Australia and Taiwan use data from emergency departments [10-12], a Canadian system uses over-the-counter pharmacy sales [13,14], and in the Netherlands data from both syndromic and surrogate data sources, such as employee absence records and prescription medications dispensed by pharmacies, are included in surveillance systems [15,16]. Currently systems based on Internet searches via search engines or on queries submitted to medical websites are being developed [17,18]. In the UK, the HPA/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data collected from the NHS Direct telephone helpline [19], while the HPA/QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems [20]. The HPA Real-time Syndromic Surveillance Team is a small team that coordinates a number of syndromic surveillance systems within the HPA and takes a lead for syndromic surveillance in England [21]. This paper describes the support provided by the team to the local incident management team during a local cryptosporidiosis outbreak and shows the use of syndromic surveillance in monitoring the extent of an outbreak using the HPA/NHS Direct and HPA/QSurveillance national surveillance systems. Cryptosporidiosis Cryptosporidium is a protozoan parasite that can cause an infection in people, cattle and sometimes other animals [22]. Cryptosporidiosis is most common in children aged between one and five years, but it can affect all ages. Those with impaired immune systems are likely to be most seriously affected. Symptoms usually appear between three and 12 days after initial exposure and include watery diarrhoea, stomach pains, dehydration and fever. In its transmissible form, called an oocyst, the parasite is protected by an outer shell, which allows it to survive in the environment for a long time. Transmission occurs most often via the faeco-oral route through person-to-person or animalto-person contact, but people may also be infected by consuming contaminated water or food or by swimming in contaminated water. Although uncommon, the largest outbreaks have occurred following contamination of drinking water [23,24]. Normal chlorine disinfection procedures do not kill the oocysts, so they are removed by filtration and water companies carry out routine monitoring of treated water. Description of the incident On 25 June 2008 the local Health Protection Unit was informed by Anglian Water of an exceedence in the level of Cryptosporidium oocysts found in water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, during 19 to 24 June 2008 [25]. The reservoir supplied a population of more than 250,000 in the Northampton area. A notice advising people in the affected areas to boil all drinking water was issued on 25 June 2008 and public health messages were circulated to local health services and to the general public via the media. Those members of the public who were concerned about health risks associated with the incident were asked to ring NHS Direct for clinical advice [26]. The HPA wrote to local GPs and hospitals asking them to monitor potential patients for signs and symptoms of Cryptosporidium infection and to submit faecal specimens to the local hospital diagnostic laboratory if patients presented with diarrhoea. Samples from 34 patients where Cryptosporidium infection was identified were sent to the UK Cryptosporidium reference unit for typing. On 30 June 2008, the Cryptosporidium oocysts found in the reservoir water were confirmed as being of the rabbit genotype Cryptosporidium cuniculus [27]. Subsequently, a dead rabbit was found in a treated water tank at the water treatment works. The genotype of Cryptosporidium oocysts in the rabbit’s large bowel was indistinguishable from that of the oocysts found in the water [27]. After remediation of the water supply and distribution, the ‘boil water notice’ was lifted on 4 July and the following day the first case of cryptosporidiosis linked to the incident was identified by the reference laboratory (this case was infected with C. cuniculus). During the course of the outbreak (24 June – 18 July 2008, the dates of symptom onset in the first and last case, respectively), 23 cases of cryptosporidiosis were confirmed as being infected with C. cuniculus; one of the 23 was a secondary case. The HPA Real-time Syndromic Surveillance Team provided data in order to aid the response to this incident and the first syndromic surveillance report was circulated to the incident management team and other relevant people in the HPA on 27 June 2008. Data from the HPA/NHS Direct and HPA/QSurveillance systems were provided in a series of regular reports, initially daily and eventually weekly, until the final report on 21 August 2008. Each report included a summary interpretation and more detailed data on diarrhoea, gastroenteritis and vomiting indicators. Methods Surveillance systems HPA/NHS Direct surveillance system NHS Direct is a 24-hour nurse-led telephone helpline that provides health information and advice to the general public. Nurses use a computerised clinical decision support system – the NHS Clinical Assessment System (NHS CAS) – to handle calls. This assessment system uses approximately 200 computerised symptom-based clinical algorithms. Nurses assign the call to the most appropriate algorithm and the patient’s symptoms determine the questions asked and the action to be taken following the call (call outcome), which could be guidance on self-care or they could be referred to their GP or advised to attend a hospital emergency department. No attempt is made to provide a formal diagnosis. Daily NHS Direct data are received by the Real-time Syndromic Surveillance Team, where the number and type of calls received during the previous day are analysed and interpreted. Call proportions are calculated by age group and algorithm against the total number of calls received. HPA/QSurveillance system The HPA/QSurveillance national surveillance system was set up by the University of Nottingham, United Kingdom, and Egton Medical Information Systems (EMIS), a supplier of general practice computer systems, in collaboration with the HPA. It comprises a network of more than 3,500 general practices throughout the UK, covering more than 22 million patients (about 38% of the population [28]). Aggregated data on GP consultations for a range of indicators are automatically uploaded daily from GP practice systems to a central database. Data are routinely reported on a weekly basis; however, daily reporting is possible for specific incidents. Reports are provided at national or regional level (strategic health authority, SHA) and by local health district (primary care trust, PCT). Analysis of surveillance data NHS Direct call proportions for gastrointestinal syndromes (diarrhoea and vomiting) for the East Midlands region in England, where Northampton is situated, were examined during the outbreak (24 June – 18 July 2008) and compared with those for England and Wales. A series of control charts for diarrhoea calls are routinely used to monitor significant rises in the numbers of calls received. Control charts are calculated by assuming that calls follow a Poisson distribution with the total number of calls as an offset: a model is fitted to each region and symptom separately [29]. The model takes into account call variation caused by weekends, public holidays and the time of year – variables that can affect the number of calls received by NHS Direct. A value above the upper limit of the 99.5% confidence interval would be considered to be unusual. The seven-day moving average for diarrhoea calls was also monitored. The number and percentage of calls for diarrhoea in the East Midlands region were presented by call outcome and the number of calls in the Northampton (NN) postcode districts and in particular the number of calls in the NN11 and NN12 postcode districts, which were most affected by the incident. QSurveillance national consultation rates per 100,000 population for diarrhoea (in the age groups under five years, five years and over, and all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with rates for the same period in 2007 (data not presented). Consultation rates by region for 2008 for diarrhoea (all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with those for the East Midlands region. The gastroenteritis indicator includes all cases of diarrhoea and/or vomiting. Consultation rates and standardised incidence ratios (SIRs) – calculated using the UK as the standard population – for diarrhoea, gastroenteritis and vomiting were compared for the UK, Yorkshire and Humberside, East Midlands, Leicestershire, Northamptonshire and Rutland SHA, and Daventry and South Northants PCT, Northamptonshire Heartlands PCT and Northampton PCT. Yorkshire and Humberside was not an affected region but was included as a control. The area supplied by the Pitsford Reservoir included the three PCTs, which were all within the Leicestershire, Northamptonshire and Rutland SHA. The consultation rates and SIRs were compared for the period from week 16 to week 35 of 2008 in order to compare the rates before and after the Cryptosporidium exceedance, which took place in week 26. Estimates of excess numbers of cases of diarrhoea occurring during and following the Cryptosporidium outbreak were made by calculating the mean consultation rate over a period of five weeks before and after the incident (weeks 20–24 and weeks 31–35, respectively). For each of the three PCTs, the calculated mean rate was applied to the PCT population to estimate the number of cases that would be expected each week. The actual consultation rates for diarrhoea for weeks 25 to 30 were used to estimate the number of cases for the PCT population each week. The expected number of cases was subtracted from the estimated number of cases in the PCT population to give the estimated number of excess cases. Results HPA/NHS Direct surveillance system A peak in the number of calls for diarrhoea in the East Midlands was recorded in 25–26 June 2008, the period that coincided with the contamination incident and the associated media coverage (Figure 1). The neighbouring areas of the West Midlands, Yorkshire and the Humber, and East of England showed no increase in the number of calls for diarrhoea. The peak produced a control chart exceedance for calls for diarrhoea on 25 June 2008 (Figure 2), when the proportion of calls exceeded the upper limit of the 99.5% confidence interval. There were further confidence interval exceedances on 26 and 28 June (which were not control chart exceedances). There was no peak in calls for vomiting or control chart exceedance for these calls in the East Midlands. HPA/QSurveillance national surveillance system The East Midlands region had significantly high consultation rates for diarrhoea and gastroenteritis in week 25 (16–22 June), week 26 (23–29 June 2008, when the contamination incident was reported) and in the following four weeks. Within this region. Leicestershire, Northamptonshire and Rutland SHA had slightly raised consultation rates and significant SIRs across weeks 25 to 30 that were not seen in the neighbouring Trent SHA. At PCT level, all three of the PCTs in the area affected by the incident showed increased consultation rates for diarrhoea (Table 1) and gastroenteritis (Table 2) with SIRs significantly above the UK rate in week 26. Daventry and South Northants PCT also had a raised SIR for both indicators in week 25, and although Northamptonshire Heartlands and Northampton PCTs did not have SIRs significantly above that of the UK in week 25, the rise in consultation rates for diarrhoea and gastroenteritis began during week 25. In Northampton PCT, consultations for both diarrhoea and gastroenteritis peaked in the week following the contamination incident, week 27, returning to normal levels by week 30 (Figure 3A and 3B). A similar effect can be seen in Northampton Heartlands PCT. Daventry and South Northants PCT also showed an increase, but appeared to have consistently higher rates for both indicators. This was the area with the smallest population so the rates were more variable than in the other PCTs and we therefore interpreted these results with caution. The consultation rates for vomiting during weeks 25 to 30 in the East Midlands were not unusual at SHA or PCT level (data not presented). Discussion We have demonstrated the sensitivity of syndromic surveillance in detecting this small Cryptosporidium outbreak and the value of the surveillance in being able to describe the extent of its spread. Both the HPA/NHS Direct and HPA/QSurveillance systems showed demonstrable increases in calls and consultations for diarrhoea that were linked to the outbreak. QSurveillance consultations appeared to increase across the PCTs immediately affected but not in the surrounding area. Both the HPA/NHS Direct and HPA/QSurveillance systems showed a clear signal at the time of the incident and we were able to describe the extent of the impact on pre-primary care and primary care services. The HPA/QSurveillance system showed a rise in consultation rates for gastrointestinal symptoms that began the week before the outbreak, consistent with the period when Cryptosporidium was present in the water leaving the Pitsford Reservoir (19–24 June 2008) and with the onset of symptoms in the first outbreak case on 24 June. Although only 33 cases were identified by the outbreak investigation team, of which 23 were confirmed as having the outbreak Cryptosporidium strain, our syndromic surveillance data detected this limited outbreak. Data also suggested a more widespread increase in general gastrointestinal symptoms around the time of the outbreak, with an estimated 422 excess diarrhoea cases; these excess cases represented an increase of about 25% above normally expected levels. It is highly probable that a proportion of these excess cases may have resulted from the increased publicity surrounding the incident – for example, it is likely that media reports contributed to the large peak in calls detected by the HPA/NHS Direct surveillance system on the day the boil water notice was issued, and could also have impacted on the GP consultation rate. It has been previously shown that reporting of mumps cases is sensitive to media coverage, with a rise in clinically reported cases following newspaper reports [30]. A similar mechanism could account for some of the excess GP consultations as cases experiencing gastrointestinal symptoms may have been more likely to consult their GP, whereas in normal circumstances they would have cared for themselves at home. It is also possible that the surveillance shows outbreak-associated cases that did not come to the attention of the outbreak team, perhaps because symptoms were not sufficiently severe to warrant further investigation, or stool samples were not provided for testing. It is interesting to note that there was no demonstrable impact on the number of calls for vomiting (which is not a prominent clinical feature of cryptosporidiosis). Other common community-based pathogens such as norovirus and rotavirus were at low levels, as is normal for that time of year [31]. In this instance, public health authorities had already been alerted to a potential problem by the water company, although the extent of the outbreak was detected by syndromic surveillance. In 2003 the syndromic surveillance systems in the city of New York, United States, were able to detect an increase in diarrhoeal illness following a power outage when there was no other indication of citywide illness [32]. The New York system covers a population of nine million, but does not regularly detect localised outbreaks [33]. It has been shown previously that the HPA/NHS Direct surveillance system would be unlikely to detect a Cryptosporidium outbreak unless call volumes are high (72% chance of detection if nine-tenths of cases called NHS Direct) [29], although the value of syndromic surveillance for such outbreaks has been recognised [34]. The system detected the East Midlands Cryptosporidium outbreak that affected a smaller population than that covered by the New York system. The three PCTs affected have a combined population of around 600,000, of which just over half use GP practices reporting to QSurveillance, yet this syndromic surveillance system was able to describe an increase in consultation rates for diarrhoea and gastroenteritis around the time of the outbreak. Limitations of the data There was extensive media reporting of the incident that may have affected both the HPA/NHS Direct and HPA/QSurveillance systems and contributed to the increase in reported gastrointestinal symptoms around the time of the contamination incident. However, the rise in consultation rates for diarrhoea began before the outbreak had been detected and therefore cannot be attributed to media coverage. The HPA/NHS Direct and HPA/QSurveillance systems monitor general symptoms and so could only monitor the relevant symptoms of diarrhoea and vomiting. They are not able to detect Cryptosporidium cases, as this would require laboratory confirmation of diagnosis, so some of the estimated excess cases could be unconnected with this incident. This outbreak was discovered by other means but both the HPA/NHS Direct and HPA/ QSurveillance systems were able to describe the extent of the disease in the general population and provide reassurance that there was no widespread impact. Compared with other populations, older people and ethnic minorities are less likely to call NHS Direct [29], and although this should not prevent detection of gastrointestinal symptoms as a result of drinking water contamination as this would affect the whole population, this may reduce the signal from the system [35]. With such large surveillance systems, there will be ‘background noise’ in the data, so procedures must be in place to correctly interpret the data and set appropriate thresholds for action. Conclusion To our knowledge, this is the first time that PCT-level data from a syndromic surveillance system, the HPA/ QSurveillance national surveillance system, have been able to show the extent of such a limited outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide PCT-level data and this Cryptosporidium contamination incident demonstrates the potential usefulness of this system.

What is the concentration of the pathogens?

1522

Value of syndromic surveillance in monitoring a focal waterborne outbreak due to an unusual Cryptosporidium genotype in Northamptonshire

The United Kingdom (UK) has several national syndromic surveillance systems. The Health Protection Agency (HPA)/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data from a national telephone helpline, while the HPA/ QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems. Data from both of these systems were used to monitor a local outbreak of cryptosporidiosis that occurred following Cryptosporidium oocyst contamination of drinking water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, in June 2008. There was a peak in the number of calls to NHS Direct concerning diarrhoea that coincided with the incident. QSurveillance data for the local areas affected by the outbreak showed a significant increase in GP consultations for diarrhoea and gastroenteritis in the week of the incident but there was no increase in consultations for vomiting. A total of 33 clinical cases of cryptosporidiosis were identified in the outbreak investigation, of which 23 were confirmed as infected with the outbreak strain. However, QSurveillance data suggest that there were an estimated 422 excess diarrhoea cases during the outbreak, an increase of about 25% over baseline weekly levels. To our knowledge, this is the first time that data from a syndromic surveillance system, the HPA/QSurveillance national surveillance system, have been able to show the extent of such a small outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide data at local health district (primary care trust) level. The Cryptosporidium contamination incident described demonstrates the potential usefulness of this information, as it is unusual for syndromic surveillance systems to be able to help monitor such a small-scale outbreak. Introduction As syndromic surveillance systems usually capture data already collected for other purposes, and monitor generic symptoms and/or clinically diagnosed disease, they provide information at an earlier stage of illness (compared with laboratory-confirmed diagnoses), so that action can be taken in time to substantially reduce the impact of disease. Some systems, for example, the Royal College of General Practitioners Weekly Returns Service, are now well established, with many years of historical data that allow monitoring of longer-term disease trends [1]. They have the ability to provide early warning of, for example, seasonal rises in influenza and can trigger public health action, such as a recommendation to prescribe antiviral drugs in line with national guidance [2-4]. They can also provide reassurance to incident response teams and the general public that an incident has not caused adverse health effects – for example, following an explosion at the Buncefield oil storage depot in Hemel Hempstead, United Kingdom (UK), in 2005, syndromic surveillance confirmed that there were no unusual rises in community-based morbidity linked to the incident [5]; following the eruption of the Eyjafjallajökull volcano in Iceland in April 2010 similar assurance was given about lack of impact on community morbidity [6]. Health departments are increasingly expected to monitor health effects of natural events such as heat wave or flooding, or implement surveillance – of which syndromic surveillance plays a major role – for mass gatherings such as the Olympics or football World Cup [7-9]. Systems in France, Australia and Taiwan use data from emergency departments [10-12], a Canadian system uses over-the-counter pharmacy sales [13,14], and in the Netherlands data from both syndromic and surrogate data sources, such as employee absence records and prescription medications dispensed by pharmacies, are included in surveillance systems [15,16]. Currently systems based on Internet searches via search engines or on queries submitted to medical websites are being developed [17,18]. In the UK, the HPA/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data collected from the NHS Direct telephone helpline [19], while the HPA/QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems [20]. The HPA Real-time Syndromic Surveillance Team is a small team that coordinates a number of syndromic surveillance systems within the HPA and takes a lead for syndromic surveillance in England [21]. This paper describes the support provided by the team to the local incident management team during a local cryptosporidiosis outbreak and shows the use of syndromic surveillance in monitoring the extent of an outbreak using the HPA/NHS Direct and HPA/QSurveillance national surveillance systems. Cryptosporidiosis Cryptosporidium is a protozoan parasite that can cause an infection in people, cattle and sometimes other animals [22]. Cryptosporidiosis is most common in children aged between one and five years, but it can affect all ages. Those with impaired immune systems are likely to be most seriously affected. Symptoms usually appear between three and 12 days after initial exposure and include watery diarrhoea, stomach pains, dehydration and fever. In its transmissible form, called an oocyst, the parasite is protected by an outer shell, which allows it to survive in the environment for a long time. Transmission occurs most often via the faeco-oral route through person-to-person or animalto-person contact, but people may also be infected by consuming contaminated water or food or by swimming in contaminated water. Although uncommon, the largest outbreaks have occurred following contamination of drinking water [23,24]. Normal chlorine disinfection procedures do not kill the oocysts, so they are removed by filtration and water companies carry out routine monitoring of treated water. Description of the incident On 25 June 2008 the local Health Protection Unit was informed by Anglian Water of an exceedence in the level of Cryptosporidium oocysts found in water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, during 19 to 24 June 2008 [25]. The reservoir supplied a population of more than 250,000 in the Northampton area. A notice advising people in the affected areas to boil all drinking water was issued on 25 June 2008 and public health messages were circulated to local health services and to the general public via the media. Those members of the public who were concerned about health risks associated with the incident were asked to ring NHS Direct for clinical advice [26]. The HPA wrote to local GPs and hospitals asking them to monitor potential patients for signs and symptoms of Cryptosporidium infection and to submit faecal specimens to the local hospital diagnostic laboratory if patients presented with diarrhoea. Samples from 34 patients where Cryptosporidium infection was identified were sent to the UK Cryptosporidium reference unit for typing. On 30 June 2008, the Cryptosporidium oocysts found in the reservoir water were confirmed as being of the rabbit genotype Cryptosporidium cuniculus [27]. Subsequently, a dead rabbit was found in a treated water tank at the water treatment works. The genotype of Cryptosporidium oocysts in the rabbit’s large bowel was indistinguishable from that of the oocysts found in the water [27]. After remediation of the water supply and distribution, the ‘boil water notice’ was lifted on 4 July and the following day the first case of cryptosporidiosis linked to the incident was identified by the reference laboratory (this case was infected with C. cuniculus). During the course of the outbreak (24 June – 18 July 2008, the dates of symptom onset in the first and last case, respectively), 23 cases of cryptosporidiosis were confirmed as being infected with C. cuniculus; one of the 23 was a secondary case. The HPA Real-time Syndromic Surveillance Team provided data in order to aid the response to this incident and the first syndromic surveillance report was circulated to the incident management team and other relevant people in the HPA on 27 June 2008. Data from the HPA/NHS Direct and HPA/QSurveillance systems were provided in a series of regular reports, initially daily and eventually weekly, until the final report on 21 August 2008. Each report included a summary interpretation and more detailed data on diarrhoea, gastroenteritis and vomiting indicators. Methods Surveillance systems HPA/NHS Direct surveillance system NHS Direct is a 24-hour nurse-led telephone helpline that provides health information and advice to the general public. Nurses use a computerised clinical decision support system – the NHS Clinical Assessment System (NHS CAS) – to handle calls. This assessment system uses approximately 200 computerised symptom-based clinical algorithms. Nurses assign the call to the most appropriate algorithm and the patient’s symptoms determine the questions asked and the action to be taken following the call (call outcome), which could be guidance on self-care or they could be referred to their GP or advised to attend a hospital emergency department. No attempt is made to provide a formal diagnosis. Daily NHS Direct data are received by the Real-time Syndromic Surveillance Team, where the number and type of calls received during the previous day are analysed and interpreted. Call proportions are calculated by age group and algorithm against the total number of calls received. HPA/QSurveillance system The HPA/QSurveillance national surveillance system was set up by the University of Nottingham, United Kingdom, and Egton Medical Information Systems (EMIS), a supplier of general practice computer systems, in collaboration with the HPA. It comprises a network of more than 3,500 general practices throughout the UK, covering more than 22 million patients (about 38% of the population [28]). Aggregated data on GP consultations for a range of indicators are automatically uploaded daily from GP practice systems to a central database. Data are routinely reported on a weekly basis; however, daily reporting is possible for specific incidents. Reports are provided at national or regional level (strategic health authority, SHA) and by local health district (primary care trust, PCT). Analysis of surveillance data NHS Direct call proportions for gastrointestinal syndromes (diarrhoea and vomiting) for the East Midlands region in England, where Northampton is situated, were examined during the outbreak (24 June – 18 July 2008) and compared with those for England and Wales. A series of control charts for diarrhoea calls are routinely used to monitor significant rises in the numbers of calls received. Control charts are calculated by assuming that calls follow a Poisson distribution with the total number of calls as an offset: a model is fitted to each region and symptom separately [29]. The model takes into account call variation caused by weekends, public holidays and the time of year – variables that can affect the number of calls received by NHS Direct. A value above the upper limit of the 99.5% confidence interval would be considered to be unusual. The seven-day moving average for diarrhoea calls was also monitored. The number and percentage of calls for diarrhoea in the East Midlands region were presented by call outcome and the number of calls in the Northampton (NN) postcode districts and in particular the number of calls in the NN11 and NN12 postcode districts, which were most affected by the incident. QSurveillance national consultation rates per 100,000 population for diarrhoea (in the age groups under five years, five years and over, and all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with rates for the same period in 2007 (data not presented). Consultation rates by region for 2008 for diarrhoea (all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with those for the East Midlands region. The gastroenteritis indicator includes all cases of diarrhoea and/or vomiting. Consultation rates and standardised incidence ratios (SIRs) – calculated using the UK as the standard population – for diarrhoea, gastroenteritis and vomiting were compared for the UK, Yorkshire and Humberside, East Midlands, Leicestershire, Northamptonshire and Rutland SHA, and Daventry and South Northants PCT, Northamptonshire Heartlands PCT and Northampton PCT. Yorkshire and Humberside was not an affected region but was included as a control. The area supplied by the Pitsford Reservoir included the three PCTs, which were all within the Leicestershire, Northamptonshire and Rutland SHA. The consultation rates and SIRs were compared for the period from week 16 to week 35 of 2008 in order to compare the rates before and after the Cryptosporidium exceedance, which took place in week 26. Estimates of excess numbers of cases of diarrhoea occurring during and following the Cryptosporidium outbreak were made by calculating the mean consultation rate over a period of five weeks before and after the incident (weeks 20–24 and weeks 31–35, respectively). For each of the three PCTs, the calculated mean rate was applied to the PCT population to estimate the number of cases that would be expected each week. The actual consultation rates for diarrhoea for weeks 25 to 30 were used to estimate the number of cases for the PCT population each week. The expected number of cases was subtracted from the estimated number of cases in the PCT population to give the estimated number of excess cases. Results HPA/NHS Direct surveillance system A peak in the number of calls for diarrhoea in the East Midlands was recorded in 25–26 June 2008, the period that coincided with the contamination incident and the associated media coverage (Figure 1). The neighbouring areas of the West Midlands, Yorkshire and the Humber, and East of England showed no increase in the number of calls for diarrhoea. The peak produced a control chart exceedance for calls for diarrhoea on 25 June 2008 (Figure 2), when the proportion of calls exceeded the upper limit of the 99.5% confidence interval. There were further confidence interval exceedances on 26 and 28 June (which were not control chart exceedances). There was no peak in calls for vomiting or control chart exceedance for these calls in the East Midlands. HPA/QSurveillance national surveillance system The East Midlands region had significantly high consultation rates for diarrhoea and gastroenteritis in week 25 (16–22 June), week 26 (23–29 June 2008, when the contamination incident was reported) and in the following four weeks. Within this region. Leicestershire, Northamptonshire and Rutland SHA had slightly raised consultation rates and significant SIRs across weeks 25 to 30 that were not seen in the neighbouring Trent SHA. At PCT level, all three of the PCTs in the area affected by the incident showed increased consultation rates for diarrhoea (Table 1) and gastroenteritis (Table 2) with SIRs significantly above the UK rate in week 26. Daventry and South Northants PCT also had a raised SIR for both indicators in week 25, and although Northamptonshire Heartlands and Northampton PCTs did not have SIRs significantly above that of the UK in week 25, the rise in consultation rates for diarrhoea and gastroenteritis began during week 25. In Northampton PCT, consultations for both diarrhoea and gastroenteritis peaked in the week following the contamination incident, week 27, returning to normal levels by week 30 (Figure 3A and 3B). A similar effect can be seen in Northampton Heartlands PCT. Daventry and South Northants PCT also showed an increase, but appeared to have consistently higher rates for both indicators. This was the area with the smallest population so the rates were more variable than in the other PCTs and we therefore interpreted these results with caution. The consultation rates for vomiting during weeks 25 to 30 in the East Midlands were not unusual at SHA or PCT level (data not presented). Discussion We have demonstrated the sensitivity of syndromic surveillance in detecting this small Cryptosporidium outbreak and the value of the surveillance in being able to describe the extent of its spread. Both the HPA/NHS Direct and HPA/QSurveillance systems showed demonstrable increases in calls and consultations for diarrhoea that were linked to the outbreak. QSurveillance consultations appeared to increase across the PCTs immediately affected but not in the surrounding area. Both the HPA/NHS Direct and HPA/QSurveillance systems showed a clear signal at the time of the incident and we were able to describe the extent of the impact on pre-primary care and primary care services. The HPA/QSurveillance system showed a rise in consultation rates for gastrointestinal symptoms that began the week before the outbreak, consistent with the period when Cryptosporidium was present in the water leaving the Pitsford Reservoir (19–24 June 2008) and with the onset of symptoms in the first outbreak case on 24 June. Although only 33 cases were identified by the outbreak investigation team, of which 23 were confirmed as having the outbreak Cryptosporidium strain, our syndromic surveillance data detected this limited outbreak. Data also suggested a more widespread increase in general gastrointestinal symptoms around the time of the outbreak, with an estimated 422 excess diarrhoea cases; these excess cases represented an increase of about 25% above normally expected levels. It is highly probable that a proportion of these excess cases may have resulted from the increased publicity surrounding the incident – for example, it is likely that media reports contributed to the large peak in calls detected by the HPA/NHS Direct surveillance system on the day the boil water notice was issued, and could also have impacted on the GP consultation rate. It has been previously shown that reporting of mumps cases is sensitive to media coverage, with a rise in clinically reported cases following newspaper reports [30]. A similar mechanism could account for some of the excess GP consultations as cases experiencing gastrointestinal symptoms may have been more likely to consult their GP, whereas in normal circumstances they would have cared for themselves at home. It is also possible that the surveillance shows outbreak-associated cases that did not come to the attention of the outbreak team, perhaps because symptoms were not sufficiently severe to warrant further investigation, or stool samples were not provided for testing. It is interesting to note that there was no demonstrable impact on the number of calls for vomiting (which is not a prominent clinical feature of cryptosporidiosis). Other common community-based pathogens such as norovirus and rotavirus were at low levels, as is normal for that time of year [31]. In this instance, public health authorities had already been alerted to a potential problem by the water company, although the extent of the outbreak was detected by syndromic surveillance. In 2003 the syndromic surveillance systems in the city of New York, United States, were able to detect an increase in diarrhoeal illness following a power outage when there was no other indication of citywide illness [32]. The New York system covers a population of nine million, but does not regularly detect localised outbreaks [33]. It has been shown previously that the HPA/NHS Direct surveillance system would be unlikely to detect a Cryptosporidium outbreak unless call volumes are high (72% chance of detection if nine-tenths of cases called NHS Direct) [29], although the value of syndromic surveillance for such outbreaks has been recognised [34]. The system detected the East Midlands Cryptosporidium outbreak that affected a smaller population than that covered by the New York system. The three PCTs affected have a combined population of around 600,000, of which just over half use GP practices reporting to QSurveillance, yet this syndromic surveillance system was able to describe an increase in consultation rates for diarrhoea and gastroenteritis around the time of the outbreak. Limitations of the data There was extensive media reporting of the incident that may have affected both the HPA/NHS Direct and HPA/QSurveillance systems and contributed to the increase in reported gastrointestinal symptoms around the time of the contamination incident. However, the rise in consultation rates for diarrhoea began before the outbreak had been detected and therefore cannot be attributed to media coverage. The HPA/NHS Direct and HPA/QSurveillance systems monitor general symptoms and so could only monitor the relevant symptoms of diarrhoea and vomiting. They are not able to detect Cryptosporidium cases, as this would require laboratory confirmation of diagnosis, so some of the estimated excess cases could be unconnected with this incident. This outbreak was discovered by other means but both the HPA/NHS Direct and HPA/ QSurveillance systems were able to describe the extent of the disease in the general population and provide reassurance that there was no widespread impact. Compared with other populations, older people and ethnic minorities are less likely to call NHS Direct [29], and although this should not prevent detection of gastrointestinal symptoms as a result of drinking water contamination as this would affect the whole population, this may reduce the signal from the system [35]. With such large surveillance systems, there will be ‘background noise’ in the data, so procedures must be in place to correctly interpret the data and set appropriate thresholds for action. Conclusion To our knowledge, this is the first time that PCT-level data from a syndromic surveillance system, the HPA/ QSurveillance national surveillance system, have been able to show the extent of such a limited outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide PCT-level data and this Cryptosporidium contamination incident demonstrates the potential usefulness of this system.

What steps were taken to restore the problem?

1523

Value of syndromic surveillance in monitoring a focal waterborne outbreak due to an unusual Cryptosporidium genotype in Northamptonshire

The United Kingdom (UK) has several national syndromic surveillance systems. The Health Protection Agency (HPA)/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data from a national telephone helpline, while the HPA/ QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems. Data from both of these systems were used to monitor a local outbreak of cryptosporidiosis that occurred following Cryptosporidium oocyst contamination of drinking water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, in June 2008. There was a peak in the number of calls to NHS Direct concerning diarrhoea that coincided with the incident. QSurveillance data for the local areas affected by the outbreak showed a significant increase in GP consultations for diarrhoea and gastroenteritis in the week of the incident but there was no increase in consultations for vomiting. A total of 33 clinical cases of cryptosporidiosis were identified in the outbreak investigation, of which 23 were confirmed as infected with the outbreak strain. However, QSurveillance data suggest that there were an estimated 422 excess diarrhoea cases during the outbreak, an increase of about 25% over baseline weekly levels. To our knowledge, this is the first time that data from a syndromic surveillance system, the HPA/QSurveillance national surveillance system, have been able to show the extent of such a small outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide data at local health district (primary care trust) level. The Cryptosporidium contamination incident described demonstrates the potential usefulness of this information, as it is unusual for syndromic surveillance systems to be able to help monitor such a small-scale outbreak. Introduction As syndromic surveillance systems usually capture data already collected for other purposes, and monitor generic symptoms and/or clinically diagnosed disease, they provide information at an earlier stage of illness (compared with laboratory-confirmed diagnoses), so that action can be taken in time to substantially reduce the impact of disease. Some systems, for example, the Royal College of General Practitioners Weekly Returns Service, are now well established, with many years of historical data that allow monitoring of longer-term disease trends [1]. They have the ability to provide early warning of, for example, seasonal rises in influenza and can trigger public health action, such as a recommendation to prescribe antiviral drugs in line with national guidance [2-4]. They can also provide reassurance to incident response teams and the general public that an incident has not caused adverse health effects – for example, following an explosion at the Buncefield oil storage depot in Hemel Hempstead, United Kingdom (UK), in 2005, syndromic surveillance confirmed that there were no unusual rises in community-based morbidity linked to the incident [5]; following the eruption of the Eyjafjallajökull volcano in Iceland in April 2010 similar assurance was given about lack of impact on community morbidity [6]. Health departments are increasingly expected to monitor health effects of natural events such as heat wave or flooding, or implement surveillance – of which syndromic surveillance plays a major role – for mass gatherings such as the Olympics or football World Cup [7-9]. Systems in France, Australia and Taiwan use data from emergency departments [10-12], a Canadian system uses over-the-counter pharmacy sales [13,14], and in the Netherlands data from both syndromic and surrogate data sources, such as employee absence records and prescription medications dispensed by pharmacies, are included in surveillance systems [15,16]. Currently systems based on Internet searches via search engines or on queries submitted to medical websites are being developed [17,18]. In the UK, the HPA/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data collected from the NHS Direct telephone helpline [19], while the HPA/QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems [20]. The HPA Real-time Syndromic Surveillance Team is a small team that coordinates a number of syndromic surveillance systems within the HPA and takes a lead for syndromic surveillance in England [21]. This paper describes the support provided by the team to the local incident management team during a local cryptosporidiosis outbreak and shows the use of syndromic surveillance in monitoring the extent of an outbreak using the HPA/NHS Direct and HPA/QSurveillance national surveillance systems. Cryptosporidiosis Cryptosporidium is a protozoan parasite that can cause an infection in people, cattle and sometimes other animals [22]. Cryptosporidiosis is most common in children aged between one and five years, but it can affect all ages. Those with impaired immune systems are likely to be most seriously affected. Symptoms usually appear between three and 12 days after initial exposure and include watery diarrhoea, stomach pains, dehydration and fever. In its transmissible form, called an oocyst, the parasite is protected by an outer shell, which allows it to survive in the environment for a long time. Transmission occurs most often via the faeco-oral route through person-to-person or animalto-person contact, but people may also be infected by consuming contaminated water or food or by swimming in contaminated water. Although uncommon, the largest outbreaks have occurred following contamination of drinking water [23,24]. Normal chlorine disinfection procedures do not kill the oocysts, so they are removed by filtration and water companies carry out routine monitoring of treated water. Description of the incident On 25 June 2008 the local Health Protection Unit was informed by Anglian Water of an exceedence in the level of Cryptosporidium oocysts found in water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, during 19 to 24 June 2008 [25]. The reservoir supplied a population of more than 250,000 in the Northampton area. A notice advising people in the affected areas to boil all drinking water was issued on 25 June 2008 and public health messages were circulated to local health services and to the general public via the media. Those members of the public who were concerned about health risks associated with the incident were asked to ring NHS Direct for clinical advice [26]. The HPA wrote to local GPs and hospitals asking them to monitor potential patients for signs and symptoms of Cryptosporidium infection and to submit faecal specimens to the local hospital diagnostic laboratory if patients presented with diarrhoea. Samples from 34 patients where Cryptosporidium infection was identified were sent to the UK Cryptosporidium reference unit for typing. On 30 June 2008, the Cryptosporidium oocysts found in the reservoir water were confirmed as being of the rabbit genotype Cryptosporidium cuniculus [27]. Subsequently, a dead rabbit was found in a treated water tank at the water treatment works. The genotype of Cryptosporidium oocysts in the rabbit’s large bowel was indistinguishable from that of the oocysts found in the water [27]. After remediation of the water supply and distribution, the ‘boil water notice’ was lifted on 4 July and the following day the first case of cryptosporidiosis linked to the incident was identified by the reference laboratory (this case was infected with C. cuniculus). During the course of the outbreak (24 June – 18 July 2008, the dates of symptom onset in the first and last case, respectively), 23 cases of cryptosporidiosis were confirmed as being infected with C. cuniculus; one of the 23 was a secondary case. The HPA Real-time Syndromic Surveillance Team provided data in order to aid the response to this incident and the first syndromic surveillance report was circulated to the incident management team and other relevant people in the HPA on 27 June 2008. Data from the HPA/NHS Direct and HPA/QSurveillance systems were provided in a series of regular reports, initially daily and eventually weekly, until the final report on 21 August 2008. Each report included a summary interpretation and more detailed data on diarrhoea, gastroenteritis and vomiting indicators. Methods Surveillance systems HPA/NHS Direct surveillance system NHS Direct is a 24-hour nurse-led telephone helpline that provides health information and advice to the general public. Nurses use a computerised clinical decision support system – the NHS Clinical Assessment System (NHS CAS) – to handle calls. This assessment system uses approximately 200 computerised symptom-based clinical algorithms. Nurses assign the call to the most appropriate algorithm and the patient’s symptoms determine the questions asked and the action to be taken following the call (call outcome), which could be guidance on self-care or they could be referred to their GP or advised to attend a hospital emergency department. No attempt is made to provide a formal diagnosis. Daily NHS Direct data are received by the Real-time Syndromic Surveillance Team, where the number and type of calls received during the previous day are analysed and interpreted. Call proportions are calculated by age group and algorithm against the total number of calls received. HPA/QSurveillance system The HPA/QSurveillance national surveillance system was set up by the University of Nottingham, United Kingdom, and Egton Medical Information Systems (EMIS), a supplier of general practice computer systems, in collaboration with the HPA. It comprises a network of more than 3,500 general practices throughout the UK, covering more than 22 million patients (about 38% of the population [28]). Aggregated data on GP consultations for a range of indicators are automatically uploaded daily from GP practice systems to a central database. Data are routinely reported on a weekly basis; however, daily reporting is possible for specific incidents. Reports are provided at national or regional level (strategic health authority, SHA) and by local health district (primary care trust, PCT). Analysis of surveillance data NHS Direct call proportions for gastrointestinal syndromes (diarrhoea and vomiting) for the East Midlands region in England, where Northampton is situated, were examined during the outbreak (24 June – 18 July 2008) and compared with those for England and Wales. A series of control charts for diarrhoea calls are routinely used to monitor significant rises in the numbers of calls received. Control charts are calculated by assuming that calls follow a Poisson distribution with the total number of calls as an offset: a model is fitted to each region and symptom separately [29]. The model takes into account call variation caused by weekends, public holidays and the time of year – variables that can affect the number of calls received by NHS Direct. A value above the upper limit of the 99.5% confidence interval would be considered to be unusual. The seven-day moving average for diarrhoea calls was also monitored. The number and percentage of calls for diarrhoea in the East Midlands region were presented by call outcome and the number of calls in the Northampton (NN) postcode districts and in particular the number of calls in the NN11 and NN12 postcode districts, which were most affected by the incident. QSurveillance national consultation rates per 100,000 population for diarrhoea (in the age groups under five years, five years and over, and all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with rates for the same period in 2007 (data not presented). Consultation rates by region for 2008 for diarrhoea (all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with those for the East Midlands region. The gastroenteritis indicator includes all cases of diarrhoea and/or vomiting. Consultation rates and standardised incidence ratios (SIRs) – calculated using the UK as the standard population – for diarrhoea, gastroenteritis and vomiting were compared for the UK, Yorkshire and Humberside, East Midlands, Leicestershire, Northamptonshire and Rutland SHA, and Daventry and South Northants PCT, Northamptonshire Heartlands PCT and Northampton PCT. Yorkshire and Humberside was not an affected region but was included as a control. The area supplied by the Pitsford Reservoir included the three PCTs, which were all within the Leicestershire, Northamptonshire and Rutland SHA. The consultation rates and SIRs were compared for the period from week 16 to week 35 of 2008 in order to compare the rates before and after the Cryptosporidium exceedance, which took place in week 26. Estimates of excess numbers of cases of diarrhoea occurring during and following the Cryptosporidium outbreak were made by calculating the mean consultation rate over a period of five weeks before and after the incident (weeks 20–24 and weeks 31–35, respectively). For each of the three PCTs, the calculated mean rate was applied to the PCT population to estimate the number of cases that would be expected each week. The actual consultation rates for diarrhoea for weeks 25 to 30 were used to estimate the number of cases for the PCT population each week. The expected number of cases was subtracted from the estimated number of cases in the PCT population to give the estimated number of excess cases. Results HPA/NHS Direct surveillance system A peak in the number of calls for diarrhoea in the East Midlands was recorded in 25–26 June 2008, the period that coincided with the contamination incident and the associated media coverage (Figure 1). The neighbouring areas of the West Midlands, Yorkshire and the Humber, and East of England showed no increase in the number of calls for diarrhoea. The peak produced a control chart exceedance for calls for diarrhoea on 25 June 2008 (Figure 2), when the proportion of calls exceeded the upper limit of the 99.5% confidence interval. There were further confidence interval exceedances on 26 and 28 June (which were not control chart exceedances). There was no peak in calls for vomiting or control chart exceedance for these calls in the East Midlands. HPA/QSurveillance national surveillance system The East Midlands region had significantly high consultation rates for diarrhoea and gastroenteritis in week 25 (16–22 June), week 26 (23–29 June 2008, when the contamination incident was reported) and in the following four weeks. Within this region. Leicestershire, Northamptonshire and Rutland SHA had slightly raised consultation rates and significant SIRs across weeks 25 to 30 that were not seen in the neighbouring Trent SHA. At PCT level, all three of the PCTs in the area affected by the incident showed increased consultation rates for diarrhoea (Table 1) and gastroenteritis (Table 2) with SIRs significantly above the UK rate in week 26. Daventry and South Northants PCT also had a raised SIR for both indicators in week 25, and although Northamptonshire Heartlands and Northampton PCTs did not have SIRs significantly above that of the UK in week 25, the rise in consultation rates for diarrhoea and gastroenteritis began during week 25. In Northampton PCT, consultations for both diarrhoea and gastroenteritis peaked in the week following the contamination incident, week 27, returning to normal levels by week 30 (Figure 3A and 3B). A similar effect can be seen in Northampton Heartlands PCT. Daventry and South Northants PCT also showed an increase, but appeared to have consistently higher rates for both indicators. This was the area with the smallest population so the rates were more variable than in the other PCTs and we therefore interpreted these results with caution. The consultation rates for vomiting during weeks 25 to 30 in the East Midlands were not unusual at SHA or PCT level (data not presented). Discussion We have demonstrated the sensitivity of syndromic surveillance in detecting this small Cryptosporidium outbreak and the value of the surveillance in being able to describe the extent of its spread. Both the HPA/NHS Direct and HPA/QSurveillance systems showed demonstrable increases in calls and consultations for diarrhoea that were linked to the outbreak. QSurveillance consultations appeared to increase across the PCTs immediately affected but not in the surrounding area. Both the HPA/NHS Direct and HPA/QSurveillance systems showed a clear signal at the time of the incident and we were able to describe the extent of the impact on pre-primary care and primary care services. The HPA/QSurveillance system showed a rise in consultation rates for gastrointestinal symptoms that began the week before the outbreak, consistent with the period when Cryptosporidium was present in the water leaving the Pitsford Reservoir (19–24 June 2008) and with the onset of symptoms in the first outbreak case on 24 June. Although only 33 cases were identified by the outbreak investigation team, of which 23 were confirmed as having the outbreak Cryptosporidium strain, our syndromic surveillance data detected this limited outbreak. Data also suggested a more widespread increase in general gastrointestinal symptoms around the time of the outbreak, with an estimated 422 excess diarrhoea cases; these excess cases represented an increase of about 25% above normally expected levels. It is highly probable that a proportion of these excess cases may have resulted from the increased publicity surrounding the incident – for example, it is likely that media reports contributed to the large peak in calls detected by the HPA/NHS Direct surveillance system on the day the boil water notice was issued, and could also have impacted on the GP consultation rate. It has been previously shown that reporting of mumps cases is sensitive to media coverage, with a rise in clinically reported cases following newspaper reports [30]. A similar mechanism could account for some of the excess GP consultations as cases experiencing gastrointestinal symptoms may have been more likely to consult their GP, whereas in normal circumstances they would have cared for themselves at home. It is also possible that the surveillance shows outbreak-associated cases that did not come to the attention of the outbreak team, perhaps because symptoms were not sufficiently severe to warrant further investigation, or stool samples were not provided for testing. It is interesting to note that there was no demonstrable impact on the number of calls for vomiting (which is not a prominent clinical feature of cryptosporidiosis). Other common community-based pathogens such as norovirus and rotavirus were at low levels, as is normal for that time of year [31]. In this instance, public health authorities had already been alerted to a potential problem by the water company, although the extent of the outbreak was detected by syndromic surveillance. In 2003 the syndromic surveillance systems in the city of New York, United States, were able to detect an increase in diarrhoeal illness following a power outage when there was no other indication of citywide illness [32]. The New York system covers a population of nine million, but does not regularly detect localised outbreaks [33]. It has been shown previously that the HPA/NHS Direct surveillance system would be unlikely to detect a Cryptosporidium outbreak unless call volumes are high (72% chance of detection if nine-tenths of cases called NHS Direct) [29], although the value of syndromic surveillance for such outbreaks has been recognised [34]. The system detected the East Midlands Cryptosporidium outbreak that affected a smaller population than that covered by the New York system. The three PCTs affected have a combined population of around 600,000, of which just over half use GP practices reporting to QSurveillance, yet this syndromic surveillance system was able to describe an increase in consultation rates for diarrhoea and gastroenteritis around the time of the outbreak. Limitations of the data There was extensive media reporting of the incident that may have affected both the HPA/NHS Direct and HPA/QSurveillance systems and contributed to the increase in reported gastrointestinal symptoms around the time of the contamination incident. However, the rise in consultation rates for diarrhoea began before the outbreak had been detected and therefore cannot be attributed to media coverage. The HPA/NHS Direct and HPA/QSurveillance systems monitor general symptoms and so could only monitor the relevant symptoms of diarrhoea and vomiting. They are not able to detect Cryptosporidium cases, as this would require laboratory confirmation of diagnosis, so some of the estimated excess cases could be unconnected with this incident. This outbreak was discovered by other means but both the HPA/NHS Direct and HPA/ QSurveillance systems were able to describe the extent of the disease in the general population and provide reassurance that there was no widespread impact. Compared with other populations, older people and ethnic minorities are less likely to call NHS Direct [29], and although this should not prevent detection of gastrointestinal symptoms as a result of drinking water contamination as this would affect the whole population, this may reduce the signal from the system [35]. With such large surveillance systems, there will be ‘background noise’ in the data, so procedures must be in place to correctly interpret the data and set appropriate thresholds for action. Conclusion To our knowledge, this is the first time that PCT-level data from a syndromic surveillance system, the HPA/ QSurveillance national surveillance system, have been able to show the extent of such a limited outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide PCT-level data and this Cryptosporidium contamination incident demonstrates the potential usefulness of this system.

What was done to fix the problem?

1524

Value of syndromic surveillance in monitoring a focal waterborne outbreak due to an unusual Cryptosporidium genotype in Northamptonshire

The United Kingdom (UK) has several national syndromic surveillance systems. The Health Protection Agency (HPA)/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data from a national telephone helpline, while the HPA/ QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems. Data from both of these systems were used to monitor a local outbreak of cryptosporidiosis that occurred following Cryptosporidium oocyst contamination of drinking water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, in June 2008. There was a peak in the number of calls to NHS Direct concerning diarrhoea that coincided with the incident. QSurveillance data for the local areas affected by the outbreak showed a significant increase in GP consultations for diarrhoea and gastroenteritis in the week of the incident but there was no increase in consultations for vomiting. A total of 33 clinical cases of cryptosporidiosis were identified in the outbreak investigation, of which 23 were confirmed as infected with the outbreak strain. However, QSurveillance data suggest that there were an estimated 422 excess diarrhoea cases during the outbreak, an increase of about 25% over baseline weekly levels. To our knowledge, this is the first time that data from a syndromic surveillance system, the HPA/QSurveillance national surveillance system, have been able to show the extent of such a small outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide data at local health district (primary care trust) level. The Cryptosporidium contamination incident described demonstrates the potential usefulness of this information, as it is unusual for syndromic surveillance systems to be able to help monitor such a small-scale outbreak. Introduction As syndromic surveillance systems usually capture data already collected for other purposes, and monitor generic symptoms and/or clinically diagnosed disease, they provide information at an earlier stage of illness (compared with laboratory-confirmed diagnoses), so that action can be taken in time to substantially reduce the impact of disease. Some systems, for example, the Royal College of General Practitioners Weekly Returns Service, are now well established, with many years of historical data that allow monitoring of longer-term disease trends [1]. They have the ability to provide early warning of, for example, seasonal rises in influenza and can trigger public health action, such as a recommendation to prescribe antiviral drugs in line with national guidance [2-4]. They can also provide reassurance to incident response teams and the general public that an incident has not caused adverse health effects – for example, following an explosion at the Buncefield oil storage depot in Hemel Hempstead, United Kingdom (UK), in 2005, syndromic surveillance confirmed that there were no unusual rises in community-based morbidity linked to the incident [5]; following the eruption of the Eyjafjallajökull volcano in Iceland in April 2010 similar assurance was given about lack of impact on community morbidity [6]. Health departments are increasingly expected to monitor health effects of natural events such as heat wave or flooding, or implement surveillance – of which syndromic surveillance plays a major role – for mass gatherings such as the Olympics or football World Cup [7-9]. Systems in France, Australia and Taiwan use data from emergency departments [10-12], a Canadian system uses over-the-counter pharmacy sales [13,14], and in the Netherlands data from both syndromic and surrogate data sources, such as employee absence records and prescription medications dispensed by pharmacies, are included in surveillance systems [15,16]. Currently systems based on Internet searches via search engines or on queries submitted to medical websites are being developed [17,18]. In the UK, the HPA/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data collected from the NHS Direct telephone helpline [19], while the HPA/QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems [20]. The HPA Real-time Syndromic Surveillance Team is a small team that coordinates a number of syndromic surveillance systems within the HPA and takes a lead for syndromic surveillance in England [21]. This paper describes the support provided by the team to the local incident management team during a local cryptosporidiosis outbreak and shows the use of syndromic surveillance in monitoring the extent of an outbreak using the HPA/NHS Direct and HPA/QSurveillance national surveillance systems. Cryptosporidiosis Cryptosporidium is a protozoan parasite that can cause an infection in people, cattle and sometimes other animals [22]. Cryptosporidiosis is most common in children aged between one and five years, but it can affect all ages. Those with impaired immune systems are likely to be most seriously affected. Symptoms usually appear between three and 12 days after initial exposure and include watery diarrhoea, stomach pains, dehydration and fever. In its transmissible form, called an oocyst, the parasite is protected by an outer shell, which allows it to survive in the environment for a long time. Transmission occurs most often via the faeco-oral route through person-to-person or animalto-person contact, but people may also be infected by consuming contaminated water or food or by swimming in contaminated water. Although uncommon, the largest outbreaks have occurred following contamination of drinking water [23,24]. Normal chlorine disinfection procedures do not kill the oocysts, so they are removed by filtration and water companies carry out routine monitoring of treated water. Description of the incident On 25 June 2008 the local Health Protection Unit was informed by Anglian Water of an exceedence in the level of Cryptosporidium oocysts found in water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, during 19 to 24 June 2008 [25]. The reservoir supplied a population of more than 250,000 in the Northampton area. A notice advising people in the affected areas to boil all drinking water was issued on 25 June 2008 and public health messages were circulated to local health services and to the general public via the media. Those members of the public who were concerned about health risks associated with the incident were asked to ring NHS Direct for clinical advice [26]. The HPA wrote to local GPs and hospitals asking them to monitor potential patients for signs and symptoms of Cryptosporidium infection and to submit faecal specimens to the local hospital diagnostic laboratory if patients presented with diarrhoea. Samples from 34 patients where Cryptosporidium infection was identified were sent to the UK Cryptosporidium reference unit for typing. On 30 June 2008, the Cryptosporidium oocysts found in the reservoir water were confirmed as being of the rabbit genotype Cryptosporidium cuniculus [27]. Subsequently, a dead rabbit was found in a treated water tank at the water treatment works. The genotype of Cryptosporidium oocysts in the rabbit’s large bowel was indistinguishable from that of the oocysts found in the water [27]. After remediation of the water supply and distribution, the ‘boil water notice’ was lifted on 4 July and the following day the first case of cryptosporidiosis linked to the incident was identified by the reference laboratory (this case was infected with C. cuniculus). During the course of the outbreak (24 June – 18 July 2008, the dates of symptom onset in the first and last case, respectively), 23 cases of cryptosporidiosis were confirmed as being infected with C. cuniculus; one of the 23 was a secondary case. The HPA Real-time Syndromic Surveillance Team provided data in order to aid the response to this incident and the first syndromic surveillance report was circulated to the incident management team and other relevant people in the HPA on 27 June 2008. Data from the HPA/NHS Direct and HPA/QSurveillance systems were provided in a series of regular reports, initially daily and eventually weekly, until the final report on 21 August 2008. Each report included a summary interpretation and more detailed data on diarrhoea, gastroenteritis and vomiting indicators. Methods Surveillance systems HPA/NHS Direct surveillance system NHS Direct is a 24-hour nurse-led telephone helpline that provides health information and advice to the general public. Nurses use a computerised clinical decision support system – the NHS Clinical Assessment System (NHS CAS) – to handle calls. This assessment system uses approximately 200 computerised symptom-based clinical algorithms. Nurses assign the call to the most appropriate algorithm and the patient’s symptoms determine the questions asked and the action to be taken following the call (call outcome), which could be guidance on self-care or they could be referred to their GP or advised to attend a hospital emergency department. No attempt is made to provide a formal diagnosis. Daily NHS Direct data are received by the Real-time Syndromic Surveillance Team, where the number and type of calls received during the previous day are analysed and interpreted. Call proportions are calculated by age group and algorithm against the total number of calls received. HPA/QSurveillance system The HPA/QSurveillance national surveillance system was set up by the University of Nottingham, United Kingdom, and Egton Medical Information Systems (EMIS), a supplier of general practice computer systems, in collaboration with the HPA. It comprises a network of more than 3,500 general practices throughout the UK, covering more than 22 million patients (about 38% of the population [28]). Aggregated data on GP consultations for a range of indicators are automatically uploaded daily from GP practice systems to a central database. Data are routinely reported on a weekly basis; however, daily reporting is possible for specific incidents. Reports are provided at national or regional level (strategic health authority, SHA) and by local health district (primary care trust, PCT). Analysis of surveillance data NHS Direct call proportions for gastrointestinal syndromes (diarrhoea and vomiting) for the East Midlands region in England, where Northampton is situated, were examined during the outbreak (24 June – 18 July 2008) and compared with those for England and Wales. A series of control charts for diarrhoea calls are routinely used to monitor significant rises in the numbers of calls received. Control charts are calculated by assuming that calls follow a Poisson distribution with the total number of calls as an offset: a model is fitted to each region and symptom separately [29]. The model takes into account call variation caused by weekends, public holidays and the time of year – variables that can affect the number of calls received by NHS Direct. A value above the upper limit of the 99.5% confidence interval would be considered to be unusual. The seven-day moving average for diarrhoea calls was also monitored. The number and percentage of calls for diarrhoea in the East Midlands region were presented by call outcome and the number of calls in the Northampton (NN) postcode districts and in particular the number of calls in the NN11 and NN12 postcode districts, which were most affected by the incident. QSurveillance national consultation rates per 100,000 population for diarrhoea (in the age groups under five years, five years and over, and all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with rates for the same period in 2007 (data not presented). Consultation rates by region for 2008 for diarrhoea (all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with those for the East Midlands region. The gastroenteritis indicator includes all cases of diarrhoea and/or vomiting. Consultation rates and standardised incidence ratios (SIRs) – calculated using the UK as the standard population – for diarrhoea, gastroenteritis and vomiting were compared for the UK, Yorkshire and Humberside, East Midlands, Leicestershire, Northamptonshire and Rutland SHA, and Daventry and South Northants PCT, Northamptonshire Heartlands PCT and Northampton PCT. Yorkshire and Humberside was not an affected region but was included as a control. The area supplied by the Pitsford Reservoir included the three PCTs, which were all within the Leicestershire, Northamptonshire and Rutland SHA. The consultation rates and SIRs were compared for the period from week 16 to week 35 of 2008 in order to compare the rates before and after the Cryptosporidium exceedance, which took place in week 26. Estimates of excess numbers of cases of diarrhoea occurring during and following the Cryptosporidium outbreak were made by calculating the mean consultation rate over a period of five weeks before and after the incident (weeks 20–24 and weeks 31–35, respectively). For each of the three PCTs, the calculated mean rate was applied to the PCT population to estimate the number of cases that would be expected each week. The actual consultation rates for diarrhoea for weeks 25 to 30 were used to estimate the number of cases for the PCT population each week. The expected number of cases was subtracted from the estimated number of cases in the PCT population to give the estimated number of excess cases. Results HPA/NHS Direct surveillance system A peak in the number of calls for diarrhoea in the East Midlands was recorded in 25–26 June 2008, the period that coincided with the contamination incident and the associated media coverage (Figure 1). The neighbouring areas of the West Midlands, Yorkshire and the Humber, and East of England showed no increase in the number of calls for diarrhoea. The peak produced a control chart exceedance for calls for diarrhoea on 25 June 2008 (Figure 2), when the proportion of calls exceeded the upper limit of the 99.5% confidence interval. There were further confidence interval exceedances on 26 and 28 June (which were not control chart exceedances). There was no peak in calls for vomiting or control chart exceedance for these calls in the East Midlands. HPA/QSurveillance national surveillance system The East Midlands region had significantly high consultation rates for diarrhoea and gastroenteritis in week 25 (16–22 June), week 26 (23–29 June 2008, when the contamination incident was reported) and in the following four weeks. Within this region. Leicestershire, Northamptonshire and Rutland SHA had slightly raised consultation rates and significant SIRs across weeks 25 to 30 that were not seen in the neighbouring Trent SHA. At PCT level, all three of the PCTs in the area affected by the incident showed increased consultation rates for diarrhoea (Table 1) and gastroenteritis (Table 2) with SIRs significantly above the UK rate in week 26. Daventry and South Northants PCT also had a raised SIR for both indicators in week 25, and although Northamptonshire Heartlands and Northampton PCTs did not have SIRs significantly above that of the UK in week 25, the rise in consultation rates for diarrhoea and gastroenteritis began during week 25. In Northampton PCT, consultations for both diarrhoea and gastroenteritis peaked in the week following the contamination incident, week 27, returning to normal levels by week 30 (Figure 3A and 3B). A similar effect can be seen in Northampton Heartlands PCT. Daventry and South Northants PCT also showed an increase, but appeared to have consistently higher rates for both indicators. This was the area with the smallest population so the rates were more variable than in the other PCTs and we therefore interpreted these results with caution. The consultation rates for vomiting during weeks 25 to 30 in the East Midlands were not unusual at SHA or PCT level (data not presented). Discussion We have demonstrated the sensitivity of syndromic surveillance in detecting this small Cryptosporidium outbreak and the value of the surveillance in being able to describe the extent of its spread. Both the HPA/NHS Direct and HPA/QSurveillance systems showed demonstrable increases in calls and consultations for diarrhoea that were linked to the outbreak. QSurveillance consultations appeared to increase across the PCTs immediately affected but not in the surrounding area. Both the HPA/NHS Direct and HPA/QSurveillance systems showed a clear signal at the time of the incident and we were able to describe the extent of the impact on pre-primary care and primary care services. The HPA/QSurveillance system showed a rise in consultation rates for gastrointestinal symptoms that began the week before the outbreak, consistent with the period when Cryptosporidium was present in the water leaving the Pitsford Reservoir (19–24 June 2008) and with the onset of symptoms in the first outbreak case on 24 June. Although only 33 cases were identified by the outbreak investigation team, of which 23 were confirmed as having the outbreak Cryptosporidium strain, our syndromic surveillance data detected this limited outbreak. Data also suggested a more widespread increase in general gastrointestinal symptoms around the time of the outbreak, with an estimated 422 excess diarrhoea cases; these excess cases represented an increase of about 25% above normally expected levels. It is highly probable that a proportion of these excess cases may have resulted from the increased publicity surrounding the incident – for example, it is likely that media reports contributed to the large peak in calls detected by the HPA/NHS Direct surveillance system on the day the boil water notice was issued, and could also have impacted on the GP consultation rate. It has been previously shown that reporting of mumps cases is sensitive to media coverage, with a rise in clinically reported cases following newspaper reports [30]. A similar mechanism could account for some of the excess GP consultations as cases experiencing gastrointestinal symptoms may have been more likely to consult their GP, whereas in normal circumstances they would have cared for themselves at home. It is also possible that the surveillance shows outbreak-associated cases that did not come to the attention of the outbreak team, perhaps because symptoms were not sufficiently severe to warrant further investigation, or stool samples were not provided for testing. It is interesting to note that there was no demonstrable impact on the number of calls for vomiting (which is not a prominent clinical feature of cryptosporidiosis). Other common community-based pathogens such as norovirus and rotavirus were at low levels, as is normal for that time of year [31]. In this instance, public health authorities had already been alerted to a potential problem by the water company, although the extent of the outbreak was detected by syndromic surveillance. In 2003 the syndromic surveillance systems in the city of New York, United States, were able to detect an increase in diarrhoeal illness following a power outage when there was no other indication of citywide illness [32]. The New York system covers a population of nine million, but does not regularly detect localised outbreaks [33]. It has been shown previously that the HPA/NHS Direct surveillance system would be unlikely to detect a Cryptosporidium outbreak unless call volumes are high (72% chance of detection if nine-tenths of cases called NHS Direct) [29], although the value of syndromic surveillance for such outbreaks has been recognised [34]. The system detected the East Midlands Cryptosporidium outbreak that affected a smaller population than that covered by the New York system. The three PCTs affected have a combined population of around 600,000, of which just over half use GP practices reporting to QSurveillance, yet this syndromic surveillance system was able to describe an increase in consultation rates for diarrhoea and gastroenteritis around the time of the outbreak. Limitations of the data There was extensive media reporting of the incident that may have affected both the HPA/NHS Direct and HPA/QSurveillance systems and contributed to the increase in reported gastrointestinal symptoms around the time of the contamination incident. However, the rise in consultation rates for diarrhoea began before the outbreak had been detected and therefore cannot be attributed to media coverage. The HPA/NHS Direct and HPA/QSurveillance systems monitor general symptoms and so could only monitor the relevant symptoms of diarrhoea and vomiting. They are not able to detect Cryptosporidium cases, as this would require laboratory confirmation of diagnosis, so some of the estimated excess cases could be unconnected with this incident. This outbreak was discovered by other means but both the HPA/NHS Direct and HPA/ QSurveillance systems were able to describe the extent of the disease in the general population and provide reassurance that there was no widespread impact. Compared with other populations, older people and ethnic minorities are less likely to call NHS Direct [29], and although this should not prevent detection of gastrointestinal symptoms as a result of drinking water contamination as this would affect the whole population, this may reduce the signal from the system [35]. With such large surveillance systems, there will be ‘background noise’ in the data, so procedures must be in place to correctly interpret the data and set appropriate thresholds for action. Conclusion To our knowledge, this is the first time that PCT-level data from a syndromic surveillance system, the HPA/ QSurveillance national surveillance system, have been able to show the extent of such a limited outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide PCT-level data and this Cryptosporidium contamination incident demonstrates the potential usefulness of this system.

What could have been done to prevent the event?

1525

Value of syndromic surveillance in monitoring a focal waterborne outbreak due to an unusual Cryptosporidium genotype in Northamptonshire

The United Kingdom (UK) has several national syndromic surveillance systems. The Health Protection Agency (HPA)/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data from a national telephone helpline, while the HPA/ QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems. Data from both of these systems were used to monitor a local outbreak of cryptosporidiosis that occurred following Cryptosporidium oocyst contamination of drinking water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, in June 2008. There was a peak in the number of calls to NHS Direct concerning diarrhoea that coincided with the incident. QSurveillance data for the local areas affected by the outbreak showed a significant increase in GP consultations for diarrhoea and gastroenteritis in the week of the incident but there was no increase in consultations for vomiting. A total of 33 clinical cases of cryptosporidiosis were identified in the outbreak investigation, of which 23 were confirmed as infected with the outbreak strain. However, QSurveillance data suggest that there were an estimated 422 excess diarrhoea cases during the outbreak, an increase of about 25% over baseline weekly levels. To our knowledge, this is the first time that data from a syndromic surveillance system, the HPA/QSurveillance national surveillance system, have been able to show the extent of such a small outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide data at local health district (primary care trust) level. The Cryptosporidium contamination incident described demonstrates the potential usefulness of this information, as it is unusual for syndromic surveillance systems to be able to help monitor such a small-scale outbreak. Introduction As syndromic surveillance systems usually capture data already collected for other purposes, and monitor generic symptoms and/or clinically diagnosed disease, they provide information at an earlier stage of illness (compared with laboratory-confirmed diagnoses), so that action can be taken in time to substantially reduce the impact of disease. Some systems, for example, the Royal College of General Practitioners Weekly Returns Service, are now well established, with many years of historical data that allow monitoring of longer-term disease trends [1]. They have the ability to provide early warning of, for example, seasonal rises in influenza and can trigger public health action, such as a recommendation to prescribe antiviral drugs in line with national guidance [2-4]. They can also provide reassurance to incident response teams and the general public that an incident has not caused adverse health effects – for example, following an explosion at the Buncefield oil storage depot in Hemel Hempstead, United Kingdom (UK), in 2005, syndromic surveillance confirmed that there were no unusual rises in community-based morbidity linked to the incident [5]; following the eruption of the Eyjafjallajökull volcano in Iceland in April 2010 similar assurance was given about lack of impact on community morbidity [6]. Health departments are increasingly expected to monitor health effects of natural events such as heat wave or flooding, or implement surveillance – of which syndromic surveillance plays a major role – for mass gatherings such as the Olympics or football World Cup [7-9]. Systems in France, Australia and Taiwan use data from emergency departments [10-12], a Canadian system uses over-the-counter pharmacy sales [13,14], and in the Netherlands data from both syndromic and surrogate data sources, such as employee absence records and prescription medications dispensed by pharmacies, are included in surveillance systems [15,16]. Currently systems based on Internet searches via search engines or on queries submitted to medical websites are being developed [17,18]. In the UK, the HPA/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data collected from the NHS Direct telephone helpline [19], while the HPA/QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems [20]. The HPA Real-time Syndromic Surveillance Team is a small team that coordinates a number of syndromic surveillance systems within the HPA and takes a lead for syndromic surveillance in England [21]. This paper describes the support provided by the team to the local incident management team during a local cryptosporidiosis outbreak and shows the use of syndromic surveillance in monitoring the extent of an outbreak using the HPA/NHS Direct and HPA/QSurveillance national surveillance systems. Cryptosporidiosis Cryptosporidium is a protozoan parasite that can cause an infection in people, cattle and sometimes other animals [22]. Cryptosporidiosis is most common in children aged between one and five years, but it can affect all ages. Those with impaired immune systems are likely to be most seriously affected. Symptoms usually appear between three and 12 days after initial exposure and include watery diarrhoea, stomach pains, dehydration and fever. In its transmissible form, called an oocyst, the parasite is protected by an outer shell, which allows it to survive in the environment for a long time. Transmission occurs most often via the faeco-oral route through person-to-person or animalto-person contact, but people may also be infected by consuming contaminated water or food or by swimming in contaminated water. Although uncommon, the largest outbreaks have occurred following contamination of drinking water [23,24]. Normal chlorine disinfection procedures do not kill the oocysts, so they are removed by filtration and water companies carry out routine monitoring of treated water. Description of the incident On 25 June 2008 the local Health Protection Unit was informed by Anglian Water of an exceedence in the level of Cryptosporidium oocysts found in water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, during 19 to 24 June 2008 [25]. The reservoir supplied a population of more than 250,000 in the Northampton area. A notice advising people in the affected areas to boil all drinking water was issued on 25 June 2008 and public health messages were circulated to local health services and to the general public via the media. Those members of the public who were concerned about health risks associated with the incident were asked to ring NHS Direct for clinical advice [26]. The HPA wrote to local GPs and hospitals asking them to monitor potential patients for signs and symptoms of Cryptosporidium infection and to submit faecal specimens to the local hospital diagnostic laboratory if patients presented with diarrhoea. Samples from 34 patients where Cryptosporidium infection was identified were sent to the UK Cryptosporidium reference unit for typing. On 30 June 2008, the Cryptosporidium oocysts found in the reservoir water were confirmed as being of the rabbit genotype Cryptosporidium cuniculus [27]. Subsequently, a dead rabbit was found in a treated water tank at the water treatment works. The genotype of Cryptosporidium oocysts in the rabbit’s large bowel was indistinguishable from that of the oocysts found in the water [27]. After remediation of the water supply and distribution, the ‘boil water notice’ was lifted on 4 July and the following day the first case of cryptosporidiosis linked to the incident was identified by the reference laboratory (this case was infected with C. cuniculus). During the course of the outbreak (24 June – 18 July 2008, the dates of symptom onset in the first and last case, respectively), 23 cases of cryptosporidiosis were confirmed as being infected with C. cuniculus; one of the 23 was a secondary case. The HPA Real-time Syndromic Surveillance Team provided data in order to aid the response to this incident and the first syndromic surveillance report was circulated to the incident management team and other relevant people in the HPA on 27 June 2008. Data from the HPA/NHS Direct and HPA/QSurveillance systems were provided in a series of regular reports, initially daily and eventually weekly, until the final report on 21 August 2008. Each report included a summary interpretation and more detailed data on diarrhoea, gastroenteritis and vomiting indicators. Methods Surveillance systems HPA/NHS Direct surveillance system NHS Direct is a 24-hour nurse-led telephone helpline that provides health information and advice to the general public. Nurses use a computerised clinical decision support system – the NHS Clinical Assessment System (NHS CAS) – to handle calls. This assessment system uses approximately 200 computerised symptom-based clinical algorithms. Nurses assign the call to the most appropriate algorithm and the patient’s symptoms determine the questions asked and the action to be taken following the call (call outcome), which could be guidance on self-care or they could be referred to their GP or advised to attend a hospital emergency department. No attempt is made to provide a formal diagnosis. Daily NHS Direct data are received by the Real-time Syndromic Surveillance Team, where the number and type of calls received during the previous day are analysed and interpreted. Call proportions are calculated by age group and algorithm against the total number of calls received. HPA/QSurveillance system The HPA/QSurveillance national surveillance system was set up by the University of Nottingham, United Kingdom, and Egton Medical Information Systems (EMIS), a supplier of general practice computer systems, in collaboration with the HPA. It comprises a network of more than 3,500 general practices throughout the UK, covering more than 22 million patients (about 38% of the population [28]). Aggregated data on GP consultations for a range of indicators are automatically uploaded daily from GP practice systems to a central database. Data are routinely reported on a weekly basis; however, daily reporting is possible for specific incidents. Reports are provided at national or regional level (strategic health authority, SHA) and by local health district (primary care trust, PCT). Analysis of surveillance data NHS Direct call proportions for gastrointestinal syndromes (diarrhoea and vomiting) for the East Midlands region in England, where Northampton is situated, were examined during the outbreak (24 June – 18 July 2008) and compared with those for England and Wales. A series of control charts for diarrhoea calls are routinely used to monitor significant rises in the numbers of calls received. Control charts are calculated by assuming that calls follow a Poisson distribution with the total number of calls as an offset: a model is fitted to each region and symptom separately [29]. The model takes into account call variation caused by weekends, public holidays and the time of year – variables that can affect the number of calls received by NHS Direct. A value above the upper limit of the 99.5% confidence interval would be considered to be unusual. The seven-day moving average for diarrhoea calls was also monitored. The number and percentage of calls for diarrhoea in the East Midlands region were presented by call outcome and the number of calls in the Northampton (NN) postcode districts and in particular the number of calls in the NN11 and NN12 postcode districts, which were most affected by the incident. QSurveillance national consultation rates per 100,000 population for diarrhoea (in the age groups under five years, five years and over, and all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with rates for the same period in 2007 (data not presented). Consultation rates by region for 2008 for diarrhoea (all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with those for the East Midlands region. The gastroenteritis indicator includes all cases of diarrhoea and/or vomiting. Consultation rates and standardised incidence ratios (SIRs) – calculated using the UK as the standard population – for diarrhoea, gastroenteritis and vomiting were compared for the UK, Yorkshire and Humberside, East Midlands, Leicestershire, Northamptonshire and Rutland SHA, and Daventry and South Northants PCT, Northamptonshire Heartlands PCT and Northampton PCT. Yorkshire and Humberside was not an affected region but was included as a control. The area supplied by the Pitsford Reservoir included the three PCTs, which were all within the Leicestershire, Northamptonshire and Rutland SHA. The consultation rates and SIRs were compared for the period from week 16 to week 35 of 2008 in order to compare the rates before and after the Cryptosporidium exceedance, which took place in week 26. Estimates of excess numbers of cases of diarrhoea occurring during and following the Cryptosporidium outbreak were made by calculating the mean consultation rate over a period of five weeks before and after the incident (weeks 20–24 and weeks 31–35, respectively). For each of the three PCTs, the calculated mean rate was applied to the PCT population to estimate the number of cases that would be expected each week. The actual consultation rates for diarrhoea for weeks 25 to 30 were used to estimate the number of cases for the PCT population each week. The expected number of cases was subtracted from the estimated number of cases in the PCT population to give the estimated number of excess cases. Results HPA/NHS Direct surveillance system A peak in the number of calls for diarrhoea in the East Midlands was recorded in 25–26 June 2008, the period that coincided with the contamination incident and the associated media coverage (Figure 1). The neighbouring areas of the West Midlands, Yorkshire and the Humber, and East of England showed no increase in the number of calls for diarrhoea. The peak produced a control chart exceedance for calls for diarrhoea on 25 June 2008 (Figure 2), when the proportion of calls exceeded the upper limit of the 99.5% confidence interval. There were further confidence interval exceedances on 26 and 28 June (which were not control chart exceedances). There was no peak in calls for vomiting or control chart exceedance for these calls in the East Midlands. HPA/QSurveillance national surveillance system The East Midlands region had significantly high consultation rates for diarrhoea and gastroenteritis in week 25 (16–22 June), week 26 (23–29 June 2008, when the contamination incident was reported) and in the following four weeks. Within this region. Leicestershire, Northamptonshire and Rutland SHA had slightly raised consultation rates and significant SIRs across weeks 25 to 30 that were not seen in the neighbouring Trent SHA. At PCT level, all three of the PCTs in the area affected by the incident showed increased consultation rates for diarrhoea (Table 1) and gastroenteritis (Table 2) with SIRs significantly above the UK rate in week 26. Daventry and South Northants PCT also had a raised SIR for both indicators in week 25, and although Northamptonshire Heartlands and Northampton PCTs did not have SIRs significantly above that of the UK in week 25, the rise in consultation rates for diarrhoea and gastroenteritis began during week 25. In Northampton PCT, consultations for both diarrhoea and gastroenteritis peaked in the week following the contamination incident, week 27, returning to normal levels by week 30 (Figure 3A and 3B). A similar effect can be seen in Northampton Heartlands PCT. Daventry and South Northants PCT also showed an increase, but appeared to have consistently higher rates for both indicators. This was the area with the smallest population so the rates were more variable than in the other PCTs and we therefore interpreted these results with caution. The consultation rates for vomiting during weeks 25 to 30 in the East Midlands were not unusual at SHA or PCT level (data not presented). Discussion We have demonstrated the sensitivity of syndromic surveillance in detecting this small Cryptosporidium outbreak and the value of the surveillance in being able to describe the extent of its spread. Both the HPA/NHS Direct and HPA/QSurveillance systems showed demonstrable increases in calls and consultations for diarrhoea that were linked to the outbreak. QSurveillance consultations appeared to increase across the PCTs immediately affected but not in the surrounding area. Both the HPA/NHS Direct and HPA/QSurveillance systems showed a clear signal at the time of the incident and we were able to describe the extent of the impact on pre-primary care and primary care services. The HPA/QSurveillance system showed a rise in consultation rates for gastrointestinal symptoms that began the week before the outbreak, consistent with the period when Cryptosporidium was present in the water leaving the Pitsford Reservoir (19–24 June 2008) and with the onset of symptoms in the first outbreak case on 24 June. Although only 33 cases were identified by the outbreak investigation team, of which 23 were confirmed as having the outbreak Cryptosporidium strain, our syndromic surveillance data detected this limited outbreak. Data also suggested a more widespread increase in general gastrointestinal symptoms around the time of the outbreak, with an estimated 422 excess diarrhoea cases; these excess cases represented an increase of about 25% above normally expected levels. It is highly probable that a proportion of these excess cases may have resulted from the increased publicity surrounding the incident – for example, it is likely that media reports contributed to the large peak in calls detected by the HPA/NHS Direct surveillance system on the day the boil water notice was issued, and could also have impacted on the GP consultation rate. It has been previously shown that reporting of mumps cases is sensitive to media coverage, with a rise in clinically reported cases following newspaper reports [30]. A similar mechanism could account for some of the excess GP consultations as cases experiencing gastrointestinal symptoms may have been more likely to consult their GP, whereas in normal circumstances they would have cared for themselves at home. It is also possible that the surveillance shows outbreak-associated cases that did not come to the attention of the outbreak team, perhaps because symptoms were not sufficiently severe to warrant further investigation, or stool samples were not provided for testing. It is interesting to note that there was no demonstrable impact on the number of calls for vomiting (which is not a prominent clinical feature of cryptosporidiosis). Other common community-based pathogens such as norovirus and rotavirus were at low levels, as is normal for that time of year [31]. In this instance, public health authorities had already been alerted to a potential problem by the water company, although the extent of the outbreak was detected by syndromic surveillance. In 2003 the syndromic surveillance systems in the city of New York, United States, were able to detect an increase in diarrhoeal illness following a power outage when there was no other indication of citywide illness [32]. The New York system covers a population of nine million, but does not regularly detect localised outbreaks [33]. It has been shown previously that the HPA/NHS Direct surveillance system would be unlikely to detect a Cryptosporidium outbreak unless call volumes are high (72% chance of detection if nine-tenths of cases called NHS Direct) [29], although the value of syndromic surveillance for such outbreaks has been recognised [34]. The system detected the East Midlands Cryptosporidium outbreak that affected a smaller population than that covered by the New York system. The three PCTs affected have a combined population of around 600,000, of which just over half use GP practices reporting to QSurveillance, yet this syndromic surveillance system was able to describe an increase in consultation rates for diarrhoea and gastroenteritis around the time of the outbreak. Limitations of the data There was extensive media reporting of the incident that may have affected both the HPA/NHS Direct and HPA/QSurveillance systems and contributed to the increase in reported gastrointestinal symptoms around the time of the contamination incident. However, the rise in consultation rates for diarrhoea began before the outbreak had been detected and therefore cannot be attributed to media coverage. The HPA/NHS Direct and HPA/QSurveillance systems monitor general symptoms and so could only monitor the relevant symptoms of diarrhoea and vomiting. They are not able to detect Cryptosporidium cases, as this would require laboratory confirmation of diagnosis, so some of the estimated excess cases could be unconnected with this incident. This outbreak was discovered by other means but both the HPA/NHS Direct and HPA/ QSurveillance systems were able to describe the extent of the disease in the general population and provide reassurance that there was no widespread impact. Compared with other populations, older people and ethnic minorities are less likely to call NHS Direct [29], and although this should not prevent detection of gastrointestinal symptoms as a result of drinking water contamination as this would affect the whole population, this may reduce the signal from the system [35]. With such large surveillance systems, there will be ‘background noise’ in the data, so procedures must be in place to correctly interpret the data and set appropriate thresholds for action. Conclusion To our knowledge, this is the first time that PCT-level data from a syndromic surveillance system, the HPA/ QSurveillance national surveillance system, have been able to show the extent of such a limited outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide PCT-level data and this Cryptosporidium contamination incident demonstrates the potential usefulness of this system.

How to prevent this?

1526

Value of syndromic surveillance in monitoring a focal waterborne outbreak due to an unusual Cryptosporidium genotype in Northamptonshire

The United Kingdom (UK) has several national syndromic surveillance systems. The Health Protection Agency (HPA)/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data from a national telephone helpline, while the HPA/ QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems. Data from both of these systems were used to monitor a local outbreak of cryptosporidiosis that occurred following Cryptosporidium oocyst contamination of drinking water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, in June 2008. There was a peak in the number of calls to NHS Direct concerning diarrhoea that coincided with the incident. QSurveillance data for the local areas affected by the outbreak showed a significant increase in GP consultations for diarrhoea and gastroenteritis in the week of the incident but there was no increase in consultations for vomiting. A total of 33 clinical cases of cryptosporidiosis were identified in the outbreak investigation, of which 23 were confirmed as infected with the outbreak strain. However, QSurveillance data suggest that there were an estimated 422 excess diarrhoea cases during the outbreak, an increase of about 25% over baseline weekly levels. To our knowledge, this is the first time that data from a syndromic surveillance system, the HPA/QSurveillance national surveillance system, have been able to show the extent of such a small outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide data at local health district (primary care trust) level. The Cryptosporidium contamination incident described demonstrates the potential usefulness of this information, as it is unusual for syndromic surveillance systems to be able to help monitor such a small-scale outbreak. Introduction As syndromic surveillance systems usually capture data already collected for other purposes, and monitor generic symptoms and/or clinically diagnosed disease, they provide information at an earlier stage of illness (compared with laboratory-confirmed diagnoses), so that action can be taken in time to substantially reduce the impact of disease. Some systems, for example, the Royal College of General Practitioners Weekly Returns Service, are now well established, with many years of historical data that allow monitoring of longer-term disease trends [1]. They have the ability to provide early warning of, for example, seasonal rises in influenza and can trigger public health action, such as a recommendation to prescribe antiviral drugs in line with national guidance [2-4]. They can also provide reassurance to incident response teams and the general public that an incident has not caused adverse health effects – for example, following an explosion at the Buncefield oil storage depot in Hemel Hempstead, United Kingdom (UK), in 2005, syndromic surveillance confirmed that there were no unusual rises in community-based morbidity linked to the incident [5]; following the eruption of the Eyjafjallajökull volcano in Iceland in April 2010 similar assurance was given about lack of impact on community morbidity [6]. Health departments are increasingly expected to monitor health effects of natural events such as heat wave or flooding, or implement surveillance – of which syndromic surveillance plays a major role – for mass gatherings such as the Olympics or football World Cup [7-9]. Systems in France, Australia and Taiwan use data from emergency departments [10-12], a Canadian system uses over-the-counter pharmacy sales [13,14], and in the Netherlands data from both syndromic and surrogate data sources, such as employee absence records and prescription medications dispensed by pharmacies, are included in surveillance systems [15,16]. Currently systems based on Internet searches via search engines or on queries submitted to medical websites are being developed [17,18]. In the UK, the HPA/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data collected from the NHS Direct telephone helpline [19], while the HPA/QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems [20]. The HPA Real-time Syndromic Surveillance Team is a small team that coordinates a number of syndromic surveillance systems within the HPA and takes a lead for syndromic surveillance in England [21]. This paper describes the support provided by the team to the local incident management team during a local cryptosporidiosis outbreak and shows the use of syndromic surveillance in monitoring the extent of an outbreak using the HPA/NHS Direct and HPA/QSurveillance national surveillance systems. Cryptosporidiosis Cryptosporidium is a protozoan parasite that can cause an infection in people, cattle and sometimes other animals [22]. Cryptosporidiosis is most common in children aged between one and five years, but it can affect all ages. Those with impaired immune systems are likely to be most seriously affected. Symptoms usually appear between three and 12 days after initial exposure and include watery diarrhoea, stomach pains, dehydration and fever. In its transmissible form, called an oocyst, the parasite is protected by an outer shell, which allows it to survive in the environment for a long time. Transmission occurs most often via the faeco-oral route through person-to-person or animalto-person contact, but people may also be infected by consuming contaminated water or food or by swimming in contaminated water. Although uncommon, the largest outbreaks have occurred following contamination of drinking water [23,24]. Normal chlorine disinfection procedures do not kill the oocysts, so they are removed by filtration and water companies carry out routine monitoring of treated water. Description of the incident On 25 June 2008 the local Health Protection Unit was informed by Anglian Water of an exceedence in the level of Cryptosporidium oocysts found in water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, during 19 to 24 June 2008 [25]. The reservoir supplied a population of more than 250,000 in the Northampton area. A notice advising people in the affected areas to boil all drinking water was issued on 25 June 2008 and public health messages were circulated to local health services and to the general public via the media. Those members of the public who were concerned about health risks associated with the incident were asked to ring NHS Direct for clinical advice [26]. The HPA wrote to local GPs and hospitals asking them to monitor potential patients for signs and symptoms of Cryptosporidium infection and to submit faecal specimens to the local hospital diagnostic laboratory if patients presented with diarrhoea. Samples from 34 patients where Cryptosporidium infection was identified were sent to the UK Cryptosporidium reference unit for typing. On 30 June 2008, the Cryptosporidium oocysts found in the reservoir water were confirmed as being of the rabbit genotype Cryptosporidium cuniculus [27]. Subsequently, a dead rabbit was found in a treated water tank at the water treatment works. The genotype of Cryptosporidium oocysts in the rabbit’s large bowel was indistinguishable from that of the oocysts found in the water [27]. After remediation of the water supply and distribution, the ‘boil water notice’ was lifted on 4 July and the following day the first case of cryptosporidiosis linked to the incident was identified by the reference laboratory (this case was infected with C. cuniculus). During the course of the outbreak (24 June – 18 July 2008, the dates of symptom onset in the first and last case, respectively), 23 cases of cryptosporidiosis were confirmed as being infected with C. cuniculus; one of the 23 was a secondary case. The HPA Real-time Syndromic Surveillance Team provided data in order to aid the response to this incident and the first syndromic surveillance report was circulated to the incident management team and other relevant people in the HPA on 27 June 2008. Data from the HPA/NHS Direct and HPA/QSurveillance systems were provided in a series of regular reports, initially daily and eventually weekly, until the final report on 21 August 2008. Each report included a summary interpretation and more detailed data on diarrhoea, gastroenteritis and vomiting indicators. Methods Surveillance systems HPA/NHS Direct surveillance system NHS Direct is a 24-hour nurse-led telephone helpline that provides health information and advice to the general public. Nurses use a computerised clinical decision support system – the NHS Clinical Assessment System (NHS CAS) – to handle calls. This assessment system uses approximately 200 computerised symptom-based clinical algorithms. Nurses assign the call to the most appropriate algorithm and the patient’s symptoms determine the questions asked and the action to be taken following the call (call outcome), which could be guidance on self-care or they could be referred to their GP or advised to attend a hospital emergency department. No attempt is made to provide a formal diagnosis. Daily NHS Direct data are received by the Real-time Syndromic Surveillance Team, where the number and type of calls received during the previous day are analysed and interpreted. Call proportions are calculated by age group and algorithm against the total number of calls received. HPA/QSurveillance system The HPA/QSurveillance national surveillance system was set up by the University of Nottingham, United Kingdom, and Egton Medical Information Systems (EMIS), a supplier of general practice computer systems, in collaboration with the HPA. It comprises a network of more than 3,500 general practices throughout the UK, covering more than 22 million patients (about 38% of the population [28]). Aggregated data on GP consultations for a range of indicators are automatically uploaded daily from GP practice systems to a central database. Data are routinely reported on a weekly basis; however, daily reporting is possible for specific incidents. Reports are provided at national or regional level (strategic health authority, SHA) and by local health district (primary care trust, PCT). Analysis of surveillance data NHS Direct call proportions for gastrointestinal syndromes (diarrhoea and vomiting) for the East Midlands region in England, where Northampton is situated, were examined during the outbreak (24 June – 18 July 2008) and compared with those for England and Wales. A series of control charts for diarrhoea calls are routinely used to monitor significant rises in the numbers of calls received. Control charts are calculated by assuming that calls follow a Poisson distribution with the total number of calls as an offset: a model is fitted to each region and symptom separately [29]. The model takes into account call variation caused by weekends, public holidays and the time of year – variables that can affect the number of calls received by NHS Direct. A value above the upper limit of the 99.5% confidence interval would be considered to be unusual. The seven-day moving average for diarrhoea calls was also monitored. The number and percentage of calls for diarrhoea in the East Midlands region were presented by call outcome and the number of calls in the Northampton (NN) postcode districts and in particular the number of calls in the NN11 and NN12 postcode districts, which were most affected by the incident. QSurveillance national consultation rates per 100,000 population for diarrhoea (in the age groups under five years, five years and over, and all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with rates for the same period in 2007 (data not presented). Consultation rates by region for 2008 for diarrhoea (all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with those for the East Midlands region. The gastroenteritis indicator includes all cases of diarrhoea and/or vomiting. Consultation rates and standardised incidence ratios (SIRs) – calculated using the UK as the standard population – for diarrhoea, gastroenteritis and vomiting were compared for the UK, Yorkshire and Humberside, East Midlands, Leicestershire, Northamptonshire and Rutland SHA, and Daventry and South Northants PCT, Northamptonshire Heartlands PCT and Northampton PCT. Yorkshire and Humberside was not an affected region but was included as a control. The area supplied by the Pitsford Reservoir included the three PCTs, which were all within the Leicestershire, Northamptonshire and Rutland SHA. The consultation rates and SIRs were compared for the period from week 16 to week 35 of 2008 in order to compare the rates before and after the Cryptosporidium exceedance, which took place in week 26. Estimates of excess numbers of cases of diarrhoea occurring during and following the Cryptosporidium outbreak were made by calculating the mean consultation rate over a period of five weeks before and after the incident (weeks 20–24 and weeks 31–35, respectively). For each of the three PCTs, the calculated mean rate was applied to the PCT population to estimate the number of cases that would be expected each week. The actual consultation rates for diarrhoea for weeks 25 to 30 were used to estimate the number of cases for the PCT population each week. The expected number of cases was subtracted from the estimated number of cases in the PCT population to give the estimated number of excess cases. Results HPA/NHS Direct surveillance system A peak in the number of calls for diarrhoea in the East Midlands was recorded in 25–26 June 2008, the period that coincided with the contamination incident and the associated media coverage (Figure 1). The neighbouring areas of the West Midlands, Yorkshire and the Humber, and East of England showed no increase in the number of calls for diarrhoea. The peak produced a control chart exceedance for calls for diarrhoea on 25 June 2008 (Figure 2), when the proportion of calls exceeded the upper limit of the 99.5% confidence interval. There were further confidence interval exceedances on 26 and 28 June (which were not control chart exceedances). There was no peak in calls for vomiting or control chart exceedance for these calls in the East Midlands. HPA/QSurveillance national surveillance system The East Midlands region had significantly high consultation rates for diarrhoea and gastroenteritis in week 25 (16–22 June), week 26 (23–29 June 2008, when the contamination incident was reported) and in the following four weeks. Within this region. Leicestershire, Northamptonshire and Rutland SHA had slightly raised consultation rates and significant SIRs across weeks 25 to 30 that were not seen in the neighbouring Trent SHA. At PCT level, all three of the PCTs in the area affected by the incident showed increased consultation rates for diarrhoea (Table 1) and gastroenteritis (Table 2) with SIRs significantly above the UK rate in week 26. Daventry and South Northants PCT also had a raised SIR for both indicators in week 25, and although Northamptonshire Heartlands and Northampton PCTs did not have SIRs significantly above that of the UK in week 25, the rise in consultation rates for diarrhoea and gastroenteritis began during week 25. In Northampton PCT, consultations for both diarrhoea and gastroenteritis peaked in the week following the contamination incident, week 27, returning to normal levels by week 30 (Figure 3A and 3B). A similar effect can be seen in Northampton Heartlands PCT. Daventry and South Northants PCT also showed an increase, but appeared to have consistently higher rates for both indicators. This was the area with the smallest population so the rates were more variable than in the other PCTs and we therefore interpreted these results with caution. The consultation rates for vomiting during weeks 25 to 30 in the East Midlands were not unusual at SHA or PCT level (data not presented). Discussion We have demonstrated the sensitivity of syndromic surveillance in detecting this small Cryptosporidium outbreak and the value of the surveillance in being able to describe the extent of its spread. Both the HPA/NHS Direct and HPA/QSurveillance systems showed demonstrable increases in calls and consultations for diarrhoea that were linked to the outbreak. QSurveillance consultations appeared to increase across the PCTs immediately affected but not in the surrounding area. Both the HPA/NHS Direct and HPA/QSurveillance systems showed a clear signal at the time of the incident and we were able to describe the extent of the impact on pre-primary care and primary care services. The HPA/QSurveillance system showed a rise in consultation rates for gastrointestinal symptoms that began the week before the outbreak, consistent with the period when Cryptosporidium was present in the water leaving the Pitsford Reservoir (19–24 June 2008) and with the onset of symptoms in the first outbreak case on 24 June. Although only 33 cases were identified by the outbreak investigation team, of which 23 were confirmed as having the outbreak Cryptosporidium strain, our syndromic surveillance data detected this limited outbreak. Data also suggested a more widespread increase in general gastrointestinal symptoms around the time of the outbreak, with an estimated 422 excess diarrhoea cases; these excess cases represented an increase of about 25% above normally expected levels. It is highly probable that a proportion of these excess cases may have resulted from the increased publicity surrounding the incident – for example, it is likely that media reports contributed to the large peak in calls detected by the HPA/NHS Direct surveillance system on the day the boil water notice was issued, and could also have impacted on the GP consultation rate. It has been previously shown that reporting of mumps cases is sensitive to media coverage, with a rise in clinically reported cases following newspaper reports [30]. A similar mechanism could account for some of the excess GP consultations as cases experiencing gastrointestinal symptoms may have been more likely to consult their GP, whereas in normal circumstances they would have cared for themselves at home. It is also possible that the surveillance shows outbreak-associated cases that did not come to the attention of the outbreak team, perhaps because symptoms were not sufficiently severe to warrant further investigation, or stool samples were not provided for testing. It is interesting to note that there was no demonstrable impact on the number of calls for vomiting (which is not a prominent clinical feature of cryptosporidiosis). Other common community-based pathogens such as norovirus and rotavirus were at low levels, as is normal for that time of year [31]. In this instance, public health authorities had already been alerted to a potential problem by the water company, although the extent of the outbreak was detected by syndromic surveillance. In 2003 the syndromic surveillance systems in the city of New York, United States, were able to detect an increase in diarrhoeal illness following a power outage when there was no other indication of citywide illness [32]. The New York system covers a population of nine million, but does not regularly detect localised outbreaks [33]. It has been shown previously that the HPA/NHS Direct surveillance system would be unlikely to detect a Cryptosporidium outbreak unless call volumes are high (72% chance of detection if nine-tenths of cases called NHS Direct) [29], although the value of syndromic surveillance for such outbreaks has been recognised [34]. The system detected the East Midlands Cryptosporidium outbreak that affected a smaller population than that covered by the New York system. The three PCTs affected have a combined population of around 600,000, of which just over half use GP practices reporting to QSurveillance, yet this syndromic surveillance system was able to describe an increase in consultation rates for diarrhoea and gastroenteritis around the time of the outbreak. Limitations of the data There was extensive media reporting of the incident that may have affected both the HPA/NHS Direct and HPA/QSurveillance systems and contributed to the increase in reported gastrointestinal symptoms around the time of the contamination incident. However, the rise in consultation rates for diarrhoea began before the outbreak had been detected and therefore cannot be attributed to media coverage. The HPA/NHS Direct and HPA/QSurveillance systems monitor general symptoms and so could only monitor the relevant symptoms of diarrhoea and vomiting. They are not able to detect Cryptosporidium cases, as this would require laboratory confirmation of diagnosis, so some of the estimated excess cases could be unconnected with this incident. This outbreak was discovered by other means but both the HPA/NHS Direct and HPA/ QSurveillance systems were able to describe the extent of the disease in the general population and provide reassurance that there was no widespread impact. Compared with other populations, older people and ethnic minorities are less likely to call NHS Direct [29], and although this should not prevent detection of gastrointestinal symptoms as a result of drinking water contamination as this would affect the whole population, this may reduce the signal from the system [35]. With such large surveillance systems, there will be ‘background noise’ in the data, so procedures must be in place to correctly interpret the data and set appropriate thresholds for action. Conclusion To our knowledge, this is the first time that PCT-level data from a syndromic surveillance system, the HPA/ QSurveillance national surveillance system, have been able to show the extent of such a limited outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide PCT-level data and this Cryptosporidium contamination incident demonstrates the potential usefulness of this system.

What were the investigation steps?

1527

Value of syndromic surveillance in monitoring a focal waterborne outbreak due to an unusual Cryptosporidium genotype in Northamptonshire

The United Kingdom (UK) has several national syndromic surveillance systems. The Health Protection Agency (HPA)/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data from a national telephone helpline, while the HPA/ QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems. Data from both of these systems were used to monitor a local outbreak of cryptosporidiosis that occurred following Cryptosporidium oocyst contamination of drinking water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, in June 2008. There was a peak in the number of calls to NHS Direct concerning diarrhoea that coincided with the incident. QSurveillance data for the local areas affected by the outbreak showed a significant increase in GP consultations for diarrhoea and gastroenteritis in the week of the incident but there was no increase in consultations for vomiting. A total of 33 clinical cases of cryptosporidiosis were identified in the outbreak investigation, of which 23 were confirmed as infected with the outbreak strain. However, QSurveillance data suggest that there were an estimated 422 excess diarrhoea cases during the outbreak, an increase of about 25% over baseline weekly levels. To our knowledge, this is the first time that data from a syndromic surveillance system, the HPA/QSurveillance national surveillance system, have been able to show the extent of such a small outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide data at local health district (primary care trust) level. The Cryptosporidium contamination incident described demonstrates the potential usefulness of this information, as it is unusual for syndromic surveillance systems to be able to help monitor such a small-scale outbreak. Introduction As syndromic surveillance systems usually capture data already collected for other purposes, and monitor generic symptoms and/or clinically diagnosed disease, they provide information at an earlier stage of illness (compared with laboratory-confirmed diagnoses), so that action can be taken in time to substantially reduce the impact of disease. Some systems, for example, the Royal College of General Practitioners Weekly Returns Service, are now well established, with many years of historical data that allow monitoring of longer-term disease trends [1]. They have the ability to provide early warning of, for example, seasonal rises in influenza and can trigger public health action, such as a recommendation to prescribe antiviral drugs in line with national guidance [2-4]. They can also provide reassurance to incident response teams and the general public that an incident has not caused adverse health effects – for example, following an explosion at the Buncefield oil storage depot in Hemel Hempstead, United Kingdom (UK), in 2005, syndromic surveillance confirmed that there were no unusual rises in community-based morbidity linked to the incident [5]; following the eruption of the Eyjafjallajökull volcano in Iceland in April 2010 similar assurance was given about lack of impact on community morbidity [6]. Health departments are increasingly expected to monitor health effects of natural events such as heat wave or flooding, or implement surveillance – of which syndromic surveillance plays a major role – for mass gatherings such as the Olympics or football World Cup [7-9]. Systems in France, Australia and Taiwan use data from emergency departments [10-12], a Canadian system uses over-the-counter pharmacy sales [13,14], and in the Netherlands data from both syndromic and surrogate data sources, such as employee absence records and prescription medications dispensed by pharmacies, are included in surveillance systems [15,16]. Currently systems based on Internet searches via search engines or on queries submitted to medical websites are being developed [17,18]. In the UK, the HPA/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data collected from the NHS Direct telephone helpline [19], while the HPA/QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems [20]. The HPA Real-time Syndromic Surveillance Team is a small team that coordinates a number of syndromic surveillance systems within the HPA and takes a lead for syndromic surveillance in England [21]. This paper describes the support provided by the team to the local incident management team during a local cryptosporidiosis outbreak and shows the use of syndromic surveillance in monitoring the extent of an outbreak using the HPA/NHS Direct and HPA/QSurveillance national surveillance systems. Cryptosporidiosis Cryptosporidium is a protozoan parasite that can cause an infection in people, cattle and sometimes other animals [22]. Cryptosporidiosis is most common in children aged between one and five years, but it can affect all ages. Those with impaired immune systems are likely to be most seriously affected. Symptoms usually appear between three and 12 days after initial exposure and include watery diarrhoea, stomach pains, dehydration and fever. In its transmissible form, called an oocyst, the parasite is protected by an outer shell, which allows it to survive in the environment for a long time. Transmission occurs most often via the faeco-oral route through person-to-person or animalto-person contact, but people may also be infected by consuming contaminated water or food or by swimming in contaminated water. Although uncommon, the largest outbreaks have occurred following contamination of drinking water [23,24]. Normal chlorine disinfection procedures do not kill the oocysts, so they are removed by filtration and water companies carry out routine monitoring of treated water. Description of the incident On 25 June 2008 the local Health Protection Unit was informed by Anglian Water of an exceedence in the level of Cryptosporidium oocysts found in water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, during 19 to 24 June 2008 [25]. The reservoir supplied a population of more than 250,000 in the Northampton area. A notice advising people in the affected areas to boil all drinking water was issued on 25 June 2008 and public health messages were circulated to local health services and to the general public via the media. Those members of the public who were concerned about health risks associated with the incident were asked to ring NHS Direct for clinical advice [26]. The HPA wrote to local GPs and hospitals asking them to monitor potential patients for signs and symptoms of Cryptosporidium infection and to submit faecal specimens to the local hospital diagnostic laboratory if patients presented with diarrhoea. Samples from 34 patients where Cryptosporidium infection was identified were sent to the UK Cryptosporidium reference unit for typing. On 30 June 2008, the Cryptosporidium oocysts found in the reservoir water were confirmed as being of the rabbit genotype Cryptosporidium cuniculus [27]. Subsequently, a dead rabbit was found in a treated water tank at the water treatment works. The genotype of Cryptosporidium oocysts in the rabbit’s large bowel was indistinguishable from that of the oocysts found in the water [27]. After remediation of the water supply and distribution, the ‘boil water notice’ was lifted on 4 July and the following day the first case of cryptosporidiosis linked to the incident was identified by the reference laboratory (this case was infected with C. cuniculus). During the course of the outbreak (24 June – 18 July 2008, the dates of symptom onset in the first and last case, respectively), 23 cases of cryptosporidiosis were confirmed as being infected with C. cuniculus; one of the 23 was a secondary case. The HPA Real-time Syndromic Surveillance Team provided data in order to aid the response to this incident and the first syndromic surveillance report was circulated to the incident management team and other relevant people in the HPA on 27 June 2008. Data from the HPA/NHS Direct and HPA/QSurveillance systems were provided in a series of regular reports, initially daily and eventually weekly, until the final report on 21 August 2008. Each report included a summary interpretation and more detailed data on diarrhoea, gastroenteritis and vomiting indicators. Methods Surveillance systems HPA/NHS Direct surveillance system NHS Direct is a 24-hour nurse-led telephone helpline that provides health information and advice to the general public. Nurses use a computerised clinical decision support system – the NHS Clinical Assessment System (NHS CAS) – to handle calls. This assessment system uses approximately 200 computerised symptom-based clinical algorithms. Nurses assign the call to the most appropriate algorithm and the patient’s symptoms determine the questions asked and the action to be taken following the call (call outcome), which could be guidance on self-care or they could be referred to their GP or advised to attend a hospital emergency department. No attempt is made to provide a formal diagnosis. Daily NHS Direct data are received by the Real-time Syndromic Surveillance Team, where the number and type of calls received during the previous day are analysed and interpreted. Call proportions are calculated by age group and algorithm against the total number of calls received. HPA/QSurveillance system The HPA/QSurveillance national surveillance system was set up by the University of Nottingham, United Kingdom, and Egton Medical Information Systems (EMIS), a supplier of general practice computer systems, in collaboration with the HPA. It comprises a network of more than 3,500 general practices throughout the UK, covering more than 22 million patients (about 38% of the population [28]). Aggregated data on GP consultations for a range of indicators are automatically uploaded daily from GP practice systems to a central database. Data are routinely reported on a weekly basis; however, daily reporting is possible for specific incidents. Reports are provided at national or regional level (strategic health authority, SHA) and by local health district (primary care trust, PCT). Analysis of surveillance data NHS Direct call proportions for gastrointestinal syndromes (diarrhoea and vomiting) for the East Midlands region in England, where Northampton is situated, were examined during the outbreak (24 June – 18 July 2008) and compared with those for England and Wales. A series of control charts for diarrhoea calls are routinely used to monitor significant rises in the numbers of calls received. Control charts are calculated by assuming that calls follow a Poisson distribution with the total number of calls as an offset: a model is fitted to each region and symptom separately [29]. The model takes into account call variation caused by weekends, public holidays and the time of year – variables that can affect the number of calls received by NHS Direct. A value above the upper limit of the 99.5% confidence interval would be considered to be unusual. The seven-day moving average for diarrhoea calls was also monitored. The number and percentage of calls for diarrhoea in the East Midlands region were presented by call outcome and the number of calls in the Northampton (NN) postcode districts and in particular the number of calls in the NN11 and NN12 postcode districts, which were most affected by the incident. QSurveillance national consultation rates per 100,000 population for diarrhoea (in the age groups under five years, five years and over, and all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with rates for the same period in 2007 (data not presented). Consultation rates by region for 2008 for diarrhoea (all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with those for the East Midlands region. The gastroenteritis indicator includes all cases of diarrhoea and/or vomiting. Consultation rates and standardised incidence ratios (SIRs) – calculated using the UK as the standard population – for diarrhoea, gastroenteritis and vomiting were compared for the UK, Yorkshire and Humberside, East Midlands, Leicestershire, Northamptonshire and Rutland SHA, and Daventry and South Northants PCT, Northamptonshire Heartlands PCT and Northampton PCT. Yorkshire and Humberside was not an affected region but was included as a control. The area supplied by the Pitsford Reservoir included the three PCTs, which were all within the Leicestershire, Northamptonshire and Rutland SHA. The consultation rates and SIRs were compared for the period from week 16 to week 35 of 2008 in order to compare the rates before and after the Cryptosporidium exceedance, which took place in week 26. Estimates of excess numbers of cases of diarrhoea occurring during and following the Cryptosporidium outbreak were made by calculating the mean consultation rate over a period of five weeks before and after the incident (weeks 20–24 and weeks 31–35, respectively). For each of the three PCTs, the calculated mean rate was applied to the PCT population to estimate the number of cases that would be expected each week. The actual consultation rates for diarrhoea for weeks 25 to 30 were used to estimate the number of cases for the PCT population each week. The expected number of cases was subtracted from the estimated number of cases in the PCT population to give the estimated number of excess cases. Results HPA/NHS Direct surveillance system A peak in the number of calls for diarrhoea in the East Midlands was recorded in 25–26 June 2008, the period that coincided with the contamination incident and the associated media coverage (Figure 1). The neighbouring areas of the West Midlands, Yorkshire and the Humber, and East of England showed no increase in the number of calls for diarrhoea. The peak produced a control chart exceedance for calls for diarrhoea on 25 June 2008 (Figure 2), when the proportion of calls exceeded the upper limit of the 99.5% confidence interval. There were further confidence interval exceedances on 26 and 28 June (which were not control chart exceedances). There was no peak in calls for vomiting or control chart exceedance for these calls in the East Midlands. HPA/QSurveillance national surveillance system The East Midlands region had significantly high consultation rates for diarrhoea and gastroenteritis in week 25 (16–22 June), week 26 (23–29 June 2008, when the contamination incident was reported) and in the following four weeks. Within this region. Leicestershire, Northamptonshire and Rutland SHA had slightly raised consultation rates and significant SIRs across weeks 25 to 30 that were not seen in the neighbouring Trent SHA. At PCT level, all three of the PCTs in the area affected by the incident showed increased consultation rates for diarrhoea (Table 1) and gastroenteritis (Table 2) with SIRs significantly above the UK rate in week 26. Daventry and South Northants PCT also had a raised SIR for both indicators in week 25, and although Northamptonshire Heartlands and Northampton PCTs did not have SIRs significantly above that of the UK in week 25, the rise in consultation rates for diarrhoea and gastroenteritis began during week 25. In Northampton PCT, consultations for both diarrhoea and gastroenteritis peaked in the week following the contamination incident, week 27, returning to normal levels by week 30 (Figure 3A and 3B). A similar effect can be seen in Northampton Heartlands PCT. Daventry and South Northants PCT also showed an increase, but appeared to have consistently higher rates for both indicators. This was the area with the smallest population so the rates were more variable than in the other PCTs and we therefore interpreted these results with caution. The consultation rates for vomiting during weeks 25 to 30 in the East Midlands were not unusual at SHA or PCT level (data not presented). Discussion We have demonstrated the sensitivity of syndromic surveillance in detecting this small Cryptosporidium outbreak and the value of the surveillance in being able to describe the extent of its spread. Both the HPA/NHS Direct and HPA/QSurveillance systems showed demonstrable increases in calls and consultations for diarrhoea that were linked to the outbreak. QSurveillance consultations appeared to increase across the PCTs immediately affected but not in the surrounding area. Both the HPA/NHS Direct and HPA/QSurveillance systems showed a clear signal at the time of the incident and we were able to describe the extent of the impact on pre-primary care and primary care services. The HPA/QSurveillance system showed a rise in consultation rates for gastrointestinal symptoms that began the week before the outbreak, consistent with the period when Cryptosporidium was present in the water leaving the Pitsford Reservoir (19–24 June 2008) and with the onset of symptoms in the first outbreak case on 24 June. Although only 33 cases were identified by the outbreak investigation team, of which 23 were confirmed as having the outbreak Cryptosporidium strain, our syndromic surveillance data detected this limited outbreak. Data also suggested a more widespread increase in general gastrointestinal symptoms around the time of the outbreak, with an estimated 422 excess diarrhoea cases; these excess cases represented an increase of about 25% above normally expected levels. It is highly probable that a proportion of these excess cases may have resulted from the increased publicity surrounding the incident – for example, it is likely that media reports contributed to the large peak in calls detected by the HPA/NHS Direct surveillance system on the day the boil water notice was issued, and could also have impacted on the GP consultation rate. It has been previously shown that reporting of mumps cases is sensitive to media coverage, with a rise in clinically reported cases following newspaper reports [30]. A similar mechanism could account for some of the excess GP consultations as cases experiencing gastrointestinal symptoms may have been more likely to consult their GP, whereas in normal circumstances they would have cared for themselves at home. It is also possible that the surveillance shows outbreak-associated cases that did not come to the attention of the outbreak team, perhaps because symptoms were not sufficiently severe to warrant further investigation, or stool samples were not provided for testing. It is interesting to note that there was no demonstrable impact on the number of calls for vomiting (which is not a prominent clinical feature of cryptosporidiosis). Other common community-based pathogens such as norovirus and rotavirus were at low levels, as is normal for that time of year [31]. In this instance, public health authorities had already been alerted to a potential problem by the water company, although the extent of the outbreak was detected by syndromic surveillance. In 2003 the syndromic surveillance systems in the city of New York, United States, were able to detect an increase in diarrhoeal illness following a power outage when there was no other indication of citywide illness [32]. The New York system covers a population of nine million, but does not regularly detect localised outbreaks [33]. It has been shown previously that the HPA/NHS Direct surveillance system would be unlikely to detect a Cryptosporidium outbreak unless call volumes are high (72% chance of detection if nine-tenths of cases called NHS Direct) [29], although the value of syndromic surveillance for such outbreaks has been recognised [34]. The system detected the East Midlands Cryptosporidium outbreak that affected a smaller population than that covered by the New York system. The three PCTs affected have a combined population of around 600,000, of which just over half use GP practices reporting to QSurveillance, yet this syndromic surveillance system was able to describe an increase in consultation rates for diarrhoea and gastroenteritis around the time of the outbreak. Limitations of the data There was extensive media reporting of the incident that may have affected both the HPA/NHS Direct and HPA/QSurveillance systems and contributed to the increase in reported gastrointestinal symptoms around the time of the contamination incident. However, the rise in consultation rates for diarrhoea began before the outbreak had been detected and therefore cannot be attributed to media coverage. The HPA/NHS Direct and HPA/QSurveillance systems monitor general symptoms and so could only monitor the relevant symptoms of diarrhoea and vomiting. They are not able to detect Cryptosporidium cases, as this would require laboratory confirmation of diagnosis, so some of the estimated excess cases could be unconnected with this incident. This outbreak was discovered by other means but both the HPA/NHS Direct and HPA/ QSurveillance systems were able to describe the extent of the disease in the general population and provide reassurance that there was no widespread impact. Compared with other populations, older people and ethnic minorities are less likely to call NHS Direct [29], and although this should not prevent detection of gastrointestinal symptoms as a result of drinking water contamination as this would affect the whole population, this may reduce the signal from the system [35]. With such large surveillance systems, there will be ‘background noise’ in the data, so procedures must be in place to correctly interpret the data and set appropriate thresholds for action. Conclusion To our knowledge, this is the first time that PCT-level data from a syndromic surveillance system, the HPA/ QSurveillance national surveillance system, have been able to show the extent of such a limited outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide PCT-level data and this Cryptosporidium contamination incident demonstrates the potential usefulness of this system.

What did the investigation find?

1528

Value of syndromic surveillance in monitoring a focal waterborne outbreak due to an unusual Cryptosporidium genotype in Northamptonshire

The United Kingdom (UK) has several national syndromic surveillance systems. The Health Protection Agency (HPA)/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data from a national telephone helpline, while the HPA/ QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems. Data from both of these systems were used to monitor a local outbreak of cryptosporidiosis that occurred following Cryptosporidium oocyst contamination of drinking water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, in June 2008. There was a peak in the number of calls to NHS Direct concerning diarrhoea that coincided with the incident. QSurveillance data for the local areas affected by the outbreak showed a significant increase in GP consultations for diarrhoea and gastroenteritis in the week of the incident but there was no increase in consultations for vomiting. A total of 33 clinical cases of cryptosporidiosis were identified in the outbreak investigation, of which 23 were confirmed as infected with the outbreak strain. However, QSurveillance data suggest that there were an estimated 422 excess diarrhoea cases during the outbreak, an increase of about 25% over baseline weekly levels. To our knowledge, this is the first time that data from a syndromic surveillance system, the HPA/QSurveillance national surveillance system, have been able to show the extent of such a small outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide data at local health district (primary care trust) level. The Cryptosporidium contamination incident described demonstrates the potential usefulness of this information, as it is unusual for syndromic surveillance systems to be able to help monitor such a small-scale outbreak. Introduction As syndromic surveillance systems usually capture data already collected for other purposes, and monitor generic symptoms and/or clinically diagnosed disease, they provide information at an earlier stage of illness (compared with laboratory-confirmed diagnoses), so that action can be taken in time to substantially reduce the impact of disease. Some systems, for example, the Royal College of General Practitioners Weekly Returns Service, are now well established, with many years of historical data that allow monitoring of longer-term disease trends [1]. They have the ability to provide early warning of, for example, seasonal rises in influenza and can trigger public health action, such as a recommendation to prescribe antiviral drugs in line with national guidance [2-4]. They can also provide reassurance to incident response teams and the general public that an incident has not caused adverse health effects – for example, following an explosion at the Buncefield oil storage depot in Hemel Hempstead, United Kingdom (UK), in 2005, syndromic surveillance confirmed that there were no unusual rises in community-based morbidity linked to the incident [5]; following the eruption of the Eyjafjallajökull volcano in Iceland in April 2010 similar assurance was given about lack of impact on community morbidity [6]. Health departments are increasingly expected to monitor health effects of natural events such as heat wave or flooding, or implement surveillance – of which syndromic surveillance plays a major role – for mass gatherings such as the Olympics or football World Cup [7-9]. Systems in France, Australia and Taiwan use data from emergency departments [10-12], a Canadian system uses over-the-counter pharmacy sales [13,14], and in the Netherlands data from both syndromic and surrogate data sources, such as employee absence records and prescription medications dispensed by pharmacies, are included in surveillance systems [15,16]. Currently systems based on Internet searches via search engines or on queries submitted to medical websites are being developed [17,18]. In the UK, the HPA/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data collected from the NHS Direct telephone helpline [19], while the HPA/QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems [20]. The HPA Real-time Syndromic Surveillance Team is a small team that coordinates a number of syndromic surveillance systems within the HPA and takes a lead for syndromic surveillance in England [21]. This paper describes the support provided by the team to the local incident management team during a local cryptosporidiosis outbreak and shows the use of syndromic surveillance in monitoring the extent of an outbreak using the HPA/NHS Direct and HPA/QSurveillance national surveillance systems. Cryptosporidiosis Cryptosporidium is a protozoan parasite that can cause an infection in people, cattle and sometimes other animals [22]. Cryptosporidiosis is most common in children aged between one and five years, but it can affect all ages. Those with impaired immune systems are likely to be most seriously affected. Symptoms usually appear between three and 12 days after initial exposure and include watery diarrhoea, stomach pains, dehydration and fever. In its transmissible form, called an oocyst, the parasite is protected by an outer shell, which allows it to survive in the environment for a long time. Transmission occurs most often via the faeco-oral route through person-to-person or animalto-person contact, but people may also be infected by consuming contaminated water or food or by swimming in contaminated water. Although uncommon, the largest outbreaks have occurred following contamination of drinking water [23,24]. Normal chlorine disinfection procedures do not kill the oocysts, so they are removed by filtration and water companies carry out routine monitoring of treated water. Description of the incident On 25 June 2008 the local Health Protection Unit was informed by Anglian Water of an exceedence in the level of Cryptosporidium oocysts found in water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, during 19 to 24 June 2008 [25]. The reservoir supplied a population of more than 250,000 in the Northampton area. A notice advising people in the affected areas to boil all drinking water was issued on 25 June 2008 and public health messages were circulated to local health services and to the general public via the media. Those members of the public who were concerned about health risks associated with the incident were asked to ring NHS Direct for clinical advice [26]. The HPA wrote to local GPs and hospitals asking them to monitor potential patients for signs and symptoms of Cryptosporidium infection and to submit faecal specimens to the local hospital diagnostic laboratory if patients presented with diarrhoea. Samples from 34 patients where Cryptosporidium infection was identified were sent to the UK Cryptosporidium reference unit for typing. On 30 June 2008, the Cryptosporidium oocysts found in the reservoir water were confirmed as being of the rabbit genotype Cryptosporidium cuniculus [27]. Subsequently, a dead rabbit was found in a treated water tank at the water treatment works. The genotype of Cryptosporidium oocysts in the rabbit’s large bowel was indistinguishable from that of the oocysts found in the water [27]. After remediation of the water supply and distribution, the ‘boil water notice’ was lifted on 4 July and the following day the first case of cryptosporidiosis linked to the incident was identified by the reference laboratory (this case was infected with C. cuniculus). During the course of the outbreak (24 June – 18 July 2008, the dates of symptom onset in the first and last case, respectively), 23 cases of cryptosporidiosis were confirmed as being infected with C. cuniculus; one of the 23 was a secondary case. The HPA Real-time Syndromic Surveillance Team provided data in order to aid the response to this incident and the first syndromic surveillance report was circulated to the incident management team and other relevant people in the HPA on 27 June 2008. Data from the HPA/NHS Direct and HPA/QSurveillance systems were provided in a series of regular reports, initially daily and eventually weekly, until the final report on 21 August 2008. Each report included a summary interpretation and more detailed data on diarrhoea, gastroenteritis and vomiting indicators. Methods Surveillance systems HPA/NHS Direct surveillance system NHS Direct is a 24-hour nurse-led telephone helpline that provides health information and advice to the general public. Nurses use a computerised clinical decision support system – the NHS Clinical Assessment System (NHS CAS) – to handle calls. This assessment system uses approximately 200 computerised symptom-based clinical algorithms. Nurses assign the call to the most appropriate algorithm and the patient’s symptoms determine the questions asked and the action to be taken following the call (call outcome), which could be guidance on self-care or they could be referred to their GP or advised to attend a hospital emergency department. No attempt is made to provide a formal diagnosis. Daily NHS Direct data are received by the Real-time Syndromic Surveillance Team, where the number and type of calls received during the previous day are analysed and interpreted. Call proportions are calculated by age group and algorithm against the total number of calls received. HPA/QSurveillance system The HPA/QSurveillance national surveillance system was set up by the University of Nottingham, United Kingdom, and Egton Medical Information Systems (EMIS), a supplier of general practice computer systems, in collaboration with the HPA. It comprises a network of more than 3,500 general practices throughout the UK, covering more than 22 million patients (about 38% of the population [28]). Aggregated data on GP consultations for a range of indicators are automatically uploaded daily from GP practice systems to a central database. Data are routinely reported on a weekly basis; however, daily reporting is possible for specific incidents. Reports are provided at national or regional level (strategic health authority, SHA) and by local health district (primary care trust, PCT). Analysis of surveillance data NHS Direct call proportions for gastrointestinal syndromes (diarrhoea and vomiting) for the East Midlands region in England, where Northampton is situated, were examined during the outbreak (24 June – 18 July 2008) and compared with those for England and Wales. A series of control charts for diarrhoea calls are routinely used to monitor significant rises in the numbers of calls received. Control charts are calculated by assuming that calls follow a Poisson distribution with the total number of calls as an offset: a model is fitted to each region and symptom separately [29]. The model takes into account call variation caused by weekends, public holidays and the time of year – variables that can affect the number of calls received by NHS Direct. A value above the upper limit of the 99.5% confidence interval would be considered to be unusual. The seven-day moving average for diarrhoea calls was also monitored. The number and percentage of calls for diarrhoea in the East Midlands region were presented by call outcome and the number of calls in the Northampton (NN) postcode districts and in particular the number of calls in the NN11 and NN12 postcode districts, which were most affected by the incident. QSurveillance national consultation rates per 100,000 population for diarrhoea (in the age groups under five years, five years and over, and all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with rates for the same period in 2007 (data not presented). Consultation rates by region for 2008 for diarrhoea (all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with those for the East Midlands region. The gastroenteritis indicator includes all cases of diarrhoea and/or vomiting. Consultation rates and standardised incidence ratios (SIRs) – calculated using the UK as the standard population – for diarrhoea, gastroenteritis and vomiting were compared for the UK, Yorkshire and Humberside, East Midlands, Leicestershire, Northamptonshire and Rutland SHA, and Daventry and South Northants PCT, Northamptonshire Heartlands PCT and Northampton PCT. Yorkshire and Humberside was not an affected region but was included as a control. The area supplied by the Pitsford Reservoir included the three PCTs, which were all within the Leicestershire, Northamptonshire and Rutland SHA. The consultation rates and SIRs were compared for the period from week 16 to week 35 of 2008 in order to compare the rates before and after the Cryptosporidium exceedance, which took place in week 26. Estimates of excess numbers of cases of diarrhoea occurring during and following the Cryptosporidium outbreak were made by calculating the mean consultation rate over a period of five weeks before and after the incident (weeks 20–24 and weeks 31–35, respectively). For each of the three PCTs, the calculated mean rate was applied to the PCT population to estimate the number of cases that would be expected each week. The actual consultation rates for diarrhoea for weeks 25 to 30 were used to estimate the number of cases for the PCT population each week. The expected number of cases was subtracted from the estimated number of cases in the PCT population to give the estimated number of excess cases. Results HPA/NHS Direct surveillance system A peak in the number of calls for diarrhoea in the East Midlands was recorded in 25–26 June 2008, the period that coincided with the contamination incident and the associated media coverage (Figure 1). The neighbouring areas of the West Midlands, Yorkshire and the Humber, and East of England showed no increase in the number of calls for diarrhoea. The peak produced a control chart exceedance for calls for diarrhoea on 25 June 2008 (Figure 2), when the proportion of calls exceeded the upper limit of the 99.5% confidence interval. There were further confidence interval exceedances on 26 and 28 June (which were not control chart exceedances). There was no peak in calls for vomiting or control chart exceedance for these calls in the East Midlands. HPA/QSurveillance national surveillance system The East Midlands region had significantly high consultation rates for diarrhoea and gastroenteritis in week 25 (16–22 June), week 26 (23–29 June 2008, when the contamination incident was reported) and in the following four weeks. Within this region. Leicestershire, Northamptonshire and Rutland SHA had slightly raised consultation rates and significant SIRs across weeks 25 to 30 that were not seen in the neighbouring Trent SHA. At PCT level, all three of the PCTs in the area affected by the incident showed increased consultation rates for diarrhoea (Table 1) and gastroenteritis (Table 2) with SIRs significantly above the UK rate in week 26. Daventry and South Northants PCT also had a raised SIR for both indicators in week 25, and although Northamptonshire Heartlands and Northampton PCTs did not have SIRs significantly above that of the UK in week 25, the rise in consultation rates for diarrhoea and gastroenteritis began during week 25. In Northampton PCT, consultations for both diarrhoea and gastroenteritis peaked in the week following the contamination incident, week 27, returning to normal levels by week 30 (Figure 3A and 3B). A similar effect can be seen in Northampton Heartlands PCT. Daventry and South Northants PCT also showed an increase, but appeared to have consistently higher rates for both indicators. This was the area with the smallest population so the rates were more variable than in the other PCTs and we therefore interpreted these results with caution. The consultation rates for vomiting during weeks 25 to 30 in the East Midlands were not unusual at SHA or PCT level (data not presented). Discussion We have demonstrated the sensitivity of syndromic surveillance in detecting this small Cryptosporidium outbreak and the value of the surveillance in being able to describe the extent of its spread. Both the HPA/NHS Direct and HPA/QSurveillance systems showed demonstrable increases in calls and consultations for diarrhoea that were linked to the outbreak. QSurveillance consultations appeared to increase across the PCTs immediately affected but not in the surrounding area. Both the HPA/NHS Direct and HPA/QSurveillance systems showed a clear signal at the time of the incident and we were able to describe the extent of the impact on pre-primary care and primary care services. The HPA/QSurveillance system showed a rise in consultation rates for gastrointestinal symptoms that began the week before the outbreak, consistent with the period when Cryptosporidium was present in the water leaving the Pitsford Reservoir (19–24 June 2008) and with the onset of symptoms in the first outbreak case on 24 June. Although only 33 cases were identified by the outbreak investigation team, of which 23 were confirmed as having the outbreak Cryptosporidium strain, our syndromic surveillance data detected this limited outbreak. Data also suggested a more widespread increase in general gastrointestinal symptoms around the time of the outbreak, with an estimated 422 excess diarrhoea cases; these excess cases represented an increase of about 25% above normally expected levels. It is highly probable that a proportion of these excess cases may have resulted from the increased publicity surrounding the incident – for example, it is likely that media reports contributed to the large peak in calls detected by the HPA/NHS Direct surveillance system on the day the boil water notice was issued, and could also have impacted on the GP consultation rate. It has been previously shown that reporting of mumps cases is sensitive to media coverage, with a rise in clinically reported cases following newspaper reports [30]. A similar mechanism could account for some of the excess GP consultations as cases experiencing gastrointestinal symptoms may have been more likely to consult their GP, whereas in normal circumstances they would have cared for themselves at home. It is also possible that the surveillance shows outbreak-associated cases that did not come to the attention of the outbreak team, perhaps because symptoms were not sufficiently severe to warrant further investigation, or stool samples were not provided for testing. It is interesting to note that there was no demonstrable impact on the number of calls for vomiting (which is not a prominent clinical feature of cryptosporidiosis). Other common community-based pathogens such as norovirus and rotavirus were at low levels, as is normal for that time of year [31]. In this instance, public health authorities had already been alerted to a potential problem by the water company, although the extent of the outbreak was detected by syndromic surveillance. In 2003 the syndromic surveillance systems in the city of New York, United States, were able to detect an increase in diarrhoeal illness following a power outage when there was no other indication of citywide illness [32]. The New York system covers a population of nine million, but does not regularly detect localised outbreaks [33]. It has been shown previously that the HPA/NHS Direct surveillance system would be unlikely to detect a Cryptosporidium outbreak unless call volumes are high (72% chance of detection if nine-tenths of cases called NHS Direct) [29], although the value of syndromic surveillance for such outbreaks has been recognised [34]. The system detected the East Midlands Cryptosporidium outbreak that affected a smaller population than that covered by the New York system. The three PCTs affected have a combined population of around 600,000, of which just over half use GP practices reporting to QSurveillance, yet this syndromic surveillance system was able to describe an increase in consultation rates for diarrhoea and gastroenteritis around the time of the outbreak. Limitations of the data There was extensive media reporting of the incident that may have affected both the HPA/NHS Direct and HPA/QSurveillance systems and contributed to the increase in reported gastrointestinal symptoms around the time of the contamination incident. However, the rise in consultation rates for diarrhoea began before the outbreak had been detected and therefore cannot be attributed to media coverage. The HPA/NHS Direct and HPA/QSurveillance systems monitor general symptoms and so could only monitor the relevant symptoms of diarrhoea and vomiting. They are not able to detect Cryptosporidium cases, as this would require laboratory confirmation of diagnosis, so some of the estimated excess cases could be unconnected with this incident. This outbreak was discovered by other means but both the HPA/NHS Direct and HPA/ QSurveillance systems were able to describe the extent of the disease in the general population and provide reassurance that there was no widespread impact. Compared with other populations, older people and ethnic minorities are less likely to call NHS Direct [29], and although this should not prevent detection of gastrointestinal symptoms as a result of drinking water contamination as this would affect the whole population, this may reduce the signal from the system [35]. With such large surveillance systems, there will be ‘background noise’ in the data, so procedures must be in place to correctly interpret the data and set appropriate thresholds for action. Conclusion To our knowledge, this is the first time that PCT-level data from a syndromic surveillance system, the HPA/ QSurveillance national surveillance system, have been able to show the extent of such a limited outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide PCT-level data and this Cryptosporidium contamination incident demonstrates the potential usefulness of this system.

How was the infrastructure affected?

1529

Value of syndromic surveillance in monitoring a focal waterborne outbreak due to an unusual Cryptosporidium genotype in Northamptonshire

The United Kingdom (UK) has several national syndromic surveillance systems. The Health Protection Agency (HPA)/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data from a national telephone helpline, while the HPA/ QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems. Data from both of these systems were used to monitor a local outbreak of cryptosporidiosis that occurred following Cryptosporidium oocyst contamination of drinking water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, in June 2008. There was a peak in the number of calls to NHS Direct concerning diarrhoea that coincided with the incident. QSurveillance data for the local areas affected by the outbreak showed a significant increase in GP consultations for diarrhoea and gastroenteritis in the week of the incident but there was no increase in consultations for vomiting. A total of 33 clinical cases of cryptosporidiosis were identified in the outbreak investigation, of which 23 were confirmed as infected with the outbreak strain. However, QSurveillance data suggest that there were an estimated 422 excess diarrhoea cases during the outbreak, an increase of about 25% over baseline weekly levels. To our knowledge, this is the first time that data from a syndromic surveillance system, the HPA/QSurveillance national surveillance system, have been able to show the extent of such a small outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide data at local health district (primary care trust) level. The Cryptosporidium contamination incident described demonstrates the potential usefulness of this information, as it is unusual for syndromic surveillance systems to be able to help monitor such a small-scale outbreak. Introduction As syndromic surveillance systems usually capture data already collected for other purposes, and monitor generic symptoms and/or clinically diagnosed disease, they provide information at an earlier stage of illness (compared with laboratory-confirmed diagnoses), so that action can be taken in time to substantially reduce the impact of disease. Some systems, for example, the Royal College of General Practitioners Weekly Returns Service, are now well established, with many years of historical data that allow monitoring of longer-term disease trends [1]. They have the ability to provide early warning of, for example, seasonal rises in influenza and can trigger public health action, such as a recommendation to prescribe antiviral drugs in line with national guidance [2-4]. They can also provide reassurance to incident response teams and the general public that an incident has not caused adverse health effects – for example, following an explosion at the Buncefield oil storage depot in Hemel Hempstead, United Kingdom (UK), in 2005, syndromic surveillance confirmed that there were no unusual rises in community-based morbidity linked to the incident [5]; following the eruption of the Eyjafjallajökull volcano in Iceland in April 2010 similar assurance was given about lack of impact on community morbidity [6]. Health departments are increasingly expected to monitor health effects of natural events such as heat wave or flooding, or implement surveillance – of which syndromic surveillance plays a major role – for mass gatherings such as the Olympics or football World Cup [7-9]. Systems in France, Australia and Taiwan use data from emergency departments [10-12], a Canadian system uses over-the-counter pharmacy sales [13,14], and in the Netherlands data from both syndromic and surrogate data sources, such as employee absence records and prescription medications dispensed by pharmacies, are included in surveillance systems [15,16]. Currently systems based on Internet searches via search engines or on queries submitted to medical websites are being developed [17,18]. In the UK, the HPA/NHS Direct syndromic surveillance system uses pre-diagnostic syndromic data collected from the NHS Direct telephone helpline [19], while the HPA/QSurveillance national surveillance system uses clinical diagnosis data extracted from general practitioner (GP)-based clinical information systems [20]. The HPA Real-time Syndromic Surveillance Team is a small team that coordinates a number of syndromic surveillance systems within the HPA and takes a lead for syndromic surveillance in England [21]. This paper describes the support provided by the team to the local incident management team during a local cryptosporidiosis outbreak and shows the use of syndromic surveillance in monitoring the extent of an outbreak using the HPA/NHS Direct and HPA/QSurveillance national surveillance systems. Cryptosporidiosis Cryptosporidium is a protozoan parasite that can cause an infection in people, cattle and sometimes other animals [22]. Cryptosporidiosis is most common in children aged between one and five years, but it can affect all ages. Those with impaired immune systems are likely to be most seriously affected. Symptoms usually appear between three and 12 days after initial exposure and include watery diarrhoea, stomach pains, dehydration and fever. In its transmissible form, called an oocyst, the parasite is protected by an outer shell, which allows it to survive in the environment for a long time. Transmission occurs most often via the faeco-oral route through person-to-person or animalto-person contact, but people may also be infected by consuming contaminated water or food or by swimming in contaminated water. Although uncommon, the largest outbreaks have occurred following contamination of drinking water [23,24]. Normal chlorine disinfection procedures do not kill the oocysts, so they are removed by filtration and water companies carry out routine monitoring of treated water. Description of the incident On 25 June 2008 the local Health Protection Unit was informed by Anglian Water of an exceedence in the level of Cryptosporidium oocysts found in water supplied from the Pitsford Reservoir in Northamptonshire, United Kingdom, during 19 to 24 June 2008 [25]. The reservoir supplied a population of more than 250,000 in the Northampton area. A notice advising people in the affected areas to boil all drinking water was issued on 25 June 2008 and public health messages were circulated to local health services and to the general public via the media. Those members of the public who were concerned about health risks associated with the incident were asked to ring NHS Direct for clinical advice [26]. The HPA wrote to local GPs and hospitals asking them to monitor potential patients for signs and symptoms of Cryptosporidium infection and to submit faecal specimens to the local hospital diagnostic laboratory if patients presented with diarrhoea. Samples from 34 patients where Cryptosporidium infection was identified were sent to the UK Cryptosporidium reference unit for typing. On 30 June 2008, the Cryptosporidium oocysts found in the reservoir water were confirmed as being of the rabbit genotype Cryptosporidium cuniculus [27]. Subsequently, a dead rabbit was found in a treated water tank at the water treatment works. The genotype of Cryptosporidium oocysts in the rabbit’s large bowel was indistinguishable from that of the oocysts found in the water [27]. After remediation of the water supply and distribution, the ‘boil water notice’ was lifted on 4 July and the following day the first case of cryptosporidiosis linked to the incident was identified by the reference laboratory (this case was infected with C. cuniculus). During the course of the outbreak (24 June – 18 July 2008, the dates of symptom onset in the first and last case, respectively), 23 cases of cryptosporidiosis were confirmed as being infected with C. cuniculus; one of the 23 was a secondary case. The HPA Real-time Syndromic Surveillance Team provided data in order to aid the response to this incident and the first syndromic surveillance report was circulated to the incident management team and other relevant people in the HPA on 27 June 2008. Data from the HPA/NHS Direct and HPA/QSurveillance systems were provided in a series of regular reports, initially daily and eventually weekly, until the final report on 21 August 2008. Each report included a summary interpretation and more detailed data on diarrhoea, gastroenteritis and vomiting indicators. Methods Surveillance systems HPA/NHS Direct surveillance system NHS Direct is a 24-hour nurse-led telephone helpline that provides health information and advice to the general public. Nurses use a computerised clinical decision support system – the NHS Clinical Assessment System (NHS CAS) – to handle calls. This assessment system uses approximately 200 computerised symptom-based clinical algorithms. Nurses assign the call to the most appropriate algorithm and the patient’s symptoms determine the questions asked and the action to be taken following the call (call outcome), which could be guidance on self-care or they could be referred to their GP or advised to attend a hospital emergency department. No attempt is made to provide a formal diagnosis. Daily NHS Direct data are received by the Real-time Syndromic Surveillance Team, where the number and type of calls received during the previous day are analysed and interpreted. Call proportions are calculated by age group and algorithm against the total number of calls received. HPA/QSurveillance system The HPA/QSurveillance national surveillance system was set up by the University of Nottingham, United Kingdom, and Egton Medical Information Systems (EMIS), a supplier of general practice computer systems, in collaboration with the HPA. It comprises a network of more than 3,500 general practices throughout the UK, covering more than 22 million patients (about 38% of the population [28]). Aggregated data on GP consultations for a range of indicators are automatically uploaded daily from GP practice systems to a central database. Data are routinely reported on a weekly basis; however, daily reporting is possible for specific incidents. Reports are provided at national or regional level (strategic health authority, SHA) and by local health district (primary care trust, PCT). Analysis of surveillance data NHS Direct call proportions for gastrointestinal syndromes (diarrhoea and vomiting) for the East Midlands region in England, where Northampton is situated, were examined during the outbreak (24 June – 18 July 2008) and compared with those for England and Wales. A series of control charts for diarrhoea calls are routinely used to monitor significant rises in the numbers of calls received. Control charts are calculated by assuming that calls follow a Poisson distribution with the total number of calls as an offset: a model is fitted to each region and symptom separately [29]. The model takes into account call variation caused by weekends, public holidays and the time of year – variables that can affect the number of calls received by NHS Direct. A value above the upper limit of the 99.5% confidence interval would be considered to be unusual. The seven-day moving average for diarrhoea calls was also monitored. The number and percentage of calls for diarrhoea in the East Midlands region were presented by call outcome and the number of calls in the Northampton (NN) postcode districts and in particular the number of calls in the NN11 and NN12 postcode districts, which were most affected by the incident. QSurveillance national consultation rates per 100,000 population for diarrhoea (in the age groups under five years, five years and over, and all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with rates for the same period in 2007 (data not presented). Consultation rates by region for 2008 for diarrhoea (all ages), gastroenteritis (all ages) and vomiting (all ages) were compared with those for the East Midlands region. The gastroenteritis indicator includes all cases of diarrhoea and/or vomiting. Consultation rates and standardised incidence ratios (SIRs) – calculated using the UK as the standard population – for diarrhoea, gastroenteritis and vomiting were compared for the UK, Yorkshire and Humberside, East Midlands, Leicestershire, Northamptonshire and Rutland SHA, and Daventry and South Northants PCT, Northamptonshire Heartlands PCT and Northampton PCT. Yorkshire and Humberside was not an affected region but was included as a control. The area supplied by the Pitsford Reservoir included the three PCTs, which were all within the Leicestershire, Northamptonshire and Rutland SHA. The consultation rates and SIRs were compared for the period from week 16 to week 35 of 2008 in order to compare the rates before and after the Cryptosporidium exceedance, which took place in week 26. Estimates of excess numbers of cases of diarrhoea occurring during and following the Cryptosporidium outbreak were made by calculating the mean consultation rate over a period of five weeks before and after the incident (weeks 20–24 and weeks 31–35, respectively). For each of the three PCTs, the calculated mean rate was applied to the PCT population to estimate the number of cases that would be expected each week. The actual consultation rates for diarrhoea for weeks 25 to 30 were used to estimate the number of cases for the PCT population each week. The expected number of cases was subtracted from the estimated number of cases in the PCT population to give the estimated number of excess cases. Results HPA/NHS Direct surveillance system A peak in the number of calls for diarrhoea in the East Midlands was recorded in 25–26 June 2008, the period that coincided with the contamination incident and the associated media coverage (Figure 1). The neighbouring areas of the West Midlands, Yorkshire and the Humber, and East of England showed no increase in the number of calls for diarrhoea. The peak produced a control chart exceedance for calls for diarrhoea on 25 June 2008 (Figure 2), when the proportion of calls exceeded the upper limit of the 99.5% confidence interval. There were further confidence interval exceedances on 26 and 28 June (which were not control chart exceedances). There was no peak in calls for vomiting or control chart exceedance for these calls in the East Midlands. HPA/QSurveillance national surveillance system The East Midlands region had significantly high consultation rates for diarrhoea and gastroenteritis in week 25 (16–22 June), week 26 (23–29 June 2008, when the contamination incident was reported) and in the following four weeks. Within this region. Leicestershire, Northamptonshire and Rutland SHA had slightly raised consultation rates and significant SIRs across weeks 25 to 30 that were not seen in the neighbouring Trent SHA. At PCT level, all three of the PCTs in the area affected by the incident showed increased consultation rates for diarrhoea (Table 1) and gastroenteritis (Table 2) with SIRs significantly above the UK rate in week 26. Daventry and South Northants PCT also had a raised SIR for both indicators in week 25, and although Northamptonshire Heartlands and Northampton PCTs did not have SIRs significantly above that of the UK in week 25, the rise in consultation rates for diarrhoea and gastroenteritis began during week 25. In Northampton PCT, consultations for both diarrhoea and gastroenteritis peaked in the week following the contamination incident, week 27, returning to normal levels by week 30 (Figure 3A and 3B). A similar effect can be seen in Northampton Heartlands PCT. Daventry and South Northants PCT also showed an increase, but appeared to have consistently higher rates for both indicators. This was the area with the smallest population so the rates were more variable than in the other PCTs and we therefore interpreted these results with caution. The consultation rates for vomiting during weeks 25 to 30 in the East Midlands were not unusual at SHA or PCT level (data not presented). Discussion We have demonstrated the sensitivity of syndromic surveillance in detecting this small Cryptosporidium outbreak and the value of the surveillance in being able to describe the extent of its spread. Both the HPA/NHS Direct and HPA/QSurveillance systems showed demonstrable increases in calls and consultations for diarrhoea that were linked to the outbreak. QSurveillance consultations appeared to increase across the PCTs immediately affected but not in the surrounding area. Both the HPA/NHS Direct and HPA/QSurveillance systems showed a clear signal at the time of the incident and we were able to describe the extent of the impact on pre-primary care and primary care services. The HPA/QSurveillance system showed a rise in consultation rates for gastrointestinal symptoms that began the week before the outbreak, consistent with the period when Cryptosporidium was present in the water leaving the Pitsford Reservoir (19–24 June 2008) and with the onset of symptoms in the first outbreak case on 24 June. Although only 33 cases were identified by the outbreak investigation team, of which 23 were confirmed as having the outbreak Cryptosporidium strain, our syndromic surveillance data detected this limited outbreak. Data also suggested a more widespread increase in general gastrointestinal symptoms around the time of the outbreak, with an estimated 422 excess diarrhoea cases; these excess cases represented an increase of about 25% above normally expected levels. It is highly probable that a proportion of these excess cases may have resulted from the increased publicity surrounding the incident – for example, it is likely that media reports contributed to the large peak in calls detected by the HPA/NHS Direct surveillance system on the day the boil water notice was issued, and could also have impacted on the GP consultation rate. It has been previously shown that reporting of mumps cases is sensitive to media coverage, with a rise in clinically reported cases following newspaper reports [30]. A similar mechanism could account for some of the excess GP consultations as cases experiencing gastrointestinal symptoms may have been more likely to consult their GP, whereas in normal circumstances they would have cared for themselves at home. It is also possible that the surveillance shows outbreak-associated cases that did not come to the attention of the outbreak team, perhaps because symptoms were not sufficiently severe to warrant further investigation, or stool samples were not provided for testing. It is interesting to note that there was no demonstrable impact on the number of calls for vomiting (which is not a prominent clinical feature of cryptosporidiosis). Other common community-based pathogens such as norovirus and rotavirus were at low levels, as is normal for that time of year [31]. In this instance, public health authorities had already been alerted to a potential problem by the water company, although the extent of the outbreak was detected by syndromic surveillance. In 2003 the syndromic surveillance systems in the city of New York, United States, were able to detect an increase in diarrhoeal illness following a power outage when there was no other indication of citywide illness [32]. The New York system covers a population of nine million, but does not regularly detect localised outbreaks [33]. It has been shown previously that the HPA/NHS Direct surveillance system would be unlikely to detect a Cryptosporidium outbreak unless call volumes are high (72% chance of detection if nine-tenths of cases called NHS Direct) [29], although the value of syndromic surveillance for such outbreaks has been recognised [34]. The system detected the East Midlands Cryptosporidium outbreak that affected a smaller population than that covered by the New York system. The three PCTs affected have a combined population of around 600,000, of which just over half use GP practices reporting to QSurveillance, yet this syndromic surveillance system was able to describe an increase in consultation rates for diarrhoea and gastroenteritis around the time of the outbreak. Limitations of the data There was extensive media reporting of the incident that may have affected both the HPA/NHS Direct and HPA/QSurveillance systems and contributed to the increase in reported gastrointestinal symptoms around the time of the contamination incident. However, the rise in consultation rates for diarrhoea began before the outbreak had been detected and therefore cannot be attributed to media coverage. The HPA/NHS Direct and HPA/QSurveillance systems monitor general symptoms and so could only monitor the relevant symptoms of diarrhoea and vomiting. They are not able to detect Cryptosporidium cases, as this would require laboratory confirmation of diagnosis, so some of the estimated excess cases could be unconnected with this incident. This outbreak was discovered by other means but both the HPA/NHS Direct and HPA/ QSurveillance systems were able to describe the extent of the disease in the general population and provide reassurance that there was no widespread impact. Compared with other populations, older people and ethnic minorities are less likely to call NHS Direct [29], and although this should not prevent detection of gastrointestinal symptoms as a result of drinking water contamination as this would affect the whole population, this may reduce the signal from the system [35]. With such large surveillance systems, there will be ‘background noise’ in the data, so procedures must be in place to correctly interpret the data and set appropriate thresholds for action. Conclusion To our knowledge, this is the first time that PCT-level data from a syndromic surveillance system, the HPA/ QSurveillance national surveillance system, have been able to show the extent of such a limited outbreak at a local level. QSurveillance, which covers about 38% of the UK population, is currently the only GP database that is able to provide PCT-level data and this Cryptosporidium contamination incident demonstrates the potential usefulness of this system.

What were the infrastructure complaints?

1530

Waterborne Norovirus outbreak at a seaside resort likely originating from municipal water distribution system failure

Abstract In May 2016 a Norovirus (NoV) gastroenteritis outbreak involved a high school class visiting a seaside resort near Taormina (Mascali, Sicily). Twenty-four students and a teacher were affected and 17 of them showed symptoms on the second day of the journey, while the others got ill within the following 2 days. Symptoms included vomiting, diarrhoea and fever, and 12 students required hospitalisation. Stool samples tested positive for NoV genome by Real-Time polymerase chain reaction assay in all 25 symptomatic subjects. The GII.P2/GII.2 NoV genotype was linked to the outbreak by ORF1/ORF2 sequence analysis. The epidemiological features of the outbreak were consistent with food/waterborne followed by person-to-person and/ or vomit transmission. Food consumed at a shared lunch on the first day of the trip was associated to illness and drinking un-bottled tap water was also considered as a risk factor. The analysis of water samples revealed the presence of bacterial indicators of faecal contamination in the water used in the resort as well as in other areas of the municipal water network, linking the NoV gastroenteritis outbreak to tap water pollution from sewage leakage. From a single water sample, an amplicon whose sequence corresponded to the capsid genotype recovered from patients could be obtained. Introduction Noroviruses (NoVs) are recognised as a leading cause of sporadic cases and outbreaks of acute gastroenteritis (AGE) across all age groups [1, 2]. They are transmitted primarily through the faecal-oral route either by direct person–person spread and with consumption of contaminated food or water [1, 3, 4]. NoVs are non-enveloped RNA viruses belonging to the Caliciviridae family and are classified into seven distinct genogroups (GI-GVII) on the basis of the major capsid protein VP1 sequence. GI, GII and GIV NoVs infect humans and they have been divided into more than 30 genotypes based on the sequences of polymerase (ORF1) and capsid proteins (ORF2) [5]. NoV GII.4 genotype has been associated with the vast majority of NoV related cases of AGE worldwide although many genotypes, belonging to both GI and GII genogroups are involved in foodborne and waterborne outbreaks [6, 7]. NoVs are highly contagious owing to their high stability in the environment and their ability to infect at low doses. In fact, just 10–100 virions of NoV may be sufficient to cause AGE [8, 9]. NoVs are considered the major cause of water-related disease in semi-closed institutions, including restaurants, schools, hospitals, nursing homes and cruise ships [10–12]. To confirm NoV as a cause of outbreaks, the presence of four epidemiological features of disease, vomiting in more than half of affected persons, a mean incubation period of 24/48 h, a mean duration of illness of 12–60 h and the absence of bacterial pathogens in stool culture, should be evaluated [13]. Waterborne NoV outbreaks can be caused by contaminated drinking water despite correct chlorine concentration and in the absence of coliform bacteria [10, 14–17]. The aim of this study was to investigate an outbreak of NoV gastroenteritis associated with drinking water involving a high school class travelling from Palermo (West Sicily) to Acireale (East Sicily) on an educational tour in May 2016. Stool samples were collected soon after the onset of AGE symptoms to ascertain the aetiology of illness. The source of infection was investigated collecting epidemiological data through an online questionnaire and, in order to explore a possible transmission route, laboratory tests on water from the municipal network were performed. Material and methods Epidemiological investigation From 20 to 22 May 2016 an AGE outbreak occurred in Mascali, a small town near Taormina, involving a school class of 28 students, 13 girls and 15 boys aged between 14 and 18 years and a 54-year-old female teacher, during a trip to participate in a theatre festival in Acireale. A total of 25/29 persons were affected by nausea, vomiting, abdominal cramps, diarrhoea and fever while they were staying at a seaside resort in Mascali, on the way to Acireale. To obtain data on symptoms and the correlated risk factors all participants filled an online questionnaire investigating personal data (age and sex), travel and accommodation details and food consumption over 48 h before symptoms presentation, time of onset and duration of symptoms, therapy and/or hospitalisation required and history of contact with people affected by diarrhoea and/or vomiting (Table 1). The severity of illness was analysed by the Vesikari severity scale allowing to distinguish non-severe (score <11) and severe (score ⩾11) AGE (Table 1). Vesikari score was calculated from maximum number of episodes/day and duration of both diarrhoea and vomiting (score 0–3 for each of the 4 parameters), temperature (score 0–3), level of dehydration (score 0–3) and treatment required (score 0–2) [18]. A case-control study was conducted to establish if food consumed before the onset of symptoms was associated with the disease. A case was defined as a student attending the tour and reporting diarrhoea and/or vomiting within 36 h from the first shared meal during the tour, while students who did not present any of the previous symptoms or those showing the first symptoms with a delay of ⩾48 h (secondary cases) were used as controls. Variables were studied using odds ratio (OR) and its 95% CI and statistical associations were determined using the χ2 test, with the level of statistical significance set at 0.05 (Table 2). NoV detection in stool samples A stool sample was obtained from each of the 29 travellers, soon after (3–5 days) the onset of the outbreak. No stool samples could be obtained from the food-handlers working at the resort. All samples were tested for NoV at the Enteric Viruses Laboratory of the University of Palermo, member of the Italian Study Group for Enteric Viruses (ISGEV), by a Real-Time polymerase chain reaction (RT-PCR) assay that allows to differentiate GI and GII NoVs [19]. Total RNA was extracted from stool samples using QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) and cDNA was generated by MMLV reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were considered positive for NoVs when cycle threshold (ct) of RT-PCR was ⩽30 and lowpositive when 30<ct⩽35. For all NoV-positive samples two genomic fragments were amplified by conventional RT-PCR using Superscript III One-Step RT-PCR system with Platinum Taq DNA polymerase (Invitrogen) and primers JV12a/JV13b, targeting 326 bps of the RdRp region A [20] and either COG1F/ GISKR or COG2F/GIISKR, amplifying a 315 bps 5′portion of GI or GII VP1 protein [19, 21]. Environmental investigation From May through August 2016, a total of 32 tap water samples were collected from the Mascali municipal network. In particular, seven samples were collected at the resort where the outbreak occurred and from its water entry (R), two in the city centre of Mascali (M), 15 in the village of Fondachello (F), on the seaside, three in Carrabba village (C), far from the coast, and two and three, respectively, from the water well ‘Carlino’ (W) and the source ‘Bufardo and Torregrossa’ (S) which feed the municipal system and provide drinking water to the Mascali area (Fig. 1, Table 3). One litre water samples were sent by the local health authorities to the SIALAB laboratory in Avola, Siracusa and 100 ml aliquots were screened for coliforms, E. coli and enterococci as faecal indicators, by standard bacteriological methods according to UNI EN ISO 9308-1:2014. Unfortunately, although the suspect of a water source for the viral outbreak was quickly communicated, only in August and November 2016, following the implementation of control measures, local authorities provided water samplings of appropriate volume to perform viral tests (10 l). Five water samples of 10 l each were collected from different sampling sites in F and R areas for viral screening (Table 3) and were concentrated according to the Italian National Institute of Health (Istituto Superiore di Sanità, ISS) protocols [22] through an ultra-filtration system (Sartoflow® Slice 200 Benchtop Crossflow System, Sartorius AG, Goettingen, Germany), using appropriate membranes (SG Hydrosart 10 kDa) pretreated with 300 ml of 3% Beef Extract (BE) at pH 7, by recirculating it for 10 min in the system, at a pressure not higher than 1.30 bar. Elution and recovery of viruses attached to the membranes were obtained by recirculating 200 ml of 3% BE at pH 9.5 for 3 h until a 10–12 ml volume of eluate was obtained. The pH of the eluate was then checked with litmus paper and if necessary brought to the value of 7 with HCl. Concentrated water samples were simultaneously tested for NoV in three laboratories: Enteric Viruses Laboratory, University of Palermo; Istituto Zooprofilattico Sperimentale della Sicilia (IZS), Palermo; and Istituto Superiore di Sanità (ISS), Rome. Total RNA was extracted from 500 and 560 µl of a concentrated water sample using BioMèrieux NucliSENS® Magnetic Extraction System (BioMèrieux) and QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) respectively, according to manufacturer’s instructions. Bonded RNA was eluted by 100 µl of Elution Buffer. Amplifications were performed at least in duplicate on whole and 1:10 diluted extracts. A Real-Time RT-PCR assay according to ISO/TS 15216- 2:2013, two in-house Real-Time RT-PCR protocols and the One-Step RT-PCR methods already used for amplifying ORF1 and ORF2 fragments in NoV-positive faecal samples were performed [19–21, 23]. Nucleotide sequences analysis and phylogenetic analysis PCR amplicons were purified by commercial columns (CleanSweep PCR Purification, Thermo Fisher Scientific, Vilnius, Lithuania) and sequenced by Sanger method and all nucleotide sequences were aligned using ClustalW. To identify NoV genotypes, RdRp and VP1 sequences were submitted to the RIVM NoV genotyping tool (http://www.rivm.nl/mpf/norovirus/typingtool). Phylogenetic trees were constructed using MEGA 7 and Neighbour-Joining method (bootstrap 1000 replicates). Sequences from stool and water samples are available at NCBI with the following accession numbers: MG517455 for the major capsid protein obtained from a water sample; MG517456 and MG517461, respectively, for the major capsid protein and RNA-dependent-RNA-polymerase, as representatives of the sequences obtained from the stool samples of those involved in the outbreak. Control measures After outbreak notification, on 21 May, the resort was closed for 2 days. On 25 May, following the first results of the bacteriological analyses on municipal water samples, the use of municipal water for human consumption was temporarily restricted (use of water was permitted only for cleaning purposes) in the Carrabba and Fondachello villages and in the area surrounding the resort. From 25 May the reservoir collecting the water from the Carlino well and its chlorination device were submitted to maintenance work and a new chlorination device was installed on the conduit bringing the water from the Bufardo source. Actually, the Bufardo source had been connected to the municipal network in 2003 but unfortunately a dedicated chlorination station had not been planned at that time. Therefore, unchlorinated water entered the system near Fondachello village on the seaside (Fig. 1), possibly diluting chlorine concentrations in water treated at the Carlino reservoir. Moreover, the exact path of the branch of the water network situated downstream the Bufardo source water connection was not marked in the maps of the municipality (Fig. 1, red line), indicating that this section had never been submitted to maintenance. The area surrounding the unmaintained section of the water pipeline was scattered with unauthorised cottages that were not connected to the sewerage systems. Likely, untreated waste from cesspools was unloaded directly into the soil, generating a risk of contamination of the municipal water system through leakage. No work had been made on the water pipelines before the outbreak but sections of the network had been temporarily closed during the winter season on the seaside (Fondachello). Local health authorities advised the connection of all households in the area to the sewerage system and the replacement of the water system section without maintenance. Health authorities also raised attention on AGE cases notification at all local Emergency Departments and Family Practitioners but there was no further evidence of AGE outbreaks in the months following the intervention. Results Data collected through the online questionnaire compiled 7–10 days after the outbreak revealed that none of the 28 students and the teacher taking part in the school trip showed gastroenteritis symptoms at the beginning of the journey and they had no contact with people who suffered from diarrhoea and/or vomiting in the previous days. Travellers had their first shared meal on 19 May (Day 1), 2016, at lunch on their arrival to the resort in Mascali. The list of foods included in the menu is shown in Table 2. Conversely, the dinner on Day 1 was not shared and the students went to dine in three different fast food restaurants in Acireale. On 20 May (Day 2), 17 students (58.6% of the travellers) presented with vomit and/or diarrhoea, falling ill 26–36 h after lunch on Day 1, strongly suggesting that exposure occurred on the first shared meal. Some students also vomited in the presence of the other travellers on the bus bringing them to Acireale. Five more students were symptomatic in the early hours of the following day (21 May, Day 3, from 0:30 am to 8:00 am), as a continuation of the wave of illnesses of the previous day. The teacher and one more student fell ill at 8:00 pm and 11:00 pm of Day 3, respectively, for a second wave of illnesses occurring >48 h from the first shared meal and <36 h from the first cases, with the last student falling sick on 22 May (Day 4) at 3:00 am (Fig. 2, Table 1). On the whole, the mean incubation period was 33.2 h. Based on the online survey, the severity of illness was evaluated by the Vesikari severity scale, grouping symptomatic travellers into severe (11/25 = 44%) and non-severe (14/25 = 56%) cases (Table 1). Vomiting and fever were the predominant symptoms, affecting 22 out of 25 (88%) symptomatic subjects, while diarrhoea was present in 14 out of 25 (56%). Hospitalisation for rehydration therapy was necessary for 13 students (91% [10/11] of severe and 21% [3/14] of non-severe cases). Emergency room (ER) or family practitioner visits were necessary for 21 students and 13 (62%) of them initiated an antibiotic therapy. Two students indicated that other family members had gastroenteric symptoms more than 24 h after their return home, on Day 3, 4 and 5. All those involved in the outbreak recovered within 4 days, with an average duration of illness of 38.4 h. A case-control study of the risk associated with eating foods at lunch on Day 1 or with attending different restaurants at dinner on Day 1 was attempted. A total of 19 subjects filled the section of the online survey concerning food consumption. Fourteen of the responders were among the 22 cases falling ill within 36 h from the first shared meal (first wave), while the remaining five responders were three travellers with supposed secondary infections (showing symptoms with a delay ⩾48 h; second wave) and two asymptomatic students that were used as controls. No statistically significant associations with any food items served at lunch on Day 1 was identified (P < 0.05; Table 2). However, all food consumed on that occasion was to some extent associated with the risk of disease, the highest OR being obtained with chicken (OR: 20.7143, P: 0.0678; Table 2). Actually, most of the cases ate all foods included in the menu, while one of the two asymptomatic controls did not participate in the meal and the other had a pasta dish only. No statistically significant association was found for attending a particular fast food restaurant at the dinner on Day 1 (OR: 2.5, P: 0.3635). The Real-Time RT-PCR assay allowed to detect GII genogroup NoV genome in all the 25 samples from symptomatic subjects, although two of them showed high Ct values, indicating lower shedding of viral genomes (Table 1). The samples from four asymptomatic students were negative for both NoV genogroups. By sequence analysis, a NoV genotype could be defined for 20/ 25 NoV-positive stool samples. All NoV strains from travellers involved in the outbreaks belonged to the same genotype, GII.P2/GII.2, showing >99% nucleotide identities (nt id) in both fragments (ORF1 and ORF2) analysed. Thirty-two water samples were collected and analysed for bacterial contamination (Table 3). Seven samples were taken from the resort (site R) and the remaining 25 from five other sites (C, F, M, S, W), representative of the entire municipal water network. The majority (10/12) of the water samples collected immediately after the onset of the outbreak, between 21 May and 16 June 2016, at the resort (site R) and at sites F and C, located close to the resort, on the seaside and in the interior, respectively, were positive for coliform bacteria. More than 1000 colony-forming units (CFU)/ 100 ml (up to 2700 CFU/100 ml) were detected in four samples, while their absence is required in 100 ml by the Standard of Water Quality in Italy and seven of 11 samples tested were also Escherichia coli-positive (up to 900 CFU/100 ml). Enterococci were searched in 11 samples and three of them were found positive for this faecal contamination indicator. From 7 July to 16 August 2016, after control measures on the water network were undertaken, 13 water samples were collected for follow-up laboratory examination of the water system at F, R and C sites. All of them were negative for bacterial indicators of faecal contamination. No faecal bacteria were identified in the seven samples collected from M, W and S sites, including the well and the source that fed the water network, at any sampling time, before and after implementing the control measures. The research of NoV genome was conducted on concentrated water samples collected on 1 August from R and F sites (Table 3). NoV genome was detected in a single sample collected from the water entry of the resort involved in the outbreak (site R). Actually, only two out of 26 One-Step RT-PCR repetitions were positive for NoV viral capsid gene. The two ORF2-positive results were obtained at the University of Palermo Laboratory using both 1:10 diluted and undiluted extracted RNA. The sequence analysis of the partial ORF2 portion (289 bp) indicated the presence of the same GII.2 capsid genotype detected in the stool samples of the students involved in the outbreak, showing 98.9% nt id. No ORF1 NoV sequence could be obtained from the NoV-positive water sample. Follow-up analyses for NoV genome, conducted on water samples collected on 20 November both at R and F sites found no positive samples. By sequence alignment, the NoV sequences obtained from Mascali outbreak samples were phylogenetically closely related in both ORF1 and ORF2 (99 and 98.9% of nt id, respectively) to a GII.P2/GII.2 strain responsible for a sporadic paediatric gastroenteritis case occurred in Palermo in January 2016 (PA 36/16, accession numbers MG517464 and MG517460, for ORF1 and ORF2, respectively) and previously detected at the Enteric Viruses Laboratory, University of Palermo. Whereas, three GII.P2/GII.2 Italian strains (PA 280/11, PA 366/11 and PA 3/ 14, accession numbers MG517462, MG517465 and MG517463, for ORF1 and MG517457 to −59, for ORF2, respectively), detected in Palermo in 2011 and 2014, were more distantly associated with the outbreak strain both in RdRp and VP1 (90% to 97.4% nt id and 96.1% to 98.1% nt id, respectively, in ORF1 and ORF2). Discussion Epidemiological features, such as the presence of vomiting in more than half (88%) of symptomatic subjects, the 33.2 h average incubation period and the 38.4 h average duration of illness, suggested a NoV aetiology for the investigated outbreak [13]. The detection of NoV genome in all 25 faecal samples from symptomatic subjects fully complies with the requirement of detection of at least one positive sample out of four tested that is considered to be sufficient to ascertain the aetiology of a NoV outbreak [24]. Moreover, molecular typing indicated that a single viral strain, belonging to genotype GII.P2/GII.2, infected all symptomatic subjects involved in the outbreak. NoVs are highly infectious and can easily spread from person to person but also through contaminated food or water. Therefore, settings such as health care institutions, schools and touristic resorts are mainly affected, since close contact and shared food services facilitate transmission. In Mascali, two waves of transmission were observed over three days, being consistent with food/waterborne followed by person-to-person and/or airborne (vomit) transmission (Fig. 2). The first wave of cases presented on 20 May 2016 after a single shared meal, while the last case was detected on 22 May. Exposure to vomit may have been responsible for some of the secondary cases occurred after the first wave of illness. Vomiting produces aerosols that can be inhaled and swallowed. The extremely low infectious dose of NoV makes the inhalation of vomit aerosols an effective transmission route in settings where there is close contact, like on a bus trip [25]. All food consumed in the only shared lunch on 19 May at the resort was associated with the risk of disease (Table 2), but the association was inconclusive since all cases shared the same dishes included in the lunch menu on Day 1. Although many foods were potentially involved, our attention focused on the tap water. In fact, trivial information from local health authorities reported that local population avoided tap water consumption and relied on bottled water for drinking and cooking. Many small outbreaks of gastroenteritis involving restaurants and family groups had been reported in the same area, especially at the beginning of the summer season, when vacation houses and recreational activities were re-opened after a period of low water demand. Illness caused by NoV is common and if the level of water contamination is low, the number of cases remains limited and water can persist as unrecognised source of infection. If polluted water was served for drinking and used for cooking and washing dishes in the Mascali resort at Day 1 a generalised contamination of food and surfaces could have occurred. The potential role of the foodhandlers as a source of contamination could not be analysed in this study but should also be taken into account. If water contamination has already been present in the area for a long time, infections may have occurred among the local food-handlers. The analysis of water samples taken immediately after the outbreak revealed the presence of bacterial indicators of faecal contamination in the water used in the resort, with the only exception of a sample of filter purified water used both for cooking and as drinking water. When extending the sampling to other areas of the municipal water network, many other sites, mostly on the seaside, showed faecal pollution, which demonstrates a poor protection of the water network from faecal leakage (Fig. 1). NoV contamination of waters generally originates from sewage since these enteric pathogens, as other enteric viruses affecting humans, are mostly species-specific and are shed at high concentrations in the stools of patients. When the analyses on the stools of the students made it clear that the outbreak had been caused by NoV, the viral genome was also searched in appropriate volumes of water samples, but this happened when correction measures on the water network had already been taken. Nevertheless, a single water sample, obtained 2 months after the outbreak (1 August 2016), still contained genome fragments that could be correlated to the same NoV genotype causing the outbreak. Although the finding of NoV genome does not demonstrate that entire infecting viral particles were present in the water, it could indicate that a NoV belonging to the same genotype involved in the outbreak was still circulating in the local population several weeks after the outbreak and it was someway able to penetrate the water network, probably indicating continuing faecal contamination through leakage from sewage systems. NoV is particularly resistant in the environment and also to chlorine, therefore transmission through chlorinated water is not exceptional [9, 12, 26]. Drinking water from municipal supplies fed by lake or groundwater has already been identified in the past as the source of waterborne NoV outbreaks in Italy [27–29] and NoV is now considered one of the pathogens causing the largest number of affected consumers in drinking waterborne outbreaks [12]. NoV can persist in water longer than 15 days [30] and it is resistant to chlorine disinfection when free chlorine levels or pre-treatment processes are inadequate [31, 32]. Moreover, microfiltration filters labelled as certified by NSF Standards 53 or 58 (with a pore size ranging from 0.05 to 5 µm) are not effective in removing viruses and can even decrease chlorine concentration in filtered water [33]. Actually, the owner of the resort involved in our outbreak was encouraged to serve tap water as drinking water at the restaurant relying on the effectiveness of such a filter purifier. Unfortunately, we did not have a tap water sample from the restaurant of the resort taken before correction measures and whose volume was appropriate for virus detection. Therefore, we could not demonstrate the failure to eliminate viral pollution of the microfiltration system. In Mascali, correction measures on the water network consisted in the maintenance of the old chlorination device and in the addition of a new one at a key point located upstream and close to suspected illegal sewage dumping (Fig. 1), but as poor protection of the pipeline from sewage leakage risked to reproduce faecal pollution episodes, renovation of the municipal water system was also advised, as well as systematic control of illegal sewage dumping. The additional chlorination device was apparently the most effective intervention, since right after its activation faecal bacterial pollution indicators disappeared. However, NoV genome was detected in a water sample taken 6 weeks after the last detection of bacterial indicators. Unfortunately, detection of faecal contamination of drinking water is only based on the use of bacterial indicators despite no correlation between levels of enteric bacteria and enteric viruses has been found [34–36]. According to ISGEV enteric virus circulation surveillance data, in 2016 the circulation of GII.P2/GII.2 NoV strains in Palermo, home town of the travellers involved in the outbreak, was the highest ever observed from the starting of local NoV surveillance in 2004. GII.P2/GII.2 was the fourth most common genotype in 2016 (12.2% prevalence), following GII.P4/GII.4 Sydney 2012 (28.4%), GII.P16/GII.2 (18.9%) and GII.Pe/GII.4 Sydney 2012 (17.6%). GII.P2/GII.2 strains were mostly observed starting from July 2016, leaving space for the possibility that their increased circulation was linked to the introduction of the ‘Mascali clone’. However, a couple of GII.P2/GII.2 strains (including PA36/16) genetically close to the ‘Mascali clone’ had been already detected in January 2016 in Palermo and GII.P2/GII.2 NoV had sporadically been observed also in 2011 and 2014. Moreover, the circulation of GII.P2/GII.2 NoVs has been increasing worldwide over the last period [37]. Therefore, the unusually high circulation observed in Palermo might just reflect the global epidemiology of GII.P2/GII.2 strains. Noteworthy, the older GII.P2/GII.2 NoV strains detected in Palermo were closely related in the ORF2 (98.9% nt id) to the GII.P16/GII.2 Nashville strain (KY865307) which is supposed to have generated the recombinant GII.P16/GII.4 clone recently circulating epidemically worldwide [37, 38] and were more distantly associated (95.3% nt id) to the GII.2 capsid sequence (AB662868) of another putative GII.P16/GII.4 ancestor from Japan [39]. The increasing frequency of detection of waterborne NoV outbreaks calls for the introduction of routine monitoring for viruses of drinking waters and for the definition of affordable and reliable indicators for the presence of NoV in drinking waters. The use of ultrafiltration systems as a concentration method in order to recover viruses from environmental waters combined with RT-PCR to detect viral genomes has been demonstrated to grant a recovery rate of >50% of multiple viruses [40] and is at present one of the most sensitive approaches for the detection of NoVs, although the discrimination between viable and nonviable viruses remains a problem. The burden of NoV outbreaks, either food associated or waterborne, is still poorly defined due to the lack of specific diagnostic tools in many clinical laboratories. NoV epidemics should always be suspected when the main symptom is vomiting and the duration of disease is short (<48 h). Waterborne transmission is suggested by the rapid spread of disease and the high attack rate. Care should be taken to perform sensitive diagnostic tests on a sufficient number of stool samples in order to identify NoV as the causative agent of enteritis outbreaks, possibly preferring genome screening over antigen detection. Only the systematic detection of NoV outbreaks can guide in the choice of the safety procedures that could be most appropriate in containing such microbiological hazard.

What happened?

1531

Waterborne Norovirus outbreak at a seaside resort likely originating from municipal water distribution system failure

Abstract In May 2016 a Norovirus (NoV) gastroenteritis outbreak involved a high school class visiting a seaside resort near Taormina (Mascali, Sicily). Twenty-four students and a teacher were affected and 17 of them showed symptoms on the second day of the journey, while the others got ill within the following 2 days. Symptoms included vomiting, diarrhoea and fever, and 12 students required hospitalisation. Stool samples tested positive for NoV genome by Real-Time polymerase chain reaction assay in all 25 symptomatic subjects. The GII.P2/GII.2 NoV genotype was linked to the outbreak by ORF1/ORF2 sequence analysis. The epidemiological features of the outbreak were consistent with food/waterborne followed by person-to-person and/ or vomit transmission. Food consumed at a shared lunch on the first day of the trip was associated to illness and drinking un-bottled tap water was also considered as a risk factor. The analysis of water samples revealed the presence of bacterial indicators of faecal contamination in the water used in the resort as well as in other areas of the municipal water network, linking the NoV gastroenteritis outbreak to tap water pollution from sewage leakage. From a single water sample, an amplicon whose sequence corresponded to the capsid genotype recovered from patients could be obtained. Introduction Noroviruses (NoVs) are recognised as a leading cause of sporadic cases and outbreaks of acute gastroenteritis (AGE) across all age groups [1, 2]. They are transmitted primarily through the faecal-oral route either by direct person–person spread and with consumption of contaminated food or water [1, 3, 4]. NoVs are non-enveloped RNA viruses belonging to the Caliciviridae family and are classified into seven distinct genogroups (GI-GVII) on the basis of the major capsid protein VP1 sequence. GI, GII and GIV NoVs infect humans and they have been divided into more than 30 genotypes based on the sequences of polymerase (ORF1) and capsid proteins (ORF2) [5]. NoV GII.4 genotype has been associated with the vast majority of NoV related cases of AGE worldwide although many genotypes, belonging to both GI and GII genogroups are involved in foodborne and waterborne outbreaks [6, 7]. NoVs are highly contagious owing to their high stability in the environment and their ability to infect at low doses. In fact, just 10–100 virions of NoV may be sufficient to cause AGE [8, 9]. NoVs are considered the major cause of water-related disease in semi-closed institutions, including restaurants, schools, hospitals, nursing homes and cruise ships [10–12]. To confirm NoV as a cause of outbreaks, the presence of four epidemiological features of disease, vomiting in more than half of affected persons, a mean incubation period of 24/48 h, a mean duration of illness of 12–60 h and the absence of bacterial pathogens in stool culture, should be evaluated [13]. Waterborne NoV outbreaks can be caused by contaminated drinking water despite correct chlorine concentration and in the absence of coliform bacteria [10, 14–17]. The aim of this study was to investigate an outbreak of NoV gastroenteritis associated with drinking water involving a high school class travelling from Palermo (West Sicily) to Acireale (East Sicily) on an educational tour in May 2016. Stool samples were collected soon after the onset of AGE symptoms to ascertain the aetiology of illness. The source of infection was investigated collecting epidemiological data through an online questionnaire and, in order to explore a possible transmission route, laboratory tests on water from the municipal network were performed. Material and methods Epidemiological investigation From 20 to 22 May 2016 an AGE outbreak occurred in Mascali, a small town near Taormina, involving a school class of 28 students, 13 girls and 15 boys aged between 14 and 18 years and a 54-year-old female teacher, during a trip to participate in a theatre festival in Acireale. A total of 25/29 persons were affected by nausea, vomiting, abdominal cramps, diarrhoea and fever while they were staying at a seaside resort in Mascali, on the way to Acireale. To obtain data on symptoms and the correlated risk factors all participants filled an online questionnaire investigating personal data (age and sex), travel and accommodation details and food consumption over 48 h before symptoms presentation, time of onset and duration of symptoms, therapy and/or hospitalisation required and history of contact with people affected by diarrhoea and/or vomiting (Table 1). The severity of illness was analysed by the Vesikari severity scale allowing to distinguish non-severe (score <11) and severe (score ⩾11) AGE (Table 1). Vesikari score was calculated from maximum number of episodes/day and duration of both diarrhoea and vomiting (score 0–3 for each of the 4 parameters), temperature (score 0–3), level of dehydration (score 0–3) and treatment required (score 0–2) [18]. A case-control study was conducted to establish if food consumed before the onset of symptoms was associated with the disease. A case was defined as a student attending the tour and reporting diarrhoea and/or vomiting within 36 h from the first shared meal during the tour, while students who did not present any of the previous symptoms or those showing the first symptoms with a delay of ⩾48 h (secondary cases) were used as controls. Variables were studied using odds ratio (OR) and its 95% CI and statistical associations were determined using the χ2 test, with the level of statistical significance set at 0.05 (Table 2). NoV detection in stool samples A stool sample was obtained from each of the 29 travellers, soon after (3–5 days) the onset of the outbreak. No stool samples could be obtained from the food-handlers working at the resort. All samples were tested for NoV at the Enteric Viruses Laboratory of the University of Palermo, member of the Italian Study Group for Enteric Viruses (ISGEV), by a Real-Time polymerase chain reaction (RT-PCR) assay that allows to differentiate GI and GII NoVs [19]. Total RNA was extracted from stool samples using QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) and cDNA was generated by MMLV reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were considered positive for NoVs when cycle threshold (ct) of RT-PCR was ⩽30 and lowpositive when 30<ct⩽35. For all NoV-positive samples two genomic fragments were amplified by conventional RT-PCR using Superscript III One-Step RT-PCR system with Platinum Taq DNA polymerase (Invitrogen) and primers JV12a/JV13b, targeting 326 bps of the RdRp region A [20] and either COG1F/ GISKR or COG2F/GIISKR, amplifying a 315 bps 5′portion of GI or GII VP1 protein [19, 21]. Environmental investigation From May through August 2016, a total of 32 tap water samples were collected from the Mascali municipal network. In particular, seven samples were collected at the resort where the outbreak occurred and from its water entry (R), two in the city centre of Mascali (M), 15 in the village of Fondachello (F), on the seaside, three in Carrabba village (C), far from the coast, and two and three, respectively, from the water well ‘Carlino’ (W) and the source ‘Bufardo and Torregrossa’ (S) which feed the municipal system and provide drinking water to the Mascali area (Fig. 1, Table 3). One litre water samples were sent by the local health authorities to the SIALAB laboratory in Avola, Siracusa and 100 ml aliquots were screened for coliforms, E. coli and enterococci as faecal indicators, by standard bacteriological methods according to UNI EN ISO 9308-1:2014. Unfortunately, although the suspect of a water source for the viral outbreak was quickly communicated, only in August and November 2016, following the implementation of control measures, local authorities provided water samplings of appropriate volume to perform viral tests (10 l). Five water samples of 10 l each were collected from different sampling sites in F and R areas for viral screening (Table 3) and were concentrated according to the Italian National Institute of Health (Istituto Superiore di Sanità, ISS) protocols [22] through an ultra-filtration system (Sartoflow® Slice 200 Benchtop Crossflow System, Sartorius AG, Goettingen, Germany), using appropriate membranes (SG Hydrosart 10 kDa) pretreated with 300 ml of 3% Beef Extract (BE) at pH 7, by recirculating it for 10 min in the system, at a pressure not higher than 1.30 bar. Elution and recovery of viruses attached to the membranes were obtained by recirculating 200 ml of 3% BE at pH 9.5 for 3 h until a 10–12 ml volume of eluate was obtained. The pH of the eluate was then checked with litmus paper and if necessary brought to the value of 7 with HCl. Concentrated water samples were simultaneously tested for NoV in three laboratories: Enteric Viruses Laboratory, University of Palermo; Istituto Zooprofilattico Sperimentale della Sicilia (IZS), Palermo; and Istituto Superiore di Sanità (ISS), Rome. Total RNA was extracted from 500 and 560 µl of a concentrated water sample using BioMèrieux NucliSENS® Magnetic Extraction System (BioMèrieux) and QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) respectively, according to manufacturer’s instructions. Bonded RNA was eluted by 100 µl of Elution Buffer. Amplifications were performed at least in duplicate on whole and 1:10 diluted extracts. A Real-Time RT-PCR assay according to ISO/TS 15216- 2:2013, two in-house Real-Time RT-PCR protocols and the One-Step RT-PCR methods already used for amplifying ORF1 and ORF2 fragments in NoV-positive faecal samples were performed [19–21, 23]. Nucleotide sequences analysis and phylogenetic analysis PCR amplicons were purified by commercial columns (CleanSweep PCR Purification, Thermo Fisher Scientific, Vilnius, Lithuania) and sequenced by Sanger method and all nucleotide sequences were aligned using ClustalW. To identify NoV genotypes, RdRp and VP1 sequences were submitted to the RIVM NoV genotyping tool (http://www.rivm.nl/mpf/norovirus/typingtool). Phylogenetic trees were constructed using MEGA 7 and Neighbour-Joining method (bootstrap 1000 replicates). Sequences from stool and water samples are available at NCBI with the following accession numbers: MG517455 for the major capsid protein obtained from a water sample; MG517456 and MG517461, respectively, for the major capsid protein and RNA-dependent-RNA-polymerase, as representatives of the sequences obtained from the stool samples of those involved in the outbreak. Control measures After outbreak notification, on 21 May, the resort was closed for 2 days. On 25 May, following the first results of the bacteriological analyses on municipal water samples, the use of municipal water for human consumption was temporarily restricted (use of water was permitted only for cleaning purposes) in the Carrabba and Fondachello villages and in the area surrounding the resort. From 25 May the reservoir collecting the water from the Carlino well and its chlorination device were submitted to maintenance work and a new chlorination device was installed on the conduit bringing the water from the Bufardo source. Actually, the Bufardo source had been connected to the municipal network in 2003 but unfortunately a dedicated chlorination station had not been planned at that time. Therefore, unchlorinated water entered the system near Fondachello village on the seaside (Fig. 1), possibly diluting chlorine concentrations in water treated at the Carlino reservoir. Moreover, the exact path of the branch of the water network situated downstream the Bufardo source water connection was not marked in the maps of the municipality (Fig. 1, red line), indicating that this section had never been submitted to maintenance. The area surrounding the unmaintained section of the water pipeline was scattered with unauthorised cottages that were not connected to the sewerage systems. Likely, untreated waste from cesspools was unloaded directly into the soil, generating a risk of contamination of the municipal water system through leakage. No work had been made on the water pipelines before the outbreak but sections of the network had been temporarily closed during the winter season on the seaside (Fondachello). Local health authorities advised the connection of all households in the area to the sewerage system and the replacement of the water system section without maintenance. Health authorities also raised attention on AGE cases notification at all local Emergency Departments and Family Practitioners but there was no further evidence of AGE outbreaks in the months following the intervention. Results Data collected through the online questionnaire compiled 7–10 days after the outbreak revealed that none of the 28 students and the teacher taking part in the school trip showed gastroenteritis symptoms at the beginning of the journey and they had no contact with people who suffered from diarrhoea and/or vomiting in the previous days. Travellers had their first shared meal on 19 May (Day 1), 2016, at lunch on their arrival to the resort in Mascali. The list of foods included in the menu is shown in Table 2. Conversely, the dinner on Day 1 was not shared and the students went to dine in three different fast food restaurants in Acireale. On 20 May (Day 2), 17 students (58.6% of the travellers) presented with vomit and/or diarrhoea, falling ill 26–36 h after lunch on Day 1, strongly suggesting that exposure occurred on the first shared meal. Some students also vomited in the presence of the other travellers on the bus bringing them to Acireale. Five more students were symptomatic in the early hours of the following day (21 May, Day 3, from 0:30 am to 8:00 am), as a continuation of the wave of illnesses of the previous day. The teacher and one more student fell ill at 8:00 pm and 11:00 pm of Day 3, respectively, for a second wave of illnesses occurring >48 h from the first shared meal and <36 h from the first cases, with the last student falling sick on 22 May (Day 4) at 3:00 am (Fig. 2, Table 1). On the whole, the mean incubation period was 33.2 h. Based on the online survey, the severity of illness was evaluated by the Vesikari severity scale, grouping symptomatic travellers into severe (11/25 = 44%) and non-severe (14/25 = 56%) cases (Table 1). Vomiting and fever were the predominant symptoms, affecting 22 out of 25 (88%) symptomatic subjects, while diarrhoea was present in 14 out of 25 (56%). Hospitalisation for rehydration therapy was necessary for 13 students (91% [10/11] of severe and 21% [3/14] of non-severe cases). Emergency room (ER) or family practitioner visits were necessary for 21 students and 13 (62%) of them initiated an antibiotic therapy. Two students indicated that other family members had gastroenteric symptoms more than 24 h after their return home, on Day 3, 4 and 5. All those involved in the outbreak recovered within 4 days, with an average duration of illness of 38.4 h. A case-control study of the risk associated with eating foods at lunch on Day 1 or with attending different restaurants at dinner on Day 1 was attempted. A total of 19 subjects filled the section of the online survey concerning food consumption. Fourteen of the responders were among the 22 cases falling ill within 36 h from the first shared meal (first wave), while the remaining five responders were three travellers with supposed secondary infections (showing symptoms with a delay ⩾48 h; second wave) and two asymptomatic students that were used as controls. No statistically significant associations with any food items served at lunch on Day 1 was identified (P < 0.05; Table 2). However, all food consumed on that occasion was to some extent associated with the risk of disease, the highest OR being obtained with chicken (OR: 20.7143, P: 0.0678; Table 2). Actually, most of the cases ate all foods included in the menu, while one of the two asymptomatic controls did not participate in the meal and the other had a pasta dish only. No statistically significant association was found for attending a particular fast food restaurant at the dinner on Day 1 (OR: 2.5, P: 0.3635). The Real-Time RT-PCR assay allowed to detect GII genogroup NoV genome in all the 25 samples from symptomatic subjects, although two of them showed high Ct values, indicating lower shedding of viral genomes (Table 1). The samples from four asymptomatic students were negative for both NoV genogroups. By sequence analysis, a NoV genotype could be defined for 20/ 25 NoV-positive stool samples. All NoV strains from travellers involved in the outbreaks belonged to the same genotype, GII.P2/GII.2, showing >99% nucleotide identities (nt id) in both fragments (ORF1 and ORF2) analysed. Thirty-two water samples were collected and analysed for bacterial contamination (Table 3). Seven samples were taken from the resort (site R) and the remaining 25 from five other sites (C, F, M, S, W), representative of the entire municipal water network. The majority (10/12) of the water samples collected immediately after the onset of the outbreak, between 21 May and 16 June 2016, at the resort (site R) and at sites F and C, located close to the resort, on the seaside and in the interior, respectively, were positive for coliform bacteria. More than 1000 colony-forming units (CFU)/ 100 ml (up to 2700 CFU/100 ml) were detected in four samples, while their absence is required in 100 ml by the Standard of Water Quality in Italy and seven of 11 samples tested were also Escherichia coli-positive (up to 900 CFU/100 ml). Enterococci were searched in 11 samples and three of them were found positive for this faecal contamination indicator. From 7 July to 16 August 2016, after control measures on the water network were undertaken, 13 water samples were collected for follow-up laboratory examination of the water system at F, R and C sites. All of them were negative for bacterial indicators of faecal contamination. No faecal bacteria were identified in the seven samples collected from M, W and S sites, including the well and the source that fed the water network, at any sampling time, before and after implementing the control measures. The research of NoV genome was conducted on concentrated water samples collected on 1 August from R and F sites (Table 3). NoV genome was detected in a single sample collected from the water entry of the resort involved in the outbreak (site R). Actually, only two out of 26 One-Step RT-PCR repetitions were positive for NoV viral capsid gene. The two ORF2-positive results were obtained at the University of Palermo Laboratory using both 1:10 diluted and undiluted extracted RNA. The sequence analysis of the partial ORF2 portion (289 bp) indicated the presence of the same GII.2 capsid genotype detected in the stool samples of the students involved in the outbreak, showing 98.9% nt id. No ORF1 NoV sequence could be obtained from the NoV-positive water sample. Follow-up analyses for NoV genome, conducted on water samples collected on 20 November both at R and F sites found no positive samples. By sequence alignment, the NoV sequences obtained from Mascali outbreak samples were phylogenetically closely related in both ORF1 and ORF2 (99 and 98.9% of nt id, respectively) to a GII.P2/GII.2 strain responsible for a sporadic paediatric gastroenteritis case occurred in Palermo in January 2016 (PA 36/16, accession numbers MG517464 and MG517460, for ORF1 and ORF2, respectively) and previously detected at the Enteric Viruses Laboratory, University of Palermo. Whereas, three GII.P2/GII.2 Italian strains (PA 280/11, PA 366/11 and PA 3/ 14, accession numbers MG517462, MG517465 and MG517463, for ORF1 and MG517457 to −59, for ORF2, respectively), detected in Palermo in 2011 and 2014, were more distantly associated with the outbreak strain both in RdRp and VP1 (90% to 97.4% nt id and 96.1% to 98.1% nt id, respectively, in ORF1 and ORF2). Discussion Epidemiological features, such as the presence of vomiting in more than half (88%) of symptomatic subjects, the 33.2 h average incubation period and the 38.4 h average duration of illness, suggested a NoV aetiology for the investigated outbreak [13]. The detection of NoV genome in all 25 faecal samples from symptomatic subjects fully complies with the requirement of detection of at least one positive sample out of four tested that is considered to be sufficient to ascertain the aetiology of a NoV outbreak [24]. Moreover, molecular typing indicated that a single viral strain, belonging to genotype GII.P2/GII.2, infected all symptomatic subjects involved in the outbreak. NoVs are highly infectious and can easily spread from person to person but also through contaminated food or water. Therefore, settings such as health care institutions, schools and touristic resorts are mainly affected, since close contact and shared food services facilitate transmission. In Mascali, two waves of transmission were observed over three days, being consistent with food/waterborne followed by person-to-person and/or airborne (vomit) transmission (Fig. 2). The first wave of cases presented on 20 May 2016 after a single shared meal, while the last case was detected on 22 May. Exposure to vomit may have been responsible for some of the secondary cases occurred after the first wave of illness. Vomiting produces aerosols that can be inhaled and swallowed. The extremely low infectious dose of NoV makes the inhalation of vomit aerosols an effective transmission route in settings where there is close contact, like on a bus trip [25]. All food consumed in the only shared lunch on 19 May at the resort was associated with the risk of disease (Table 2), but the association was inconclusive since all cases shared the same dishes included in the lunch menu on Day 1. Although many foods were potentially involved, our attention focused on the tap water. In fact, trivial information from local health authorities reported that local population avoided tap water consumption and relied on bottled water for drinking and cooking. Many small outbreaks of gastroenteritis involving restaurants and family groups had been reported in the same area, especially at the beginning of the summer season, when vacation houses and recreational activities were re-opened after a period of low water demand. Illness caused by NoV is common and if the level of water contamination is low, the number of cases remains limited and water can persist as unrecognised source of infection. If polluted water was served for drinking and used for cooking and washing dishes in the Mascali resort at Day 1 a generalised contamination of food and surfaces could have occurred. The potential role of the foodhandlers as a source of contamination could not be analysed in this study but should also be taken into account. If water contamination has already been present in the area for a long time, infections may have occurred among the local food-handlers. The analysis of water samples taken immediately after the outbreak revealed the presence of bacterial indicators of faecal contamination in the water used in the resort, with the only exception of a sample of filter purified water used both for cooking and as drinking water. When extending the sampling to other areas of the municipal water network, many other sites, mostly on the seaside, showed faecal pollution, which demonstrates a poor protection of the water network from faecal leakage (Fig. 1). NoV contamination of waters generally originates from sewage since these enteric pathogens, as other enteric viruses affecting humans, are mostly species-specific and are shed at high concentrations in the stools of patients. When the analyses on the stools of the students made it clear that the outbreak had been caused by NoV, the viral genome was also searched in appropriate volumes of water samples, but this happened when correction measures on the water network had already been taken. Nevertheless, a single water sample, obtained 2 months after the outbreak (1 August 2016), still contained genome fragments that could be correlated to the same NoV genotype causing the outbreak. Although the finding of NoV genome does not demonstrate that entire infecting viral particles were present in the water, it could indicate that a NoV belonging to the same genotype involved in the outbreak was still circulating in the local population several weeks after the outbreak and it was someway able to penetrate the water network, probably indicating continuing faecal contamination through leakage from sewage systems. NoV is particularly resistant in the environment and also to chlorine, therefore transmission through chlorinated water is not exceptional [9, 12, 26]. Drinking water from municipal supplies fed by lake or groundwater has already been identified in the past as the source of waterborne NoV outbreaks in Italy [27–29] and NoV is now considered one of the pathogens causing the largest number of affected consumers in drinking waterborne outbreaks [12]. NoV can persist in water longer than 15 days [30] and it is resistant to chlorine disinfection when free chlorine levels or pre-treatment processes are inadequate [31, 32]. Moreover, microfiltration filters labelled as certified by NSF Standards 53 or 58 (with a pore size ranging from 0.05 to 5 µm) are not effective in removing viruses and can even decrease chlorine concentration in filtered water [33]. Actually, the owner of the resort involved in our outbreak was encouraged to serve tap water as drinking water at the restaurant relying on the effectiveness of such a filter purifier. Unfortunately, we did not have a tap water sample from the restaurant of the resort taken before correction measures and whose volume was appropriate for virus detection. Therefore, we could not demonstrate the failure to eliminate viral pollution of the microfiltration system. In Mascali, correction measures on the water network consisted in the maintenance of the old chlorination device and in the addition of a new one at a key point located upstream and close to suspected illegal sewage dumping (Fig. 1), but as poor protection of the pipeline from sewage leakage risked to reproduce faecal pollution episodes, renovation of the municipal water system was also advised, as well as systematic control of illegal sewage dumping. The additional chlorination device was apparently the most effective intervention, since right after its activation faecal bacterial pollution indicators disappeared. However, NoV genome was detected in a water sample taken 6 weeks after the last detection of bacterial indicators. Unfortunately, detection of faecal contamination of drinking water is only based on the use of bacterial indicators despite no correlation between levels of enteric bacteria and enteric viruses has been found [34–36]. According to ISGEV enteric virus circulation surveillance data, in 2016 the circulation of GII.P2/GII.2 NoV strains in Palermo, home town of the travellers involved in the outbreak, was the highest ever observed from the starting of local NoV surveillance in 2004. GII.P2/GII.2 was the fourth most common genotype in 2016 (12.2% prevalence), following GII.P4/GII.4 Sydney 2012 (28.4%), GII.P16/GII.2 (18.9%) and GII.Pe/GII.4 Sydney 2012 (17.6%). GII.P2/GII.2 strains were mostly observed starting from July 2016, leaving space for the possibility that their increased circulation was linked to the introduction of the ‘Mascali clone’. However, a couple of GII.P2/GII.2 strains (including PA36/16) genetically close to the ‘Mascali clone’ had been already detected in January 2016 in Palermo and GII.P2/GII.2 NoV had sporadically been observed also in 2011 and 2014. Moreover, the circulation of GII.P2/GII.2 NoVs has been increasing worldwide over the last period [37]. Therefore, the unusually high circulation observed in Palermo might just reflect the global epidemiology of GII.P2/GII.2 strains. Noteworthy, the older GII.P2/GII.2 NoV strains detected in Palermo were closely related in the ORF2 (98.9% nt id) to the GII.P16/GII.2 Nashville strain (KY865307) which is supposed to have generated the recombinant GII.P16/GII.4 clone recently circulating epidemically worldwide [37, 38] and were more distantly associated (95.3% nt id) to the GII.2 capsid sequence (AB662868) of another putative GII.P16/GII.4 ancestor from Japan [39]. The increasing frequency of detection of waterborne NoV outbreaks calls for the introduction of routine monitoring for viruses of drinking waters and for the definition of affordable and reliable indicators for the presence of NoV in drinking waters. The use of ultrafiltration systems as a concentration method in order to recover viruses from environmental waters combined with RT-PCR to detect viral genomes has been demonstrated to grant a recovery rate of >50% of multiple viruses [40] and is at present one of the most sensitive approaches for the detection of NoVs, although the discrimination between viable and nonviable viruses remains a problem. The burden of NoV outbreaks, either food associated or waterborne, is still poorly defined due to the lack of specific diagnostic tools in many clinical laboratories. NoV epidemics should always be suspected when the main symptom is vomiting and the duration of disease is short (<48 h). Waterborne transmission is suggested by the rapid spread of disease and the high attack rate. Care should be taken to perform sensitive diagnostic tests on a sufficient number of stool samples in order to identify NoV as the causative agent of enteritis outbreaks, possibly preferring genome screening over antigen detection. Only the systematic detection of NoV outbreaks can guide in the choice of the safety procedures that could be most appropriate in containing such microbiological hazard.

What was the event?

1532

Waterborne Norovirus outbreak at a seaside resort likely originating from municipal water distribution system failure

Abstract In May 2016 a Norovirus (NoV) gastroenteritis outbreak involved a high school class visiting a seaside resort near Taormina (Mascali, Sicily). Twenty-four students and a teacher were affected and 17 of them showed symptoms on the second day of the journey, while the others got ill within the following 2 days. Symptoms included vomiting, diarrhoea and fever, and 12 students required hospitalisation. Stool samples tested positive for NoV genome by Real-Time polymerase chain reaction assay in all 25 symptomatic subjects. The GII.P2/GII.2 NoV genotype was linked to the outbreak by ORF1/ORF2 sequence analysis. The epidemiological features of the outbreak were consistent with food/waterborne followed by person-to-person and/ or vomit transmission. Food consumed at a shared lunch on the first day of the trip was associated to illness and drinking un-bottled tap water was also considered as a risk factor. The analysis of water samples revealed the presence of bacterial indicators of faecal contamination in the water used in the resort as well as in other areas of the municipal water network, linking the NoV gastroenteritis outbreak to tap water pollution from sewage leakage. From a single water sample, an amplicon whose sequence corresponded to the capsid genotype recovered from patients could be obtained. Introduction Noroviruses (NoVs) are recognised as a leading cause of sporadic cases and outbreaks of acute gastroenteritis (AGE) across all age groups [1, 2]. They are transmitted primarily through the faecal-oral route either by direct person–person spread and with consumption of contaminated food or water [1, 3, 4]. NoVs are non-enveloped RNA viruses belonging to the Caliciviridae family and are classified into seven distinct genogroups (GI-GVII) on the basis of the major capsid protein VP1 sequence. GI, GII and GIV NoVs infect humans and they have been divided into more than 30 genotypes based on the sequences of polymerase (ORF1) and capsid proteins (ORF2) [5]. NoV GII.4 genotype has been associated with the vast majority of NoV related cases of AGE worldwide although many genotypes, belonging to both GI and GII genogroups are involved in foodborne and waterborne outbreaks [6, 7]. NoVs are highly contagious owing to their high stability in the environment and their ability to infect at low doses. In fact, just 10–100 virions of NoV may be sufficient to cause AGE [8, 9]. NoVs are considered the major cause of water-related disease in semi-closed institutions, including restaurants, schools, hospitals, nursing homes and cruise ships [10–12]. To confirm NoV as a cause of outbreaks, the presence of four epidemiological features of disease, vomiting in more than half of affected persons, a mean incubation period of 24/48 h, a mean duration of illness of 12–60 h and the absence of bacterial pathogens in stool culture, should be evaluated [13]. Waterborne NoV outbreaks can be caused by contaminated drinking water despite correct chlorine concentration and in the absence of coliform bacteria [10, 14–17]. The aim of this study was to investigate an outbreak of NoV gastroenteritis associated with drinking water involving a high school class travelling from Palermo (West Sicily) to Acireale (East Sicily) on an educational tour in May 2016. Stool samples were collected soon after the onset of AGE symptoms to ascertain the aetiology of illness. The source of infection was investigated collecting epidemiological data through an online questionnaire and, in order to explore a possible transmission route, laboratory tests on water from the municipal network were performed. Material and methods Epidemiological investigation From 20 to 22 May 2016 an AGE outbreak occurred in Mascali, a small town near Taormina, involving a school class of 28 students, 13 girls and 15 boys aged between 14 and 18 years and a 54-year-old female teacher, during a trip to participate in a theatre festival in Acireale. A total of 25/29 persons were affected by nausea, vomiting, abdominal cramps, diarrhoea and fever while they were staying at a seaside resort in Mascali, on the way to Acireale. To obtain data on symptoms and the correlated risk factors all participants filled an online questionnaire investigating personal data (age and sex), travel and accommodation details and food consumption over 48 h before symptoms presentation, time of onset and duration of symptoms, therapy and/or hospitalisation required and history of contact with people affected by diarrhoea and/or vomiting (Table 1). The severity of illness was analysed by the Vesikari severity scale allowing to distinguish non-severe (score <11) and severe (score ⩾11) AGE (Table 1). Vesikari score was calculated from maximum number of episodes/day and duration of both diarrhoea and vomiting (score 0–3 for each of the 4 parameters), temperature (score 0–3), level of dehydration (score 0–3) and treatment required (score 0–2) [18]. A case-control study was conducted to establish if food consumed before the onset of symptoms was associated with the disease. A case was defined as a student attending the tour and reporting diarrhoea and/or vomiting within 36 h from the first shared meal during the tour, while students who did not present any of the previous symptoms or those showing the first symptoms with a delay of ⩾48 h (secondary cases) were used as controls. Variables were studied using odds ratio (OR) and its 95% CI and statistical associations were determined using the χ2 test, with the level of statistical significance set at 0.05 (Table 2). NoV detection in stool samples A stool sample was obtained from each of the 29 travellers, soon after (3–5 days) the onset of the outbreak. No stool samples could be obtained from the food-handlers working at the resort. All samples were tested for NoV at the Enteric Viruses Laboratory of the University of Palermo, member of the Italian Study Group for Enteric Viruses (ISGEV), by a Real-Time polymerase chain reaction (RT-PCR) assay that allows to differentiate GI and GII NoVs [19]. Total RNA was extracted from stool samples using QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) and cDNA was generated by MMLV reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were considered positive for NoVs when cycle threshold (ct) of RT-PCR was ⩽30 and lowpositive when 30<ct⩽35. For all NoV-positive samples two genomic fragments were amplified by conventional RT-PCR using Superscript III One-Step RT-PCR system with Platinum Taq DNA polymerase (Invitrogen) and primers JV12a/JV13b, targeting 326 bps of the RdRp region A [20] and either COG1F/ GISKR or COG2F/GIISKR, amplifying a 315 bps 5′portion of GI or GII VP1 protein [19, 21]. Environmental investigation From May through August 2016, a total of 32 tap water samples were collected from the Mascali municipal network. In particular, seven samples were collected at the resort where the outbreak occurred and from its water entry (R), two in the city centre of Mascali (M), 15 in the village of Fondachello (F), on the seaside, three in Carrabba village (C), far from the coast, and two and three, respectively, from the water well ‘Carlino’ (W) and the source ‘Bufardo and Torregrossa’ (S) which feed the municipal system and provide drinking water to the Mascali area (Fig. 1, Table 3). One litre water samples were sent by the local health authorities to the SIALAB laboratory in Avola, Siracusa and 100 ml aliquots were screened for coliforms, E. coli and enterococci as faecal indicators, by standard bacteriological methods according to UNI EN ISO 9308-1:2014. Unfortunately, although the suspect of a water source for the viral outbreak was quickly communicated, only in August and November 2016, following the implementation of control measures, local authorities provided water samplings of appropriate volume to perform viral tests (10 l). Five water samples of 10 l each were collected from different sampling sites in F and R areas for viral screening (Table 3) and were concentrated according to the Italian National Institute of Health (Istituto Superiore di Sanità, ISS) protocols [22] through an ultra-filtration system (Sartoflow® Slice 200 Benchtop Crossflow System, Sartorius AG, Goettingen, Germany), using appropriate membranes (SG Hydrosart 10 kDa) pretreated with 300 ml of 3% Beef Extract (BE) at pH 7, by recirculating it for 10 min in the system, at a pressure not higher than 1.30 bar. Elution and recovery of viruses attached to the membranes were obtained by recirculating 200 ml of 3% BE at pH 9.5 for 3 h until a 10–12 ml volume of eluate was obtained. The pH of the eluate was then checked with litmus paper and if necessary brought to the value of 7 with HCl. Concentrated water samples were simultaneously tested for NoV in three laboratories: Enteric Viruses Laboratory, University of Palermo; Istituto Zooprofilattico Sperimentale della Sicilia (IZS), Palermo; and Istituto Superiore di Sanità (ISS), Rome. Total RNA was extracted from 500 and 560 µl of a concentrated water sample using BioMèrieux NucliSENS® Magnetic Extraction System (BioMèrieux) and QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) respectively, according to manufacturer’s instructions. Bonded RNA was eluted by 100 µl of Elution Buffer. Amplifications were performed at least in duplicate on whole and 1:10 diluted extracts. A Real-Time RT-PCR assay according to ISO/TS 15216- 2:2013, two in-house Real-Time RT-PCR protocols and the One-Step RT-PCR methods already used for amplifying ORF1 and ORF2 fragments in NoV-positive faecal samples were performed [19–21, 23]. Nucleotide sequences analysis and phylogenetic analysis PCR amplicons were purified by commercial columns (CleanSweep PCR Purification, Thermo Fisher Scientific, Vilnius, Lithuania) and sequenced by Sanger method and all nucleotide sequences were aligned using ClustalW. To identify NoV genotypes, RdRp and VP1 sequences were submitted to the RIVM NoV genotyping tool (http://www.rivm.nl/mpf/norovirus/typingtool). Phylogenetic trees were constructed using MEGA 7 and Neighbour-Joining method (bootstrap 1000 replicates). Sequences from stool and water samples are available at NCBI with the following accession numbers: MG517455 for the major capsid protein obtained from a water sample; MG517456 and MG517461, respectively, for the major capsid protein and RNA-dependent-RNA-polymerase, as representatives of the sequences obtained from the stool samples of those involved in the outbreak. Control measures After outbreak notification, on 21 May, the resort was closed for 2 days. On 25 May, following the first results of the bacteriological analyses on municipal water samples, the use of municipal water for human consumption was temporarily restricted (use of water was permitted only for cleaning purposes) in the Carrabba and Fondachello villages and in the area surrounding the resort. From 25 May the reservoir collecting the water from the Carlino well and its chlorination device were submitted to maintenance work and a new chlorination device was installed on the conduit bringing the water from the Bufardo source. Actually, the Bufardo source had been connected to the municipal network in 2003 but unfortunately a dedicated chlorination station had not been planned at that time. Therefore, unchlorinated water entered the system near Fondachello village on the seaside (Fig. 1), possibly diluting chlorine concentrations in water treated at the Carlino reservoir. Moreover, the exact path of the branch of the water network situated downstream the Bufardo source water connection was not marked in the maps of the municipality (Fig. 1, red line), indicating that this section had never been submitted to maintenance. The area surrounding the unmaintained section of the water pipeline was scattered with unauthorised cottages that were not connected to the sewerage systems. Likely, untreated waste from cesspools was unloaded directly into the soil, generating a risk of contamination of the municipal water system through leakage. No work had been made on the water pipelines before the outbreak but sections of the network had been temporarily closed during the winter season on the seaside (Fondachello). Local health authorities advised the connection of all households in the area to the sewerage system and the replacement of the water system section without maintenance. Health authorities also raised attention on AGE cases notification at all local Emergency Departments and Family Practitioners but there was no further evidence of AGE outbreaks in the months following the intervention. Results Data collected through the online questionnaire compiled 7–10 days after the outbreak revealed that none of the 28 students and the teacher taking part in the school trip showed gastroenteritis symptoms at the beginning of the journey and they had no contact with people who suffered from diarrhoea and/or vomiting in the previous days. Travellers had their first shared meal on 19 May (Day 1), 2016, at lunch on their arrival to the resort in Mascali. The list of foods included in the menu is shown in Table 2. Conversely, the dinner on Day 1 was not shared and the students went to dine in three different fast food restaurants in Acireale. On 20 May (Day 2), 17 students (58.6% of the travellers) presented with vomit and/or diarrhoea, falling ill 26–36 h after lunch on Day 1, strongly suggesting that exposure occurred on the first shared meal. Some students also vomited in the presence of the other travellers on the bus bringing them to Acireale. Five more students were symptomatic in the early hours of the following day (21 May, Day 3, from 0:30 am to 8:00 am), as a continuation of the wave of illnesses of the previous day. The teacher and one more student fell ill at 8:00 pm and 11:00 pm of Day 3, respectively, for a second wave of illnesses occurring >48 h from the first shared meal and <36 h from the first cases, with the last student falling sick on 22 May (Day 4) at 3:00 am (Fig. 2, Table 1). On the whole, the mean incubation period was 33.2 h. Based on the online survey, the severity of illness was evaluated by the Vesikari severity scale, grouping symptomatic travellers into severe (11/25 = 44%) and non-severe (14/25 = 56%) cases (Table 1). Vomiting and fever were the predominant symptoms, affecting 22 out of 25 (88%) symptomatic subjects, while diarrhoea was present in 14 out of 25 (56%). Hospitalisation for rehydration therapy was necessary for 13 students (91% [10/11] of severe and 21% [3/14] of non-severe cases). Emergency room (ER) or family practitioner visits were necessary for 21 students and 13 (62%) of them initiated an antibiotic therapy. Two students indicated that other family members had gastroenteric symptoms more than 24 h after their return home, on Day 3, 4 and 5. All those involved in the outbreak recovered within 4 days, with an average duration of illness of 38.4 h. A case-control study of the risk associated with eating foods at lunch on Day 1 or with attending different restaurants at dinner on Day 1 was attempted. A total of 19 subjects filled the section of the online survey concerning food consumption. Fourteen of the responders were among the 22 cases falling ill within 36 h from the first shared meal (first wave), while the remaining five responders were three travellers with supposed secondary infections (showing symptoms with a delay ⩾48 h; second wave) and two asymptomatic students that were used as controls. No statistically significant associations with any food items served at lunch on Day 1 was identified (P < 0.05; Table 2). However, all food consumed on that occasion was to some extent associated with the risk of disease, the highest OR being obtained with chicken (OR: 20.7143, P: 0.0678; Table 2). Actually, most of the cases ate all foods included in the menu, while one of the two asymptomatic controls did not participate in the meal and the other had a pasta dish only. No statistically significant association was found for attending a particular fast food restaurant at the dinner on Day 1 (OR: 2.5, P: 0.3635). The Real-Time RT-PCR assay allowed to detect GII genogroup NoV genome in all the 25 samples from symptomatic subjects, although two of them showed high Ct values, indicating lower shedding of viral genomes (Table 1). The samples from four asymptomatic students were negative for both NoV genogroups. By sequence analysis, a NoV genotype could be defined for 20/ 25 NoV-positive stool samples. All NoV strains from travellers involved in the outbreaks belonged to the same genotype, GII.P2/GII.2, showing >99% nucleotide identities (nt id) in both fragments (ORF1 and ORF2) analysed. Thirty-two water samples were collected and analysed for bacterial contamination (Table 3). Seven samples were taken from the resort (site R) and the remaining 25 from five other sites (C, F, M, S, W), representative of the entire municipal water network. The majority (10/12) of the water samples collected immediately after the onset of the outbreak, between 21 May and 16 June 2016, at the resort (site R) and at sites F and C, located close to the resort, on the seaside and in the interior, respectively, were positive for coliform bacteria. More than 1000 colony-forming units (CFU)/ 100 ml (up to 2700 CFU/100 ml) were detected in four samples, while their absence is required in 100 ml by the Standard of Water Quality in Italy and seven of 11 samples tested were also Escherichia coli-positive (up to 900 CFU/100 ml). Enterococci were searched in 11 samples and three of them were found positive for this faecal contamination indicator. From 7 July to 16 August 2016, after control measures on the water network were undertaken, 13 water samples were collected for follow-up laboratory examination of the water system at F, R and C sites. All of them were negative for bacterial indicators of faecal contamination. No faecal bacteria were identified in the seven samples collected from M, W and S sites, including the well and the source that fed the water network, at any sampling time, before and after implementing the control measures. The research of NoV genome was conducted on concentrated water samples collected on 1 August from R and F sites (Table 3). NoV genome was detected in a single sample collected from the water entry of the resort involved in the outbreak (site R). Actually, only two out of 26 One-Step RT-PCR repetitions were positive for NoV viral capsid gene. The two ORF2-positive results were obtained at the University of Palermo Laboratory using both 1:10 diluted and undiluted extracted RNA. The sequence analysis of the partial ORF2 portion (289 bp) indicated the presence of the same GII.2 capsid genotype detected in the stool samples of the students involved in the outbreak, showing 98.9% nt id. No ORF1 NoV sequence could be obtained from the NoV-positive water sample. Follow-up analyses for NoV genome, conducted on water samples collected on 20 November both at R and F sites found no positive samples. By sequence alignment, the NoV sequences obtained from Mascali outbreak samples were phylogenetically closely related in both ORF1 and ORF2 (99 and 98.9% of nt id, respectively) to a GII.P2/GII.2 strain responsible for a sporadic paediatric gastroenteritis case occurred in Palermo in January 2016 (PA 36/16, accession numbers MG517464 and MG517460, for ORF1 and ORF2, respectively) and previously detected at the Enteric Viruses Laboratory, University of Palermo. Whereas, three GII.P2/GII.2 Italian strains (PA 280/11, PA 366/11 and PA 3/ 14, accession numbers MG517462, MG517465 and MG517463, for ORF1 and MG517457 to −59, for ORF2, respectively), detected in Palermo in 2011 and 2014, were more distantly associated with the outbreak strain both in RdRp and VP1 (90% to 97.4% nt id and 96.1% to 98.1% nt id, respectively, in ORF1 and ORF2). Discussion Epidemiological features, such as the presence of vomiting in more than half (88%) of symptomatic subjects, the 33.2 h average incubation period and the 38.4 h average duration of illness, suggested a NoV aetiology for the investigated outbreak [13]. The detection of NoV genome in all 25 faecal samples from symptomatic subjects fully complies with the requirement of detection of at least one positive sample out of four tested that is considered to be sufficient to ascertain the aetiology of a NoV outbreak [24]. Moreover, molecular typing indicated that a single viral strain, belonging to genotype GII.P2/GII.2, infected all symptomatic subjects involved in the outbreak. NoVs are highly infectious and can easily spread from person to person but also through contaminated food or water. Therefore, settings such as health care institutions, schools and touristic resorts are mainly affected, since close contact and shared food services facilitate transmission. In Mascali, two waves of transmission were observed over three days, being consistent with food/waterborne followed by person-to-person and/or airborne (vomit) transmission (Fig. 2). The first wave of cases presented on 20 May 2016 after a single shared meal, while the last case was detected on 22 May. Exposure to vomit may have been responsible for some of the secondary cases occurred after the first wave of illness. Vomiting produces aerosols that can be inhaled and swallowed. The extremely low infectious dose of NoV makes the inhalation of vomit aerosols an effective transmission route in settings where there is close contact, like on a bus trip [25]. All food consumed in the only shared lunch on 19 May at the resort was associated with the risk of disease (Table 2), but the association was inconclusive since all cases shared the same dishes included in the lunch menu on Day 1. Although many foods were potentially involved, our attention focused on the tap water. In fact, trivial information from local health authorities reported that local population avoided tap water consumption and relied on bottled water for drinking and cooking. Many small outbreaks of gastroenteritis involving restaurants and family groups had been reported in the same area, especially at the beginning of the summer season, when vacation houses and recreational activities were re-opened after a period of low water demand. Illness caused by NoV is common and if the level of water contamination is low, the number of cases remains limited and water can persist as unrecognised source of infection. If polluted water was served for drinking and used for cooking and washing dishes in the Mascali resort at Day 1 a generalised contamination of food and surfaces could have occurred. The potential role of the foodhandlers as a source of contamination could not be analysed in this study but should also be taken into account. If water contamination has already been present in the area for a long time, infections may have occurred among the local food-handlers. The analysis of water samples taken immediately after the outbreak revealed the presence of bacterial indicators of faecal contamination in the water used in the resort, with the only exception of a sample of filter purified water used both for cooking and as drinking water. When extending the sampling to other areas of the municipal water network, many other sites, mostly on the seaside, showed faecal pollution, which demonstrates a poor protection of the water network from faecal leakage (Fig. 1). NoV contamination of waters generally originates from sewage since these enteric pathogens, as other enteric viruses affecting humans, are mostly species-specific and are shed at high concentrations in the stools of patients. When the analyses on the stools of the students made it clear that the outbreak had been caused by NoV, the viral genome was also searched in appropriate volumes of water samples, but this happened when correction measures on the water network had already been taken. Nevertheless, a single water sample, obtained 2 months after the outbreak (1 August 2016), still contained genome fragments that could be correlated to the same NoV genotype causing the outbreak. Although the finding of NoV genome does not demonstrate that entire infecting viral particles were present in the water, it could indicate that a NoV belonging to the same genotype involved in the outbreak was still circulating in the local population several weeks after the outbreak and it was someway able to penetrate the water network, probably indicating continuing faecal contamination through leakage from sewage systems. NoV is particularly resistant in the environment and also to chlorine, therefore transmission through chlorinated water is not exceptional [9, 12, 26]. Drinking water from municipal supplies fed by lake or groundwater has already been identified in the past as the source of waterborne NoV outbreaks in Italy [27–29] and NoV is now considered one of the pathogens causing the largest number of affected consumers in drinking waterborne outbreaks [12]. NoV can persist in water longer than 15 days [30] and it is resistant to chlorine disinfection when free chlorine levels or pre-treatment processes are inadequate [31, 32]. Moreover, microfiltration filters labelled as certified by NSF Standards 53 or 58 (with a pore size ranging from 0.05 to 5 µm) are not effective in removing viruses and can even decrease chlorine concentration in filtered water [33]. Actually, the owner of the resort involved in our outbreak was encouraged to serve tap water as drinking water at the restaurant relying on the effectiveness of such a filter purifier. Unfortunately, we did not have a tap water sample from the restaurant of the resort taken before correction measures and whose volume was appropriate for virus detection. Therefore, we could not demonstrate the failure to eliminate viral pollution of the microfiltration system. In Mascali, correction measures on the water network consisted in the maintenance of the old chlorination device and in the addition of a new one at a key point located upstream and close to suspected illegal sewage dumping (Fig. 1), but as poor protection of the pipeline from sewage leakage risked to reproduce faecal pollution episodes, renovation of the municipal water system was also advised, as well as systematic control of illegal sewage dumping. The additional chlorination device was apparently the most effective intervention, since right after its activation faecal bacterial pollution indicators disappeared. However, NoV genome was detected in a water sample taken 6 weeks after the last detection of bacterial indicators. Unfortunately, detection of faecal contamination of drinking water is only based on the use of bacterial indicators despite no correlation between levels of enteric bacteria and enteric viruses has been found [34–36]. According to ISGEV enteric virus circulation surveillance data, in 2016 the circulation of GII.P2/GII.2 NoV strains in Palermo, home town of the travellers involved in the outbreak, was the highest ever observed from the starting of local NoV surveillance in 2004. GII.P2/GII.2 was the fourth most common genotype in 2016 (12.2% prevalence), following GII.P4/GII.4 Sydney 2012 (28.4%), GII.P16/GII.2 (18.9%) and GII.Pe/GII.4 Sydney 2012 (17.6%). GII.P2/GII.2 strains were mostly observed starting from July 2016, leaving space for the possibility that their increased circulation was linked to the introduction of the ‘Mascali clone’. However, a couple of GII.P2/GII.2 strains (including PA36/16) genetically close to the ‘Mascali clone’ had been already detected in January 2016 in Palermo and GII.P2/GII.2 NoV had sporadically been observed also in 2011 and 2014. Moreover, the circulation of GII.P2/GII.2 NoVs has been increasing worldwide over the last period [37]. Therefore, the unusually high circulation observed in Palermo might just reflect the global epidemiology of GII.P2/GII.2 strains. Noteworthy, the older GII.P2/GII.2 NoV strains detected in Palermo were closely related in the ORF2 (98.9% nt id) to the GII.P16/GII.2 Nashville strain (KY865307) which is supposed to have generated the recombinant GII.P16/GII.4 clone recently circulating epidemically worldwide [37, 38] and were more distantly associated (95.3% nt id) to the GII.2 capsid sequence (AB662868) of another putative GII.P16/GII.4 ancestor from Japan [39]. The increasing frequency of detection of waterborne NoV outbreaks calls for the introduction of routine monitoring for viruses of drinking waters and for the definition of affordable and reliable indicators for the presence of NoV in drinking waters. The use of ultrafiltration systems as a concentration method in order to recover viruses from environmental waters combined with RT-PCR to detect viral genomes has been demonstrated to grant a recovery rate of >50% of multiple viruses [40] and is at present one of the most sensitive approaches for the detection of NoVs, although the discrimination between viable and nonviable viruses remains a problem. The burden of NoV outbreaks, either food associated or waterborne, is still poorly defined due to the lack of specific diagnostic tools in many clinical laboratories. NoV epidemics should always be suspected when the main symptom is vomiting and the duration of disease is short (<48 h). Waterborne transmission is suggested by the rapid spread of disease and the high attack rate. Care should be taken to perform sensitive diagnostic tests on a sufficient number of stool samples in order to identify NoV as the causative agent of enteritis outbreaks, possibly preferring genome screening over antigen detection. Only the systematic detection of NoV outbreaks can guide in the choice of the safety procedures that could be most appropriate in containing such microbiological hazard.

When did this happen?

1533

Waterborne Norovirus outbreak at a seaside resort likely originating from municipal water distribution system failure

Abstract In May 2016 a Norovirus (NoV) gastroenteritis outbreak involved a high school class visiting a seaside resort near Taormina (Mascali, Sicily). Twenty-four students and a teacher were affected and 17 of them showed symptoms on the second day of the journey, while the others got ill within the following 2 days. Symptoms included vomiting, diarrhoea and fever, and 12 students required hospitalisation. Stool samples tested positive for NoV genome by Real-Time polymerase chain reaction assay in all 25 symptomatic subjects. The GII.P2/GII.2 NoV genotype was linked to the outbreak by ORF1/ORF2 sequence analysis. The epidemiological features of the outbreak were consistent with food/waterborne followed by person-to-person and/ or vomit transmission. Food consumed at a shared lunch on the first day of the trip was associated to illness and drinking un-bottled tap water was also considered as a risk factor. The analysis of water samples revealed the presence of bacterial indicators of faecal contamination in the water used in the resort as well as in other areas of the municipal water network, linking the NoV gastroenteritis outbreak to tap water pollution from sewage leakage. From a single water sample, an amplicon whose sequence corresponded to the capsid genotype recovered from patients could be obtained. Introduction Noroviruses (NoVs) are recognised as a leading cause of sporadic cases and outbreaks of acute gastroenteritis (AGE) across all age groups [1, 2]. They are transmitted primarily through the faecal-oral route either by direct person–person spread and with consumption of contaminated food or water [1, 3, 4]. NoVs are non-enveloped RNA viruses belonging to the Caliciviridae family and are classified into seven distinct genogroups (GI-GVII) on the basis of the major capsid protein VP1 sequence. GI, GII and GIV NoVs infect humans and they have been divided into more than 30 genotypes based on the sequences of polymerase (ORF1) and capsid proteins (ORF2) [5]. NoV GII.4 genotype has been associated with the vast majority of NoV related cases of AGE worldwide although many genotypes, belonging to both GI and GII genogroups are involved in foodborne and waterborne outbreaks [6, 7]. NoVs are highly contagious owing to their high stability in the environment and their ability to infect at low doses. In fact, just 10–100 virions of NoV may be sufficient to cause AGE [8, 9]. NoVs are considered the major cause of water-related disease in semi-closed institutions, including restaurants, schools, hospitals, nursing homes and cruise ships [10–12]. To confirm NoV as a cause of outbreaks, the presence of four epidemiological features of disease, vomiting in more than half of affected persons, a mean incubation period of 24/48 h, a mean duration of illness of 12–60 h and the absence of bacterial pathogens in stool culture, should be evaluated [13]. Waterborne NoV outbreaks can be caused by contaminated drinking water despite correct chlorine concentration and in the absence of coliform bacteria [10, 14–17]. The aim of this study was to investigate an outbreak of NoV gastroenteritis associated with drinking water involving a high school class travelling from Palermo (West Sicily) to Acireale (East Sicily) on an educational tour in May 2016. Stool samples were collected soon after the onset of AGE symptoms to ascertain the aetiology of illness. The source of infection was investigated collecting epidemiological data through an online questionnaire and, in order to explore a possible transmission route, laboratory tests on water from the municipal network were performed. Material and methods Epidemiological investigation From 20 to 22 May 2016 an AGE outbreak occurred in Mascali, a small town near Taormina, involving a school class of 28 students, 13 girls and 15 boys aged between 14 and 18 years and a 54-year-old female teacher, during a trip to participate in a theatre festival in Acireale. A total of 25/29 persons were affected by nausea, vomiting, abdominal cramps, diarrhoea and fever while they were staying at a seaside resort in Mascali, on the way to Acireale. To obtain data on symptoms and the correlated risk factors all participants filled an online questionnaire investigating personal data (age and sex), travel and accommodation details and food consumption over 48 h before symptoms presentation, time of onset and duration of symptoms, therapy and/or hospitalisation required and history of contact with people affected by diarrhoea and/or vomiting (Table 1). The severity of illness was analysed by the Vesikari severity scale allowing to distinguish non-severe (score <11) and severe (score ⩾11) AGE (Table 1). Vesikari score was calculated from maximum number of episodes/day and duration of both diarrhoea and vomiting (score 0–3 for each of the 4 parameters), temperature (score 0–3), level of dehydration (score 0–3) and treatment required (score 0–2) [18]. A case-control study was conducted to establish if food consumed before the onset of symptoms was associated with the disease. A case was defined as a student attending the tour and reporting diarrhoea and/or vomiting within 36 h from the first shared meal during the tour, while students who did not present any of the previous symptoms or those showing the first symptoms with a delay of ⩾48 h (secondary cases) were used as controls. Variables were studied using odds ratio (OR) and its 95% CI and statistical associations were determined using the χ2 test, with the level of statistical significance set at 0.05 (Table 2). NoV detection in stool samples A stool sample was obtained from each of the 29 travellers, soon after (3–5 days) the onset of the outbreak. No stool samples could be obtained from the food-handlers working at the resort. All samples were tested for NoV at the Enteric Viruses Laboratory of the University of Palermo, member of the Italian Study Group for Enteric Viruses (ISGEV), by a Real-Time polymerase chain reaction (RT-PCR) assay that allows to differentiate GI and GII NoVs [19]. Total RNA was extracted from stool samples using QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) and cDNA was generated by MMLV reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were considered positive for NoVs when cycle threshold (ct) of RT-PCR was ⩽30 and lowpositive when 30<ct⩽35. For all NoV-positive samples two genomic fragments were amplified by conventional RT-PCR using Superscript III One-Step RT-PCR system with Platinum Taq DNA polymerase (Invitrogen) and primers JV12a/JV13b, targeting 326 bps of the RdRp region A [20] and either COG1F/ GISKR or COG2F/GIISKR, amplifying a 315 bps 5′portion of GI or GII VP1 protein [19, 21]. Environmental investigation From May through August 2016, a total of 32 tap water samples were collected from the Mascali municipal network. In particular, seven samples were collected at the resort where the outbreak occurred and from its water entry (R), two in the city centre of Mascali (M), 15 in the village of Fondachello (F), on the seaside, three in Carrabba village (C), far from the coast, and two and three, respectively, from the water well ‘Carlino’ (W) and the source ‘Bufardo and Torregrossa’ (S) which feed the municipal system and provide drinking water to the Mascali area (Fig. 1, Table 3). One litre water samples were sent by the local health authorities to the SIALAB laboratory in Avola, Siracusa and 100 ml aliquots were screened for coliforms, E. coli and enterococci as faecal indicators, by standard bacteriological methods according to UNI EN ISO 9308-1:2014. Unfortunately, although the suspect of a water source for the viral outbreak was quickly communicated, only in August and November 2016, following the implementation of control measures, local authorities provided water samplings of appropriate volume to perform viral tests (10 l). Five water samples of 10 l each were collected from different sampling sites in F and R areas for viral screening (Table 3) and were concentrated according to the Italian National Institute of Health (Istituto Superiore di Sanità, ISS) protocols [22] through an ultra-filtration system (Sartoflow® Slice 200 Benchtop Crossflow System, Sartorius AG, Goettingen, Germany), using appropriate membranes (SG Hydrosart 10 kDa) pretreated with 300 ml of 3% Beef Extract (BE) at pH 7, by recirculating it for 10 min in the system, at a pressure not higher than 1.30 bar. Elution and recovery of viruses attached to the membranes were obtained by recirculating 200 ml of 3% BE at pH 9.5 for 3 h until a 10–12 ml volume of eluate was obtained. The pH of the eluate was then checked with litmus paper and if necessary brought to the value of 7 with HCl. Concentrated water samples were simultaneously tested for NoV in three laboratories: Enteric Viruses Laboratory, University of Palermo; Istituto Zooprofilattico Sperimentale della Sicilia (IZS), Palermo; and Istituto Superiore di Sanità (ISS), Rome. Total RNA was extracted from 500 and 560 µl of a concentrated water sample using BioMèrieux NucliSENS® Magnetic Extraction System (BioMèrieux) and QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) respectively, according to manufacturer’s instructions. Bonded RNA was eluted by 100 µl of Elution Buffer. Amplifications were performed at least in duplicate on whole and 1:10 diluted extracts. A Real-Time RT-PCR assay according to ISO/TS 15216- 2:2013, two in-house Real-Time RT-PCR protocols and the One-Step RT-PCR methods already used for amplifying ORF1 and ORF2 fragments in NoV-positive faecal samples were performed [19–21, 23]. Nucleotide sequences analysis and phylogenetic analysis PCR amplicons were purified by commercial columns (CleanSweep PCR Purification, Thermo Fisher Scientific, Vilnius, Lithuania) and sequenced by Sanger method and all nucleotide sequences were aligned using ClustalW. To identify NoV genotypes, RdRp and VP1 sequences were submitted to the RIVM NoV genotyping tool (http://www.rivm.nl/mpf/norovirus/typingtool). Phylogenetic trees were constructed using MEGA 7 and Neighbour-Joining method (bootstrap 1000 replicates). Sequences from stool and water samples are available at NCBI with the following accession numbers: MG517455 for the major capsid protein obtained from a water sample; MG517456 and MG517461, respectively, for the major capsid protein and RNA-dependent-RNA-polymerase, as representatives of the sequences obtained from the stool samples of those involved in the outbreak. Control measures After outbreak notification, on 21 May, the resort was closed for 2 days. On 25 May, following the first results of the bacteriological analyses on municipal water samples, the use of municipal water for human consumption was temporarily restricted (use of water was permitted only for cleaning purposes) in the Carrabba and Fondachello villages and in the area surrounding the resort. From 25 May the reservoir collecting the water from the Carlino well and its chlorination device were submitted to maintenance work and a new chlorination device was installed on the conduit bringing the water from the Bufardo source. Actually, the Bufardo source had been connected to the municipal network in 2003 but unfortunately a dedicated chlorination station had not been planned at that time. Therefore, unchlorinated water entered the system near Fondachello village on the seaside (Fig. 1), possibly diluting chlorine concentrations in water treated at the Carlino reservoir. Moreover, the exact path of the branch of the water network situated downstream the Bufardo source water connection was not marked in the maps of the municipality (Fig. 1, red line), indicating that this section had never been submitted to maintenance. The area surrounding the unmaintained section of the water pipeline was scattered with unauthorised cottages that were not connected to the sewerage systems. Likely, untreated waste from cesspools was unloaded directly into the soil, generating a risk of contamination of the municipal water system through leakage. No work had been made on the water pipelines before the outbreak but sections of the network had been temporarily closed during the winter season on the seaside (Fondachello). Local health authorities advised the connection of all households in the area to the sewerage system and the replacement of the water system section without maintenance. Health authorities also raised attention on AGE cases notification at all local Emergency Departments and Family Practitioners but there was no further evidence of AGE outbreaks in the months following the intervention. Results Data collected through the online questionnaire compiled 7–10 days after the outbreak revealed that none of the 28 students and the teacher taking part in the school trip showed gastroenteritis symptoms at the beginning of the journey and they had no contact with people who suffered from diarrhoea and/or vomiting in the previous days. Travellers had their first shared meal on 19 May (Day 1), 2016, at lunch on their arrival to the resort in Mascali. The list of foods included in the menu is shown in Table 2. Conversely, the dinner on Day 1 was not shared and the students went to dine in three different fast food restaurants in Acireale. On 20 May (Day 2), 17 students (58.6% of the travellers) presented with vomit and/or diarrhoea, falling ill 26–36 h after lunch on Day 1, strongly suggesting that exposure occurred on the first shared meal. Some students also vomited in the presence of the other travellers on the bus bringing them to Acireale. Five more students were symptomatic in the early hours of the following day (21 May, Day 3, from 0:30 am to 8:00 am), as a continuation of the wave of illnesses of the previous day. The teacher and one more student fell ill at 8:00 pm and 11:00 pm of Day 3, respectively, for a second wave of illnesses occurring >48 h from the first shared meal and <36 h from the first cases, with the last student falling sick on 22 May (Day 4) at 3:00 am (Fig. 2, Table 1). On the whole, the mean incubation period was 33.2 h. Based on the online survey, the severity of illness was evaluated by the Vesikari severity scale, grouping symptomatic travellers into severe (11/25 = 44%) and non-severe (14/25 = 56%) cases (Table 1). Vomiting and fever were the predominant symptoms, affecting 22 out of 25 (88%) symptomatic subjects, while diarrhoea was present in 14 out of 25 (56%). Hospitalisation for rehydration therapy was necessary for 13 students (91% [10/11] of severe and 21% [3/14] of non-severe cases). Emergency room (ER) or family practitioner visits were necessary for 21 students and 13 (62%) of them initiated an antibiotic therapy. Two students indicated that other family members had gastroenteric symptoms more than 24 h after their return home, on Day 3, 4 and 5. All those involved in the outbreak recovered within 4 days, with an average duration of illness of 38.4 h. A case-control study of the risk associated with eating foods at lunch on Day 1 or with attending different restaurants at dinner on Day 1 was attempted. A total of 19 subjects filled the section of the online survey concerning food consumption. Fourteen of the responders were among the 22 cases falling ill within 36 h from the first shared meal (first wave), while the remaining five responders were three travellers with supposed secondary infections (showing symptoms with a delay ⩾48 h; second wave) and two asymptomatic students that were used as controls. No statistically significant associations with any food items served at lunch on Day 1 was identified (P < 0.05; Table 2). However, all food consumed on that occasion was to some extent associated with the risk of disease, the highest OR being obtained with chicken (OR: 20.7143, P: 0.0678; Table 2). Actually, most of the cases ate all foods included in the menu, while one of the two asymptomatic controls did not participate in the meal and the other had a pasta dish only. No statistically significant association was found for attending a particular fast food restaurant at the dinner on Day 1 (OR: 2.5, P: 0.3635). The Real-Time RT-PCR assay allowed to detect GII genogroup NoV genome in all the 25 samples from symptomatic subjects, although two of them showed high Ct values, indicating lower shedding of viral genomes (Table 1). The samples from four asymptomatic students were negative for both NoV genogroups. By sequence analysis, a NoV genotype could be defined for 20/ 25 NoV-positive stool samples. All NoV strains from travellers involved in the outbreaks belonged to the same genotype, GII.P2/GII.2, showing >99% nucleotide identities (nt id) in both fragments (ORF1 and ORF2) analysed. Thirty-two water samples were collected and analysed for bacterial contamination (Table 3). Seven samples were taken from the resort (site R) and the remaining 25 from five other sites (C, F, M, S, W), representative of the entire municipal water network. The majority (10/12) of the water samples collected immediately after the onset of the outbreak, between 21 May and 16 June 2016, at the resort (site R) and at sites F and C, located close to the resort, on the seaside and in the interior, respectively, were positive for coliform bacteria. More than 1000 colony-forming units (CFU)/ 100 ml (up to 2700 CFU/100 ml) were detected in four samples, while their absence is required in 100 ml by the Standard of Water Quality in Italy and seven of 11 samples tested were also Escherichia coli-positive (up to 900 CFU/100 ml). Enterococci were searched in 11 samples and three of them were found positive for this faecal contamination indicator. From 7 July to 16 August 2016, after control measures on the water network were undertaken, 13 water samples were collected for follow-up laboratory examination of the water system at F, R and C sites. All of them were negative for bacterial indicators of faecal contamination. No faecal bacteria were identified in the seven samples collected from M, W and S sites, including the well and the source that fed the water network, at any sampling time, before and after implementing the control measures. The research of NoV genome was conducted on concentrated water samples collected on 1 August from R and F sites (Table 3). NoV genome was detected in a single sample collected from the water entry of the resort involved in the outbreak (site R). Actually, only two out of 26 One-Step RT-PCR repetitions were positive for NoV viral capsid gene. The two ORF2-positive results were obtained at the University of Palermo Laboratory using both 1:10 diluted and undiluted extracted RNA. The sequence analysis of the partial ORF2 portion (289 bp) indicated the presence of the same GII.2 capsid genotype detected in the stool samples of the students involved in the outbreak, showing 98.9% nt id. No ORF1 NoV sequence could be obtained from the NoV-positive water sample. Follow-up analyses for NoV genome, conducted on water samples collected on 20 November both at R and F sites found no positive samples. By sequence alignment, the NoV sequences obtained from Mascali outbreak samples were phylogenetically closely related in both ORF1 and ORF2 (99 and 98.9% of nt id, respectively) to a GII.P2/GII.2 strain responsible for a sporadic paediatric gastroenteritis case occurred in Palermo in January 2016 (PA 36/16, accession numbers MG517464 and MG517460, for ORF1 and ORF2, respectively) and previously detected at the Enteric Viruses Laboratory, University of Palermo. Whereas, three GII.P2/GII.2 Italian strains (PA 280/11, PA 366/11 and PA 3/ 14, accession numbers MG517462, MG517465 and MG517463, for ORF1 and MG517457 to −59, for ORF2, respectively), detected in Palermo in 2011 and 2014, were more distantly associated with the outbreak strain both in RdRp and VP1 (90% to 97.4% nt id and 96.1% to 98.1% nt id, respectively, in ORF1 and ORF2). Discussion Epidemiological features, such as the presence of vomiting in more than half (88%) of symptomatic subjects, the 33.2 h average incubation period and the 38.4 h average duration of illness, suggested a NoV aetiology for the investigated outbreak [13]. The detection of NoV genome in all 25 faecal samples from symptomatic subjects fully complies with the requirement of detection of at least one positive sample out of four tested that is considered to be sufficient to ascertain the aetiology of a NoV outbreak [24]. Moreover, molecular typing indicated that a single viral strain, belonging to genotype GII.P2/GII.2, infected all symptomatic subjects involved in the outbreak. NoVs are highly infectious and can easily spread from person to person but also through contaminated food or water. Therefore, settings such as health care institutions, schools and touristic resorts are mainly affected, since close contact and shared food services facilitate transmission. In Mascali, two waves of transmission were observed over three days, being consistent with food/waterborne followed by person-to-person and/or airborne (vomit) transmission (Fig. 2). The first wave of cases presented on 20 May 2016 after a single shared meal, while the last case was detected on 22 May. Exposure to vomit may have been responsible for some of the secondary cases occurred after the first wave of illness. Vomiting produces aerosols that can be inhaled and swallowed. The extremely low infectious dose of NoV makes the inhalation of vomit aerosols an effective transmission route in settings where there is close contact, like on a bus trip [25]. All food consumed in the only shared lunch on 19 May at the resort was associated with the risk of disease (Table 2), but the association was inconclusive since all cases shared the same dishes included in the lunch menu on Day 1. Although many foods were potentially involved, our attention focused on the tap water. In fact, trivial information from local health authorities reported that local population avoided tap water consumption and relied on bottled water for drinking and cooking. Many small outbreaks of gastroenteritis involving restaurants and family groups had been reported in the same area, especially at the beginning of the summer season, when vacation houses and recreational activities were re-opened after a period of low water demand. Illness caused by NoV is common and if the level of water contamination is low, the number of cases remains limited and water can persist as unrecognised source of infection. If polluted water was served for drinking and used for cooking and washing dishes in the Mascali resort at Day 1 a generalised contamination of food and surfaces could have occurred. The potential role of the foodhandlers as a source of contamination could not be analysed in this study but should also be taken into account. If water contamination has already been present in the area for a long time, infections may have occurred among the local food-handlers. The analysis of water samples taken immediately after the outbreak revealed the presence of bacterial indicators of faecal contamination in the water used in the resort, with the only exception of a sample of filter purified water used both for cooking and as drinking water. When extending the sampling to other areas of the municipal water network, many other sites, mostly on the seaside, showed faecal pollution, which demonstrates a poor protection of the water network from faecal leakage (Fig. 1). NoV contamination of waters generally originates from sewage since these enteric pathogens, as other enteric viruses affecting humans, are mostly species-specific and are shed at high concentrations in the stools of patients. When the analyses on the stools of the students made it clear that the outbreak had been caused by NoV, the viral genome was also searched in appropriate volumes of water samples, but this happened when correction measures on the water network had already been taken. Nevertheless, a single water sample, obtained 2 months after the outbreak (1 August 2016), still contained genome fragments that could be correlated to the same NoV genotype causing the outbreak. Although the finding of NoV genome does not demonstrate that entire infecting viral particles were present in the water, it could indicate that a NoV belonging to the same genotype involved in the outbreak was still circulating in the local population several weeks after the outbreak and it was someway able to penetrate the water network, probably indicating continuing faecal contamination through leakage from sewage systems. NoV is particularly resistant in the environment and also to chlorine, therefore transmission through chlorinated water is not exceptional [9, 12, 26]. Drinking water from municipal supplies fed by lake or groundwater has already been identified in the past as the source of waterborne NoV outbreaks in Italy [27–29] and NoV is now considered one of the pathogens causing the largest number of affected consumers in drinking waterborne outbreaks [12]. NoV can persist in water longer than 15 days [30] and it is resistant to chlorine disinfection when free chlorine levels or pre-treatment processes are inadequate [31, 32]. Moreover, microfiltration filters labelled as certified by NSF Standards 53 or 58 (with a pore size ranging from 0.05 to 5 µm) are not effective in removing viruses and can even decrease chlorine concentration in filtered water [33]. Actually, the owner of the resort involved in our outbreak was encouraged to serve tap water as drinking water at the restaurant relying on the effectiveness of such a filter purifier. Unfortunately, we did not have a tap water sample from the restaurant of the resort taken before correction measures and whose volume was appropriate for virus detection. Therefore, we could not demonstrate the failure to eliminate viral pollution of the microfiltration system. In Mascali, correction measures on the water network consisted in the maintenance of the old chlorination device and in the addition of a new one at a key point located upstream and close to suspected illegal sewage dumping (Fig. 1), but as poor protection of the pipeline from sewage leakage risked to reproduce faecal pollution episodes, renovation of the municipal water system was also advised, as well as systematic control of illegal sewage dumping. The additional chlorination device was apparently the most effective intervention, since right after its activation faecal bacterial pollution indicators disappeared. However, NoV genome was detected in a water sample taken 6 weeks after the last detection of bacterial indicators. Unfortunately, detection of faecal contamination of drinking water is only based on the use of bacterial indicators despite no correlation between levels of enteric bacteria and enteric viruses has been found [34–36]. According to ISGEV enteric virus circulation surveillance data, in 2016 the circulation of GII.P2/GII.2 NoV strains in Palermo, home town of the travellers involved in the outbreak, was the highest ever observed from the starting of local NoV surveillance in 2004. GII.P2/GII.2 was the fourth most common genotype in 2016 (12.2% prevalence), following GII.P4/GII.4 Sydney 2012 (28.4%), GII.P16/GII.2 (18.9%) and GII.Pe/GII.4 Sydney 2012 (17.6%). GII.P2/GII.2 strains were mostly observed starting from July 2016, leaving space for the possibility that their increased circulation was linked to the introduction of the ‘Mascali clone’. However, a couple of GII.P2/GII.2 strains (including PA36/16) genetically close to the ‘Mascali clone’ had been already detected in January 2016 in Palermo and GII.P2/GII.2 NoV had sporadically been observed also in 2011 and 2014. Moreover, the circulation of GII.P2/GII.2 NoVs has been increasing worldwide over the last period [37]. Therefore, the unusually high circulation observed in Palermo might just reflect the global epidemiology of GII.P2/GII.2 strains. Noteworthy, the older GII.P2/GII.2 NoV strains detected in Palermo were closely related in the ORF2 (98.9% nt id) to the GII.P16/GII.2 Nashville strain (KY865307) which is supposed to have generated the recombinant GII.P16/GII.4 clone recently circulating epidemically worldwide [37, 38] and were more distantly associated (95.3% nt id) to the GII.2 capsid sequence (AB662868) of another putative GII.P16/GII.4 ancestor from Japan [39]. The increasing frequency of detection of waterborne NoV outbreaks calls for the introduction of routine monitoring for viruses of drinking waters and for the definition of affordable and reliable indicators for the presence of NoV in drinking waters. The use of ultrafiltration systems as a concentration method in order to recover viruses from environmental waters combined with RT-PCR to detect viral genomes has been demonstrated to grant a recovery rate of >50% of multiple viruses [40] and is at present one of the most sensitive approaches for the detection of NoVs, although the discrimination between viable and nonviable viruses remains a problem. The burden of NoV outbreaks, either food associated or waterborne, is still poorly defined due to the lack of specific diagnostic tools in many clinical laboratories. NoV epidemics should always be suspected when the main symptom is vomiting and the duration of disease is short (<48 h). Waterborne transmission is suggested by the rapid spread of disease and the high attack rate. Care should be taken to perform sensitive diagnostic tests on a sufficient number of stool samples in order to identify NoV as the causative agent of enteritis outbreaks, possibly preferring genome screening over antigen detection. Only the systematic detection of NoV outbreaks can guide in the choice of the safety procedures that could be most appropriate in containing such microbiological hazard.

When did this event start?

1534

Waterborne Norovirus outbreak at a seaside resort likely originating from municipal water distribution system failure

Abstract In May 2016 a Norovirus (NoV) gastroenteritis outbreak involved a high school class visiting a seaside resort near Taormina (Mascali, Sicily). Twenty-four students and a teacher were affected and 17 of them showed symptoms on the second day of the journey, while the others got ill within the following 2 days. Symptoms included vomiting, diarrhoea and fever, and 12 students required hospitalisation. Stool samples tested positive for NoV genome by Real-Time polymerase chain reaction assay in all 25 symptomatic subjects. The GII.P2/GII.2 NoV genotype was linked to the outbreak by ORF1/ORF2 sequence analysis. The epidemiological features of the outbreak were consistent with food/waterborne followed by person-to-person and/ or vomit transmission. Food consumed at a shared lunch on the first day of the trip was associated to illness and drinking un-bottled tap water was also considered as a risk factor. The analysis of water samples revealed the presence of bacterial indicators of faecal contamination in the water used in the resort as well as in other areas of the municipal water network, linking the NoV gastroenteritis outbreak to tap water pollution from sewage leakage. From a single water sample, an amplicon whose sequence corresponded to the capsid genotype recovered from patients could be obtained. Introduction Noroviruses (NoVs) are recognised as a leading cause of sporadic cases and outbreaks of acute gastroenteritis (AGE) across all age groups [1, 2]. They are transmitted primarily through the faecal-oral route either by direct person–person spread and with consumption of contaminated food or water [1, 3, 4]. NoVs are non-enveloped RNA viruses belonging to the Caliciviridae family and are classified into seven distinct genogroups (GI-GVII) on the basis of the major capsid protein VP1 sequence. GI, GII and GIV NoVs infect humans and they have been divided into more than 30 genotypes based on the sequences of polymerase (ORF1) and capsid proteins (ORF2) [5]. NoV GII.4 genotype has been associated with the vast majority of NoV related cases of AGE worldwide although many genotypes, belonging to both GI and GII genogroups are involved in foodborne and waterborne outbreaks [6, 7]. NoVs are highly contagious owing to their high stability in the environment and their ability to infect at low doses. In fact, just 10–100 virions of NoV may be sufficient to cause AGE [8, 9]. NoVs are considered the major cause of water-related disease in semi-closed institutions, including restaurants, schools, hospitals, nursing homes and cruise ships [10–12]. To confirm NoV as a cause of outbreaks, the presence of four epidemiological features of disease, vomiting in more than half of affected persons, a mean incubation period of 24/48 h, a mean duration of illness of 12–60 h and the absence of bacterial pathogens in stool culture, should be evaluated [13]. Waterborne NoV outbreaks can be caused by contaminated drinking water despite correct chlorine concentration and in the absence of coliform bacteria [10, 14–17]. The aim of this study was to investigate an outbreak of NoV gastroenteritis associated with drinking water involving a high school class travelling from Palermo (West Sicily) to Acireale (East Sicily) on an educational tour in May 2016. Stool samples were collected soon after the onset of AGE symptoms to ascertain the aetiology of illness. The source of infection was investigated collecting epidemiological data through an online questionnaire and, in order to explore a possible transmission route, laboratory tests on water from the municipal network were performed. Material and methods Epidemiological investigation From 20 to 22 May 2016 an AGE outbreak occurred in Mascali, a small town near Taormina, involving a school class of 28 students, 13 girls and 15 boys aged between 14 and 18 years and a 54-year-old female teacher, during a trip to participate in a theatre festival in Acireale. A total of 25/29 persons were affected by nausea, vomiting, abdominal cramps, diarrhoea and fever while they were staying at a seaside resort in Mascali, on the way to Acireale. To obtain data on symptoms and the correlated risk factors all participants filled an online questionnaire investigating personal data (age and sex), travel and accommodation details and food consumption over 48 h before symptoms presentation, time of onset and duration of symptoms, therapy and/or hospitalisation required and history of contact with people affected by diarrhoea and/or vomiting (Table 1). The severity of illness was analysed by the Vesikari severity scale allowing to distinguish non-severe (score <11) and severe (score ⩾11) AGE (Table 1). Vesikari score was calculated from maximum number of episodes/day and duration of both diarrhoea and vomiting (score 0–3 for each of the 4 parameters), temperature (score 0–3), level of dehydration (score 0–3) and treatment required (score 0–2) [18]. A case-control study was conducted to establish if food consumed before the onset of symptoms was associated with the disease. A case was defined as a student attending the tour and reporting diarrhoea and/or vomiting within 36 h from the first shared meal during the tour, while students who did not present any of the previous symptoms or those showing the first symptoms with a delay of ⩾48 h (secondary cases) were used as controls. Variables were studied using odds ratio (OR) and its 95% CI and statistical associations were determined using the χ2 test, with the level of statistical significance set at 0.05 (Table 2). NoV detection in stool samples A stool sample was obtained from each of the 29 travellers, soon after (3–5 days) the onset of the outbreak. No stool samples could be obtained from the food-handlers working at the resort. All samples were tested for NoV at the Enteric Viruses Laboratory of the University of Palermo, member of the Italian Study Group for Enteric Viruses (ISGEV), by a Real-Time polymerase chain reaction (RT-PCR) assay that allows to differentiate GI and GII NoVs [19]. Total RNA was extracted from stool samples using QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) and cDNA was generated by MMLV reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were considered positive for NoVs when cycle threshold (ct) of RT-PCR was ⩽30 and lowpositive when 30<ct⩽35. For all NoV-positive samples two genomic fragments were amplified by conventional RT-PCR using Superscript III One-Step RT-PCR system with Platinum Taq DNA polymerase (Invitrogen) and primers JV12a/JV13b, targeting 326 bps of the RdRp region A [20] and either COG1F/ GISKR or COG2F/GIISKR, amplifying a 315 bps 5′portion of GI or GII VP1 protein [19, 21]. Environmental investigation From May through August 2016, a total of 32 tap water samples were collected from the Mascali municipal network. In particular, seven samples were collected at the resort where the outbreak occurred and from its water entry (R), two in the city centre of Mascali (M), 15 in the village of Fondachello (F), on the seaside, three in Carrabba village (C), far from the coast, and two and three, respectively, from the water well ‘Carlino’ (W) and the source ‘Bufardo and Torregrossa’ (S) which feed the municipal system and provide drinking water to the Mascali area (Fig. 1, Table 3). One litre water samples were sent by the local health authorities to the SIALAB laboratory in Avola, Siracusa and 100 ml aliquots were screened for coliforms, E. coli and enterococci as faecal indicators, by standard bacteriological methods according to UNI EN ISO 9308-1:2014. Unfortunately, although the suspect of a water source for the viral outbreak was quickly communicated, only in August and November 2016, following the implementation of control measures, local authorities provided water samplings of appropriate volume to perform viral tests (10 l). Five water samples of 10 l each were collected from different sampling sites in F and R areas for viral screening (Table 3) and were concentrated according to the Italian National Institute of Health (Istituto Superiore di Sanità, ISS) protocols [22] through an ultra-filtration system (Sartoflow® Slice 200 Benchtop Crossflow System, Sartorius AG, Goettingen, Germany), using appropriate membranes (SG Hydrosart 10 kDa) pretreated with 300 ml of 3% Beef Extract (BE) at pH 7, by recirculating it for 10 min in the system, at a pressure not higher than 1.30 bar. Elution and recovery of viruses attached to the membranes were obtained by recirculating 200 ml of 3% BE at pH 9.5 for 3 h until a 10–12 ml volume of eluate was obtained. The pH of the eluate was then checked with litmus paper and if necessary brought to the value of 7 with HCl. Concentrated water samples were simultaneously tested for NoV in three laboratories: Enteric Viruses Laboratory, University of Palermo; Istituto Zooprofilattico Sperimentale della Sicilia (IZS), Palermo; and Istituto Superiore di Sanità (ISS), Rome. Total RNA was extracted from 500 and 560 µl of a concentrated water sample using BioMèrieux NucliSENS® Magnetic Extraction System (BioMèrieux) and QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) respectively, according to manufacturer’s instructions. Bonded RNA was eluted by 100 µl of Elution Buffer. Amplifications were performed at least in duplicate on whole and 1:10 diluted extracts. A Real-Time RT-PCR assay according to ISO/TS 15216- 2:2013, two in-house Real-Time RT-PCR protocols and the One-Step RT-PCR methods already used for amplifying ORF1 and ORF2 fragments in NoV-positive faecal samples were performed [19–21, 23]. Nucleotide sequences analysis and phylogenetic analysis PCR amplicons were purified by commercial columns (CleanSweep PCR Purification, Thermo Fisher Scientific, Vilnius, Lithuania) and sequenced by Sanger method and all nucleotide sequences were aligned using ClustalW. To identify NoV genotypes, RdRp and VP1 sequences were submitted to the RIVM NoV genotyping tool (http://www.rivm.nl/mpf/norovirus/typingtool). Phylogenetic trees were constructed using MEGA 7 and Neighbour-Joining method (bootstrap 1000 replicates). Sequences from stool and water samples are available at NCBI with the following accession numbers: MG517455 for the major capsid protein obtained from a water sample; MG517456 and MG517461, respectively, for the major capsid protein and RNA-dependent-RNA-polymerase, as representatives of the sequences obtained from the stool samples of those involved in the outbreak. Control measures After outbreak notification, on 21 May, the resort was closed for 2 days. On 25 May, following the first results of the bacteriological analyses on municipal water samples, the use of municipal water for human consumption was temporarily restricted (use of water was permitted only for cleaning purposes) in the Carrabba and Fondachello villages and in the area surrounding the resort. From 25 May the reservoir collecting the water from the Carlino well and its chlorination device were submitted to maintenance work and a new chlorination device was installed on the conduit bringing the water from the Bufardo source. Actually, the Bufardo source had been connected to the municipal network in 2003 but unfortunately a dedicated chlorination station had not been planned at that time. Therefore, unchlorinated water entered the system near Fondachello village on the seaside (Fig. 1), possibly diluting chlorine concentrations in water treated at the Carlino reservoir. Moreover, the exact path of the branch of the water network situated downstream the Bufardo source water connection was not marked in the maps of the municipality (Fig. 1, red line), indicating that this section had never been submitted to maintenance. The area surrounding the unmaintained section of the water pipeline was scattered with unauthorised cottages that were not connected to the sewerage systems. Likely, untreated waste from cesspools was unloaded directly into the soil, generating a risk of contamination of the municipal water system through leakage. No work had been made on the water pipelines before the outbreak but sections of the network had been temporarily closed during the winter season on the seaside (Fondachello). Local health authorities advised the connection of all households in the area to the sewerage system and the replacement of the water system section without maintenance. Health authorities also raised attention on AGE cases notification at all local Emergency Departments and Family Practitioners but there was no further evidence of AGE outbreaks in the months following the intervention. Results Data collected through the online questionnaire compiled 7–10 days after the outbreak revealed that none of the 28 students and the teacher taking part in the school trip showed gastroenteritis symptoms at the beginning of the journey and they had no contact with people who suffered from diarrhoea and/or vomiting in the previous days. Travellers had their first shared meal on 19 May (Day 1), 2016, at lunch on their arrival to the resort in Mascali. The list of foods included in the menu is shown in Table 2. Conversely, the dinner on Day 1 was not shared and the students went to dine in three different fast food restaurants in Acireale. On 20 May (Day 2), 17 students (58.6% of the travellers) presented with vomit and/or diarrhoea, falling ill 26–36 h after lunch on Day 1, strongly suggesting that exposure occurred on the first shared meal. Some students also vomited in the presence of the other travellers on the bus bringing them to Acireale. Five more students were symptomatic in the early hours of the following day (21 May, Day 3, from 0:30 am to 8:00 am), as a continuation of the wave of illnesses of the previous day. The teacher and one more student fell ill at 8:00 pm and 11:00 pm of Day 3, respectively, for a second wave of illnesses occurring >48 h from the first shared meal and <36 h from the first cases, with the last student falling sick on 22 May (Day 4) at 3:00 am (Fig. 2, Table 1). On the whole, the mean incubation period was 33.2 h. Based on the online survey, the severity of illness was evaluated by the Vesikari severity scale, grouping symptomatic travellers into severe (11/25 = 44%) and non-severe (14/25 = 56%) cases (Table 1). Vomiting and fever were the predominant symptoms, affecting 22 out of 25 (88%) symptomatic subjects, while diarrhoea was present in 14 out of 25 (56%). Hospitalisation for rehydration therapy was necessary for 13 students (91% [10/11] of severe and 21% [3/14] of non-severe cases). Emergency room (ER) or family practitioner visits were necessary for 21 students and 13 (62%) of them initiated an antibiotic therapy. Two students indicated that other family members had gastroenteric symptoms more than 24 h after their return home, on Day 3, 4 and 5. All those involved in the outbreak recovered within 4 days, with an average duration of illness of 38.4 h. A case-control study of the risk associated with eating foods at lunch on Day 1 or with attending different restaurants at dinner on Day 1 was attempted. A total of 19 subjects filled the section of the online survey concerning food consumption. Fourteen of the responders were among the 22 cases falling ill within 36 h from the first shared meal (first wave), while the remaining five responders were three travellers with supposed secondary infections (showing symptoms with a delay ⩾48 h; second wave) and two asymptomatic students that were used as controls. No statistically significant associations with any food items served at lunch on Day 1 was identified (P < 0.05; Table 2). However, all food consumed on that occasion was to some extent associated with the risk of disease, the highest OR being obtained with chicken (OR: 20.7143, P: 0.0678; Table 2). Actually, most of the cases ate all foods included in the menu, while one of the two asymptomatic controls did not participate in the meal and the other had a pasta dish only. No statistically significant association was found for attending a particular fast food restaurant at the dinner on Day 1 (OR: 2.5, P: 0.3635). The Real-Time RT-PCR assay allowed to detect GII genogroup NoV genome in all the 25 samples from symptomatic subjects, although two of them showed high Ct values, indicating lower shedding of viral genomes (Table 1). The samples from four asymptomatic students were negative for both NoV genogroups. By sequence analysis, a NoV genotype could be defined for 20/ 25 NoV-positive stool samples. All NoV strains from travellers involved in the outbreaks belonged to the same genotype, GII.P2/GII.2, showing >99% nucleotide identities (nt id) in both fragments (ORF1 and ORF2) analysed. Thirty-two water samples were collected and analysed for bacterial contamination (Table 3). Seven samples were taken from the resort (site R) and the remaining 25 from five other sites (C, F, M, S, W), representative of the entire municipal water network. The majority (10/12) of the water samples collected immediately after the onset of the outbreak, between 21 May and 16 June 2016, at the resort (site R) and at sites F and C, located close to the resort, on the seaside and in the interior, respectively, were positive for coliform bacteria. More than 1000 colony-forming units (CFU)/ 100 ml (up to 2700 CFU/100 ml) were detected in four samples, while their absence is required in 100 ml by the Standard of Water Quality in Italy and seven of 11 samples tested were also Escherichia coli-positive (up to 900 CFU/100 ml). Enterococci were searched in 11 samples and three of them were found positive for this faecal contamination indicator. From 7 July to 16 August 2016, after control measures on the water network were undertaken, 13 water samples were collected for follow-up laboratory examination of the water system at F, R and C sites. All of them were negative for bacterial indicators of faecal contamination. No faecal bacteria were identified in the seven samples collected from M, W and S sites, including the well and the source that fed the water network, at any sampling time, before and after implementing the control measures. The research of NoV genome was conducted on concentrated water samples collected on 1 August from R and F sites (Table 3). NoV genome was detected in a single sample collected from the water entry of the resort involved in the outbreak (site R). Actually, only two out of 26 One-Step RT-PCR repetitions were positive for NoV viral capsid gene. The two ORF2-positive results were obtained at the University of Palermo Laboratory using both 1:10 diluted and undiluted extracted RNA. The sequence analysis of the partial ORF2 portion (289 bp) indicated the presence of the same GII.2 capsid genotype detected in the stool samples of the students involved in the outbreak, showing 98.9% nt id. No ORF1 NoV sequence could be obtained from the NoV-positive water sample. Follow-up analyses for NoV genome, conducted on water samples collected on 20 November both at R and F sites found no positive samples. By sequence alignment, the NoV sequences obtained from Mascali outbreak samples were phylogenetically closely related in both ORF1 and ORF2 (99 and 98.9% of nt id, respectively) to a GII.P2/GII.2 strain responsible for a sporadic paediatric gastroenteritis case occurred in Palermo in January 2016 (PA 36/16, accession numbers MG517464 and MG517460, for ORF1 and ORF2, respectively) and previously detected at the Enteric Viruses Laboratory, University of Palermo. Whereas, three GII.P2/GII.2 Italian strains (PA 280/11, PA 366/11 and PA 3/ 14, accession numbers MG517462, MG517465 and MG517463, for ORF1 and MG517457 to −59, for ORF2, respectively), detected in Palermo in 2011 and 2014, were more distantly associated with the outbreak strain both in RdRp and VP1 (90% to 97.4% nt id and 96.1% to 98.1% nt id, respectively, in ORF1 and ORF2). Discussion Epidemiological features, such as the presence of vomiting in more than half (88%) of symptomatic subjects, the 33.2 h average incubation period and the 38.4 h average duration of illness, suggested a NoV aetiology for the investigated outbreak [13]. The detection of NoV genome in all 25 faecal samples from symptomatic subjects fully complies with the requirement of detection of at least one positive sample out of four tested that is considered to be sufficient to ascertain the aetiology of a NoV outbreak [24]. Moreover, molecular typing indicated that a single viral strain, belonging to genotype GII.P2/GII.2, infected all symptomatic subjects involved in the outbreak. NoVs are highly infectious and can easily spread from person to person but also through contaminated food or water. Therefore, settings such as health care institutions, schools and touristic resorts are mainly affected, since close contact and shared food services facilitate transmission. In Mascali, two waves of transmission were observed over three days, being consistent with food/waterborne followed by person-to-person and/or airborne (vomit) transmission (Fig. 2). The first wave of cases presented on 20 May 2016 after a single shared meal, while the last case was detected on 22 May. Exposure to vomit may have been responsible for some of the secondary cases occurred after the first wave of illness. Vomiting produces aerosols that can be inhaled and swallowed. The extremely low infectious dose of NoV makes the inhalation of vomit aerosols an effective transmission route in settings where there is close contact, like on a bus trip [25]. All food consumed in the only shared lunch on 19 May at the resort was associated with the risk of disease (Table 2), but the association was inconclusive since all cases shared the same dishes included in the lunch menu on Day 1. Although many foods were potentially involved, our attention focused on the tap water. In fact, trivial information from local health authorities reported that local population avoided tap water consumption and relied on bottled water for drinking and cooking. Many small outbreaks of gastroenteritis involving restaurants and family groups had been reported in the same area, especially at the beginning of the summer season, when vacation houses and recreational activities were re-opened after a period of low water demand. Illness caused by NoV is common and if the level of water contamination is low, the number of cases remains limited and water can persist as unrecognised source of infection. If polluted water was served for drinking and used for cooking and washing dishes in the Mascali resort at Day 1 a generalised contamination of food and surfaces could have occurred. The potential role of the foodhandlers as a source of contamination could not be analysed in this study but should also be taken into account. If water contamination has already been present in the area for a long time, infections may have occurred among the local food-handlers. The analysis of water samples taken immediately after the outbreak revealed the presence of bacterial indicators of faecal contamination in the water used in the resort, with the only exception of a sample of filter purified water used both for cooking and as drinking water. When extending the sampling to other areas of the municipal water network, many other sites, mostly on the seaside, showed faecal pollution, which demonstrates a poor protection of the water network from faecal leakage (Fig. 1). NoV contamination of waters generally originates from sewage since these enteric pathogens, as other enteric viruses affecting humans, are mostly species-specific and are shed at high concentrations in the stools of patients. When the analyses on the stools of the students made it clear that the outbreak had been caused by NoV, the viral genome was also searched in appropriate volumes of water samples, but this happened when correction measures on the water network had already been taken. Nevertheless, a single water sample, obtained 2 months after the outbreak (1 August 2016), still contained genome fragments that could be correlated to the same NoV genotype causing the outbreak. Although the finding of NoV genome does not demonstrate that entire infecting viral particles were present in the water, it could indicate that a NoV belonging to the same genotype involved in the outbreak was still circulating in the local population several weeks after the outbreak and it was someway able to penetrate the water network, probably indicating continuing faecal contamination through leakage from sewage systems. NoV is particularly resistant in the environment and also to chlorine, therefore transmission through chlorinated water is not exceptional [9, 12, 26]. Drinking water from municipal supplies fed by lake or groundwater has already been identified in the past as the source of waterborne NoV outbreaks in Italy [27–29] and NoV is now considered one of the pathogens causing the largest number of affected consumers in drinking waterborne outbreaks [12]. NoV can persist in water longer than 15 days [30] and it is resistant to chlorine disinfection when free chlorine levels or pre-treatment processes are inadequate [31, 32]. Moreover, microfiltration filters labelled as certified by NSF Standards 53 or 58 (with a pore size ranging from 0.05 to 5 µm) are not effective in removing viruses and can even decrease chlorine concentration in filtered water [33]. Actually, the owner of the resort involved in our outbreak was encouraged to serve tap water as drinking water at the restaurant relying on the effectiveness of such a filter purifier. Unfortunately, we did not have a tap water sample from the restaurant of the resort taken before correction measures and whose volume was appropriate for virus detection. Therefore, we could not demonstrate the failure to eliminate viral pollution of the microfiltration system. In Mascali, correction measures on the water network consisted in the maintenance of the old chlorination device and in the addition of a new one at a key point located upstream and close to suspected illegal sewage dumping (Fig. 1), but as poor protection of the pipeline from sewage leakage risked to reproduce faecal pollution episodes, renovation of the municipal water system was also advised, as well as systematic control of illegal sewage dumping. The additional chlorination device was apparently the most effective intervention, since right after its activation faecal bacterial pollution indicators disappeared. However, NoV genome was detected in a water sample taken 6 weeks after the last detection of bacterial indicators. Unfortunately, detection of faecal contamination of drinking water is only based on the use of bacterial indicators despite no correlation between levels of enteric bacteria and enteric viruses has been found [34–36]. According to ISGEV enteric virus circulation surveillance data, in 2016 the circulation of GII.P2/GII.2 NoV strains in Palermo, home town of the travellers involved in the outbreak, was the highest ever observed from the starting of local NoV surveillance in 2004. GII.P2/GII.2 was the fourth most common genotype in 2016 (12.2% prevalence), following GII.P4/GII.4 Sydney 2012 (28.4%), GII.P16/GII.2 (18.9%) and GII.Pe/GII.4 Sydney 2012 (17.6%). GII.P2/GII.2 strains were mostly observed starting from July 2016, leaving space for the possibility that their increased circulation was linked to the introduction of the ‘Mascali clone’. However, a couple of GII.P2/GII.2 strains (including PA36/16) genetically close to the ‘Mascali clone’ had been already detected in January 2016 in Palermo and GII.P2/GII.2 NoV had sporadically been observed also in 2011 and 2014. Moreover, the circulation of GII.P2/GII.2 NoVs has been increasing worldwide over the last period [37]. Therefore, the unusually high circulation observed in Palermo might just reflect the global epidemiology of GII.P2/GII.2 strains. Noteworthy, the older GII.P2/GII.2 NoV strains detected in Palermo were closely related in the ORF2 (98.9% nt id) to the GII.P16/GII.2 Nashville strain (KY865307) which is supposed to have generated the recombinant GII.P16/GII.4 clone recently circulating epidemically worldwide [37, 38] and were more distantly associated (95.3% nt id) to the GII.2 capsid sequence (AB662868) of another putative GII.P16/GII.4 ancestor from Japan [39]. The increasing frequency of detection of waterborne NoV outbreaks calls for the introduction of routine monitoring for viruses of drinking waters and for the definition of affordable and reliable indicators for the presence of NoV in drinking waters. The use of ultrafiltration systems as a concentration method in order to recover viruses from environmental waters combined with RT-PCR to detect viral genomes has been demonstrated to grant a recovery rate of >50% of multiple viruses [40] and is at present one of the most sensitive approaches for the detection of NoVs, although the discrimination between viable and nonviable viruses remains a problem. The burden of NoV outbreaks, either food associated or waterborne, is still poorly defined due to the lack of specific diagnostic tools in many clinical laboratories. NoV epidemics should always be suspected when the main symptom is vomiting and the duration of disease is short (<48 h). Waterborne transmission is suggested by the rapid spread of disease and the high attack rate. Care should be taken to perform sensitive diagnostic tests on a sufficient number of stool samples in order to identify NoV as the causative agent of enteritis outbreaks, possibly preferring genome screening over antigen detection. Only the systematic detection of NoV outbreaks can guide in the choice of the safety procedures that could be most appropriate in containing such microbiological hazard.

What is the date of this event?

1535

Waterborne Norovirus outbreak at a seaside resort likely originating from municipal water distribution system failure

Abstract In May 2016 a Norovirus (NoV) gastroenteritis outbreak involved a high school class visiting a seaside resort near Taormina (Mascali, Sicily). Twenty-four students and a teacher were affected and 17 of them showed symptoms on the second day of the journey, while the others got ill within the following 2 days. Symptoms included vomiting, diarrhoea and fever, and 12 students required hospitalisation. Stool samples tested positive for NoV genome by Real-Time polymerase chain reaction assay in all 25 symptomatic subjects. The GII.P2/GII.2 NoV genotype was linked to the outbreak by ORF1/ORF2 sequence analysis. The epidemiological features of the outbreak were consistent with food/waterborne followed by person-to-person and/ or vomit transmission. Food consumed at a shared lunch on the first day of the trip was associated to illness and drinking un-bottled tap water was also considered as a risk factor. The analysis of water samples revealed the presence of bacterial indicators of faecal contamination in the water used in the resort as well as in other areas of the municipal water network, linking the NoV gastroenteritis outbreak to tap water pollution from sewage leakage. From a single water sample, an amplicon whose sequence corresponded to the capsid genotype recovered from patients could be obtained. Introduction Noroviruses (NoVs) are recognised as a leading cause of sporadic cases and outbreaks of acute gastroenteritis (AGE) across all age groups [1, 2]. They are transmitted primarily through the faecal-oral route either by direct person–person spread and with consumption of contaminated food or water [1, 3, 4]. NoVs are non-enveloped RNA viruses belonging to the Caliciviridae family and are classified into seven distinct genogroups (GI-GVII) on the basis of the major capsid protein VP1 sequence. GI, GII and GIV NoVs infect humans and they have been divided into more than 30 genotypes based on the sequences of polymerase (ORF1) and capsid proteins (ORF2) [5]. NoV GII.4 genotype has been associated with the vast majority of NoV related cases of AGE worldwide although many genotypes, belonging to both GI and GII genogroups are involved in foodborne and waterborne outbreaks [6, 7]. NoVs are highly contagious owing to their high stability in the environment and their ability to infect at low doses. In fact, just 10–100 virions of NoV may be sufficient to cause AGE [8, 9]. NoVs are considered the major cause of water-related disease in semi-closed institutions, including restaurants, schools, hospitals, nursing homes and cruise ships [10–12]. To confirm NoV as a cause of outbreaks, the presence of four epidemiological features of disease, vomiting in more than half of affected persons, a mean incubation period of 24/48 h, a mean duration of illness of 12–60 h and the absence of bacterial pathogens in stool culture, should be evaluated [13]. Waterborne NoV outbreaks can be caused by contaminated drinking water despite correct chlorine concentration and in the absence of coliform bacteria [10, 14–17]. The aim of this study was to investigate an outbreak of NoV gastroenteritis associated with drinking water involving a high school class travelling from Palermo (West Sicily) to Acireale (East Sicily) on an educational tour in May 2016. Stool samples were collected soon after the onset of AGE symptoms to ascertain the aetiology of illness. The source of infection was investigated collecting epidemiological data through an online questionnaire and, in order to explore a possible transmission route, laboratory tests on water from the municipal network were performed. Material and methods Epidemiological investigation From 20 to 22 May 2016 an AGE outbreak occurred in Mascali, a small town near Taormina, involving a school class of 28 students, 13 girls and 15 boys aged between 14 and 18 years and a 54-year-old female teacher, during a trip to participate in a theatre festival in Acireale. A total of 25/29 persons were affected by nausea, vomiting, abdominal cramps, diarrhoea and fever while they were staying at a seaside resort in Mascali, on the way to Acireale. To obtain data on symptoms and the correlated risk factors all participants filled an online questionnaire investigating personal data (age and sex), travel and accommodation details and food consumption over 48 h before symptoms presentation, time of onset and duration of symptoms, therapy and/or hospitalisation required and history of contact with people affected by diarrhoea and/or vomiting (Table 1). The severity of illness was analysed by the Vesikari severity scale allowing to distinguish non-severe (score <11) and severe (score ⩾11) AGE (Table 1). Vesikari score was calculated from maximum number of episodes/day and duration of both diarrhoea and vomiting (score 0–3 for each of the 4 parameters), temperature (score 0–3), level of dehydration (score 0–3) and treatment required (score 0–2) [18]. A case-control study was conducted to establish if food consumed before the onset of symptoms was associated with the disease. A case was defined as a student attending the tour and reporting diarrhoea and/or vomiting within 36 h from the first shared meal during the tour, while students who did not present any of the previous symptoms or those showing the first symptoms with a delay of ⩾48 h (secondary cases) were used as controls. Variables were studied using odds ratio (OR) and its 95% CI and statistical associations were determined using the χ2 test, with the level of statistical significance set at 0.05 (Table 2). NoV detection in stool samples A stool sample was obtained from each of the 29 travellers, soon after (3–5 days) the onset of the outbreak. No stool samples could be obtained from the food-handlers working at the resort. All samples were tested for NoV at the Enteric Viruses Laboratory of the University of Palermo, member of the Italian Study Group for Enteric Viruses (ISGEV), by a Real-Time polymerase chain reaction (RT-PCR) assay that allows to differentiate GI and GII NoVs [19]. Total RNA was extracted from stool samples using QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) and cDNA was generated by MMLV reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were considered positive for NoVs when cycle threshold (ct) of RT-PCR was ⩽30 and lowpositive when 30<ct⩽35. For all NoV-positive samples two genomic fragments were amplified by conventional RT-PCR using Superscript III One-Step RT-PCR system with Platinum Taq DNA polymerase (Invitrogen) and primers JV12a/JV13b, targeting 326 bps of the RdRp region A [20] and either COG1F/ GISKR or COG2F/GIISKR, amplifying a 315 bps 5′portion of GI or GII VP1 protein [19, 21]. Environmental investigation From May through August 2016, a total of 32 tap water samples were collected from the Mascali municipal network. In particular, seven samples were collected at the resort where the outbreak occurred and from its water entry (R), two in the city centre of Mascali (M), 15 in the village of Fondachello (F), on the seaside, three in Carrabba village (C), far from the coast, and two and three, respectively, from the water well ‘Carlino’ (W) and the source ‘Bufardo and Torregrossa’ (S) which feed the municipal system and provide drinking water to the Mascali area (Fig. 1, Table 3). One litre water samples were sent by the local health authorities to the SIALAB laboratory in Avola, Siracusa and 100 ml aliquots were screened for coliforms, E. coli and enterococci as faecal indicators, by standard bacteriological methods according to UNI EN ISO 9308-1:2014. Unfortunately, although the suspect of a water source for the viral outbreak was quickly communicated, only in August and November 2016, following the implementation of control measures, local authorities provided water samplings of appropriate volume to perform viral tests (10 l). Five water samples of 10 l each were collected from different sampling sites in F and R areas for viral screening (Table 3) and were concentrated according to the Italian National Institute of Health (Istituto Superiore di Sanità, ISS) protocols [22] through an ultra-filtration system (Sartoflow® Slice 200 Benchtop Crossflow System, Sartorius AG, Goettingen, Germany), using appropriate membranes (SG Hydrosart 10 kDa) pretreated with 300 ml of 3% Beef Extract (BE) at pH 7, by recirculating it for 10 min in the system, at a pressure not higher than 1.30 bar. Elution and recovery of viruses attached to the membranes were obtained by recirculating 200 ml of 3% BE at pH 9.5 for 3 h until a 10–12 ml volume of eluate was obtained. The pH of the eluate was then checked with litmus paper and if necessary brought to the value of 7 with HCl. Concentrated water samples were simultaneously tested for NoV in three laboratories: Enteric Viruses Laboratory, University of Palermo; Istituto Zooprofilattico Sperimentale della Sicilia (IZS), Palermo; and Istituto Superiore di Sanità (ISS), Rome. Total RNA was extracted from 500 and 560 µl of a concentrated water sample using BioMèrieux NucliSENS® Magnetic Extraction System (BioMèrieux) and QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) respectively, according to manufacturer’s instructions. Bonded RNA was eluted by 100 µl of Elution Buffer. Amplifications were performed at least in duplicate on whole and 1:10 diluted extracts. A Real-Time RT-PCR assay according to ISO/TS 15216- 2:2013, two in-house Real-Time RT-PCR protocols and the One-Step RT-PCR methods already used for amplifying ORF1 and ORF2 fragments in NoV-positive faecal samples were performed [19–21, 23]. Nucleotide sequences analysis and phylogenetic analysis PCR amplicons were purified by commercial columns (CleanSweep PCR Purification, Thermo Fisher Scientific, Vilnius, Lithuania) and sequenced by Sanger method and all nucleotide sequences were aligned using ClustalW. To identify NoV genotypes, RdRp and VP1 sequences were submitted to the RIVM NoV genotyping tool (http://www.rivm.nl/mpf/norovirus/typingtool). Phylogenetic trees were constructed using MEGA 7 and Neighbour-Joining method (bootstrap 1000 replicates). Sequences from stool and water samples are available at NCBI with the following accession numbers: MG517455 for the major capsid protein obtained from a water sample; MG517456 and MG517461, respectively, for the major capsid protein and RNA-dependent-RNA-polymerase, as representatives of the sequences obtained from the stool samples of those involved in the outbreak. Control measures After outbreak notification, on 21 May, the resort was closed for 2 days. On 25 May, following the first results of the bacteriological analyses on municipal water samples, the use of municipal water for human consumption was temporarily restricted (use of water was permitted only for cleaning purposes) in the Carrabba and Fondachello villages and in the area surrounding the resort. From 25 May the reservoir collecting the water from the Carlino well and its chlorination device were submitted to maintenance work and a new chlorination device was installed on the conduit bringing the water from the Bufardo source. Actually, the Bufardo source had been connected to the municipal network in 2003 but unfortunately a dedicated chlorination station had not been planned at that time. Therefore, unchlorinated water entered the system near Fondachello village on the seaside (Fig. 1), possibly diluting chlorine concentrations in water treated at the Carlino reservoir. Moreover, the exact path of the branch of the water network situated downstream the Bufardo source water connection was not marked in the maps of the municipality (Fig. 1, red line), indicating that this section had never been submitted to maintenance. The area surrounding the unmaintained section of the water pipeline was scattered with unauthorised cottages that were not connected to the sewerage systems. Likely, untreated waste from cesspools was unloaded directly into the soil, generating a risk of contamination of the municipal water system through leakage. No work had been made on the water pipelines before the outbreak but sections of the network had been temporarily closed during the winter season on the seaside (Fondachello). Local health authorities advised the connection of all households in the area to the sewerage system and the replacement of the water system section without maintenance. Health authorities also raised attention on AGE cases notification at all local Emergency Departments and Family Practitioners but there was no further evidence of AGE outbreaks in the months following the intervention. Results Data collected through the online questionnaire compiled 7–10 days after the outbreak revealed that none of the 28 students and the teacher taking part in the school trip showed gastroenteritis symptoms at the beginning of the journey and they had no contact with people who suffered from diarrhoea and/or vomiting in the previous days. Travellers had their first shared meal on 19 May (Day 1), 2016, at lunch on their arrival to the resort in Mascali. The list of foods included in the menu is shown in Table 2. Conversely, the dinner on Day 1 was not shared and the students went to dine in three different fast food restaurants in Acireale. On 20 May (Day 2), 17 students (58.6% of the travellers) presented with vomit and/or diarrhoea, falling ill 26–36 h after lunch on Day 1, strongly suggesting that exposure occurred on the first shared meal. Some students also vomited in the presence of the other travellers on the bus bringing them to Acireale. Five more students were symptomatic in the early hours of the following day (21 May, Day 3, from 0:30 am to 8:00 am), as a continuation of the wave of illnesses of the previous day. The teacher and one more student fell ill at 8:00 pm and 11:00 pm of Day 3, respectively, for a second wave of illnesses occurring >48 h from the first shared meal and <36 h from the first cases, with the last student falling sick on 22 May (Day 4) at 3:00 am (Fig. 2, Table 1). On the whole, the mean incubation period was 33.2 h. Based on the online survey, the severity of illness was evaluated by the Vesikari severity scale, grouping symptomatic travellers into severe (11/25 = 44%) and non-severe (14/25 = 56%) cases (Table 1). Vomiting and fever were the predominant symptoms, affecting 22 out of 25 (88%) symptomatic subjects, while diarrhoea was present in 14 out of 25 (56%). Hospitalisation for rehydration therapy was necessary for 13 students (91% [10/11] of severe and 21% [3/14] of non-severe cases). Emergency room (ER) or family practitioner visits were necessary for 21 students and 13 (62%) of them initiated an antibiotic therapy. Two students indicated that other family members had gastroenteric symptoms more than 24 h after their return home, on Day 3, 4 and 5. All those involved in the outbreak recovered within 4 days, with an average duration of illness of 38.4 h. A case-control study of the risk associated with eating foods at lunch on Day 1 or with attending different restaurants at dinner on Day 1 was attempted. A total of 19 subjects filled the section of the online survey concerning food consumption. Fourteen of the responders were among the 22 cases falling ill within 36 h from the first shared meal (first wave), while the remaining five responders were three travellers with supposed secondary infections (showing symptoms with a delay ⩾48 h; second wave) and two asymptomatic students that were used as controls. No statistically significant associations with any food items served at lunch on Day 1 was identified (P < 0.05; Table 2). However, all food consumed on that occasion was to some extent associated with the risk of disease, the highest OR being obtained with chicken (OR: 20.7143, P: 0.0678; Table 2). Actually, most of the cases ate all foods included in the menu, while one of the two asymptomatic controls did not participate in the meal and the other had a pasta dish only. No statistically significant association was found for attending a particular fast food restaurant at the dinner on Day 1 (OR: 2.5, P: 0.3635). The Real-Time RT-PCR assay allowed to detect GII genogroup NoV genome in all the 25 samples from symptomatic subjects, although two of them showed high Ct values, indicating lower shedding of viral genomes (Table 1). The samples from four asymptomatic students were negative for both NoV genogroups. By sequence analysis, a NoV genotype could be defined for 20/ 25 NoV-positive stool samples. All NoV strains from travellers involved in the outbreaks belonged to the same genotype, GII.P2/GII.2, showing >99% nucleotide identities (nt id) in both fragments (ORF1 and ORF2) analysed. Thirty-two water samples were collected and analysed for bacterial contamination (Table 3). Seven samples were taken from the resort (site R) and the remaining 25 from five other sites (C, F, M, S, W), representative of the entire municipal water network. The majority (10/12) of the water samples collected immediately after the onset of the outbreak, between 21 May and 16 June 2016, at the resort (site R) and at sites F and C, located close to the resort, on the seaside and in the interior, respectively, were positive for coliform bacteria. More than 1000 colony-forming units (CFU)/ 100 ml (up to 2700 CFU/100 ml) were detected in four samples, while their absence is required in 100 ml by the Standard of Water Quality in Italy and seven of 11 samples tested were also Escherichia coli-positive (up to 900 CFU/100 ml). Enterococci were searched in 11 samples and three of them were found positive for this faecal contamination indicator. From 7 July to 16 August 2016, after control measures on the water network were undertaken, 13 water samples were collected for follow-up laboratory examination of the water system at F, R and C sites. All of them were negative for bacterial indicators of faecal contamination. No faecal bacteria were identified in the seven samples collected from M, W and S sites, including the well and the source that fed the water network, at any sampling time, before and after implementing the control measures. The research of NoV genome was conducted on concentrated water samples collected on 1 August from R and F sites (Table 3). NoV genome was detected in a single sample collected from the water entry of the resort involved in the outbreak (site R). Actually, only two out of 26 One-Step RT-PCR repetitions were positive for NoV viral capsid gene. The two ORF2-positive results were obtained at the University of Palermo Laboratory using both 1:10 diluted and undiluted extracted RNA. The sequence analysis of the partial ORF2 portion (289 bp) indicated the presence of the same GII.2 capsid genotype detected in the stool samples of the students involved in the outbreak, showing 98.9% nt id. No ORF1 NoV sequence could be obtained from the NoV-positive water sample. Follow-up analyses for NoV genome, conducted on water samples collected on 20 November both at R and F sites found no positive samples. By sequence alignment, the NoV sequences obtained from Mascali outbreak samples were phylogenetically closely related in both ORF1 and ORF2 (99 and 98.9% of nt id, respectively) to a GII.P2/GII.2 strain responsible for a sporadic paediatric gastroenteritis case occurred in Palermo in January 2016 (PA 36/16, accession numbers MG517464 and MG517460, for ORF1 and ORF2, respectively) and previously detected at the Enteric Viruses Laboratory, University of Palermo. Whereas, three GII.P2/GII.2 Italian strains (PA 280/11, PA 366/11 and PA 3/ 14, accession numbers MG517462, MG517465 and MG517463, for ORF1 and MG517457 to −59, for ORF2, respectively), detected in Palermo in 2011 and 2014, were more distantly associated with the outbreak strain both in RdRp and VP1 (90% to 97.4% nt id and 96.1% to 98.1% nt id, respectively, in ORF1 and ORF2). Discussion Epidemiological features, such as the presence of vomiting in more than half (88%) of symptomatic subjects, the 33.2 h average incubation period and the 38.4 h average duration of illness, suggested a NoV aetiology for the investigated outbreak [13]. The detection of NoV genome in all 25 faecal samples from symptomatic subjects fully complies with the requirement of detection of at least one positive sample out of four tested that is considered to be sufficient to ascertain the aetiology of a NoV outbreak [24]. Moreover, molecular typing indicated that a single viral strain, belonging to genotype GII.P2/GII.2, infected all symptomatic subjects involved in the outbreak. NoVs are highly infectious and can easily spread from person to person but also through contaminated food or water. Therefore, settings such as health care institutions, schools and touristic resorts are mainly affected, since close contact and shared food services facilitate transmission. In Mascali, two waves of transmission were observed over three days, being consistent with food/waterborne followed by person-to-person and/or airborne (vomit) transmission (Fig. 2). The first wave of cases presented on 20 May 2016 after a single shared meal, while the last case was detected on 22 May. Exposure to vomit may have been responsible for some of the secondary cases occurred after the first wave of illness. Vomiting produces aerosols that can be inhaled and swallowed. The extremely low infectious dose of NoV makes the inhalation of vomit aerosols an effective transmission route in settings where there is close contact, like on a bus trip [25]. All food consumed in the only shared lunch on 19 May at the resort was associated with the risk of disease (Table 2), but the association was inconclusive since all cases shared the same dishes included in the lunch menu on Day 1. Although many foods were potentially involved, our attention focused on the tap water. In fact, trivial information from local health authorities reported that local population avoided tap water consumption and relied on bottled water for drinking and cooking. Many small outbreaks of gastroenteritis involving restaurants and family groups had been reported in the same area, especially at the beginning of the summer season, when vacation houses and recreational activities were re-opened after a period of low water demand. Illness caused by NoV is common and if the level of water contamination is low, the number of cases remains limited and water can persist as unrecognised source of infection. If polluted water was served for drinking and used for cooking and washing dishes in the Mascali resort at Day 1 a generalised contamination of food and surfaces could have occurred. The potential role of the foodhandlers as a source of contamination could not be analysed in this study but should also be taken into account. If water contamination has already been present in the area for a long time, infections may have occurred among the local food-handlers. The analysis of water samples taken immediately after the outbreak revealed the presence of bacterial indicators of faecal contamination in the water used in the resort, with the only exception of a sample of filter purified water used both for cooking and as drinking water. When extending the sampling to other areas of the municipal water network, many other sites, mostly on the seaside, showed faecal pollution, which demonstrates a poor protection of the water network from faecal leakage (Fig. 1). NoV contamination of waters generally originates from sewage since these enteric pathogens, as other enteric viruses affecting humans, are mostly species-specific and are shed at high concentrations in the stools of patients. When the analyses on the stools of the students made it clear that the outbreak had been caused by NoV, the viral genome was also searched in appropriate volumes of water samples, but this happened when correction measures on the water network had already been taken. Nevertheless, a single water sample, obtained 2 months after the outbreak (1 August 2016), still contained genome fragments that could be correlated to the same NoV genotype causing the outbreak. Although the finding of NoV genome does not demonstrate that entire infecting viral particles were present in the water, it could indicate that a NoV belonging to the same genotype involved in the outbreak was still circulating in the local population several weeks after the outbreak and it was someway able to penetrate the water network, probably indicating continuing faecal contamination through leakage from sewage systems. NoV is particularly resistant in the environment and also to chlorine, therefore transmission through chlorinated water is not exceptional [9, 12, 26]. Drinking water from municipal supplies fed by lake or groundwater has already been identified in the past as the source of waterborne NoV outbreaks in Italy [27–29] and NoV is now considered one of the pathogens causing the largest number of affected consumers in drinking waterborne outbreaks [12]. NoV can persist in water longer than 15 days [30] and it is resistant to chlorine disinfection when free chlorine levels or pre-treatment processes are inadequate [31, 32]. Moreover, microfiltration filters labelled as certified by NSF Standards 53 or 58 (with a pore size ranging from 0.05 to 5 µm) are not effective in removing viruses and can even decrease chlorine concentration in filtered water [33]. Actually, the owner of the resort involved in our outbreak was encouraged to serve tap water as drinking water at the restaurant relying on the effectiveness of such a filter purifier. Unfortunately, we did not have a tap water sample from the restaurant of the resort taken before correction measures and whose volume was appropriate for virus detection. Therefore, we could not demonstrate the failure to eliminate viral pollution of the microfiltration system. In Mascali, correction measures on the water network consisted in the maintenance of the old chlorination device and in the addition of a new one at a key point located upstream and close to suspected illegal sewage dumping (Fig. 1), but as poor protection of the pipeline from sewage leakage risked to reproduce faecal pollution episodes, renovation of the municipal water system was also advised, as well as systematic control of illegal sewage dumping. The additional chlorination device was apparently the most effective intervention, since right after its activation faecal bacterial pollution indicators disappeared. However, NoV genome was detected in a water sample taken 6 weeks after the last detection of bacterial indicators. Unfortunately, detection of faecal contamination of drinking water is only based on the use of bacterial indicators despite no correlation between levels of enteric bacteria and enteric viruses has been found [34–36]. According to ISGEV enteric virus circulation surveillance data, in 2016 the circulation of GII.P2/GII.2 NoV strains in Palermo, home town of the travellers involved in the outbreak, was the highest ever observed from the starting of local NoV surveillance in 2004. GII.P2/GII.2 was the fourth most common genotype in 2016 (12.2% prevalence), following GII.P4/GII.4 Sydney 2012 (28.4%), GII.P16/GII.2 (18.9%) and GII.Pe/GII.4 Sydney 2012 (17.6%). GII.P2/GII.2 strains were mostly observed starting from July 2016, leaving space for the possibility that their increased circulation was linked to the introduction of the ‘Mascali clone’. However, a couple of GII.P2/GII.2 strains (including PA36/16) genetically close to the ‘Mascali clone’ had been already detected in January 2016 in Palermo and GII.P2/GII.2 NoV had sporadically been observed also in 2011 and 2014. Moreover, the circulation of GII.P2/GII.2 NoVs has been increasing worldwide over the last period [37]. Therefore, the unusually high circulation observed in Palermo might just reflect the global epidemiology of GII.P2/GII.2 strains. Noteworthy, the older GII.P2/GII.2 NoV strains detected in Palermo were closely related in the ORF2 (98.9% nt id) to the GII.P16/GII.2 Nashville strain (KY865307) which is supposed to have generated the recombinant GII.P16/GII.4 clone recently circulating epidemically worldwide [37, 38] and were more distantly associated (95.3% nt id) to the GII.2 capsid sequence (AB662868) of another putative GII.P16/GII.4 ancestor from Japan [39]. The increasing frequency of detection of waterborne NoV outbreaks calls for the introduction of routine monitoring for viruses of drinking waters and for the definition of affordable and reliable indicators for the presence of NoV in drinking waters. The use of ultrafiltration systems as a concentration method in order to recover viruses from environmental waters combined with RT-PCR to detect viral genomes has been demonstrated to grant a recovery rate of >50% of multiple viruses [40] and is at present one of the most sensitive approaches for the detection of NoVs, although the discrimination between viable and nonviable viruses remains a problem. The burden of NoV outbreaks, either food associated or waterborne, is still poorly defined due to the lack of specific diagnostic tools in many clinical laboratories. NoV epidemics should always be suspected when the main symptom is vomiting and the duration of disease is short (<48 h). Waterborne transmission is suggested by the rapid spread of disease and the high attack rate. Care should be taken to perform sensitive diagnostic tests on a sufficient number of stool samples in order to identify NoV as the causative agent of enteritis outbreaks, possibly preferring genome screening over antigen detection. Only the systematic detection of NoV outbreaks can guide in the choice of the safety procedures that could be most appropriate in containing such microbiological hazard.

How long was the event?

1536

Waterborne Norovirus outbreak at a seaside resort likely originating from municipal water distribution system failure

Abstract In May 2016 a Norovirus (NoV) gastroenteritis outbreak involved a high school class visiting a seaside resort near Taormina (Mascali, Sicily). Twenty-four students and a teacher were affected and 17 of them showed symptoms on the second day of the journey, while the others got ill within the following 2 days. Symptoms included vomiting, diarrhoea and fever, and 12 students required hospitalisation. Stool samples tested positive for NoV genome by Real-Time polymerase chain reaction assay in all 25 symptomatic subjects. The GII.P2/GII.2 NoV genotype was linked to the outbreak by ORF1/ORF2 sequence analysis. The epidemiological features of the outbreak were consistent with food/waterborne followed by person-to-person and/ or vomit transmission. Food consumed at a shared lunch on the first day of the trip was associated to illness and drinking un-bottled tap water was also considered as a risk factor. The analysis of water samples revealed the presence of bacterial indicators of faecal contamination in the water used in the resort as well as in other areas of the municipal water network, linking the NoV gastroenteritis outbreak to tap water pollution from sewage leakage. From a single water sample, an amplicon whose sequence corresponded to the capsid genotype recovered from patients could be obtained. Introduction Noroviruses (NoVs) are recognised as a leading cause of sporadic cases and outbreaks of acute gastroenteritis (AGE) across all age groups [1, 2]. They are transmitted primarily through the faecal-oral route either by direct person–person spread and with consumption of contaminated food or water [1, 3, 4]. NoVs are non-enveloped RNA viruses belonging to the Caliciviridae family and are classified into seven distinct genogroups (GI-GVII) on the basis of the major capsid protein VP1 sequence. GI, GII and GIV NoVs infect humans and they have been divided into more than 30 genotypes based on the sequences of polymerase (ORF1) and capsid proteins (ORF2) [5]. NoV GII.4 genotype has been associated with the vast majority of NoV related cases of AGE worldwide although many genotypes, belonging to both GI and GII genogroups are involved in foodborne and waterborne outbreaks [6, 7]. NoVs are highly contagious owing to their high stability in the environment and their ability to infect at low doses. In fact, just 10–100 virions of NoV may be sufficient to cause AGE [8, 9]. NoVs are considered the major cause of water-related disease in semi-closed institutions, including restaurants, schools, hospitals, nursing homes and cruise ships [10–12]. To confirm NoV as a cause of outbreaks, the presence of four epidemiological features of disease, vomiting in more than half of affected persons, a mean incubation period of 24/48 h, a mean duration of illness of 12–60 h and the absence of bacterial pathogens in stool culture, should be evaluated [13]. Waterborne NoV outbreaks can be caused by contaminated drinking water despite correct chlorine concentration and in the absence of coliform bacteria [10, 14–17]. The aim of this study was to investigate an outbreak of NoV gastroenteritis associated with drinking water involving a high school class travelling from Palermo (West Sicily) to Acireale (East Sicily) on an educational tour in May 2016. Stool samples were collected soon after the onset of AGE symptoms to ascertain the aetiology of illness. The source of infection was investigated collecting epidemiological data through an online questionnaire and, in order to explore a possible transmission route, laboratory tests on water from the municipal network were performed. Material and methods Epidemiological investigation From 20 to 22 May 2016 an AGE outbreak occurred in Mascali, a small town near Taormina, involving a school class of 28 students, 13 girls and 15 boys aged between 14 and 18 years and a 54-year-old female teacher, during a trip to participate in a theatre festival in Acireale. A total of 25/29 persons were affected by nausea, vomiting, abdominal cramps, diarrhoea and fever while they were staying at a seaside resort in Mascali, on the way to Acireale. To obtain data on symptoms and the correlated risk factors all participants filled an online questionnaire investigating personal data (age and sex), travel and accommodation details and food consumption over 48 h before symptoms presentation, time of onset and duration of symptoms, therapy and/or hospitalisation required and history of contact with people affected by diarrhoea and/or vomiting (Table 1). The severity of illness was analysed by the Vesikari severity scale allowing to distinguish non-severe (score <11) and severe (score ⩾11) AGE (Table 1). Vesikari score was calculated from maximum number of episodes/day and duration of both diarrhoea and vomiting (score 0–3 for each of the 4 parameters), temperature (score 0–3), level of dehydration (score 0–3) and treatment required (score 0–2) [18]. A case-control study was conducted to establish if food consumed before the onset of symptoms was associated with the disease. A case was defined as a student attending the tour and reporting diarrhoea and/or vomiting within 36 h from the first shared meal during the tour, while students who did not present any of the previous symptoms or those showing the first symptoms with a delay of ⩾48 h (secondary cases) were used as controls. Variables were studied using odds ratio (OR) and its 95% CI and statistical associations were determined using the χ2 test, with the level of statistical significance set at 0.05 (Table 2). NoV detection in stool samples A stool sample was obtained from each of the 29 travellers, soon after (3–5 days) the onset of the outbreak. No stool samples could be obtained from the food-handlers working at the resort. All samples were tested for NoV at the Enteric Viruses Laboratory of the University of Palermo, member of the Italian Study Group for Enteric Viruses (ISGEV), by a Real-Time polymerase chain reaction (RT-PCR) assay that allows to differentiate GI and GII NoVs [19]. Total RNA was extracted from stool samples using QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) and cDNA was generated by MMLV reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were considered positive for NoVs when cycle threshold (ct) of RT-PCR was ⩽30 and lowpositive when 30<ct⩽35. For all NoV-positive samples two genomic fragments were amplified by conventional RT-PCR using Superscript III One-Step RT-PCR system with Platinum Taq DNA polymerase (Invitrogen) and primers JV12a/JV13b, targeting 326 bps of the RdRp region A [20] and either COG1F/ GISKR or COG2F/GIISKR, amplifying a 315 bps 5′portion of GI or GII VP1 protein [19, 21]. Environmental investigation From May through August 2016, a total of 32 tap water samples were collected from the Mascali municipal network. In particular, seven samples were collected at the resort where the outbreak occurred and from its water entry (R), two in the city centre of Mascali (M), 15 in the village of Fondachello (F), on the seaside, three in Carrabba village (C), far from the coast, and two and three, respectively, from the water well ‘Carlino’ (W) and the source ‘Bufardo and Torregrossa’ (S) which feed the municipal system and provide drinking water to the Mascali area (Fig. 1, Table 3). One litre water samples were sent by the local health authorities to the SIALAB laboratory in Avola, Siracusa and 100 ml aliquots were screened for coliforms, E. coli and enterococci as faecal indicators, by standard bacteriological methods according to UNI EN ISO 9308-1:2014. Unfortunately, although the suspect of a water source for the viral outbreak was quickly communicated, only in August and November 2016, following the implementation of control measures, local authorities provided water samplings of appropriate volume to perform viral tests (10 l). Five water samples of 10 l each were collected from different sampling sites in F and R areas for viral screening (Table 3) and were concentrated according to the Italian National Institute of Health (Istituto Superiore di Sanità, ISS) protocols [22] through an ultra-filtration system (Sartoflow® Slice 200 Benchtop Crossflow System, Sartorius AG, Goettingen, Germany), using appropriate membranes (SG Hydrosart 10 kDa) pretreated with 300 ml of 3% Beef Extract (BE) at pH 7, by recirculating it for 10 min in the system, at a pressure not higher than 1.30 bar. Elution and recovery of viruses attached to the membranes were obtained by recirculating 200 ml of 3% BE at pH 9.5 for 3 h until a 10–12 ml volume of eluate was obtained. The pH of the eluate was then checked with litmus paper and if necessary brought to the value of 7 with HCl. Concentrated water samples were simultaneously tested for NoV in three laboratories: Enteric Viruses Laboratory, University of Palermo; Istituto Zooprofilattico Sperimentale della Sicilia (IZS), Palermo; and Istituto Superiore di Sanità (ISS), Rome. Total RNA was extracted from 500 and 560 µl of a concentrated water sample using BioMèrieux NucliSENS® Magnetic Extraction System (BioMèrieux) and QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) respectively, according to manufacturer’s instructions. Bonded RNA was eluted by 100 µl of Elution Buffer. Amplifications were performed at least in duplicate on whole and 1:10 diluted extracts. A Real-Time RT-PCR assay according to ISO/TS 15216- 2:2013, two in-house Real-Time RT-PCR protocols and the One-Step RT-PCR methods already used for amplifying ORF1 and ORF2 fragments in NoV-positive faecal samples were performed [19–21, 23]. Nucleotide sequences analysis and phylogenetic analysis PCR amplicons were purified by commercial columns (CleanSweep PCR Purification, Thermo Fisher Scientific, Vilnius, Lithuania) and sequenced by Sanger method and all nucleotide sequences were aligned using ClustalW. To identify NoV genotypes, RdRp and VP1 sequences were submitted to the RIVM NoV genotyping tool (http://www.rivm.nl/mpf/norovirus/typingtool). Phylogenetic trees were constructed using MEGA 7 and Neighbour-Joining method (bootstrap 1000 replicates). Sequences from stool and water samples are available at NCBI with the following accession numbers: MG517455 for the major capsid protein obtained from a water sample; MG517456 and MG517461, respectively, for the major capsid protein and RNA-dependent-RNA-polymerase, as representatives of the sequences obtained from the stool samples of those involved in the outbreak. Control measures After outbreak notification, on 21 May, the resort was closed for 2 days. On 25 May, following the first results of the bacteriological analyses on municipal water samples, the use of municipal water for human consumption was temporarily restricted (use of water was permitted only for cleaning purposes) in the Carrabba and Fondachello villages and in the area surrounding the resort. From 25 May the reservoir collecting the water from the Carlino well and its chlorination device were submitted to maintenance work and a new chlorination device was installed on the conduit bringing the water from the Bufardo source. Actually, the Bufardo source had been connected to the municipal network in 2003 but unfortunately a dedicated chlorination station had not been planned at that time. Therefore, unchlorinated water entered the system near Fondachello village on the seaside (Fig. 1), possibly diluting chlorine concentrations in water treated at the Carlino reservoir. Moreover, the exact path of the branch of the water network situated downstream the Bufardo source water connection was not marked in the maps of the municipality (Fig. 1, red line), indicating that this section had never been submitted to maintenance. The area surrounding the unmaintained section of the water pipeline was scattered with unauthorised cottages that were not connected to the sewerage systems. Likely, untreated waste from cesspools was unloaded directly into the soil, generating a risk of contamination of the municipal water system through leakage. No work had been made on the water pipelines before the outbreak but sections of the network had been temporarily closed during the winter season on the seaside (Fondachello). Local health authorities advised the connection of all households in the area to the sewerage system and the replacement of the water system section without maintenance. Health authorities also raised attention on AGE cases notification at all local Emergency Departments and Family Practitioners but there was no further evidence of AGE outbreaks in the months following the intervention. Results Data collected through the online questionnaire compiled 7–10 days after the outbreak revealed that none of the 28 students and the teacher taking part in the school trip showed gastroenteritis symptoms at the beginning of the journey and they had no contact with people who suffered from diarrhoea and/or vomiting in the previous days. Travellers had their first shared meal on 19 May (Day 1), 2016, at lunch on their arrival to the resort in Mascali. The list of foods included in the menu is shown in Table 2. Conversely, the dinner on Day 1 was not shared and the students went to dine in three different fast food restaurants in Acireale. On 20 May (Day 2), 17 students (58.6% of the travellers) presented with vomit and/or diarrhoea, falling ill 26–36 h after lunch on Day 1, strongly suggesting that exposure occurred on the first shared meal. Some students also vomited in the presence of the other travellers on the bus bringing them to Acireale. Five more students were symptomatic in the early hours of the following day (21 May, Day 3, from 0:30 am to 8:00 am), as a continuation of the wave of illnesses of the previous day. The teacher and one more student fell ill at 8:00 pm and 11:00 pm of Day 3, respectively, for a second wave of illnesses occurring >48 h from the first shared meal and <36 h from the first cases, with the last student falling sick on 22 May (Day 4) at 3:00 am (Fig. 2, Table 1). On the whole, the mean incubation period was 33.2 h. Based on the online survey, the severity of illness was evaluated by the Vesikari severity scale, grouping symptomatic travellers into severe (11/25 = 44%) and non-severe (14/25 = 56%) cases (Table 1). Vomiting and fever were the predominant symptoms, affecting 22 out of 25 (88%) symptomatic subjects, while diarrhoea was present in 14 out of 25 (56%). Hospitalisation for rehydration therapy was necessary for 13 students (91% [10/11] of severe and 21% [3/14] of non-severe cases). Emergency room (ER) or family practitioner visits were necessary for 21 students and 13 (62%) of them initiated an antibiotic therapy. Two students indicated that other family members had gastroenteric symptoms more than 24 h after their return home, on Day 3, 4 and 5. All those involved in the outbreak recovered within 4 days, with an average duration of illness of 38.4 h. A case-control study of the risk associated with eating foods at lunch on Day 1 or with attending different restaurants at dinner on Day 1 was attempted. A total of 19 subjects filled the section of the online survey concerning food consumption. Fourteen of the responders were among the 22 cases falling ill within 36 h from the first shared meal (first wave), while the remaining five responders were three travellers with supposed secondary infections (showing symptoms with a delay ⩾48 h; second wave) and two asymptomatic students that were used as controls. No statistically significant associations with any food items served at lunch on Day 1 was identified (P < 0.05; Table 2). However, all food consumed on that occasion was to some extent associated with the risk of disease, the highest OR being obtained with chicken (OR: 20.7143, P: 0.0678; Table 2). Actually, most of the cases ate all foods included in the menu, while one of the two asymptomatic controls did not participate in the meal and the other had a pasta dish only. No statistically significant association was found for attending a particular fast food restaurant at the dinner on Day 1 (OR: 2.5, P: 0.3635). The Real-Time RT-PCR assay allowed to detect GII genogroup NoV genome in all the 25 samples from symptomatic subjects, although two of them showed high Ct values, indicating lower shedding of viral genomes (Table 1). The samples from four asymptomatic students were negative for both NoV genogroups. By sequence analysis, a NoV genotype could be defined for 20/ 25 NoV-positive stool samples. All NoV strains from travellers involved in the outbreaks belonged to the same genotype, GII.P2/GII.2, showing >99% nucleotide identities (nt id) in both fragments (ORF1 and ORF2) analysed. Thirty-two water samples were collected and analysed for bacterial contamination (Table 3). Seven samples were taken from the resort (site R) and the remaining 25 from five other sites (C, F, M, S, W), representative of the entire municipal water network. The majority (10/12) of the water samples collected immediately after the onset of the outbreak, between 21 May and 16 June 2016, at the resort (site R) and at sites F and C, located close to the resort, on the seaside and in the interior, respectively, were positive for coliform bacteria. More than 1000 colony-forming units (CFU)/ 100 ml (up to 2700 CFU/100 ml) were detected in four samples, while their absence is required in 100 ml by the Standard of Water Quality in Italy and seven of 11 samples tested were also Escherichia coli-positive (up to 900 CFU/100 ml). Enterococci were searched in 11 samples and three of them were found positive for this faecal contamination indicator. From 7 July to 16 August 2016, after control measures on the water network were undertaken, 13 water samples were collected for follow-up laboratory examination of the water system at F, R and C sites. All of them were negative for bacterial indicators of faecal contamination. No faecal bacteria were identified in the seven samples collected from M, W and S sites, including the well and the source that fed the water network, at any sampling time, before and after implementing the control measures. The research of NoV genome was conducted on concentrated water samples collected on 1 August from R and F sites (Table 3). NoV genome was detected in a single sample collected from the water entry of the resort involved in the outbreak (site R). Actually, only two out of 26 One-Step RT-PCR repetitions were positive for NoV viral capsid gene. The two ORF2-positive results were obtained at the University of Palermo Laboratory using both 1:10 diluted and undiluted extracted RNA. The sequence analysis of the partial ORF2 portion (289 bp) indicated the presence of the same GII.2 capsid genotype detected in the stool samples of the students involved in the outbreak, showing 98.9% nt id. No ORF1 NoV sequence could be obtained from the NoV-positive water sample. Follow-up analyses for NoV genome, conducted on water samples collected on 20 November both at R and F sites found no positive samples. By sequence alignment, the NoV sequences obtained from Mascali outbreak samples were phylogenetically closely related in both ORF1 and ORF2 (99 and 98.9% of nt id, respectively) to a GII.P2/GII.2 strain responsible for a sporadic paediatric gastroenteritis case occurred in Palermo in January 2016 (PA 36/16, accession numbers MG517464 and MG517460, for ORF1 and ORF2, respectively) and previously detected at the Enteric Viruses Laboratory, University of Palermo. Whereas, three GII.P2/GII.2 Italian strains (PA 280/11, PA 366/11 and PA 3/ 14, accession numbers MG517462, MG517465 and MG517463, for ORF1 and MG517457 to −59, for ORF2, respectively), detected in Palermo in 2011 and 2014, were more distantly associated with the outbreak strain both in RdRp and VP1 (90% to 97.4% nt id and 96.1% to 98.1% nt id, respectively, in ORF1 and ORF2). Discussion Epidemiological features, such as the presence of vomiting in more than half (88%) of symptomatic subjects, the 33.2 h average incubation period and the 38.4 h average duration of illness, suggested a NoV aetiology for the investigated outbreak [13]. The detection of NoV genome in all 25 faecal samples from symptomatic subjects fully complies with the requirement of detection of at least one positive sample out of four tested that is considered to be sufficient to ascertain the aetiology of a NoV outbreak [24]. Moreover, molecular typing indicated that a single viral strain, belonging to genotype GII.P2/GII.2, infected all symptomatic subjects involved in the outbreak. NoVs are highly infectious and can easily spread from person to person but also through contaminated food or water. Therefore, settings such as health care institutions, schools and touristic resorts are mainly affected, since close contact and shared food services facilitate transmission. In Mascali, two waves of transmission were observed over three days, being consistent with food/waterborne followed by person-to-person and/or airborne (vomit) transmission (Fig. 2). The first wave of cases presented on 20 May 2016 after a single shared meal, while the last case was detected on 22 May. Exposure to vomit may have been responsible for some of the secondary cases occurred after the first wave of illness. Vomiting produces aerosols that can be inhaled and swallowed. The extremely low infectious dose of NoV makes the inhalation of vomit aerosols an effective transmission route in settings where there is close contact, like on a bus trip [25]. All food consumed in the only shared lunch on 19 May at the resort was associated with the risk of disease (Table 2), but the association was inconclusive since all cases shared the same dishes included in the lunch menu on Day 1. Although many foods were potentially involved, our attention focused on the tap water. In fact, trivial information from local health authorities reported that local population avoided tap water consumption and relied on bottled water for drinking and cooking. Many small outbreaks of gastroenteritis involving restaurants and family groups had been reported in the same area, especially at the beginning of the summer season, when vacation houses and recreational activities were re-opened after a period of low water demand. Illness caused by NoV is common and if the level of water contamination is low, the number of cases remains limited and water can persist as unrecognised source of infection. If polluted water was served for drinking and used for cooking and washing dishes in the Mascali resort at Day 1 a generalised contamination of food and surfaces could have occurred. The potential role of the foodhandlers as a source of contamination could not be analysed in this study but should also be taken into account. If water contamination has already been present in the area for a long time, infections may have occurred among the local food-handlers. The analysis of water samples taken immediately after the outbreak revealed the presence of bacterial indicators of faecal contamination in the water used in the resort, with the only exception of a sample of filter purified water used both for cooking and as drinking water. When extending the sampling to other areas of the municipal water network, many other sites, mostly on the seaside, showed faecal pollution, which demonstrates a poor protection of the water network from faecal leakage (Fig. 1). NoV contamination of waters generally originates from sewage since these enteric pathogens, as other enteric viruses affecting humans, are mostly species-specific and are shed at high concentrations in the stools of patients. When the analyses on the stools of the students made it clear that the outbreak had been caused by NoV, the viral genome was also searched in appropriate volumes of water samples, but this happened when correction measures on the water network had already been taken. Nevertheless, a single water sample, obtained 2 months after the outbreak (1 August 2016), still contained genome fragments that could be correlated to the same NoV genotype causing the outbreak. Although the finding of NoV genome does not demonstrate that entire infecting viral particles were present in the water, it could indicate that a NoV belonging to the same genotype involved in the outbreak was still circulating in the local population several weeks after the outbreak and it was someway able to penetrate the water network, probably indicating continuing faecal contamination through leakage from sewage systems. NoV is particularly resistant in the environment and also to chlorine, therefore transmission through chlorinated water is not exceptional [9, 12, 26]. Drinking water from municipal supplies fed by lake or groundwater has already been identified in the past as the source of waterborne NoV outbreaks in Italy [27–29] and NoV is now considered one of the pathogens causing the largest number of affected consumers in drinking waterborne outbreaks [12]. NoV can persist in water longer than 15 days [30] and it is resistant to chlorine disinfection when free chlorine levels or pre-treatment processes are inadequate [31, 32]. Moreover, microfiltration filters labelled as certified by NSF Standards 53 or 58 (with a pore size ranging from 0.05 to 5 µm) are not effective in removing viruses and can even decrease chlorine concentration in filtered water [33]. Actually, the owner of the resort involved in our outbreak was encouraged to serve tap water as drinking water at the restaurant relying on the effectiveness of such a filter purifier. Unfortunately, we did not have a tap water sample from the restaurant of the resort taken before correction measures and whose volume was appropriate for virus detection. Therefore, we could not demonstrate the failure to eliminate viral pollution of the microfiltration system. In Mascali, correction measures on the water network consisted in the maintenance of the old chlorination device and in the addition of a new one at a key point located upstream and close to suspected illegal sewage dumping (Fig. 1), but as poor protection of the pipeline from sewage leakage risked to reproduce faecal pollution episodes, renovation of the municipal water system was also advised, as well as systematic control of illegal sewage dumping. The additional chlorination device was apparently the most effective intervention, since right after its activation faecal bacterial pollution indicators disappeared. However, NoV genome was detected in a water sample taken 6 weeks after the last detection of bacterial indicators. Unfortunately, detection of faecal contamination of drinking water is only based on the use of bacterial indicators despite no correlation between levels of enteric bacteria and enteric viruses has been found [34–36]. According to ISGEV enteric virus circulation surveillance data, in 2016 the circulation of GII.P2/GII.2 NoV strains in Palermo, home town of the travellers involved in the outbreak, was the highest ever observed from the starting of local NoV surveillance in 2004. GII.P2/GII.2 was the fourth most common genotype in 2016 (12.2% prevalence), following GII.P4/GII.4 Sydney 2012 (28.4%), GII.P16/GII.2 (18.9%) and GII.Pe/GII.4 Sydney 2012 (17.6%). GII.P2/GII.2 strains were mostly observed starting from July 2016, leaving space for the possibility that their increased circulation was linked to the introduction of the ‘Mascali clone’. However, a couple of GII.P2/GII.2 strains (including PA36/16) genetically close to the ‘Mascali clone’ had been already detected in January 2016 in Palermo and GII.P2/GII.2 NoV had sporadically been observed also in 2011 and 2014. Moreover, the circulation of GII.P2/GII.2 NoVs has been increasing worldwide over the last period [37]. Therefore, the unusually high circulation observed in Palermo might just reflect the global epidemiology of GII.P2/GII.2 strains. Noteworthy, the older GII.P2/GII.2 NoV strains detected in Palermo were closely related in the ORF2 (98.9% nt id) to the GII.P16/GII.2 Nashville strain (KY865307) which is supposed to have generated the recombinant GII.P16/GII.4 clone recently circulating epidemically worldwide [37, 38] and were more distantly associated (95.3% nt id) to the GII.2 capsid sequence (AB662868) of another putative GII.P16/GII.4 ancestor from Japan [39]. The increasing frequency of detection of waterborne NoV outbreaks calls for the introduction of routine monitoring for viruses of drinking waters and for the definition of affordable and reliable indicators for the presence of NoV in drinking waters. The use of ultrafiltration systems as a concentration method in order to recover viruses from environmental waters combined with RT-PCR to detect viral genomes has been demonstrated to grant a recovery rate of >50% of multiple viruses [40] and is at present one of the most sensitive approaches for the detection of NoVs, although the discrimination between viable and nonviable viruses remains a problem. The burden of NoV outbreaks, either food associated or waterborne, is still poorly defined due to the lack of specific diagnostic tools in many clinical laboratories. NoV epidemics should always be suspected when the main symptom is vomiting and the duration of disease is short (<48 h). Waterborne transmission is suggested by the rapid spread of disease and the high attack rate. Care should be taken to perform sensitive diagnostic tests on a sufficient number of stool samples in order to identify NoV as the causative agent of enteritis outbreaks, possibly preferring genome screening over antigen detection. Only the systematic detection of NoV outbreaks can guide in the choice of the safety procedures that could be most appropriate in containing such microbiological hazard.

How long did the event last?

1537

Waterborne Norovirus outbreak at a seaside resort likely originating from municipal water distribution system failure

Abstract In May 2016 a Norovirus (NoV) gastroenteritis outbreak involved a high school class visiting a seaside resort near Taormina (Mascali, Sicily). Twenty-four students and a teacher were affected and 17 of them showed symptoms on the second day of the journey, while the others got ill within the following 2 days. Symptoms included vomiting, diarrhoea and fever, and 12 students required hospitalisation. Stool samples tested positive for NoV genome by Real-Time polymerase chain reaction assay in all 25 symptomatic subjects. The GII.P2/GII.2 NoV genotype was linked to the outbreak by ORF1/ORF2 sequence analysis. The epidemiological features of the outbreak were consistent with food/waterborne followed by person-to-person and/ or vomit transmission. Food consumed at a shared lunch on the first day of the trip was associated to illness and drinking un-bottled tap water was also considered as a risk factor. The analysis of water samples revealed the presence of bacterial indicators of faecal contamination in the water used in the resort as well as in other areas of the municipal water network, linking the NoV gastroenteritis outbreak to tap water pollution from sewage leakage. From a single water sample, an amplicon whose sequence corresponded to the capsid genotype recovered from patients could be obtained. Introduction Noroviruses (NoVs) are recognised as a leading cause of sporadic cases and outbreaks of acute gastroenteritis (AGE) across all age groups [1, 2]. They are transmitted primarily through the faecal-oral route either by direct person–person spread and with consumption of contaminated food or water [1, 3, 4]. NoVs are non-enveloped RNA viruses belonging to the Caliciviridae family and are classified into seven distinct genogroups (GI-GVII) on the basis of the major capsid protein VP1 sequence. GI, GII and GIV NoVs infect humans and they have been divided into more than 30 genotypes based on the sequences of polymerase (ORF1) and capsid proteins (ORF2) [5]. NoV GII.4 genotype has been associated with the vast majority of NoV related cases of AGE worldwide although many genotypes, belonging to both GI and GII genogroups are involved in foodborne and waterborne outbreaks [6, 7]. NoVs are highly contagious owing to their high stability in the environment and their ability to infect at low doses. In fact, just 10–100 virions of NoV may be sufficient to cause AGE [8, 9]. NoVs are considered the major cause of water-related disease in semi-closed institutions, including restaurants, schools, hospitals, nursing homes and cruise ships [10–12]. To confirm NoV as a cause of outbreaks, the presence of four epidemiological features of disease, vomiting in more than half of affected persons, a mean incubation period of 24/48 h, a mean duration of illness of 12–60 h and the absence of bacterial pathogens in stool culture, should be evaluated [13]. Waterborne NoV outbreaks can be caused by contaminated drinking water despite correct chlorine concentration and in the absence of coliform bacteria [10, 14–17]. The aim of this study was to investigate an outbreak of NoV gastroenteritis associated with drinking water involving a high school class travelling from Palermo (West Sicily) to Acireale (East Sicily) on an educational tour in May 2016. Stool samples were collected soon after the onset of AGE symptoms to ascertain the aetiology of illness. The source of infection was investigated collecting epidemiological data through an online questionnaire and, in order to explore a possible transmission route, laboratory tests on water from the municipal network were performed. Material and methods Epidemiological investigation From 20 to 22 May 2016 an AGE outbreak occurred in Mascali, a small town near Taormina, involving a school class of 28 students, 13 girls and 15 boys aged between 14 and 18 years and a 54-year-old female teacher, during a trip to participate in a theatre festival in Acireale. A total of 25/29 persons were affected by nausea, vomiting, abdominal cramps, diarrhoea and fever while they were staying at a seaside resort in Mascali, on the way to Acireale. To obtain data on symptoms and the correlated risk factors all participants filled an online questionnaire investigating personal data (age and sex), travel and accommodation details and food consumption over 48 h before symptoms presentation, time of onset and duration of symptoms, therapy and/or hospitalisation required and history of contact with people affected by diarrhoea and/or vomiting (Table 1). The severity of illness was analysed by the Vesikari severity scale allowing to distinguish non-severe (score <11) and severe (score ⩾11) AGE (Table 1). Vesikari score was calculated from maximum number of episodes/day and duration of both diarrhoea and vomiting (score 0–3 for each of the 4 parameters), temperature (score 0–3), level of dehydration (score 0–3) and treatment required (score 0–2) [18]. A case-control study was conducted to establish if food consumed before the onset of symptoms was associated with the disease. A case was defined as a student attending the tour and reporting diarrhoea and/or vomiting within 36 h from the first shared meal during the tour, while students who did not present any of the previous symptoms or those showing the first symptoms with a delay of ⩾48 h (secondary cases) were used as controls. Variables were studied using odds ratio (OR) and its 95% CI and statistical associations were determined using the χ2 test, with the level of statistical significance set at 0.05 (Table 2). NoV detection in stool samples A stool sample was obtained from each of the 29 travellers, soon after (3–5 days) the onset of the outbreak. No stool samples could be obtained from the food-handlers working at the resort. All samples were tested for NoV at the Enteric Viruses Laboratory of the University of Palermo, member of the Italian Study Group for Enteric Viruses (ISGEV), by a Real-Time polymerase chain reaction (RT-PCR) assay that allows to differentiate GI and GII NoVs [19]. Total RNA was extracted from stool samples using QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) and cDNA was generated by MMLV reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were considered positive for NoVs when cycle threshold (ct) of RT-PCR was ⩽30 and lowpositive when 30<ct⩽35. For all NoV-positive samples two genomic fragments were amplified by conventional RT-PCR using Superscript III One-Step RT-PCR system with Platinum Taq DNA polymerase (Invitrogen) and primers JV12a/JV13b, targeting 326 bps of the RdRp region A [20] and either COG1F/ GISKR or COG2F/GIISKR, amplifying a 315 bps 5′portion of GI or GII VP1 protein [19, 21]. Environmental investigation From May through August 2016, a total of 32 tap water samples were collected from the Mascali municipal network. In particular, seven samples were collected at the resort where the outbreak occurred and from its water entry (R), two in the city centre of Mascali (M), 15 in the village of Fondachello (F), on the seaside, three in Carrabba village (C), far from the coast, and two and three, respectively, from the water well ‘Carlino’ (W) and the source ‘Bufardo and Torregrossa’ (S) which feed the municipal system and provide drinking water to the Mascali area (Fig. 1, Table 3). One litre water samples were sent by the local health authorities to the SIALAB laboratory in Avola, Siracusa and 100 ml aliquots were screened for coliforms, E. coli and enterococci as faecal indicators, by standard bacteriological methods according to UNI EN ISO 9308-1:2014. Unfortunately, although the suspect of a water source for the viral outbreak was quickly communicated, only in August and November 2016, following the implementation of control measures, local authorities provided water samplings of appropriate volume to perform viral tests (10 l). Five water samples of 10 l each were collected from different sampling sites in F and R areas for viral screening (Table 3) and were concentrated according to the Italian National Institute of Health (Istituto Superiore di Sanità, ISS) protocols [22] through an ultra-filtration system (Sartoflow® Slice 200 Benchtop Crossflow System, Sartorius AG, Goettingen, Germany), using appropriate membranes (SG Hydrosart 10 kDa) pretreated with 300 ml of 3% Beef Extract (BE) at pH 7, by recirculating it for 10 min in the system, at a pressure not higher than 1.30 bar. Elution and recovery of viruses attached to the membranes were obtained by recirculating 200 ml of 3% BE at pH 9.5 for 3 h until a 10–12 ml volume of eluate was obtained. The pH of the eluate was then checked with litmus paper and if necessary brought to the value of 7 with HCl. Concentrated water samples were simultaneously tested for NoV in three laboratories: Enteric Viruses Laboratory, University of Palermo; Istituto Zooprofilattico Sperimentale della Sicilia (IZS), Palermo; and Istituto Superiore di Sanità (ISS), Rome. Total RNA was extracted from 500 and 560 µl of a concentrated water sample using BioMèrieux NucliSENS® Magnetic Extraction System (BioMèrieux) and QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) respectively, according to manufacturer’s instructions. Bonded RNA was eluted by 100 µl of Elution Buffer. Amplifications were performed at least in duplicate on whole and 1:10 diluted extracts. A Real-Time RT-PCR assay according to ISO/TS 15216- 2:2013, two in-house Real-Time RT-PCR protocols and the One-Step RT-PCR methods already used for amplifying ORF1 and ORF2 fragments in NoV-positive faecal samples were performed [19–21, 23]. Nucleotide sequences analysis and phylogenetic analysis PCR amplicons were purified by commercial columns (CleanSweep PCR Purification, Thermo Fisher Scientific, Vilnius, Lithuania) and sequenced by Sanger method and all nucleotide sequences were aligned using ClustalW. To identify NoV genotypes, RdRp and VP1 sequences were submitted to the RIVM NoV genotyping tool (http://www.rivm.nl/mpf/norovirus/typingtool). Phylogenetic trees were constructed using MEGA 7 and Neighbour-Joining method (bootstrap 1000 replicates). Sequences from stool and water samples are available at NCBI with the following accession numbers: MG517455 for the major capsid protein obtained from a water sample; MG517456 and MG517461, respectively, for the major capsid protein and RNA-dependent-RNA-polymerase, as representatives of the sequences obtained from the stool samples of those involved in the outbreak. Control measures After outbreak notification, on 21 May, the resort was closed for 2 days. On 25 May, following the first results of the bacteriological analyses on municipal water samples, the use of municipal water for human consumption was temporarily restricted (use of water was permitted only for cleaning purposes) in the Carrabba and Fondachello villages and in the area surrounding the resort. From 25 May the reservoir collecting the water from the Carlino well and its chlorination device were submitted to maintenance work and a new chlorination device was installed on the conduit bringing the water from the Bufardo source. Actually, the Bufardo source had been connected to the municipal network in 2003 but unfortunately a dedicated chlorination station had not been planned at that time. Therefore, unchlorinated water entered the system near Fondachello village on the seaside (Fig. 1), possibly diluting chlorine concentrations in water treated at the Carlino reservoir. Moreover, the exact path of the branch of the water network situated downstream the Bufardo source water connection was not marked in the maps of the municipality (Fig. 1, red line), indicating that this section had never been submitted to maintenance. The area surrounding the unmaintained section of the water pipeline was scattered with unauthorised cottages that were not connected to the sewerage systems. Likely, untreated waste from cesspools was unloaded directly into the soil, generating a risk of contamination of the municipal water system through leakage. No work had been made on the water pipelines before the outbreak but sections of the network had been temporarily closed during the winter season on the seaside (Fondachello). Local health authorities advised the connection of all households in the area to the sewerage system and the replacement of the water system section without maintenance. Health authorities also raised attention on AGE cases notification at all local Emergency Departments and Family Practitioners but there was no further evidence of AGE outbreaks in the months following the intervention. Results Data collected through the online questionnaire compiled 7–10 days after the outbreak revealed that none of the 28 students and the teacher taking part in the school trip showed gastroenteritis symptoms at the beginning of the journey and they had no contact with people who suffered from diarrhoea and/or vomiting in the previous days. Travellers had their first shared meal on 19 May (Day 1), 2016, at lunch on their arrival to the resort in Mascali. The list of foods included in the menu is shown in Table 2. Conversely, the dinner on Day 1 was not shared and the students went to dine in three different fast food restaurants in Acireale. On 20 May (Day 2), 17 students (58.6% of the travellers) presented with vomit and/or diarrhoea, falling ill 26–36 h after lunch on Day 1, strongly suggesting that exposure occurred on the first shared meal. Some students also vomited in the presence of the other travellers on the bus bringing them to Acireale. Five more students were symptomatic in the early hours of the following day (21 May, Day 3, from 0:30 am to 8:00 am), as a continuation of the wave of illnesses of the previous day. The teacher and one more student fell ill at 8:00 pm and 11:00 pm of Day 3, respectively, for a second wave of illnesses occurring >48 h from the first shared meal and <36 h from the first cases, with the last student falling sick on 22 May (Day 4) at 3:00 am (Fig. 2, Table 1). On the whole, the mean incubation period was 33.2 h. Based on the online survey, the severity of illness was evaluated by the Vesikari severity scale, grouping symptomatic travellers into severe (11/25 = 44%) and non-severe (14/25 = 56%) cases (Table 1). Vomiting and fever were the predominant symptoms, affecting 22 out of 25 (88%) symptomatic subjects, while diarrhoea was present in 14 out of 25 (56%). Hospitalisation for rehydration therapy was necessary for 13 students (91% [10/11] of severe and 21% [3/14] of non-severe cases). Emergency room (ER) or family practitioner visits were necessary for 21 students and 13 (62%) of them initiated an antibiotic therapy. Two students indicated that other family members had gastroenteric symptoms more than 24 h after their return home, on Day 3, 4 and 5. All those involved in the outbreak recovered within 4 days, with an average duration of illness of 38.4 h. A case-control study of the risk associated with eating foods at lunch on Day 1 or with attending different restaurants at dinner on Day 1 was attempted. A total of 19 subjects filled the section of the online survey concerning food consumption. Fourteen of the responders were among the 22 cases falling ill within 36 h from the first shared meal (first wave), while the remaining five responders were three travellers with supposed secondary infections (showing symptoms with a delay ⩾48 h; second wave) and two asymptomatic students that were used as controls. No statistically significant associations with any food items served at lunch on Day 1 was identified (P < 0.05; Table 2). However, all food consumed on that occasion was to some extent associated with the risk of disease, the highest OR being obtained with chicken (OR: 20.7143, P: 0.0678; Table 2). Actually, most of the cases ate all foods included in the menu, while one of the two asymptomatic controls did not participate in the meal and the other had a pasta dish only. No statistically significant association was found for attending a particular fast food restaurant at the dinner on Day 1 (OR: 2.5, P: 0.3635). The Real-Time RT-PCR assay allowed to detect GII genogroup NoV genome in all the 25 samples from symptomatic subjects, although two of them showed high Ct values, indicating lower shedding of viral genomes (Table 1). The samples from four asymptomatic students were negative for both NoV genogroups. By sequence analysis, a NoV genotype could be defined for 20/ 25 NoV-positive stool samples. All NoV strains from travellers involved in the outbreaks belonged to the same genotype, GII.P2/GII.2, showing >99% nucleotide identities (nt id) in both fragments (ORF1 and ORF2) analysed. Thirty-two water samples were collected and analysed for bacterial contamination (Table 3). Seven samples were taken from the resort (site R) and the remaining 25 from five other sites (C, F, M, S, W), representative of the entire municipal water network. The majority (10/12) of the water samples collected immediately after the onset of the outbreak, between 21 May and 16 June 2016, at the resort (site R) and at sites F and C, located close to the resort, on the seaside and in the interior, respectively, were positive for coliform bacteria. More than 1000 colony-forming units (CFU)/ 100 ml (up to 2700 CFU/100 ml) were detected in four samples, while their absence is required in 100 ml by the Standard of Water Quality in Italy and seven of 11 samples tested were also Escherichia coli-positive (up to 900 CFU/100 ml). Enterococci were searched in 11 samples and three of them were found positive for this faecal contamination indicator. From 7 July to 16 August 2016, after control measures on the water network were undertaken, 13 water samples were collected for follow-up laboratory examination of the water system at F, R and C sites. All of them were negative for bacterial indicators of faecal contamination. No faecal bacteria were identified in the seven samples collected from M, W and S sites, including the well and the source that fed the water network, at any sampling time, before and after implementing the control measures. The research of NoV genome was conducted on concentrated water samples collected on 1 August from R and F sites (Table 3). NoV genome was detected in a single sample collected from the water entry of the resort involved in the outbreak (site R). Actually, only two out of 26 One-Step RT-PCR repetitions were positive for NoV viral capsid gene. The two ORF2-positive results were obtained at the University of Palermo Laboratory using both 1:10 diluted and undiluted extracted RNA. The sequence analysis of the partial ORF2 portion (289 bp) indicated the presence of the same GII.2 capsid genotype detected in the stool samples of the students involved in the outbreak, showing 98.9% nt id. No ORF1 NoV sequence could be obtained from the NoV-positive water sample. Follow-up analyses for NoV genome, conducted on water samples collected on 20 November both at R and F sites found no positive samples. By sequence alignment, the NoV sequences obtained from Mascali outbreak samples were phylogenetically closely related in both ORF1 and ORF2 (99 and 98.9% of nt id, respectively) to a GII.P2/GII.2 strain responsible for a sporadic paediatric gastroenteritis case occurred in Palermo in January 2016 (PA 36/16, accession numbers MG517464 and MG517460, for ORF1 and ORF2, respectively) and previously detected at the Enteric Viruses Laboratory, University of Palermo. Whereas, three GII.P2/GII.2 Italian strains (PA 280/11, PA 366/11 and PA 3/ 14, accession numbers MG517462, MG517465 and MG517463, for ORF1 and MG517457 to −59, for ORF2, respectively), detected in Palermo in 2011 and 2014, were more distantly associated with the outbreak strain both in RdRp and VP1 (90% to 97.4% nt id and 96.1% to 98.1% nt id, respectively, in ORF1 and ORF2). Discussion Epidemiological features, such as the presence of vomiting in more than half (88%) of symptomatic subjects, the 33.2 h average incubation period and the 38.4 h average duration of illness, suggested a NoV aetiology for the investigated outbreak [13]. The detection of NoV genome in all 25 faecal samples from symptomatic subjects fully complies with the requirement of detection of at least one positive sample out of four tested that is considered to be sufficient to ascertain the aetiology of a NoV outbreak [24]. Moreover, molecular typing indicated that a single viral strain, belonging to genotype GII.P2/GII.2, infected all symptomatic subjects involved in the outbreak. NoVs are highly infectious and can easily spread from person to person but also through contaminated food or water. Therefore, settings such as health care institutions, schools and touristic resorts are mainly affected, since close contact and shared food services facilitate transmission. In Mascali, two waves of transmission were observed over three days, being consistent with food/waterborne followed by person-to-person and/or airborne (vomit) transmission (Fig. 2). The first wave of cases presented on 20 May 2016 after a single shared meal, while the last case was detected on 22 May. Exposure to vomit may have been responsible for some of the secondary cases occurred after the first wave of illness. Vomiting produces aerosols that can be inhaled and swallowed. The extremely low infectious dose of NoV makes the inhalation of vomit aerosols an effective transmission route in settings where there is close contact, like on a bus trip [25]. All food consumed in the only shared lunch on 19 May at the resort was associated with the risk of disease (Table 2), but the association was inconclusive since all cases shared the same dishes included in the lunch menu on Day 1. Although many foods were potentially involved, our attention focused on the tap water. In fact, trivial information from local health authorities reported that local population avoided tap water consumption and relied on bottled water for drinking and cooking. Many small outbreaks of gastroenteritis involving restaurants and family groups had been reported in the same area, especially at the beginning of the summer season, when vacation houses and recreational activities were re-opened after a period of low water demand. Illness caused by NoV is common and if the level of water contamination is low, the number of cases remains limited and water can persist as unrecognised source of infection. If polluted water was served for drinking and used for cooking and washing dishes in the Mascali resort at Day 1 a generalised contamination of food and surfaces could have occurred. The potential role of the foodhandlers as a source of contamination could not be analysed in this study but should also be taken into account. If water contamination has already been present in the area for a long time, infections may have occurred among the local food-handlers. The analysis of water samples taken immediately after the outbreak revealed the presence of bacterial indicators of faecal contamination in the water used in the resort, with the only exception of a sample of filter purified water used both for cooking and as drinking water. When extending the sampling to other areas of the municipal water network, many other sites, mostly on the seaside, showed faecal pollution, which demonstrates a poor protection of the water network from faecal leakage (Fig. 1). NoV contamination of waters generally originates from sewage since these enteric pathogens, as other enteric viruses affecting humans, are mostly species-specific and are shed at high concentrations in the stools of patients. When the analyses on the stools of the students made it clear that the outbreak had been caused by NoV, the viral genome was also searched in appropriate volumes of water samples, but this happened when correction measures on the water network had already been taken. Nevertheless, a single water sample, obtained 2 months after the outbreak (1 August 2016), still contained genome fragments that could be correlated to the same NoV genotype causing the outbreak. Although the finding of NoV genome does not demonstrate that entire infecting viral particles were present in the water, it could indicate that a NoV belonging to the same genotype involved in the outbreak was still circulating in the local population several weeks after the outbreak and it was someway able to penetrate the water network, probably indicating continuing faecal contamination through leakage from sewage systems. NoV is particularly resistant in the environment and also to chlorine, therefore transmission through chlorinated water is not exceptional [9, 12, 26]. Drinking water from municipal supplies fed by lake or groundwater has already been identified in the past as the source of waterborne NoV outbreaks in Italy [27–29] and NoV is now considered one of the pathogens causing the largest number of affected consumers in drinking waterborne outbreaks [12]. NoV can persist in water longer than 15 days [30] and it is resistant to chlorine disinfection when free chlorine levels or pre-treatment processes are inadequate [31, 32]. Moreover, microfiltration filters labelled as certified by NSF Standards 53 or 58 (with a pore size ranging from 0.05 to 5 µm) are not effective in removing viruses and can even decrease chlorine concentration in filtered water [33]. Actually, the owner of the resort involved in our outbreak was encouraged to serve tap water as drinking water at the restaurant relying on the effectiveness of such a filter purifier. Unfortunately, we did not have a tap water sample from the restaurant of the resort taken before correction measures and whose volume was appropriate for virus detection. Therefore, we could not demonstrate the failure to eliminate viral pollution of the microfiltration system. In Mascali, correction measures on the water network consisted in the maintenance of the old chlorination device and in the addition of a new one at a key point located upstream and close to suspected illegal sewage dumping (Fig. 1), but as poor protection of the pipeline from sewage leakage risked to reproduce faecal pollution episodes, renovation of the municipal water system was also advised, as well as systematic control of illegal sewage dumping. The additional chlorination device was apparently the most effective intervention, since right after its activation faecal bacterial pollution indicators disappeared. However, NoV genome was detected in a water sample taken 6 weeks after the last detection of bacterial indicators. Unfortunately, detection of faecal contamination of drinking water is only based on the use of bacterial indicators despite no correlation between levels of enteric bacteria and enteric viruses has been found [34–36]. According to ISGEV enteric virus circulation surveillance data, in 2016 the circulation of GII.P2/GII.2 NoV strains in Palermo, home town of the travellers involved in the outbreak, was the highest ever observed from the starting of local NoV surveillance in 2004. GII.P2/GII.2 was the fourth most common genotype in 2016 (12.2% prevalence), following GII.P4/GII.4 Sydney 2012 (28.4%), GII.P16/GII.2 (18.9%) and GII.Pe/GII.4 Sydney 2012 (17.6%). GII.P2/GII.2 strains were mostly observed starting from July 2016, leaving space for the possibility that their increased circulation was linked to the introduction of the ‘Mascali clone’. However, a couple of GII.P2/GII.2 strains (including PA36/16) genetically close to the ‘Mascali clone’ had been already detected in January 2016 in Palermo and GII.P2/GII.2 NoV had sporadically been observed also in 2011 and 2014. Moreover, the circulation of GII.P2/GII.2 NoVs has been increasing worldwide over the last period [37]. Therefore, the unusually high circulation observed in Palermo might just reflect the global epidemiology of GII.P2/GII.2 strains. Noteworthy, the older GII.P2/GII.2 NoV strains detected in Palermo were closely related in the ORF2 (98.9% nt id) to the GII.P16/GII.2 Nashville strain (KY865307) which is supposed to have generated the recombinant GII.P16/GII.4 clone recently circulating epidemically worldwide [37, 38] and were more distantly associated (95.3% nt id) to the GII.2 capsid sequence (AB662868) of another putative GII.P16/GII.4 ancestor from Japan [39]. The increasing frequency of detection of waterborne NoV outbreaks calls for the introduction of routine monitoring for viruses of drinking waters and for the definition of affordable and reliable indicators for the presence of NoV in drinking waters. The use of ultrafiltration systems as a concentration method in order to recover viruses from environmental waters combined with RT-PCR to detect viral genomes has been demonstrated to grant a recovery rate of >50% of multiple viruses [40] and is at present one of the most sensitive approaches for the detection of NoVs, although the discrimination between viable and nonviable viruses remains a problem. The burden of NoV outbreaks, either food associated or waterborne, is still poorly defined due to the lack of specific diagnostic tools in many clinical laboratories. NoV epidemics should always be suspected when the main symptom is vomiting and the duration of disease is short (<48 h). Waterborne transmission is suggested by the rapid spread of disease and the high attack rate. Care should be taken to perform sensitive diagnostic tests on a sufficient number of stool samples in order to identify NoV as the causative agent of enteritis outbreaks, possibly preferring genome screening over antigen detection. Only the systematic detection of NoV outbreaks can guide in the choice of the safety procedures that could be most appropriate in containing such microbiological hazard.

In which street did this happen?

1538

Waterborne Norovirus outbreak at a seaside resort likely originating from municipal water distribution system failure

Abstract In May 2016 a Norovirus (NoV) gastroenteritis outbreak involved a high school class visiting a seaside resort near Taormina (Mascali, Sicily). Twenty-four students and a teacher were affected and 17 of them showed symptoms on the second day of the journey, while the others got ill within the following 2 days. Symptoms included vomiting, diarrhoea and fever, and 12 students required hospitalisation. Stool samples tested positive for NoV genome by Real-Time polymerase chain reaction assay in all 25 symptomatic subjects. The GII.P2/GII.2 NoV genotype was linked to the outbreak by ORF1/ORF2 sequence analysis. The epidemiological features of the outbreak were consistent with food/waterborne followed by person-to-person and/ or vomit transmission. Food consumed at a shared lunch on the first day of the trip was associated to illness and drinking un-bottled tap water was also considered as a risk factor. The analysis of water samples revealed the presence of bacterial indicators of faecal contamination in the water used in the resort as well as in other areas of the municipal water network, linking the NoV gastroenteritis outbreak to tap water pollution from sewage leakage. From a single water sample, an amplicon whose sequence corresponded to the capsid genotype recovered from patients could be obtained. Introduction Noroviruses (NoVs) are recognised as a leading cause of sporadic cases and outbreaks of acute gastroenteritis (AGE) across all age groups [1, 2]. They are transmitted primarily through the faecal-oral route either by direct person–person spread and with consumption of contaminated food or water [1, 3, 4]. NoVs are non-enveloped RNA viruses belonging to the Caliciviridae family and are classified into seven distinct genogroups (GI-GVII) on the basis of the major capsid protein VP1 sequence. GI, GII and GIV NoVs infect humans and they have been divided into more than 30 genotypes based on the sequences of polymerase (ORF1) and capsid proteins (ORF2) [5]. NoV GII.4 genotype has been associated with the vast majority of NoV related cases of AGE worldwide although many genotypes, belonging to both GI and GII genogroups are involved in foodborne and waterborne outbreaks [6, 7]. NoVs are highly contagious owing to their high stability in the environment and their ability to infect at low doses. In fact, just 10–100 virions of NoV may be sufficient to cause AGE [8, 9]. NoVs are considered the major cause of water-related disease in semi-closed institutions, including restaurants, schools, hospitals, nursing homes and cruise ships [10–12]. To confirm NoV as a cause of outbreaks, the presence of four epidemiological features of disease, vomiting in more than half of affected persons, a mean incubation period of 24/48 h, a mean duration of illness of 12–60 h and the absence of bacterial pathogens in stool culture, should be evaluated [13]. Waterborne NoV outbreaks can be caused by contaminated drinking water despite correct chlorine concentration and in the absence of coliform bacteria [10, 14–17]. The aim of this study was to investigate an outbreak of NoV gastroenteritis associated with drinking water involving a high school class travelling from Palermo (West Sicily) to Acireale (East Sicily) on an educational tour in May 2016. Stool samples were collected soon after the onset of AGE symptoms to ascertain the aetiology of illness. The source of infection was investigated collecting epidemiological data through an online questionnaire and, in order to explore a possible transmission route, laboratory tests on water from the municipal network were performed. Material and methods Epidemiological investigation From 20 to 22 May 2016 an AGE outbreak occurred in Mascali, a small town near Taormina, involving a school class of 28 students, 13 girls and 15 boys aged between 14 and 18 years and a 54-year-old female teacher, during a trip to participate in a theatre festival in Acireale. A total of 25/29 persons were affected by nausea, vomiting, abdominal cramps, diarrhoea and fever while they were staying at a seaside resort in Mascali, on the way to Acireale. To obtain data on symptoms and the correlated risk factors all participants filled an online questionnaire investigating personal data (age and sex), travel and accommodation details and food consumption over 48 h before symptoms presentation, time of onset and duration of symptoms, therapy and/or hospitalisation required and history of contact with people affected by diarrhoea and/or vomiting (Table 1). The severity of illness was analysed by the Vesikari severity scale allowing to distinguish non-severe (score <11) and severe (score ⩾11) AGE (Table 1). Vesikari score was calculated from maximum number of episodes/day and duration of both diarrhoea and vomiting (score 0–3 for each of the 4 parameters), temperature (score 0–3), level of dehydration (score 0–3) and treatment required (score 0–2) [18]. A case-control study was conducted to establish if food consumed before the onset of symptoms was associated with the disease. A case was defined as a student attending the tour and reporting diarrhoea and/or vomiting within 36 h from the first shared meal during the tour, while students who did not present any of the previous symptoms or those showing the first symptoms with a delay of ⩾48 h (secondary cases) were used as controls. Variables were studied using odds ratio (OR) and its 95% CI and statistical associations were determined using the χ2 test, with the level of statistical significance set at 0.05 (Table 2). NoV detection in stool samples A stool sample was obtained from each of the 29 travellers, soon after (3–5 days) the onset of the outbreak. No stool samples could be obtained from the food-handlers working at the resort. All samples were tested for NoV at the Enteric Viruses Laboratory of the University of Palermo, member of the Italian Study Group for Enteric Viruses (ISGEV), by a Real-Time polymerase chain reaction (RT-PCR) assay that allows to differentiate GI and GII NoVs [19]. Total RNA was extracted from stool samples using QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) and cDNA was generated by MMLV reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were considered positive for NoVs when cycle threshold (ct) of RT-PCR was ⩽30 and lowpositive when 30<ct⩽35. For all NoV-positive samples two genomic fragments were amplified by conventional RT-PCR using Superscript III One-Step RT-PCR system with Platinum Taq DNA polymerase (Invitrogen) and primers JV12a/JV13b, targeting 326 bps of the RdRp region A [20] and either COG1F/ GISKR or COG2F/GIISKR, amplifying a 315 bps 5′portion of GI or GII VP1 protein [19, 21]. Environmental investigation From May through August 2016, a total of 32 tap water samples were collected from the Mascali municipal network. In particular, seven samples were collected at the resort where the outbreak occurred and from its water entry (R), two in the city centre of Mascali (M), 15 in the village of Fondachello (F), on the seaside, three in Carrabba village (C), far from the coast, and two and three, respectively, from the water well ‘Carlino’ (W) and the source ‘Bufardo and Torregrossa’ (S) which feed the municipal system and provide drinking water to the Mascali area (Fig. 1, Table 3). One litre water samples were sent by the local health authorities to the SIALAB laboratory in Avola, Siracusa and 100 ml aliquots were screened for coliforms, E. coli and enterococci as faecal indicators, by standard bacteriological methods according to UNI EN ISO 9308-1:2014. Unfortunately, although the suspect of a water source for the viral outbreak was quickly communicated, only in August and November 2016, following the implementation of control measures, local authorities provided water samplings of appropriate volume to perform viral tests (10 l). Five water samples of 10 l each were collected from different sampling sites in F and R areas for viral screening (Table 3) and were concentrated according to the Italian National Institute of Health (Istituto Superiore di Sanità, ISS) protocols [22] through an ultra-filtration system (Sartoflow® Slice 200 Benchtop Crossflow System, Sartorius AG, Goettingen, Germany), using appropriate membranes (SG Hydrosart 10 kDa) pretreated with 300 ml of 3% Beef Extract (BE) at pH 7, by recirculating it for 10 min in the system, at a pressure not higher than 1.30 bar. Elution and recovery of viruses attached to the membranes were obtained by recirculating 200 ml of 3% BE at pH 9.5 for 3 h until a 10–12 ml volume of eluate was obtained. The pH of the eluate was then checked with litmus paper and if necessary brought to the value of 7 with HCl. Concentrated water samples were simultaneously tested for NoV in three laboratories: Enteric Viruses Laboratory, University of Palermo; Istituto Zooprofilattico Sperimentale della Sicilia (IZS), Palermo; and Istituto Superiore di Sanità (ISS), Rome. Total RNA was extracted from 500 and 560 µl of a concentrated water sample using BioMèrieux NucliSENS® Magnetic Extraction System (BioMèrieux) and QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) respectively, according to manufacturer’s instructions. Bonded RNA was eluted by 100 µl of Elution Buffer. Amplifications were performed at least in duplicate on whole and 1:10 diluted extracts. A Real-Time RT-PCR assay according to ISO/TS 15216- 2:2013, two in-house Real-Time RT-PCR protocols and the One-Step RT-PCR methods already used for amplifying ORF1 and ORF2 fragments in NoV-positive faecal samples were performed [19–21, 23]. Nucleotide sequences analysis and phylogenetic analysis PCR amplicons were purified by commercial columns (CleanSweep PCR Purification, Thermo Fisher Scientific, Vilnius, Lithuania) and sequenced by Sanger method and all nucleotide sequences were aligned using ClustalW. To identify NoV genotypes, RdRp and VP1 sequences were submitted to the RIVM NoV genotyping tool (http://www.rivm.nl/mpf/norovirus/typingtool). Phylogenetic trees were constructed using MEGA 7 and Neighbour-Joining method (bootstrap 1000 replicates). Sequences from stool and water samples are available at NCBI with the following accession numbers: MG517455 for the major capsid protein obtained from a water sample; MG517456 and MG517461, respectively, for the major capsid protein and RNA-dependent-RNA-polymerase, as representatives of the sequences obtained from the stool samples of those involved in the outbreak. Control measures After outbreak notification, on 21 May, the resort was closed for 2 days. On 25 May, following the first results of the bacteriological analyses on municipal water samples, the use of municipal water for human consumption was temporarily restricted (use of water was permitted only for cleaning purposes) in the Carrabba and Fondachello villages and in the area surrounding the resort. From 25 May the reservoir collecting the water from the Carlino well and its chlorination device were submitted to maintenance work and a new chlorination device was installed on the conduit bringing the water from the Bufardo source. Actually, the Bufardo source had been connected to the municipal network in 2003 but unfortunately a dedicated chlorination station had not been planned at that time. Therefore, unchlorinated water entered the system near Fondachello village on the seaside (Fig. 1), possibly diluting chlorine concentrations in water treated at the Carlino reservoir. Moreover, the exact path of the branch of the water network situated downstream the Bufardo source water connection was not marked in the maps of the municipality (Fig. 1, red line), indicating that this section had never been submitted to maintenance. The area surrounding the unmaintained section of the water pipeline was scattered with unauthorised cottages that were not connected to the sewerage systems. Likely, untreated waste from cesspools was unloaded directly into the soil, generating a risk of contamination of the municipal water system through leakage. No work had been made on the water pipelines before the outbreak but sections of the network had been temporarily closed during the winter season on the seaside (Fondachello). Local health authorities advised the connection of all households in the area to the sewerage system and the replacement of the water system section without maintenance. Health authorities also raised attention on AGE cases notification at all local Emergency Departments and Family Practitioners but there was no further evidence of AGE outbreaks in the months following the intervention. Results Data collected through the online questionnaire compiled 7–10 days after the outbreak revealed that none of the 28 students and the teacher taking part in the school trip showed gastroenteritis symptoms at the beginning of the journey and they had no contact with people who suffered from diarrhoea and/or vomiting in the previous days. Travellers had their first shared meal on 19 May (Day 1), 2016, at lunch on their arrival to the resort in Mascali. The list of foods included in the menu is shown in Table 2. Conversely, the dinner on Day 1 was not shared and the students went to dine in three different fast food restaurants in Acireale. On 20 May (Day 2), 17 students (58.6% of the travellers) presented with vomit and/or diarrhoea, falling ill 26–36 h after lunch on Day 1, strongly suggesting that exposure occurred on the first shared meal. Some students also vomited in the presence of the other travellers on the bus bringing them to Acireale. Five more students were symptomatic in the early hours of the following day (21 May, Day 3, from 0:30 am to 8:00 am), as a continuation of the wave of illnesses of the previous day. The teacher and one more student fell ill at 8:00 pm and 11:00 pm of Day 3, respectively, for a second wave of illnesses occurring >48 h from the first shared meal and <36 h from the first cases, with the last student falling sick on 22 May (Day 4) at 3:00 am (Fig. 2, Table 1). On the whole, the mean incubation period was 33.2 h. Based on the online survey, the severity of illness was evaluated by the Vesikari severity scale, grouping symptomatic travellers into severe (11/25 = 44%) and non-severe (14/25 = 56%) cases (Table 1). Vomiting and fever were the predominant symptoms, affecting 22 out of 25 (88%) symptomatic subjects, while diarrhoea was present in 14 out of 25 (56%). Hospitalisation for rehydration therapy was necessary for 13 students (91% [10/11] of severe and 21% [3/14] of non-severe cases). Emergency room (ER) or family practitioner visits were necessary for 21 students and 13 (62%) of them initiated an antibiotic therapy. Two students indicated that other family members had gastroenteric symptoms more than 24 h after their return home, on Day 3, 4 and 5. All those involved in the outbreak recovered within 4 days, with an average duration of illness of 38.4 h. A case-control study of the risk associated with eating foods at lunch on Day 1 or with attending different restaurants at dinner on Day 1 was attempted. A total of 19 subjects filled the section of the online survey concerning food consumption. Fourteen of the responders were among the 22 cases falling ill within 36 h from the first shared meal (first wave), while the remaining five responders were three travellers with supposed secondary infections (showing symptoms with a delay ⩾48 h; second wave) and two asymptomatic students that were used as controls. No statistically significant associations with any food items served at lunch on Day 1 was identified (P < 0.05; Table 2). However, all food consumed on that occasion was to some extent associated with the risk of disease, the highest OR being obtained with chicken (OR: 20.7143, P: 0.0678; Table 2). Actually, most of the cases ate all foods included in the menu, while one of the two asymptomatic controls did not participate in the meal and the other had a pasta dish only. No statistically significant association was found for attending a particular fast food restaurant at the dinner on Day 1 (OR: 2.5, P: 0.3635). The Real-Time RT-PCR assay allowed to detect GII genogroup NoV genome in all the 25 samples from symptomatic subjects, although two of them showed high Ct values, indicating lower shedding of viral genomes (Table 1). The samples from four asymptomatic students were negative for both NoV genogroups. By sequence analysis, a NoV genotype could be defined for 20/ 25 NoV-positive stool samples. All NoV strains from travellers involved in the outbreaks belonged to the same genotype, GII.P2/GII.2, showing >99% nucleotide identities (nt id) in both fragments (ORF1 and ORF2) analysed. Thirty-two water samples were collected and analysed for bacterial contamination (Table 3). Seven samples were taken from the resort (site R) and the remaining 25 from five other sites (C, F, M, S, W), representative of the entire municipal water network. The majority (10/12) of the water samples collected immediately after the onset of the outbreak, between 21 May and 16 June 2016, at the resort (site R) and at sites F and C, located close to the resort, on the seaside and in the interior, respectively, were positive for coliform bacteria. More than 1000 colony-forming units (CFU)/ 100 ml (up to 2700 CFU/100 ml) were detected in four samples, while their absence is required in 100 ml by the Standard of Water Quality in Italy and seven of 11 samples tested were also Escherichia coli-positive (up to 900 CFU/100 ml). Enterococci were searched in 11 samples and three of them were found positive for this faecal contamination indicator. From 7 July to 16 August 2016, after control measures on the water network were undertaken, 13 water samples were collected for follow-up laboratory examination of the water system at F, R and C sites. All of them were negative for bacterial indicators of faecal contamination. No faecal bacteria were identified in the seven samples collected from M, W and S sites, including the well and the source that fed the water network, at any sampling time, before and after implementing the control measures. The research of NoV genome was conducted on concentrated water samples collected on 1 August from R and F sites (Table 3). NoV genome was detected in a single sample collected from the water entry of the resort involved in the outbreak (site R). Actually, only two out of 26 One-Step RT-PCR repetitions were positive for NoV viral capsid gene. The two ORF2-positive results were obtained at the University of Palermo Laboratory using both 1:10 diluted and undiluted extracted RNA. The sequence analysis of the partial ORF2 portion (289 bp) indicated the presence of the same GII.2 capsid genotype detected in the stool samples of the students involved in the outbreak, showing 98.9% nt id. No ORF1 NoV sequence could be obtained from the NoV-positive water sample. Follow-up analyses for NoV genome, conducted on water samples collected on 20 November both at R and F sites found no positive samples. By sequence alignment, the NoV sequences obtained from Mascali outbreak samples were phylogenetically closely related in both ORF1 and ORF2 (99 and 98.9% of nt id, respectively) to a GII.P2/GII.2 strain responsible for a sporadic paediatric gastroenteritis case occurred in Palermo in January 2016 (PA 36/16, accession numbers MG517464 and MG517460, for ORF1 and ORF2, respectively) and previously detected at the Enteric Viruses Laboratory, University of Palermo. Whereas, three GII.P2/GII.2 Italian strains (PA 280/11, PA 366/11 and PA 3/ 14, accession numbers MG517462, MG517465 and MG517463, for ORF1 and MG517457 to −59, for ORF2, respectively), detected in Palermo in 2011 and 2014, were more distantly associated with the outbreak strain both in RdRp and VP1 (90% to 97.4% nt id and 96.1% to 98.1% nt id, respectively, in ORF1 and ORF2). Discussion Epidemiological features, such as the presence of vomiting in more than half (88%) of symptomatic subjects, the 33.2 h average incubation period and the 38.4 h average duration of illness, suggested a NoV aetiology for the investigated outbreak [13]. The detection of NoV genome in all 25 faecal samples from symptomatic subjects fully complies with the requirement of detection of at least one positive sample out of four tested that is considered to be sufficient to ascertain the aetiology of a NoV outbreak [24]. Moreover, molecular typing indicated that a single viral strain, belonging to genotype GII.P2/GII.2, infected all symptomatic subjects involved in the outbreak. NoVs are highly infectious and can easily spread from person to person but also through contaminated food or water. Therefore, settings such as health care institutions, schools and touristic resorts are mainly affected, since close contact and shared food services facilitate transmission. In Mascali, two waves of transmission were observed over three days, being consistent with food/waterborne followed by person-to-person and/or airborne (vomit) transmission (Fig. 2). The first wave of cases presented on 20 May 2016 after a single shared meal, while the last case was detected on 22 May. Exposure to vomit may have been responsible for some of the secondary cases occurred after the first wave of illness. Vomiting produces aerosols that can be inhaled and swallowed. The extremely low infectious dose of NoV makes the inhalation of vomit aerosols an effective transmission route in settings where there is close contact, like on a bus trip [25]. All food consumed in the only shared lunch on 19 May at the resort was associated with the risk of disease (Table 2), but the association was inconclusive since all cases shared the same dishes included in the lunch menu on Day 1. Although many foods were potentially involved, our attention focused on the tap water. In fact, trivial information from local health authorities reported that local population avoided tap water consumption and relied on bottled water for drinking and cooking. Many small outbreaks of gastroenteritis involving restaurants and family groups had been reported in the same area, especially at the beginning of the summer season, when vacation houses and recreational activities were re-opened after a period of low water demand. Illness caused by NoV is common and if the level of water contamination is low, the number of cases remains limited and water can persist as unrecognised source of infection. If polluted water was served for drinking and used for cooking and washing dishes in the Mascali resort at Day 1 a generalised contamination of food and surfaces could have occurred. The potential role of the foodhandlers as a source of contamination could not be analysed in this study but should also be taken into account. If water contamination has already been present in the area for a long time, infections may have occurred among the local food-handlers. The analysis of water samples taken immediately after the outbreak revealed the presence of bacterial indicators of faecal contamination in the water used in the resort, with the only exception of a sample of filter purified water used both for cooking and as drinking water. When extending the sampling to other areas of the municipal water network, many other sites, mostly on the seaside, showed faecal pollution, which demonstrates a poor protection of the water network from faecal leakage (Fig. 1). NoV contamination of waters generally originates from sewage since these enteric pathogens, as other enteric viruses affecting humans, are mostly species-specific and are shed at high concentrations in the stools of patients. When the analyses on the stools of the students made it clear that the outbreak had been caused by NoV, the viral genome was also searched in appropriate volumes of water samples, but this happened when correction measures on the water network had already been taken. Nevertheless, a single water sample, obtained 2 months after the outbreak (1 August 2016), still contained genome fragments that could be correlated to the same NoV genotype causing the outbreak. Although the finding of NoV genome does not demonstrate that entire infecting viral particles were present in the water, it could indicate that a NoV belonging to the same genotype involved in the outbreak was still circulating in the local population several weeks after the outbreak and it was someway able to penetrate the water network, probably indicating continuing faecal contamination through leakage from sewage systems. NoV is particularly resistant in the environment and also to chlorine, therefore transmission through chlorinated water is not exceptional [9, 12, 26]. Drinking water from municipal supplies fed by lake or groundwater has already been identified in the past as the source of waterborne NoV outbreaks in Italy [27–29] and NoV is now considered one of the pathogens causing the largest number of affected consumers in drinking waterborne outbreaks [12]. NoV can persist in water longer than 15 days [30] and it is resistant to chlorine disinfection when free chlorine levels or pre-treatment processes are inadequate [31, 32]. Moreover, microfiltration filters labelled as certified by NSF Standards 53 or 58 (with a pore size ranging from 0.05 to 5 µm) are not effective in removing viruses and can even decrease chlorine concentration in filtered water [33]. Actually, the owner of the resort involved in our outbreak was encouraged to serve tap water as drinking water at the restaurant relying on the effectiveness of such a filter purifier. Unfortunately, we did not have a tap water sample from the restaurant of the resort taken before correction measures and whose volume was appropriate for virus detection. Therefore, we could not demonstrate the failure to eliminate viral pollution of the microfiltration system. In Mascali, correction measures on the water network consisted in the maintenance of the old chlorination device and in the addition of a new one at a key point located upstream and close to suspected illegal sewage dumping (Fig. 1), but as poor protection of the pipeline from sewage leakage risked to reproduce faecal pollution episodes, renovation of the municipal water system was also advised, as well as systematic control of illegal sewage dumping. The additional chlorination device was apparently the most effective intervention, since right after its activation faecal bacterial pollution indicators disappeared. However, NoV genome was detected in a water sample taken 6 weeks after the last detection of bacterial indicators. Unfortunately, detection of faecal contamination of drinking water is only based on the use of bacterial indicators despite no correlation between levels of enteric bacteria and enteric viruses has been found [34–36]. According to ISGEV enteric virus circulation surveillance data, in 2016 the circulation of GII.P2/GII.2 NoV strains in Palermo, home town of the travellers involved in the outbreak, was the highest ever observed from the starting of local NoV surveillance in 2004. GII.P2/GII.2 was the fourth most common genotype in 2016 (12.2% prevalence), following GII.P4/GII.4 Sydney 2012 (28.4%), GII.P16/GII.2 (18.9%) and GII.Pe/GII.4 Sydney 2012 (17.6%). GII.P2/GII.2 strains were mostly observed starting from July 2016, leaving space for the possibility that their increased circulation was linked to the introduction of the ‘Mascali clone’. However, a couple of GII.P2/GII.2 strains (including PA36/16) genetically close to the ‘Mascali clone’ had been already detected in January 2016 in Palermo and GII.P2/GII.2 NoV had sporadically been observed also in 2011 and 2014. Moreover, the circulation of GII.P2/GII.2 NoVs has been increasing worldwide over the last period [37]. Therefore, the unusually high circulation observed in Palermo might just reflect the global epidemiology of GII.P2/GII.2 strains. Noteworthy, the older GII.P2/GII.2 NoV strains detected in Palermo were closely related in the ORF2 (98.9% nt id) to the GII.P16/GII.2 Nashville strain (KY865307) which is supposed to have generated the recombinant GII.P16/GII.4 clone recently circulating epidemically worldwide [37, 38] and were more distantly associated (95.3% nt id) to the GII.2 capsid sequence (AB662868) of another putative GII.P16/GII.4 ancestor from Japan [39]. The increasing frequency of detection of waterborne NoV outbreaks calls for the introduction of routine monitoring for viruses of drinking waters and for the definition of affordable and reliable indicators for the presence of NoV in drinking waters. The use of ultrafiltration systems as a concentration method in order to recover viruses from environmental waters combined with RT-PCR to detect viral genomes has been demonstrated to grant a recovery rate of >50% of multiple viruses [40] and is at present one of the most sensitive approaches for the detection of NoVs, although the discrimination between viable and nonviable viruses remains a problem. The burden of NoV outbreaks, either food associated or waterborne, is still poorly defined due to the lack of specific diagnostic tools in many clinical laboratories. NoV epidemics should always be suspected when the main symptom is vomiting and the duration of disease is short (<48 h). Waterborne transmission is suggested by the rapid spread of disease and the high attack rate. Care should be taken to perform sensitive diagnostic tests on a sufficient number of stool samples in order to identify NoV as the causative agent of enteritis outbreaks, possibly preferring genome screening over antigen detection. Only the systematic detection of NoV outbreaks can guide in the choice of the safety procedures that could be most appropriate in containing such microbiological hazard.

In which city did this happen?

1539

Waterborne Norovirus outbreak at a seaside resort likely originating from municipal water distribution system failure

Abstract In May 2016 a Norovirus (NoV) gastroenteritis outbreak involved a high school class visiting a seaside resort near Taormina (Mascali, Sicily). Twenty-four students and a teacher were affected and 17 of them showed symptoms on the second day of the journey, while the others got ill within the following 2 days. Symptoms included vomiting, diarrhoea and fever, and 12 students required hospitalisation. Stool samples tested positive for NoV genome by Real-Time polymerase chain reaction assay in all 25 symptomatic subjects. The GII.P2/GII.2 NoV genotype was linked to the outbreak by ORF1/ORF2 sequence analysis. The epidemiological features of the outbreak were consistent with food/waterborne followed by person-to-person and/ or vomit transmission. Food consumed at a shared lunch on the first day of the trip was associated to illness and drinking un-bottled tap water was also considered as a risk factor. The analysis of water samples revealed the presence of bacterial indicators of faecal contamination in the water used in the resort as well as in other areas of the municipal water network, linking the NoV gastroenteritis outbreak to tap water pollution from sewage leakage. From a single water sample, an amplicon whose sequence corresponded to the capsid genotype recovered from patients could be obtained. Introduction Noroviruses (NoVs) are recognised as a leading cause of sporadic cases and outbreaks of acute gastroenteritis (AGE) across all age groups [1, 2]. They are transmitted primarily through the faecal-oral route either by direct person–person spread and with consumption of contaminated food or water [1, 3, 4]. NoVs are non-enveloped RNA viruses belonging to the Caliciviridae family and are classified into seven distinct genogroups (GI-GVII) on the basis of the major capsid protein VP1 sequence. GI, GII and GIV NoVs infect humans and they have been divided into more than 30 genotypes based on the sequences of polymerase (ORF1) and capsid proteins (ORF2) [5]. NoV GII.4 genotype has been associated with the vast majority of NoV related cases of AGE worldwide although many genotypes, belonging to both GI and GII genogroups are involved in foodborne and waterborne outbreaks [6, 7]. NoVs are highly contagious owing to their high stability in the environment and their ability to infect at low doses. In fact, just 10–100 virions of NoV may be sufficient to cause AGE [8, 9]. NoVs are considered the major cause of water-related disease in semi-closed institutions, including restaurants, schools, hospitals, nursing homes and cruise ships [10–12]. To confirm NoV as a cause of outbreaks, the presence of four epidemiological features of disease, vomiting in more than half of affected persons, a mean incubation period of 24/48 h, a mean duration of illness of 12–60 h and the absence of bacterial pathogens in stool culture, should be evaluated [13]. Waterborne NoV outbreaks can be caused by contaminated drinking water despite correct chlorine concentration and in the absence of coliform bacteria [10, 14–17]. The aim of this study was to investigate an outbreak of NoV gastroenteritis associated with drinking water involving a high school class travelling from Palermo (West Sicily) to Acireale (East Sicily) on an educational tour in May 2016. Stool samples were collected soon after the onset of AGE symptoms to ascertain the aetiology of illness. The source of infection was investigated collecting epidemiological data through an online questionnaire and, in order to explore a possible transmission route, laboratory tests on water from the municipal network were performed. Material and methods Epidemiological investigation From 20 to 22 May 2016 an AGE outbreak occurred in Mascali, a small town near Taormina, involving a school class of 28 students, 13 girls and 15 boys aged between 14 and 18 years and a 54-year-old female teacher, during a trip to participate in a theatre festival in Acireale. A total of 25/29 persons were affected by nausea, vomiting, abdominal cramps, diarrhoea and fever while they were staying at a seaside resort in Mascali, on the way to Acireale. To obtain data on symptoms and the correlated risk factors all participants filled an online questionnaire investigating personal data (age and sex), travel and accommodation details and food consumption over 48 h before symptoms presentation, time of onset and duration of symptoms, therapy and/or hospitalisation required and history of contact with people affected by diarrhoea and/or vomiting (Table 1). The severity of illness was analysed by the Vesikari severity scale allowing to distinguish non-severe (score <11) and severe (score ⩾11) AGE (Table 1). Vesikari score was calculated from maximum number of episodes/day and duration of both diarrhoea and vomiting (score 0–3 for each of the 4 parameters), temperature (score 0–3), level of dehydration (score 0–3) and treatment required (score 0–2) [18]. A case-control study was conducted to establish if food consumed before the onset of symptoms was associated with the disease. A case was defined as a student attending the tour and reporting diarrhoea and/or vomiting within 36 h from the first shared meal during the tour, while students who did not present any of the previous symptoms or those showing the first symptoms with a delay of ⩾48 h (secondary cases) were used as controls. Variables were studied using odds ratio (OR) and its 95% CI and statistical associations were determined using the χ2 test, with the level of statistical significance set at 0.05 (Table 2). NoV detection in stool samples A stool sample was obtained from each of the 29 travellers, soon after (3–5 days) the onset of the outbreak. No stool samples could be obtained from the food-handlers working at the resort. All samples were tested for NoV at the Enteric Viruses Laboratory of the University of Palermo, member of the Italian Study Group for Enteric Viruses (ISGEV), by a Real-Time polymerase chain reaction (RT-PCR) assay that allows to differentiate GI and GII NoVs [19]. Total RNA was extracted from stool samples using QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) and cDNA was generated by MMLV reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were considered positive for NoVs when cycle threshold (ct) of RT-PCR was ⩽30 and lowpositive when 30<ct⩽35. For all NoV-positive samples two genomic fragments were amplified by conventional RT-PCR using Superscript III One-Step RT-PCR system with Platinum Taq DNA polymerase (Invitrogen) and primers JV12a/JV13b, targeting 326 bps of the RdRp region A [20] and either COG1F/ GISKR or COG2F/GIISKR, amplifying a 315 bps 5′portion of GI or GII VP1 protein [19, 21]. Environmental investigation From May through August 2016, a total of 32 tap water samples were collected from the Mascali municipal network. In particular, seven samples were collected at the resort where the outbreak occurred and from its water entry (R), two in the city centre of Mascali (M), 15 in the village of Fondachello (F), on the seaside, three in Carrabba village (C), far from the coast, and two and three, respectively, from the water well ‘Carlino’ (W) and the source ‘Bufardo and Torregrossa’ (S) which feed the municipal system and provide drinking water to the Mascali area (Fig. 1, Table 3). One litre water samples were sent by the local health authorities to the SIALAB laboratory in Avola, Siracusa and 100 ml aliquots were screened for coliforms, E. coli and enterococci as faecal indicators, by standard bacteriological methods according to UNI EN ISO 9308-1:2014. Unfortunately, although the suspect of a water source for the viral outbreak was quickly communicated, only in August and November 2016, following the implementation of control measures, local authorities provided water samplings of appropriate volume to perform viral tests (10 l). Five water samples of 10 l each were collected from different sampling sites in F and R areas for viral screening (Table 3) and were concentrated according to the Italian National Institute of Health (Istituto Superiore di Sanità, ISS) protocols [22] through an ultra-filtration system (Sartoflow® Slice 200 Benchtop Crossflow System, Sartorius AG, Goettingen, Germany), using appropriate membranes (SG Hydrosart 10 kDa) pretreated with 300 ml of 3% Beef Extract (BE) at pH 7, by recirculating it for 10 min in the system, at a pressure not higher than 1.30 bar. Elution and recovery of viruses attached to the membranes were obtained by recirculating 200 ml of 3% BE at pH 9.5 for 3 h until a 10–12 ml volume of eluate was obtained. The pH of the eluate was then checked with litmus paper and if necessary brought to the value of 7 with HCl. Concentrated water samples were simultaneously tested for NoV in three laboratories: Enteric Viruses Laboratory, University of Palermo; Istituto Zooprofilattico Sperimentale della Sicilia (IZS), Palermo; and Istituto Superiore di Sanità (ISS), Rome. Total RNA was extracted from 500 and 560 µl of a concentrated water sample using BioMèrieux NucliSENS® Magnetic Extraction System (BioMèrieux) and QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) respectively, according to manufacturer’s instructions. Bonded RNA was eluted by 100 µl of Elution Buffer. Amplifications were performed at least in duplicate on whole and 1:10 diluted extracts. A Real-Time RT-PCR assay according to ISO/TS 15216- 2:2013, two in-house Real-Time RT-PCR protocols and the One-Step RT-PCR methods already used for amplifying ORF1 and ORF2 fragments in NoV-positive faecal samples were performed [19–21, 23]. Nucleotide sequences analysis and phylogenetic analysis PCR amplicons were purified by commercial columns (CleanSweep PCR Purification, Thermo Fisher Scientific, Vilnius, Lithuania) and sequenced by Sanger method and all nucleotide sequences were aligned using ClustalW. To identify NoV genotypes, RdRp and VP1 sequences were submitted to the RIVM NoV genotyping tool (http://www.rivm.nl/mpf/norovirus/typingtool). Phylogenetic trees were constructed using MEGA 7 and Neighbour-Joining method (bootstrap 1000 replicates). Sequences from stool and water samples are available at NCBI with the following accession numbers: MG517455 for the major capsid protein obtained from a water sample; MG517456 and MG517461, respectively, for the major capsid protein and RNA-dependent-RNA-polymerase, as representatives of the sequences obtained from the stool samples of those involved in the outbreak. Control measures After outbreak notification, on 21 May, the resort was closed for 2 days. On 25 May, following the first results of the bacteriological analyses on municipal water samples, the use of municipal water for human consumption was temporarily restricted (use of water was permitted only for cleaning purposes) in the Carrabba and Fondachello villages and in the area surrounding the resort. From 25 May the reservoir collecting the water from the Carlino well and its chlorination device were submitted to maintenance work and a new chlorination device was installed on the conduit bringing the water from the Bufardo source. Actually, the Bufardo source had been connected to the municipal network in 2003 but unfortunately a dedicated chlorination station had not been planned at that time. Therefore, unchlorinated water entered the system near Fondachello village on the seaside (Fig. 1), possibly diluting chlorine concentrations in water treated at the Carlino reservoir. Moreover, the exact path of the branch of the water network situated downstream the Bufardo source water connection was not marked in the maps of the municipality (Fig. 1, red line), indicating that this section had never been submitted to maintenance. The area surrounding the unmaintained section of the water pipeline was scattered with unauthorised cottages that were not connected to the sewerage systems. Likely, untreated waste from cesspools was unloaded directly into the soil, generating a risk of contamination of the municipal water system through leakage. No work had been made on the water pipelines before the outbreak but sections of the network had been temporarily closed during the winter season on the seaside (Fondachello). Local health authorities advised the connection of all households in the area to the sewerage system and the replacement of the water system section without maintenance. Health authorities also raised attention on AGE cases notification at all local Emergency Departments and Family Practitioners but there was no further evidence of AGE outbreaks in the months following the intervention. Results Data collected through the online questionnaire compiled 7–10 days after the outbreak revealed that none of the 28 students and the teacher taking part in the school trip showed gastroenteritis symptoms at the beginning of the journey and they had no contact with people who suffered from diarrhoea and/or vomiting in the previous days. Travellers had their first shared meal on 19 May (Day 1), 2016, at lunch on their arrival to the resort in Mascali. The list of foods included in the menu is shown in Table 2. Conversely, the dinner on Day 1 was not shared and the students went to dine in three different fast food restaurants in Acireale. On 20 May (Day 2), 17 students (58.6% of the travellers) presented with vomit and/or diarrhoea, falling ill 26–36 h after lunch on Day 1, strongly suggesting that exposure occurred on the first shared meal. Some students also vomited in the presence of the other travellers on the bus bringing them to Acireale. Five more students were symptomatic in the early hours of the following day (21 May, Day 3, from 0:30 am to 8:00 am), as a continuation of the wave of illnesses of the previous day. The teacher and one more student fell ill at 8:00 pm and 11:00 pm of Day 3, respectively, for a second wave of illnesses occurring >48 h from the first shared meal and <36 h from the first cases, with the last student falling sick on 22 May (Day 4) at 3:00 am (Fig. 2, Table 1). On the whole, the mean incubation period was 33.2 h. Based on the online survey, the severity of illness was evaluated by the Vesikari severity scale, grouping symptomatic travellers into severe (11/25 = 44%) and non-severe (14/25 = 56%) cases (Table 1). Vomiting and fever were the predominant symptoms, affecting 22 out of 25 (88%) symptomatic subjects, while diarrhoea was present in 14 out of 25 (56%). Hospitalisation for rehydration therapy was necessary for 13 students (91% [10/11] of severe and 21% [3/14] of non-severe cases). Emergency room (ER) or family practitioner visits were necessary for 21 students and 13 (62%) of them initiated an antibiotic therapy. Two students indicated that other family members had gastroenteric symptoms more than 24 h after their return home, on Day 3, 4 and 5. All those involved in the outbreak recovered within 4 days, with an average duration of illness of 38.4 h. A case-control study of the risk associated with eating foods at lunch on Day 1 or with attending different restaurants at dinner on Day 1 was attempted. A total of 19 subjects filled the section of the online survey concerning food consumption. Fourteen of the responders were among the 22 cases falling ill within 36 h from the first shared meal (first wave), while the remaining five responders were three travellers with supposed secondary infections (showing symptoms with a delay ⩾48 h; second wave) and two asymptomatic students that were used as controls. No statistically significant associations with any food items served at lunch on Day 1 was identified (P < 0.05; Table 2). However, all food consumed on that occasion was to some extent associated with the risk of disease, the highest OR being obtained with chicken (OR: 20.7143, P: 0.0678; Table 2). Actually, most of the cases ate all foods included in the menu, while one of the two asymptomatic controls did not participate in the meal and the other had a pasta dish only. No statistically significant association was found for attending a particular fast food restaurant at the dinner on Day 1 (OR: 2.5, P: 0.3635). The Real-Time RT-PCR assay allowed to detect GII genogroup NoV genome in all the 25 samples from symptomatic subjects, although two of them showed high Ct values, indicating lower shedding of viral genomes (Table 1). The samples from four asymptomatic students were negative for both NoV genogroups. By sequence analysis, a NoV genotype could be defined for 20/ 25 NoV-positive stool samples. All NoV strains from travellers involved in the outbreaks belonged to the same genotype, GII.P2/GII.2, showing >99% nucleotide identities (nt id) in both fragments (ORF1 and ORF2) analysed. Thirty-two water samples were collected and analysed for bacterial contamination (Table 3). Seven samples were taken from the resort (site R) and the remaining 25 from five other sites (C, F, M, S, W), representative of the entire municipal water network. The majority (10/12) of the water samples collected immediately after the onset of the outbreak, between 21 May and 16 June 2016, at the resort (site R) and at sites F and C, located close to the resort, on the seaside and in the interior, respectively, were positive for coliform bacteria. More than 1000 colony-forming units (CFU)/ 100 ml (up to 2700 CFU/100 ml) were detected in four samples, while their absence is required in 100 ml by the Standard of Water Quality in Italy and seven of 11 samples tested were also Escherichia coli-positive (up to 900 CFU/100 ml). Enterococci were searched in 11 samples and three of them were found positive for this faecal contamination indicator. From 7 July to 16 August 2016, after control measures on the water network were undertaken, 13 water samples were collected for follow-up laboratory examination of the water system at F, R and C sites. All of them were negative for bacterial indicators of faecal contamination. No faecal bacteria were identified in the seven samples collected from M, W and S sites, including the well and the source that fed the water network, at any sampling time, before and after implementing the control measures. The research of NoV genome was conducted on concentrated water samples collected on 1 August from R and F sites (Table 3). NoV genome was detected in a single sample collected from the water entry of the resort involved in the outbreak (site R). Actually, only two out of 26 One-Step RT-PCR repetitions were positive for NoV viral capsid gene. The two ORF2-positive results were obtained at the University of Palermo Laboratory using both 1:10 diluted and undiluted extracted RNA. The sequence analysis of the partial ORF2 portion (289 bp) indicated the presence of the same GII.2 capsid genotype detected in the stool samples of the students involved in the outbreak, showing 98.9% nt id. No ORF1 NoV sequence could be obtained from the NoV-positive water sample. Follow-up analyses for NoV genome, conducted on water samples collected on 20 November both at R and F sites found no positive samples. By sequence alignment, the NoV sequences obtained from Mascali outbreak samples were phylogenetically closely related in both ORF1 and ORF2 (99 and 98.9% of nt id, respectively) to a GII.P2/GII.2 strain responsible for a sporadic paediatric gastroenteritis case occurred in Palermo in January 2016 (PA 36/16, accession numbers MG517464 and MG517460, for ORF1 and ORF2, respectively) and previously detected at the Enteric Viruses Laboratory, University of Palermo. Whereas, three GII.P2/GII.2 Italian strains (PA 280/11, PA 366/11 and PA 3/ 14, accession numbers MG517462, MG517465 and MG517463, for ORF1 and MG517457 to −59, for ORF2, respectively), detected in Palermo in 2011 and 2014, were more distantly associated with the outbreak strain both in RdRp and VP1 (90% to 97.4% nt id and 96.1% to 98.1% nt id, respectively, in ORF1 and ORF2). Discussion Epidemiological features, such as the presence of vomiting in more than half (88%) of symptomatic subjects, the 33.2 h average incubation period and the 38.4 h average duration of illness, suggested a NoV aetiology for the investigated outbreak [13]. The detection of NoV genome in all 25 faecal samples from symptomatic subjects fully complies with the requirement of detection of at least one positive sample out of four tested that is considered to be sufficient to ascertain the aetiology of a NoV outbreak [24]. Moreover, molecular typing indicated that a single viral strain, belonging to genotype GII.P2/GII.2, infected all symptomatic subjects involved in the outbreak. NoVs are highly infectious and can easily spread from person to person but also through contaminated food or water. Therefore, settings such as health care institutions, schools and touristic resorts are mainly affected, since close contact and shared food services facilitate transmission. In Mascali, two waves of transmission were observed over three days, being consistent with food/waterborne followed by person-to-person and/or airborne (vomit) transmission (Fig. 2). The first wave of cases presented on 20 May 2016 after a single shared meal, while the last case was detected on 22 May. Exposure to vomit may have been responsible for some of the secondary cases occurred after the first wave of illness. Vomiting produces aerosols that can be inhaled and swallowed. The extremely low infectious dose of NoV makes the inhalation of vomit aerosols an effective transmission route in settings where there is close contact, like on a bus trip [25]. All food consumed in the only shared lunch on 19 May at the resort was associated with the risk of disease (Table 2), but the association was inconclusive since all cases shared the same dishes included in the lunch menu on Day 1. Although many foods were potentially involved, our attention focused on the tap water. In fact, trivial information from local health authorities reported that local population avoided tap water consumption and relied on bottled water for drinking and cooking. Many small outbreaks of gastroenteritis involving restaurants and family groups had been reported in the same area, especially at the beginning of the summer season, when vacation houses and recreational activities were re-opened after a period of low water demand. Illness caused by NoV is common and if the level of water contamination is low, the number of cases remains limited and water can persist as unrecognised source of infection. If polluted water was served for drinking and used for cooking and washing dishes in the Mascali resort at Day 1 a generalised contamination of food and surfaces could have occurred. The potential role of the foodhandlers as a source of contamination could not be analysed in this study but should also be taken into account. If water contamination has already been present in the area for a long time, infections may have occurred among the local food-handlers. The analysis of water samples taken immediately after the outbreak revealed the presence of bacterial indicators of faecal contamination in the water used in the resort, with the only exception of a sample of filter purified water used both for cooking and as drinking water. When extending the sampling to other areas of the municipal water network, many other sites, mostly on the seaside, showed faecal pollution, which demonstrates a poor protection of the water network from faecal leakage (Fig. 1). NoV contamination of waters generally originates from sewage since these enteric pathogens, as other enteric viruses affecting humans, are mostly species-specific and are shed at high concentrations in the stools of patients. When the analyses on the stools of the students made it clear that the outbreak had been caused by NoV, the viral genome was also searched in appropriate volumes of water samples, but this happened when correction measures on the water network had already been taken. Nevertheless, a single water sample, obtained 2 months after the outbreak (1 August 2016), still contained genome fragments that could be correlated to the same NoV genotype causing the outbreak. Although the finding of NoV genome does not demonstrate that entire infecting viral particles were present in the water, it could indicate that a NoV belonging to the same genotype involved in the outbreak was still circulating in the local population several weeks after the outbreak and it was someway able to penetrate the water network, probably indicating continuing faecal contamination through leakage from sewage systems. NoV is particularly resistant in the environment and also to chlorine, therefore transmission through chlorinated water is not exceptional [9, 12, 26]. Drinking water from municipal supplies fed by lake or groundwater has already been identified in the past as the source of waterborne NoV outbreaks in Italy [27–29] and NoV is now considered one of the pathogens causing the largest number of affected consumers in drinking waterborne outbreaks [12]. NoV can persist in water longer than 15 days [30] and it is resistant to chlorine disinfection when free chlorine levels or pre-treatment processes are inadequate [31, 32]. Moreover, microfiltration filters labelled as certified by NSF Standards 53 or 58 (with a pore size ranging from 0.05 to 5 µm) are not effective in removing viruses and can even decrease chlorine concentration in filtered water [33]. Actually, the owner of the resort involved in our outbreak was encouraged to serve tap water as drinking water at the restaurant relying on the effectiveness of such a filter purifier. Unfortunately, we did not have a tap water sample from the restaurant of the resort taken before correction measures and whose volume was appropriate for virus detection. Therefore, we could not demonstrate the failure to eliminate viral pollution of the microfiltration system. In Mascali, correction measures on the water network consisted in the maintenance of the old chlorination device and in the addition of a new one at a key point located upstream and close to suspected illegal sewage dumping (Fig. 1), but as poor protection of the pipeline from sewage leakage risked to reproduce faecal pollution episodes, renovation of the municipal water system was also advised, as well as systematic control of illegal sewage dumping. The additional chlorination device was apparently the most effective intervention, since right after its activation faecal bacterial pollution indicators disappeared. However, NoV genome was detected in a water sample taken 6 weeks after the last detection of bacterial indicators. Unfortunately, detection of faecal contamination of drinking water is only based on the use of bacterial indicators despite no correlation between levels of enteric bacteria and enteric viruses has been found [34–36]. According to ISGEV enteric virus circulation surveillance data, in 2016 the circulation of GII.P2/GII.2 NoV strains in Palermo, home town of the travellers involved in the outbreak, was the highest ever observed from the starting of local NoV surveillance in 2004. GII.P2/GII.2 was the fourth most common genotype in 2016 (12.2% prevalence), following GII.P4/GII.4 Sydney 2012 (28.4%), GII.P16/GII.2 (18.9%) and GII.Pe/GII.4 Sydney 2012 (17.6%). GII.P2/GII.2 strains were mostly observed starting from July 2016, leaving space for the possibility that their increased circulation was linked to the introduction of the ‘Mascali clone’. However, a couple of GII.P2/GII.2 strains (including PA36/16) genetically close to the ‘Mascali clone’ had been already detected in January 2016 in Palermo and GII.P2/GII.2 NoV had sporadically been observed also in 2011 and 2014. Moreover, the circulation of GII.P2/GII.2 NoVs has been increasing worldwide over the last period [37]. Therefore, the unusually high circulation observed in Palermo might just reflect the global epidemiology of GII.P2/GII.2 strains. Noteworthy, the older GII.P2/GII.2 NoV strains detected in Palermo were closely related in the ORF2 (98.9% nt id) to the GII.P16/GII.2 Nashville strain (KY865307) which is supposed to have generated the recombinant GII.P16/GII.4 clone recently circulating epidemically worldwide [37, 38] and were more distantly associated (95.3% nt id) to the GII.2 capsid sequence (AB662868) of another putative GII.P16/GII.4 ancestor from Japan [39]. The increasing frequency of detection of waterborne NoV outbreaks calls for the introduction of routine monitoring for viruses of drinking waters and for the definition of affordable and reliable indicators for the presence of NoV in drinking waters. The use of ultrafiltration systems as a concentration method in order to recover viruses from environmental waters combined with RT-PCR to detect viral genomes has been demonstrated to grant a recovery rate of >50% of multiple viruses [40] and is at present one of the most sensitive approaches for the detection of NoVs, although the discrimination between viable and nonviable viruses remains a problem. The burden of NoV outbreaks, either food associated or waterborne, is still poorly defined due to the lack of specific diagnostic tools in many clinical laboratories. NoV epidemics should always be suspected when the main symptom is vomiting and the duration of disease is short (<48 h). Waterborne transmission is suggested by the rapid spread of disease and the high attack rate. Care should be taken to perform sensitive diagnostic tests on a sufficient number of stool samples in order to identify NoV as the causative agent of enteritis outbreaks, possibly preferring genome screening over antigen detection. Only the systematic detection of NoV outbreaks can guide in the choice of the safety procedures that could be most appropriate in containing such microbiological hazard.

In which region did this happen?

1540

Waterborne Norovirus outbreak at a seaside resort likely originating from municipal water distribution system failure

Abstract In May 2016 a Norovirus (NoV) gastroenteritis outbreak involved a high school class visiting a seaside resort near Taormina (Mascali, Sicily). Twenty-four students and a teacher were affected and 17 of them showed symptoms on the second day of the journey, while the others got ill within the following 2 days. Symptoms included vomiting, diarrhoea and fever, and 12 students required hospitalisation. Stool samples tested positive for NoV genome by Real-Time polymerase chain reaction assay in all 25 symptomatic subjects. The GII.P2/GII.2 NoV genotype was linked to the outbreak by ORF1/ORF2 sequence analysis. The epidemiological features of the outbreak were consistent with food/waterborne followed by person-to-person and/ or vomit transmission. Food consumed at a shared lunch on the first day of the trip was associated to illness and drinking un-bottled tap water was also considered as a risk factor. The analysis of water samples revealed the presence of bacterial indicators of faecal contamination in the water used in the resort as well as in other areas of the municipal water network, linking the NoV gastroenteritis outbreak to tap water pollution from sewage leakage. From a single water sample, an amplicon whose sequence corresponded to the capsid genotype recovered from patients could be obtained. Introduction Noroviruses (NoVs) are recognised as a leading cause of sporadic cases and outbreaks of acute gastroenteritis (AGE) across all age groups [1, 2]. They are transmitted primarily through the faecal-oral route either by direct person–person spread and with consumption of contaminated food or water [1, 3, 4]. NoVs are non-enveloped RNA viruses belonging to the Caliciviridae family and are classified into seven distinct genogroups (GI-GVII) on the basis of the major capsid protein VP1 sequence. GI, GII and GIV NoVs infect humans and they have been divided into more than 30 genotypes based on the sequences of polymerase (ORF1) and capsid proteins (ORF2) [5]. NoV GII.4 genotype has been associated with the vast majority of NoV related cases of AGE worldwide although many genotypes, belonging to both GI and GII genogroups are involved in foodborne and waterborne outbreaks [6, 7]. NoVs are highly contagious owing to their high stability in the environment and their ability to infect at low doses. In fact, just 10–100 virions of NoV may be sufficient to cause AGE [8, 9]. NoVs are considered the major cause of water-related disease in semi-closed institutions, including restaurants, schools, hospitals, nursing homes and cruise ships [10–12]. To confirm NoV as a cause of outbreaks, the presence of four epidemiological features of disease, vomiting in more than half of affected persons, a mean incubation period of 24/48 h, a mean duration of illness of 12–60 h and the absence of bacterial pathogens in stool culture, should be evaluated [13]. Waterborne NoV outbreaks can be caused by contaminated drinking water despite correct chlorine concentration and in the absence of coliform bacteria [10, 14–17]. The aim of this study was to investigate an outbreak of NoV gastroenteritis associated with drinking water involving a high school class travelling from Palermo (West Sicily) to Acireale (East Sicily) on an educational tour in May 2016. Stool samples were collected soon after the onset of AGE symptoms to ascertain the aetiology of illness. The source of infection was investigated collecting epidemiological data through an online questionnaire and, in order to explore a possible transmission route, laboratory tests on water from the municipal network were performed. Material and methods Epidemiological investigation From 20 to 22 May 2016 an AGE outbreak occurred in Mascali, a small town near Taormina, involving a school class of 28 students, 13 girls and 15 boys aged between 14 and 18 years and a 54-year-old female teacher, during a trip to participate in a theatre festival in Acireale. A total of 25/29 persons were affected by nausea, vomiting, abdominal cramps, diarrhoea and fever while they were staying at a seaside resort in Mascali, on the way to Acireale. To obtain data on symptoms and the correlated risk factors all participants filled an online questionnaire investigating personal data (age and sex), travel and accommodation details and food consumption over 48 h before symptoms presentation, time of onset and duration of symptoms, therapy and/or hospitalisation required and history of contact with people affected by diarrhoea and/or vomiting (Table 1). The severity of illness was analysed by the Vesikari severity scale allowing to distinguish non-severe (score <11) and severe (score ⩾11) AGE (Table 1). Vesikari score was calculated from maximum number of episodes/day and duration of both diarrhoea and vomiting (score 0–3 for each of the 4 parameters), temperature (score 0–3), level of dehydration (score 0–3) and treatment required (score 0–2) [18]. A case-control study was conducted to establish if food consumed before the onset of symptoms was associated with the disease. A case was defined as a student attending the tour and reporting diarrhoea and/or vomiting within 36 h from the first shared meal during the tour, while students who did not present any of the previous symptoms or those showing the first symptoms with a delay of ⩾48 h (secondary cases) were used as controls. Variables were studied using odds ratio (OR) and its 95% CI and statistical associations were determined using the χ2 test, with the level of statistical significance set at 0.05 (Table 2). NoV detection in stool samples A stool sample was obtained from each of the 29 travellers, soon after (3–5 days) the onset of the outbreak. No stool samples could be obtained from the food-handlers working at the resort. All samples were tested for NoV at the Enteric Viruses Laboratory of the University of Palermo, member of the Italian Study Group for Enteric Viruses (ISGEV), by a Real-Time polymerase chain reaction (RT-PCR) assay that allows to differentiate GI and GII NoVs [19]. Total RNA was extracted from stool samples using QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) and cDNA was generated by MMLV reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were considered positive for NoVs when cycle threshold (ct) of RT-PCR was ⩽30 and lowpositive when 30<ct⩽35. For all NoV-positive samples two genomic fragments were amplified by conventional RT-PCR using Superscript III One-Step RT-PCR system with Platinum Taq DNA polymerase (Invitrogen) and primers JV12a/JV13b, targeting 326 bps of the RdRp region A [20] and either COG1F/ GISKR or COG2F/GIISKR, amplifying a 315 bps 5′portion of GI or GII VP1 protein [19, 21]. Environmental investigation From May through August 2016, a total of 32 tap water samples were collected from the Mascali municipal network. In particular, seven samples were collected at the resort where the outbreak occurred and from its water entry (R), two in the city centre of Mascali (M), 15 in the village of Fondachello (F), on the seaside, three in Carrabba village (C), far from the coast, and two and three, respectively, from the water well ‘Carlino’ (W) and the source ‘Bufardo and Torregrossa’ (S) which feed the municipal system and provide drinking water to the Mascali area (Fig. 1, Table 3). One litre water samples were sent by the local health authorities to the SIALAB laboratory in Avola, Siracusa and 100 ml aliquots were screened for coliforms, E. coli and enterococci as faecal indicators, by standard bacteriological methods according to UNI EN ISO 9308-1:2014. Unfortunately, although the suspect of a water source for the viral outbreak was quickly communicated, only in August and November 2016, following the implementation of control measures, local authorities provided water samplings of appropriate volume to perform viral tests (10 l). Five water samples of 10 l each were collected from different sampling sites in F and R areas for viral screening (Table 3) and were concentrated according to the Italian National Institute of Health (Istituto Superiore di Sanità, ISS) protocols [22] through an ultra-filtration system (Sartoflow® Slice 200 Benchtop Crossflow System, Sartorius AG, Goettingen, Germany), using appropriate membranes (SG Hydrosart 10 kDa) pretreated with 300 ml of 3% Beef Extract (BE) at pH 7, by recirculating it for 10 min in the system, at a pressure not higher than 1.30 bar. Elution and recovery of viruses attached to the membranes were obtained by recirculating 200 ml of 3% BE at pH 9.5 for 3 h until a 10–12 ml volume of eluate was obtained. The pH of the eluate was then checked with litmus paper and if necessary brought to the value of 7 with HCl. Concentrated water samples were simultaneously tested for NoV in three laboratories: Enteric Viruses Laboratory, University of Palermo; Istituto Zooprofilattico Sperimentale della Sicilia (IZS), Palermo; and Istituto Superiore di Sanità (ISS), Rome. Total RNA was extracted from 500 and 560 µl of a concentrated water sample using BioMèrieux NucliSENS® Magnetic Extraction System (BioMèrieux) and QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) respectively, according to manufacturer’s instructions. Bonded RNA was eluted by 100 µl of Elution Buffer. Amplifications were performed at least in duplicate on whole and 1:10 diluted extracts. A Real-Time RT-PCR assay according to ISO/TS 15216- 2:2013, two in-house Real-Time RT-PCR protocols and the One-Step RT-PCR methods already used for amplifying ORF1 and ORF2 fragments in NoV-positive faecal samples were performed [19–21, 23]. Nucleotide sequences analysis and phylogenetic analysis PCR amplicons were purified by commercial columns (CleanSweep PCR Purification, Thermo Fisher Scientific, Vilnius, Lithuania) and sequenced by Sanger method and all nucleotide sequences were aligned using ClustalW. To identify NoV genotypes, RdRp and VP1 sequences were submitted to the RIVM NoV genotyping tool (http://www.rivm.nl/mpf/norovirus/typingtool). Phylogenetic trees were constructed using MEGA 7 and Neighbour-Joining method (bootstrap 1000 replicates). Sequences from stool and water samples are available at NCBI with the following accession numbers: MG517455 for the major capsid protein obtained from a water sample; MG517456 and MG517461, respectively, for the major capsid protein and RNA-dependent-RNA-polymerase, as representatives of the sequences obtained from the stool samples of those involved in the outbreak. Control measures After outbreak notification, on 21 May, the resort was closed for 2 days. On 25 May, following the first results of the bacteriological analyses on municipal water samples, the use of municipal water for human consumption was temporarily restricted (use of water was permitted only for cleaning purposes) in the Carrabba and Fondachello villages and in the area surrounding the resort. From 25 May the reservoir collecting the water from the Carlino well and its chlorination device were submitted to maintenance work and a new chlorination device was installed on the conduit bringing the water from the Bufardo source. Actually, the Bufardo source had been connected to the municipal network in 2003 but unfortunately a dedicated chlorination station had not been planned at that time. Therefore, unchlorinated water entered the system near Fondachello village on the seaside (Fig. 1), possibly diluting chlorine concentrations in water treated at the Carlino reservoir. Moreover, the exact path of the branch of the water network situated downstream the Bufardo source water connection was not marked in the maps of the municipality (Fig. 1, red line), indicating that this section had never been submitted to maintenance. The area surrounding the unmaintained section of the water pipeline was scattered with unauthorised cottages that were not connected to the sewerage systems. Likely, untreated waste from cesspools was unloaded directly into the soil, generating a risk of contamination of the municipal water system through leakage. No work had been made on the water pipelines before the outbreak but sections of the network had been temporarily closed during the winter season on the seaside (Fondachello). Local health authorities advised the connection of all households in the area to the sewerage system and the replacement of the water system section without maintenance. Health authorities also raised attention on AGE cases notification at all local Emergency Departments and Family Practitioners but there was no further evidence of AGE outbreaks in the months following the intervention. Results Data collected through the online questionnaire compiled 7–10 days after the outbreak revealed that none of the 28 students and the teacher taking part in the school trip showed gastroenteritis symptoms at the beginning of the journey and they had no contact with people who suffered from diarrhoea and/or vomiting in the previous days. Travellers had their first shared meal on 19 May (Day 1), 2016, at lunch on their arrival to the resort in Mascali. The list of foods included in the menu is shown in Table 2. Conversely, the dinner on Day 1 was not shared and the students went to dine in three different fast food restaurants in Acireale. On 20 May (Day 2), 17 students (58.6% of the travellers) presented with vomit and/or diarrhoea, falling ill 26–36 h after lunch on Day 1, strongly suggesting that exposure occurred on the first shared meal. Some students also vomited in the presence of the other travellers on the bus bringing them to Acireale. Five more students were symptomatic in the early hours of the following day (21 May, Day 3, from 0:30 am to 8:00 am), as a continuation of the wave of illnesses of the previous day. The teacher and one more student fell ill at 8:00 pm and 11:00 pm of Day 3, respectively, for a second wave of illnesses occurring >48 h from the first shared meal and <36 h from the first cases, with the last student falling sick on 22 May (Day 4) at 3:00 am (Fig. 2, Table 1). On the whole, the mean incubation period was 33.2 h. Based on the online survey, the severity of illness was evaluated by the Vesikari severity scale, grouping symptomatic travellers into severe (11/25 = 44%) and non-severe (14/25 = 56%) cases (Table 1). Vomiting and fever were the predominant symptoms, affecting 22 out of 25 (88%) symptomatic subjects, while diarrhoea was present in 14 out of 25 (56%). Hospitalisation for rehydration therapy was necessary for 13 students (91% [10/11] of severe and 21% [3/14] of non-severe cases). Emergency room (ER) or family practitioner visits were necessary for 21 students and 13 (62%) of them initiated an antibiotic therapy. Two students indicated that other family members had gastroenteric symptoms more than 24 h after their return home, on Day 3, 4 and 5. All those involved in the outbreak recovered within 4 days, with an average duration of illness of 38.4 h. A case-control study of the risk associated with eating foods at lunch on Day 1 or with attending different restaurants at dinner on Day 1 was attempted. A total of 19 subjects filled the section of the online survey concerning food consumption. Fourteen of the responders were among the 22 cases falling ill within 36 h from the first shared meal (first wave), while the remaining five responders were three travellers with supposed secondary infections (showing symptoms with a delay ⩾48 h; second wave) and two asymptomatic students that were used as controls. No statistically significant associations with any food items served at lunch on Day 1 was identified (P < 0.05; Table 2). However, all food consumed on that occasion was to some extent associated with the risk of disease, the highest OR being obtained with chicken (OR: 20.7143, P: 0.0678; Table 2). Actually, most of the cases ate all foods included in the menu, while one of the two asymptomatic controls did not participate in the meal and the other had a pasta dish only. No statistically significant association was found for attending a particular fast food restaurant at the dinner on Day 1 (OR: 2.5, P: 0.3635). The Real-Time RT-PCR assay allowed to detect GII genogroup NoV genome in all the 25 samples from symptomatic subjects, although two of them showed high Ct values, indicating lower shedding of viral genomes (Table 1). The samples from four asymptomatic students were negative for both NoV genogroups. By sequence analysis, a NoV genotype could be defined for 20/ 25 NoV-positive stool samples. All NoV strains from travellers involved in the outbreaks belonged to the same genotype, GII.P2/GII.2, showing >99% nucleotide identities (nt id) in both fragments (ORF1 and ORF2) analysed. Thirty-two water samples were collected and analysed for bacterial contamination (Table 3). Seven samples were taken from the resort (site R) and the remaining 25 from five other sites (C, F, M, S, W), representative of the entire municipal water network. The majority (10/12) of the water samples collected immediately after the onset of the outbreak, between 21 May and 16 June 2016, at the resort (site R) and at sites F and C, located close to the resort, on the seaside and in the interior, respectively, were positive for coliform bacteria. More than 1000 colony-forming units (CFU)/ 100 ml (up to 2700 CFU/100 ml) were detected in four samples, while their absence is required in 100 ml by the Standard of Water Quality in Italy and seven of 11 samples tested were also Escherichia coli-positive (up to 900 CFU/100 ml). Enterococci were searched in 11 samples and three of them were found positive for this faecal contamination indicator. From 7 July to 16 August 2016, after control measures on the water network were undertaken, 13 water samples were collected for follow-up laboratory examination of the water system at F, R and C sites. All of them were negative for bacterial indicators of faecal contamination. No faecal bacteria were identified in the seven samples collected from M, W and S sites, including the well and the source that fed the water network, at any sampling time, before and after implementing the control measures. The research of NoV genome was conducted on concentrated water samples collected on 1 August from R and F sites (Table 3). NoV genome was detected in a single sample collected from the water entry of the resort involved in the outbreak (site R). Actually, only two out of 26 One-Step RT-PCR repetitions were positive for NoV viral capsid gene. The two ORF2-positive results were obtained at the University of Palermo Laboratory using both 1:10 diluted and undiluted extracted RNA. The sequence analysis of the partial ORF2 portion (289 bp) indicated the presence of the same GII.2 capsid genotype detected in the stool samples of the students involved in the outbreak, showing 98.9% nt id. No ORF1 NoV sequence could be obtained from the NoV-positive water sample. Follow-up analyses for NoV genome, conducted on water samples collected on 20 November both at R and F sites found no positive samples. By sequence alignment, the NoV sequences obtained from Mascali outbreak samples were phylogenetically closely related in both ORF1 and ORF2 (99 and 98.9% of nt id, respectively) to a GII.P2/GII.2 strain responsible for a sporadic paediatric gastroenteritis case occurred in Palermo in January 2016 (PA 36/16, accession numbers MG517464 and MG517460, for ORF1 and ORF2, respectively) and previously detected at the Enteric Viruses Laboratory, University of Palermo. Whereas, three GII.P2/GII.2 Italian strains (PA 280/11, PA 366/11 and PA 3/ 14, accession numbers MG517462, MG517465 and MG517463, for ORF1 and MG517457 to −59, for ORF2, respectively), detected in Palermo in 2011 and 2014, were more distantly associated with the outbreak strain both in RdRp and VP1 (90% to 97.4% nt id and 96.1% to 98.1% nt id, respectively, in ORF1 and ORF2). Discussion Epidemiological features, such as the presence of vomiting in more than half (88%) of symptomatic subjects, the 33.2 h average incubation period and the 38.4 h average duration of illness, suggested a NoV aetiology for the investigated outbreak [13]. The detection of NoV genome in all 25 faecal samples from symptomatic subjects fully complies with the requirement of detection of at least one positive sample out of four tested that is considered to be sufficient to ascertain the aetiology of a NoV outbreak [24]. Moreover, molecular typing indicated that a single viral strain, belonging to genotype GII.P2/GII.2, infected all symptomatic subjects involved in the outbreak. NoVs are highly infectious and can easily spread from person to person but also through contaminated food or water. Therefore, settings such as health care institutions, schools and touristic resorts are mainly affected, since close contact and shared food services facilitate transmission. In Mascali, two waves of transmission were observed over three days, being consistent with food/waterborne followed by person-to-person and/or airborne (vomit) transmission (Fig. 2). The first wave of cases presented on 20 May 2016 after a single shared meal, while the last case was detected on 22 May. Exposure to vomit may have been responsible for some of the secondary cases occurred after the first wave of illness. Vomiting produces aerosols that can be inhaled and swallowed. The extremely low infectious dose of NoV makes the inhalation of vomit aerosols an effective transmission route in settings where there is close contact, like on a bus trip [25]. All food consumed in the only shared lunch on 19 May at the resort was associated with the risk of disease (Table 2), but the association was inconclusive since all cases shared the same dishes included in the lunch menu on Day 1. Although many foods were potentially involved, our attention focused on the tap water. In fact, trivial information from local health authorities reported that local population avoided tap water consumption and relied on bottled water for drinking and cooking. Many small outbreaks of gastroenteritis involving restaurants and family groups had been reported in the same area, especially at the beginning of the summer season, when vacation houses and recreational activities were re-opened after a period of low water demand. Illness caused by NoV is common and if the level of water contamination is low, the number of cases remains limited and water can persist as unrecognised source of infection. If polluted water was served for drinking and used for cooking and washing dishes in the Mascali resort at Day 1 a generalised contamination of food and surfaces could have occurred. The potential role of the foodhandlers as a source of contamination could not be analysed in this study but should also be taken into account. If water contamination has already been present in the area for a long time, infections may have occurred among the local food-handlers. The analysis of water samples taken immediately after the outbreak revealed the presence of bacterial indicators of faecal contamination in the water used in the resort, with the only exception of a sample of filter purified water used both for cooking and as drinking water. When extending the sampling to other areas of the municipal water network, many other sites, mostly on the seaside, showed faecal pollution, which demonstrates a poor protection of the water network from faecal leakage (Fig. 1). NoV contamination of waters generally originates from sewage since these enteric pathogens, as other enteric viruses affecting humans, are mostly species-specific and are shed at high concentrations in the stools of patients. When the analyses on the stools of the students made it clear that the outbreak had been caused by NoV, the viral genome was also searched in appropriate volumes of water samples, but this happened when correction measures on the water network had already been taken. Nevertheless, a single water sample, obtained 2 months after the outbreak (1 August 2016), still contained genome fragments that could be correlated to the same NoV genotype causing the outbreak. Although the finding of NoV genome does not demonstrate that entire infecting viral particles were present in the water, it could indicate that a NoV belonging to the same genotype involved in the outbreak was still circulating in the local population several weeks after the outbreak and it was someway able to penetrate the water network, probably indicating continuing faecal contamination through leakage from sewage systems. NoV is particularly resistant in the environment and also to chlorine, therefore transmission through chlorinated water is not exceptional [9, 12, 26]. Drinking water from municipal supplies fed by lake or groundwater has already been identified in the past as the source of waterborne NoV outbreaks in Italy [27–29] and NoV is now considered one of the pathogens causing the largest number of affected consumers in drinking waterborne outbreaks [12]. NoV can persist in water longer than 15 days [30] and it is resistant to chlorine disinfection when free chlorine levels or pre-treatment processes are inadequate [31, 32]. Moreover, microfiltration filters labelled as certified by NSF Standards 53 or 58 (with a pore size ranging from 0.05 to 5 µm) are not effective in removing viruses and can even decrease chlorine concentration in filtered water [33]. Actually, the owner of the resort involved in our outbreak was encouraged to serve tap water as drinking water at the restaurant relying on the effectiveness of such a filter purifier. Unfortunately, we did not have a tap water sample from the restaurant of the resort taken before correction measures and whose volume was appropriate for virus detection. Therefore, we could not demonstrate the failure to eliminate viral pollution of the microfiltration system. In Mascali, correction measures on the water network consisted in the maintenance of the old chlorination device and in the addition of a new one at a key point located upstream and close to suspected illegal sewage dumping (Fig. 1), but as poor protection of the pipeline from sewage leakage risked to reproduce faecal pollution episodes, renovation of the municipal water system was also advised, as well as systematic control of illegal sewage dumping. The additional chlorination device was apparently the most effective intervention, since right after its activation faecal bacterial pollution indicators disappeared. However, NoV genome was detected in a water sample taken 6 weeks after the last detection of bacterial indicators. Unfortunately, detection of faecal contamination of drinking water is only based on the use of bacterial indicators despite no correlation between levels of enteric bacteria and enteric viruses has been found [34–36]. According to ISGEV enteric virus circulation surveillance data, in 2016 the circulation of GII.P2/GII.2 NoV strains in Palermo, home town of the travellers involved in the outbreak, was the highest ever observed from the starting of local NoV surveillance in 2004. GII.P2/GII.2 was the fourth most common genotype in 2016 (12.2% prevalence), following GII.P4/GII.4 Sydney 2012 (28.4%), GII.P16/GII.2 (18.9%) and GII.Pe/GII.4 Sydney 2012 (17.6%). GII.P2/GII.2 strains were mostly observed starting from July 2016, leaving space for the possibility that their increased circulation was linked to the introduction of the ‘Mascali clone’. However, a couple of GII.P2/GII.2 strains (including PA36/16) genetically close to the ‘Mascali clone’ had been already detected in January 2016 in Palermo and GII.P2/GII.2 NoV had sporadically been observed also in 2011 and 2014. Moreover, the circulation of GII.P2/GII.2 NoVs has been increasing worldwide over the last period [37]. Therefore, the unusually high circulation observed in Palermo might just reflect the global epidemiology of GII.P2/GII.2 strains. Noteworthy, the older GII.P2/GII.2 NoV strains detected in Palermo were closely related in the ORF2 (98.9% nt id) to the GII.P16/GII.2 Nashville strain (KY865307) which is supposed to have generated the recombinant GII.P16/GII.4 clone recently circulating epidemically worldwide [37, 38] and were more distantly associated (95.3% nt id) to the GII.2 capsid sequence (AB662868) of another putative GII.P16/GII.4 ancestor from Japan [39]. The increasing frequency of detection of waterborne NoV outbreaks calls for the introduction of routine monitoring for viruses of drinking waters and for the definition of affordable and reliable indicators for the presence of NoV in drinking waters. The use of ultrafiltration systems as a concentration method in order to recover viruses from environmental waters combined with RT-PCR to detect viral genomes has been demonstrated to grant a recovery rate of >50% of multiple viruses [40] and is at present one of the most sensitive approaches for the detection of NoVs, although the discrimination between viable and nonviable viruses remains a problem. The burden of NoV outbreaks, either food associated or waterborne, is still poorly defined due to the lack of specific diagnostic tools in many clinical laboratories. NoV epidemics should always be suspected when the main symptom is vomiting and the duration of disease is short (<48 h). Waterborne transmission is suggested by the rapid spread of disease and the high attack rate. Care should be taken to perform sensitive diagnostic tests on a sufficient number of stool samples in order to identify NoV as the causative agent of enteritis outbreaks, possibly preferring genome screening over antigen detection. Only the systematic detection of NoV outbreaks can guide in the choice of the safety procedures that could be most appropriate in containing such microbiological hazard.

In which country did this happen?

1541

Waterborne Norovirus outbreak at a seaside resort likely originating from municipal water distribution system failure

Abstract In May 2016 a Norovirus (NoV) gastroenteritis outbreak involved a high school class visiting a seaside resort near Taormina (Mascali, Sicily). Twenty-four students and a teacher were affected and 17 of them showed symptoms on the second day of the journey, while the others got ill within the following 2 days. Symptoms included vomiting, diarrhoea and fever, and 12 students required hospitalisation. Stool samples tested positive for NoV genome by Real-Time polymerase chain reaction assay in all 25 symptomatic subjects. The GII.P2/GII.2 NoV genotype was linked to the outbreak by ORF1/ORF2 sequence analysis. The epidemiological features of the outbreak were consistent with food/waterborne followed by person-to-person and/ or vomit transmission. Food consumed at a shared lunch on the first day of the trip was associated to illness and drinking un-bottled tap water was also considered as a risk factor. The analysis of water samples revealed the presence of bacterial indicators of faecal contamination in the water used in the resort as well as in other areas of the municipal water network, linking the NoV gastroenteritis outbreak to tap water pollution from sewage leakage. From a single water sample, an amplicon whose sequence corresponded to the capsid genotype recovered from patients could be obtained. Introduction Noroviruses (NoVs) are recognised as a leading cause of sporadic cases and outbreaks of acute gastroenteritis (AGE) across all age groups [1, 2]. They are transmitted primarily through the faecal-oral route either by direct person–person spread and with consumption of contaminated food or water [1, 3, 4]. NoVs are non-enveloped RNA viruses belonging to the Caliciviridae family and are classified into seven distinct genogroups (GI-GVII) on the basis of the major capsid protein VP1 sequence. GI, GII and GIV NoVs infect humans and they have been divided into more than 30 genotypes based on the sequences of polymerase (ORF1) and capsid proteins (ORF2) [5]. NoV GII.4 genotype has been associated with the vast majority of NoV related cases of AGE worldwide although many genotypes, belonging to both GI and GII genogroups are involved in foodborne and waterborne outbreaks [6, 7]. NoVs are highly contagious owing to their high stability in the environment and their ability to infect at low doses. In fact, just 10–100 virions of NoV may be sufficient to cause AGE [8, 9]. NoVs are considered the major cause of water-related disease in semi-closed institutions, including restaurants, schools, hospitals, nursing homes and cruise ships [10–12]. To confirm NoV as a cause of outbreaks, the presence of four epidemiological features of disease, vomiting in more than half of affected persons, a mean incubation period of 24/48 h, a mean duration of illness of 12–60 h and the absence of bacterial pathogens in stool culture, should be evaluated [13]. Waterborne NoV outbreaks can be caused by contaminated drinking water despite correct chlorine concentration and in the absence of coliform bacteria [10, 14–17]. The aim of this study was to investigate an outbreak of NoV gastroenteritis associated with drinking water involving a high school class travelling from Palermo (West Sicily) to Acireale (East Sicily) on an educational tour in May 2016. Stool samples were collected soon after the onset of AGE symptoms to ascertain the aetiology of illness. The source of infection was investigated collecting epidemiological data through an online questionnaire and, in order to explore a possible transmission route, laboratory tests on water from the municipal network were performed. Material and methods Epidemiological investigation From 20 to 22 May 2016 an AGE outbreak occurred in Mascali, a small town near Taormina, involving a school class of 28 students, 13 girls and 15 boys aged between 14 and 18 years and a 54-year-old female teacher, during a trip to participate in a theatre festival in Acireale. A total of 25/29 persons were affected by nausea, vomiting, abdominal cramps, diarrhoea and fever while they were staying at a seaside resort in Mascali, on the way to Acireale. To obtain data on symptoms and the correlated risk factors all participants filled an online questionnaire investigating personal data (age and sex), travel and accommodation details and food consumption over 48 h before symptoms presentation, time of onset and duration of symptoms, therapy and/or hospitalisation required and history of contact with people affected by diarrhoea and/or vomiting (Table 1). The severity of illness was analysed by the Vesikari severity scale allowing to distinguish non-severe (score <11) and severe (score ⩾11) AGE (Table 1). Vesikari score was calculated from maximum number of episodes/day and duration of both diarrhoea and vomiting (score 0–3 for each of the 4 parameters), temperature (score 0–3), level of dehydration (score 0–3) and treatment required (score 0–2) [18]. A case-control study was conducted to establish if food consumed before the onset of symptoms was associated with the disease. A case was defined as a student attending the tour and reporting diarrhoea and/or vomiting within 36 h from the first shared meal during the tour, while students who did not present any of the previous symptoms or those showing the first symptoms with a delay of ⩾48 h (secondary cases) were used as controls. Variables were studied using odds ratio (OR) and its 95% CI and statistical associations were determined using the χ2 test, with the level of statistical significance set at 0.05 (Table 2). NoV detection in stool samples A stool sample was obtained from each of the 29 travellers, soon after (3–5 days) the onset of the outbreak. No stool samples could be obtained from the food-handlers working at the resort. All samples were tested for NoV at the Enteric Viruses Laboratory of the University of Palermo, member of the Italian Study Group for Enteric Viruses (ISGEV), by a Real-Time polymerase chain reaction (RT-PCR) assay that allows to differentiate GI and GII NoVs [19]. Total RNA was extracted from stool samples using QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) and cDNA was generated by MMLV reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were considered positive for NoVs when cycle threshold (ct) of RT-PCR was ⩽30 and lowpositive when 30<ct⩽35. For all NoV-positive samples two genomic fragments were amplified by conventional RT-PCR using Superscript III One-Step RT-PCR system with Platinum Taq DNA polymerase (Invitrogen) and primers JV12a/JV13b, targeting 326 bps of the RdRp region A [20] and either COG1F/ GISKR or COG2F/GIISKR, amplifying a 315 bps 5′portion of GI or GII VP1 protein [19, 21]. Environmental investigation From May through August 2016, a total of 32 tap water samples were collected from the Mascali municipal network. In particular, seven samples were collected at the resort where the outbreak occurred and from its water entry (R), two in the city centre of Mascali (M), 15 in the village of Fondachello (F), on the seaside, three in Carrabba village (C), far from the coast, and two and three, respectively, from the water well ‘Carlino’ (W) and the source ‘Bufardo and Torregrossa’ (S) which feed the municipal system and provide drinking water to the Mascali area (Fig. 1, Table 3). One litre water samples were sent by the local health authorities to the SIALAB laboratory in Avola, Siracusa and 100 ml aliquots were screened for coliforms, E. coli and enterococci as faecal indicators, by standard bacteriological methods according to UNI EN ISO 9308-1:2014. Unfortunately, although the suspect of a water source for the viral outbreak was quickly communicated, only in August and November 2016, following the implementation of control measures, local authorities provided water samplings of appropriate volume to perform viral tests (10 l). Five water samples of 10 l each were collected from different sampling sites in F and R areas for viral screening (Table 3) and were concentrated according to the Italian National Institute of Health (Istituto Superiore di Sanità, ISS) protocols [22] through an ultra-filtration system (Sartoflow® Slice 200 Benchtop Crossflow System, Sartorius AG, Goettingen, Germany), using appropriate membranes (SG Hydrosart 10 kDa) pretreated with 300 ml of 3% Beef Extract (BE) at pH 7, by recirculating it for 10 min in the system, at a pressure not higher than 1.30 bar. Elution and recovery of viruses attached to the membranes were obtained by recirculating 200 ml of 3% BE at pH 9.5 for 3 h until a 10–12 ml volume of eluate was obtained. The pH of the eluate was then checked with litmus paper and if necessary brought to the value of 7 with HCl. Concentrated water samples were simultaneously tested for NoV in three laboratories: Enteric Viruses Laboratory, University of Palermo; Istituto Zooprofilattico Sperimentale della Sicilia (IZS), Palermo; and Istituto Superiore di Sanità (ISS), Rome. Total RNA was extracted from 500 and 560 µl of a concentrated water sample using BioMèrieux NucliSENS® Magnetic Extraction System (BioMèrieux) and QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) respectively, according to manufacturer’s instructions. Bonded RNA was eluted by 100 µl of Elution Buffer. Amplifications were performed at least in duplicate on whole and 1:10 diluted extracts. A Real-Time RT-PCR assay according to ISO/TS 15216- 2:2013, two in-house Real-Time RT-PCR protocols and the One-Step RT-PCR methods already used for amplifying ORF1 and ORF2 fragments in NoV-positive faecal samples were performed [19–21, 23]. Nucleotide sequences analysis and phylogenetic analysis PCR amplicons were purified by commercial columns (CleanSweep PCR Purification, Thermo Fisher Scientific, Vilnius, Lithuania) and sequenced by Sanger method and all nucleotide sequences were aligned using ClustalW. To identify NoV genotypes, RdRp and VP1 sequences were submitted to the RIVM NoV genotyping tool (http://www.rivm.nl/mpf/norovirus/typingtool). Phylogenetic trees were constructed using MEGA 7 and Neighbour-Joining method (bootstrap 1000 replicates). Sequences from stool and water samples are available at NCBI with the following accession numbers: MG517455 for the major capsid protein obtained from a water sample; MG517456 and MG517461, respectively, for the major capsid protein and RNA-dependent-RNA-polymerase, as representatives of the sequences obtained from the stool samples of those involved in the outbreak. Control measures After outbreak notification, on 21 May, the resort was closed for 2 days. On 25 May, following the first results of the bacteriological analyses on municipal water samples, the use of municipal water for human consumption was temporarily restricted (use of water was permitted only for cleaning purposes) in the Carrabba and Fondachello villages and in the area surrounding the resort. From 25 May the reservoir collecting the water from the Carlino well and its chlorination device were submitted to maintenance work and a new chlorination device was installed on the conduit bringing the water from the Bufardo source. Actually, the Bufardo source had been connected to the municipal network in 2003 but unfortunately a dedicated chlorination station had not been planned at that time. Therefore, unchlorinated water entered the system near Fondachello village on the seaside (Fig. 1), possibly diluting chlorine concentrations in water treated at the Carlino reservoir. Moreover, the exact path of the branch of the water network situated downstream the Bufardo source water connection was not marked in the maps of the municipality (Fig. 1, red line), indicating that this section had never been submitted to maintenance. The area surrounding the unmaintained section of the water pipeline was scattered with unauthorised cottages that were not connected to the sewerage systems. Likely, untreated waste from cesspools was unloaded directly into the soil, generating a risk of contamination of the municipal water system through leakage. No work had been made on the water pipelines before the outbreak but sections of the network had been temporarily closed during the winter season on the seaside (Fondachello). Local health authorities advised the connection of all households in the area to the sewerage system and the replacement of the water system section without maintenance. Health authorities also raised attention on AGE cases notification at all local Emergency Departments and Family Practitioners but there was no further evidence of AGE outbreaks in the months following the intervention. Results Data collected through the online questionnaire compiled 7–10 days after the outbreak revealed that none of the 28 students and the teacher taking part in the school trip showed gastroenteritis symptoms at the beginning of the journey and they had no contact with people who suffered from diarrhoea and/or vomiting in the previous days. Travellers had their first shared meal on 19 May (Day 1), 2016, at lunch on their arrival to the resort in Mascali. The list of foods included in the menu is shown in Table 2. Conversely, the dinner on Day 1 was not shared and the students went to dine in three different fast food restaurants in Acireale. On 20 May (Day 2), 17 students (58.6% of the travellers) presented with vomit and/or diarrhoea, falling ill 26–36 h after lunch on Day 1, strongly suggesting that exposure occurred on the first shared meal. Some students also vomited in the presence of the other travellers on the bus bringing them to Acireale. Five more students were symptomatic in the early hours of the following day (21 May, Day 3, from 0:30 am to 8:00 am), as a continuation of the wave of illnesses of the previous day. The teacher and one more student fell ill at 8:00 pm and 11:00 pm of Day 3, respectively, for a second wave of illnesses occurring >48 h from the first shared meal and <36 h from the first cases, with the last student falling sick on 22 May (Day 4) at 3:00 am (Fig. 2, Table 1). On the whole, the mean incubation period was 33.2 h. Based on the online survey, the severity of illness was evaluated by the Vesikari severity scale, grouping symptomatic travellers into severe (11/25 = 44%) and non-severe (14/25 = 56%) cases (Table 1). Vomiting and fever were the predominant symptoms, affecting 22 out of 25 (88%) symptomatic subjects, while diarrhoea was present in 14 out of 25 (56%). Hospitalisation for rehydration therapy was necessary for 13 students (91% [10/11] of severe and 21% [3/14] of non-severe cases). Emergency room (ER) or family practitioner visits were necessary for 21 students and 13 (62%) of them initiated an antibiotic therapy. Two students indicated that other family members had gastroenteric symptoms more than 24 h after their return home, on Day 3, 4 and 5. All those involved in the outbreak recovered within 4 days, with an average duration of illness of 38.4 h. A case-control study of the risk associated with eating foods at lunch on Day 1 or with attending different restaurants at dinner on Day 1 was attempted. A total of 19 subjects filled the section of the online survey concerning food consumption. Fourteen of the responders were among the 22 cases falling ill within 36 h from the first shared meal (first wave), while the remaining five responders were three travellers with supposed secondary infections (showing symptoms with a delay ⩾48 h; second wave) and two asymptomatic students that were used as controls. No statistically significant associations with any food items served at lunch on Day 1 was identified (P < 0.05; Table 2). However, all food consumed on that occasion was to some extent associated with the risk of disease, the highest OR being obtained with chicken (OR: 20.7143, P: 0.0678; Table 2). Actually, most of the cases ate all foods included in the menu, while one of the two asymptomatic controls did not participate in the meal and the other had a pasta dish only. No statistically significant association was found for attending a particular fast food restaurant at the dinner on Day 1 (OR: 2.5, P: 0.3635). The Real-Time RT-PCR assay allowed to detect GII genogroup NoV genome in all the 25 samples from symptomatic subjects, although two of them showed high Ct values, indicating lower shedding of viral genomes (Table 1). The samples from four asymptomatic students were negative for both NoV genogroups. By sequence analysis, a NoV genotype could be defined for 20/ 25 NoV-positive stool samples. All NoV strains from travellers involved in the outbreaks belonged to the same genotype, GII.P2/GII.2, showing >99% nucleotide identities (nt id) in both fragments (ORF1 and ORF2) analysed. Thirty-two water samples were collected and analysed for bacterial contamination (Table 3). Seven samples were taken from the resort (site R) and the remaining 25 from five other sites (C, F, M, S, W), representative of the entire municipal water network. The majority (10/12) of the water samples collected immediately after the onset of the outbreak, between 21 May and 16 June 2016, at the resort (site R) and at sites F and C, located close to the resort, on the seaside and in the interior, respectively, were positive for coliform bacteria. More than 1000 colony-forming units (CFU)/ 100 ml (up to 2700 CFU/100 ml) were detected in four samples, while their absence is required in 100 ml by the Standard of Water Quality in Italy and seven of 11 samples tested were also Escherichia coli-positive (up to 900 CFU/100 ml). Enterococci were searched in 11 samples and three of them were found positive for this faecal contamination indicator. From 7 July to 16 August 2016, after control measures on the water network were undertaken, 13 water samples were collected for follow-up laboratory examination of the water system at F, R and C sites. All of them were negative for bacterial indicators of faecal contamination. No faecal bacteria were identified in the seven samples collected from M, W and S sites, including the well and the source that fed the water network, at any sampling time, before and after implementing the control measures. The research of NoV genome was conducted on concentrated water samples collected on 1 August from R and F sites (Table 3). NoV genome was detected in a single sample collected from the water entry of the resort involved in the outbreak (site R). Actually, only two out of 26 One-Step RT-PCR repetitions were positive for NoV viral capsid gene. The two ORF2-positive results were obtained at the University of Palermo Laboratory using both 1:10 diluted and undiluted extracted RNA. The sequence analysis of the partial ORF2 portion (289 bp) indicated the presence of the same GII.2 capsid genotype detected in the stool samples of the students involved in the outbreak, showing 98.9% nt id. No ORF1 NoV sequence could be obtained from the NoV-positive water sample. Follow-up analyses for NoV genome, conducted on water samples collected on 20 November both at R and F sites found no positive samples. By sequence alignment, the NoV sequences obtained from Mascali outbreak samples were phylogenetically closely related in both ORF1 and ORF2 (99 and 98.9% of nt id, respectively) to a GII.P2/GII.2 strain responsible for a sporadic paediatric gastroenteritis case occurred in Palermo in January 2016 (PA 36/16, accession numbers MG517464 and MG517460, for ORF1 and ORF2, respectively) and previously detected at the Enteric Viruses Laboratory, University of Palermo. Whereas, three GII.P2/GII.2 Italian strains (PA 280/11, PA 366/11 and PA 3/ 14, accession numbers MG517462, MG517465 and MG517463, for ORF1 and MG517457 to −59, for ORF2, respectively), detected in Palermo in 2011 and 2014, were more distantly associated with the outbreak strain both in RdRp and VP1 (90% to 97.4% nt id and 96.1% to 98.1% nt id, respectively, in ORF1 and ORF2). Discussion Epidemiological features, such as the presence of vomiting in more than half (88%) of symptomatic subjects, the 33.2 h average incubation period and the 38.4 h average duration of illness, suggested a NoV aetiology for the investigated outbreak [13]. The detection of NoV genome in all 25 faecal samples from symptomatic subjects fully complies with the requirement of detection of at least one positive sample out of four tested that is considered to be sufficient to ascertain the aetiology of a NoV outbreak [24]. Moreover, molecular typing indicated that a single viral strain, belonging to genotype GII.P2/GII.2, infected all symptomatic subjects involved in the outbreak. NoVs are highly infectious and can easily spread from person to person but also through contaminated food or water. Therefore, settings such as health care institutions, schools and touristic resorts are mainly affected, since close contact and shared food services facilitate transmission. In Mascali, two waves of transmission were observed over three days, being consistent with food/waterborne followed by person-to-person and/or airborne (vomit) transmission (Fig. 2). The first wave of cases presented on 20 May 2016 after a single shared meal, while the last case was detected on 22 May. Exposure to vomit may have been responsible for some of the secondary cases occurred after the first wave of illness. Vomiting produces aerosols that can be inhaled and swallowed. The extremely low infectious dose of NoV makes the inhalation of vomit aerosols an effective transmission route in settings where there is close contact, like on a bus trip [25]. All food consumed in the only shared lunch on 19 May at the resort was associated with the risk of disease (Table 2), but the association was inconclusive since all cases shared the same dishes included in the lunch menu on Day 1. Although many foods were potentially involved, our attention focused on the tap water. In fact, trivial information from local health authorities reported that local population avoided tap water consumption and relied on bottled water for drinking and cooking. Many small outbreaks of gastroenteritis involving restaurants and family groups had been reported in the same area, especially at the beginning of the summer season, when vacation houses and recreational activities were re-opened after a period of low water demand. Illness caused by NoV is common and if the level of water contamination is low, the number of cases remains limited and water can persist as unrecognised source of infection. If polluted water was served for drinking and used for cooking and washing dishes in the Mascali resort at Day 1 a generalised contamination of food and surfaces could have occurred. The potential role of the foodhandlers as a source of contamination could not be analysed in this study but should also be taken into account. If water contamination has already been present in the area for a long time, infections may have occurred among the local food-handlers. The analysis of water samples taken immediately after the outbreak revealed the presence of bacterial indicators of faecal contamination in the water used in the resort, with the only exception of a sample of filter purified water used both for cooking and as drinking water. When extending the sampling to other areas of the municipal water network, many other sites, mostly on the seaside, showed faecal pollution, which demonstrates a poor protection of the water network from faecal leakage (Fig. 1). NoV contamination of waters generally originates from sewage since these enteric pathogens, as other enteric viruses affecting humans, are mostly species-specific and are shed at high concentrations in the stools of patients. When the analyses on the stools of the students made it clear that the outbreak had been caused by NoV, the viral genome was also searched in appropriate volumes of water samples, but this happened when correction measures on the water network had already been taken. Nevertheless, a single water sample, obtained 2 months after the outbreak (1 August 2016), still contained genome fragments that could be correlated to the same NoV genotype causing the outbreak. Although the finding of NoV genome does not demonstrate that entire infecting viral particles were present in the water, it could indicate that a NoV belonging to the same genotype involved in the outbreak was still circulating in the local population several weeks after the outbreak and it was someway able to penetrate the water network, probably indicating continuing faecal contamination through leakage from sewage systems. NoV is particularly resistant in the environment and also to chlorine, therefore transmission through chlorinated water is not exceptional [9, 12, 26]. Drinking water from municipal supplies fed by lake or groundwater has already been identified in the past as the source of waterborne NoV outbreaks in Italy [27–29] and NoV is now considered one of the pathogens causing the largest number of affected consumers in drinking waterborne outbreaks [12]. NoV can persist in water longer than 15 days [30] and it is resistant to chlorine disinfection when free chlorine levels or pre-treatment processes are inadequate [31, 32]. Moreover, microfiltration filters labelled as certified by NSF Standards 53 or 58 (with a pore size ranging from 0.05 to 5 µm) are not effective in removing viruses and can even decrease chlorine concentration in filtered water [33]. Actually, the owner of the resort involved in our outbreak was encouraged to serve tap water as drinking water at the restaurant relying on the effectiveness of such a filter purifier. Unfortunately, we did not have a tap water sample from the restaurant of the resort taken before correction measures and whose volume was appropriate for virus detection. Therefore, we could not demonstrate the failure to eliminate viral pollution of the microfiltration system. In Mascali, correction measures on the water network consisted in the maintenance of the old chlorination device and in the addition of a new one at a key point located upstream and close to suspected illegal sewage dumping (Fig. 1), but as poor protection of the pipeline from sewage leakage risked to reproduce faecal pollution episodes, renovation of the municipal water system was also advised, as well as systematic control of illegal sewage dumping. The additional chlorination device was apparently the most effective intervention, since right after its activation faecal bacterial pollution indicators disappeared. However, NoV genome was detected in a water sample taken 6 weeks after the last detection of bacterial indicators. Unfortunately, detection of faecal contamination of drinking water is only based on the use of bacterial indicators despite no correlation between levels of enteric bacteria and enteric viruses has been found [34–36]. According to ISGEV enteric virus circulation surveillance data, in 2016 the circulation of GII.P2/GII.2 NoV strains in Palermo, home town of the travellers involved in the outbreak, was the highest ever observed from the starting of local NoV surveillance in 2004. GII.P2/GII.2 was the fourth most common genotype in 2016 (12.2% prevalence), following GII.P4/GII.4 Sydney 2012 (28.4%), GII.P16/GII.2 (18.9%) and GII.Pe/GII.4 Sydney 2012 (17.6%). GII.P2/GII.2 strains were mostly observed starting from July 2016, leaving space for the possibility that their increased circulation was linked to the introduction of the ‘Mascali clone’. However, a couple of GII.P2/GII.2 strains (including PA36/16) genetically close to the ‘Mascali clone’ had been already detected in January 2016 in Palermo and GII.P2/GII.2 NoV had sporadically been observed also in 2011 and 2014. Moreover, the circulation of GII.P2/GII.2 NoVs has been increasing worldwide over the last period [37]. Therefore, the unusually high circulation observed in Palermo might just reflect the global epidemiology of GII.P2/GII.2 strains. Noteworthy, the older GII.P2/GII.2 NoV strains detected in Palermo were closely related in the ORF2 (98.9% nt id) to the GII.P16/GII.2 Nashville strain (KY865307) which is supposed to have generated the recombinant GII.P16/GII.4 clone recently circulating epidemically worldwide [37, 38] and were more distantly associated (95.3% nt id) to the GII.2 capsid sequence (AB662868) of another putative GII.P16/GII.4 ancestor from Japan [39]. The increasing frequency of detection of waterborne NoV outbreaks calls for the introduction of routine monitoring for viruses of drinking waters and for the definition of affordable and reliable indicators for the presence of NoV in drinking waters. The use of ultrafiltration systems as a concentration method in order to recover viruses from environmental waters combined with RT-PCR to detect viral genomes has been demonstrated to grant a recovery rate of >50% of multiple viruses [40] and is at present one of the most sensitive approaches for the detection of NoVs, although the discrimination between viable and nonviable viruses remains a problem. The burden of NoV outbreaks, either food associated or waterborne, is still poorly defined due to the lack of specific diagnostic tools in many clinical laboratories. NoV epidemics should always be suspected when the main symptom is vomiting and the duration of disease is short (<48 h). Waterborne transmission is suggested by the rapid spread of disease and the high attack rate. Care should be taken to perform sensitive diagnostic tests on a sufficient number of stool samples in order to identify NoV as the causative agent of enteritis outbreaks, possibly preferring genome screening over antigen detection. Only the systematic detection of NoV outbreaks can guide in the choice of the safety procedures that could be most appropriate in containing such microbiological hazard.

Where did this happen?

1542

Waterborne Norovirus outbreak at a seaside resort likely originating from municipal water distribution system failure

Abstract In May 2016 a Norovirus (NoV) gastroenteritis outbreak involved a high school class visiting a seaside resort near Taormina (Mascali, Sicily). Twenty-four students and a teacher were affected and 17 of them showed symptoms on the second day of the journey, while the others got ill within the following 2 days. Symptoms included vomiting, diarrhoea and fever, and 12 students required hospitalisation. Stool samples tested positive for NoV genome by Real-Time polymerase chain reaction assay in all 25 symptomatic subjects. The GII.P2/GII.2 NoV genotype was linked to the outbreak by ORF1/ORF2 sequence analysis. The epidemiological features of the outbreak were consistent with food/waterborne followed by person-to-person and/ or vomit transmission. Food consumed at a shared lunch on the first day of the trip was associated to illness and drinking un-bottled tap water was also considered as a risk factor. The analysis of water samples revealed the presence of bacterial indicators of faecal contamination in the water used in the resort as well as in other areas of the municipal water network, linking the NoV gastroenteritis outbreak to tap water pollution from sewage leakage. From a single water sample, an amplicon whose sequence corresponded to the capsid genotype recovered from patients could be obtained. Introduction Noroviruses (NoVs) are recognised as a leading cause of sporadic cases and outbreaks of acute gastroenteritis (AGE) across all age groups [1, 2]. They are transmitted primarily through the faecal-oral route either by direct person–person spread and with consumption of contaminated food or water [1, 3, 4]. NoVs are non-enveloped RNA viruses belonging to the Caliciviridae family and are classified into seven distinct genogroups (GI-GVII) on the basis of the major capsid protein VP1 sequence. GI, GII and GIV NoVs infect humans and they have been divided into more than 30 genotypes based on the sequences of polymerase (ORF1) and capsid proteins (ORF2) [5]. NoV GII.4 genotype has been associated with the vast majority of NoV related cases of AGE worldwide although many genotypes, belonging to both GI and GII genogroups are involved in foodborne and waterborne outbreaks [6, 7]. NoVs are highly contagious owing to their high stability in the environment and their ability to infect at low doses. In fact, just 10–100 virions of NoV may be sufficient to cause AGE [8, 9]. NoVs are considered the major cause of water-related disease in semi-closed institutions, including restaurants, schools, hospitals, nursing homes and cruise ships [10–12]. To confirm NoV as a cause of outbreaks, the presence of four epidemiological features of disease, vomiting in more than half of affected persons, a mean incubation period of 24/48 h, a mean duration of illness of 12–60 h and the absence of bacterial pathogens in stool culture, should be evaluated [13]. Waterborne NoV outbreaks can be caused by contaminated drinking water despite correct chlorine concentration and in the absence of coliform bacteria [10, 14–17]. The aim of this study was to investigate an outbreak of NoV gastroenteritis associated with drinking water involving a high school class travelling from Palermo (West Sicily) to Acireale (East Sicily) on an educational tour in May 2016. Stool samples were collected soon after the onset of AGE symptoms to ascertain the aetiology of illness. The source of infection was investigated collecting epidemiological data through an online questionnaire and, in order to explore a possible transmission route, laboratory tests on water from the municipal network were performed. Material and methods Epidemiological investigation From 20 to 22 May 2016 an AGE outbreak occurred in Mascali, a small town near Taormina, involving a school class of 28 students, 13 girls and 15 boys aged between 14 and 18 years and a 54-year-old female teacher, during a trip to participate in a theatre festival in Acireale. A total of 25/29 persons were affected by nausea, vomiting, abdominal cramps, diarrhoea and fever while they were staying at a seaside resort in Mascali, on the way to Acireale. To obtain data on symptoms and the correlated risk factors all participants filled an online questionnaire investigating personal data (age and sex), travel and accommodation details and food consumption over 48 h before symptoms presentation, time of onset and duration of symptoms, therapy and/or hospitalisation required and history of contact with people affected by diarrhoea and/or vomiting (Table 1). The severity of illness was analysed by the Vesikari severity scale allowing to distinguish non-severe (score <11) and severe (score ⩾11) AGE (Table 1). Vesikari score was calculated from maximum number of episodes/day and duration of both diarrhoea and vomiting (score 0–3 for each of the 4 parameters), temperature (score 0–3), level of dehydration (score 0–3) and treatment required (score 0–2) [18]. A case-control study was conducted to establish if food consumed before the onset of symptoms was associated with the disease. A case was defined as a student attending the tour and reporting diarrhoea and/or vomiting within 36 h from the first shared meal during the tour, while students who did not present any of the previous symptoms or those showing the first symptoms with a delay of ⩾48 h (secondary cases) were used as controls. Variables were studied using odds ratio (OR) and its 95% CI and statistical associations were determined using the χ2 test, with the level of statistical significance set at 0.05 (Table 2). NoV detection in stool samples A stool sample was obtained from each of the 29 travellers, soon after (3–5 days) the onset of the outbreak. No stool samples could be obtained from the food-handlers working at the resort. All samples were tested for NoV at the Enteric Viruses Laboratory of the University of Palermo, member of the Italian Study Group for Enteric Viruses (ISGEV), by a Real-Time polymerase chain reaction (RT-PCR) assay that allows to differentiate GI and GII NoVs [19]. Total RNA was extracted from stool samples using QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) and cDNA was generated by MMLV reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were considered positive for NoVs when cycle threshold (ct) of RT-PCR was ⩽30 and lowpositive when 30<ct⩽35. For all NoV-positive samples two genomic fragments were amplified by conventional RT-PCR using Superscript III One-Step RT-PCR system with Platinum Taq DNA polymerase (Invitrogen) and primers JV12a/JV13b, targeting 326 bps of the RdRp region A [20] and either COG1F/ GISKR or COG2F/GIISKR, amplifying a 315 bps 5′portion of GI or GII VP1 protein [19, 21]. Environmental investigation From May through August 2016, a total of 32 tap water samples were collected from the Mascali municipal network. In particular, seven samples were collected at the resort where the outbreak occurred and from its water entry (R), two in the city centre of Mascali (M), 15 in the village of Fondachello (F), on the seaside, three in Carrabba village (C), far from the coast, and two and three, respectively, from the water well ‘Carlino’ (W) and the source ‘Bufardo and Torregrossa’ (S) which feed the municipal system and provide drinking water to the Mascali area (Fig. 1, Table 3). One litre water samples were sent by the local health authorities to the SIALAB laboratory in Avola, Siracusa and 100 ml aliquots were screened for coliforms, E. coli and enterococci as faecal indicators, by standard bacteriological methods according to UNI EN ISO 9308-1:2014. Unfortunately, although the suspect of a water source for the viral outbreak was quickly communicated, only in August and November 2016, following the implementation of control measures, local authorities provided water samplings of appropriate volume to perform viral tests (10 l). Five water samples of 10 l each were collected from different sampling sites in F and R areas for viral screening (Table 3) and were concentrated according to the Italian National Institute of Health (Istituto Superiore di Sanità, ISS) protocols [22] through an ultra-filtration system (Sartoflow® Slice 200 Benchtop Crossflow System, Sartorius AG, Goettingen, Germany), using appropriate membranes (SG Hydrosart 10 kDa) pretreated with 300 ml of 3% Beef Extract (BE) at pH 7, by recirculating it for 10 min in the system, at a pressure not higher than 1.30 bar. Elution and recovery of viruses attached to the membranes were obtained by recirculating 200 ml of 3% BE at pH 9.5 for 3 h until a 10–12 ml volume of eluate was obtained. The pH of the eluate was then checked with litmus paper and if necessary brought to the value of 7 with HCl. Concentrated water samples were simultaneously tested for NoV in three laboratories: Enteric Viruses Laboratory, University of Palermo; Istituto Zooprofilattico Sperimentale della Sicilia (IZS), Palermo; and Istituto Superiore di Sanità (ISS), Rome. Total RNA was extracted from 500 and 560 µl of a concentrated water sample using BioMèrieux NucliSENS® Magnetic Extraction System (BioMèrieux) and QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) respectively, according to manufacturer’s instructions. Bonded RNA was eluted by 100 µl of Elution Buffer. Amplifications were performed at least in duplicate on whole and 1:10 diluted extracts. A Real-Time RT-PCR assay according to ISO/TS 15216- 2:2013, two in-house Real-Time RT-PCR protocols and the One-Step RT-PCR methods already used for amplifying ORF1 and ORF2 fragments in NoV-positive faecal samples were performed [19–21, 23]. Nucleotide sequences analysis and phylogenetic analysis PCR amplicons were purified by commercial columns (CleanSweep PCR Purification, Thermo Fisher Scientific, Vilnius, Lithuania) and sequenced by Sanger method and all nucleotide sequences were aligned using ClustalW. To identify NoV genotypes, RdRp and VP1 sequences were submitted to the RIVM NoV genotyping tool (http://www.rivm.nl/mpf/norovirus/typingtool). Phylogenetic trees were constructed using MEGA 7 and Neighbour-Joining method (bootstrap 1000 replicates). Sequences from stool and water samples are available at NCBI with the following accession numbers: MG517455 for the major capsid protein obtained from a water sample; MG517456 and MG517461, respectively, for the major capsid protein and RNA-dependent-RNA-polymerase, as representatives of the sequences obtained from the stool samples of those involved in the outbreak. Control measures After outbreak notification, on 21 May, the resort was closed for 2 days. On 25 May, following the first results of the bacteriological analyses on municipal water samples, the use of municipal water for human consumption was temporarily restricted (use of water was permitted only for cleaning purposes) in the Carrabba and Fondachello villages and in the area surrounding the resort. From 25 May the reservoir collecting the water from the Carlino well and its chlorination device were submitted to maintenance work and a new chlorination device was installed on the conduit bringing the water from the Bufardo source. Actually, the Bufardo source had been connected to the municipal network in 2003 but unfortunately a dedicated chlorination station had not been planned at that time. Therefore, unchlorinated water entered the system near Fondachello village on the seaside (Fig. 1), possibly diluting chlorine concentrations in water treated at the Carlino reservoir. Moreover, the exact path of the branch of the water network situated downstream the Bufardo source water connection was not marked in the maps of the municipality (Fig. 1, red line), indicating that this section had never been submitted to maintenance. The area surrounding the unmaintained section of the water pipeline was scattered with unauthorised cottages that were not connected to the sewerage systems. Likely, untreated waste from cesspools was unloaded directly into the soil, generating a risk of contamination of the municipal water system through leakage. No work had been made on the water pipelines before the outbreak but sections of the network had been temporarily closed during the winter season on the seaside (Fondachello). Local health authorities advised the connection of all households in the area to the sewerage system and the replacement of the water system section without maintenance. Health authorities also raised attention on AGE cases notification at all local Emergency Departments and Family Practitioners but there was no further evidence of AGE outbreaks in the months following the intervention. Results Data collected through the online questionnaire compiled 7–10 days after the outbreak revealed that none of the 28 students and the teacher taking part in the school trip showed gastroenteritis symptoms at the beginning of the journey and they had no contact with people who suffered from diarrhoea and/or vomiting in the previous days. Travellers had their first shared meal on 19 May (Day 1), 2016, at lunch on their arrival to the resort in Mascali. The list of foods included in the menu is shown in Table 2. Conversely, the dinner on Day 1 was not shared and the students went to dine in three different fast food restaurants in Acireale. On 20 May (Day 2), 17 students (58.6% of the travellers) presented with vomit and/or diarrhoea, falling ill 26–36 h after lunch on Day 1, strongly suggesting that exposure occurred on the first shared meal. Some students also vomited in the presence of the other travellers on the bus bringing them to Acireale. Five more students were symptomatic in the early hours of the following day (21 May, Day 3, from 0:30 am to 8:00 am), as a continuation of the wave of illnesses of the previous day. The teacher and one more student fell ill at 8:00 pm and 11:00 pm of Day 3, respectively, for a second wave of illnesses occurring >48 h from the first shared meal and <36 h from the first cases, with the last student falling sick on 22 May (Day 4) at 3:00 am (Fig. 2, Table 1). On the whole, the mean incubation period was 33.2 h. Based on the online survey, the severity of illness was evaluated by the Vesikari severity scale, grouping symptomatic travellers into severe (11/25 = 44%) and non-severe (14/25 = 56%) cases (Table 1). Vomiting and fever were the predominant symptoms, affecting 22 out of 25 (88%) symptomatic subjects, while diarrhoea was present in 14 out of 25 (56%). Hospitalisation for rehydration therapy was necessary for 13 students (91% [10/11] of severe and 21% [3/14] of non-severe cases). Emergency room (ER) or family practitioner visits were necessary for 21 students and 13 (62%) of them initiated an antibiotic therapy. Two students indicated that other family members had gastroenteric symptoms more than 24 h after their return home, on Day 3, 4 and 5. All those involved in the outbreak recovered within 4 days, with an average duration of illness of 38.4 h. A case-control study of the risk associated with eating foods at lunch on Day 1 or with attending different restaurants at dinner on Day 1 was attempted. A total of 19 subjects filled the section of the online survey concerning food consumption. Fourteen of the responders were among the 22 cases falling ill within 36 h from the first shared meal (first wave), while the remaining five responders were three travellers with supposed secondary infections (showing symptoms with a delay ⩾48 h; second wave) and two asymptomatic students that were used as controls. No statistically significant associations with any food items served at lunch on Day 1 was identified (P < 0.05; Table 2). However, all food consumed on that occasion was to some extent associated with the risk of disease, the highest OR being obtained with chicken (OR: 20.7143, P: 0.0678; Table 2). Actually, most of the cases ate all foods included in the menu, while one of the two asymptomatic controls did not participate in the meal and the other had a pasta dish only. No statistically significant association was found for attending a particular fast food restaurant at the dinner on Day 1 (OR: 2.5, P: 0.3635). The Real-Time RT-PCR assay allowed to detect GII genogroup NoV genome in all the 25 samples from symptomatic subjects, although two of them showed high Ct values, indicating lower shedding of viral genomes (Table 1). The samples from four asymptomatic students were negative for both NoV genogroups. By sequence analysis, a NoV genotype could be defined for 20/ 25 NoV-positive stool samples. All NoV strains from travellers involved in the outbreaks belonged to the same genotype, GII.P2/GII.2, showing >99% nucleotide identities (nt id) in both fragments (ORF1 and ORF2) analysed. Thirty-two water samples were collected and analysed for bacterial contamination (Table 3). Seven samples were taken from the resort (site R) and the remaining 25 from five other sites (C, F, M, S, W), representative of the entire municipal water network. The majority (10/12) of the water samples collected immediately after the onset of the outbreak, between 21 May and 16 June 2016, at the resort (site R) and at sites F and C, located close to the resort, on the seaside and in the interior, respectively, were positive for coliform bacteria. More than 1000 colony-forming units (CFU)/ 100 ml (up to 2700 CFU/100 ml) were detected in four samples, while their absence is required in 100 ml by the Standard of Water Quality in Italy and seven of 11 samples tested were also Escherichia coli-positive (up to 900 CFU/100 ml). Enterococci were searched in 11 samples and three of them were found positive for this faecal contamination indicator. From 7 July to 16 August 2016, after control measures on the water network were undertaken, 13 water samples were collected for follow-up laboratory examination of the water system at F, R and C sites. All of them were negative for bacterial indicators of faecal contamination. No faecal bacteria were identified in the seven samples collected from M, W and S sites, including the well and the source that fed the water network, at any sampling time, before and after implementing the control measures. The research of NoV genome was conducted on concentrated water samples collected on 1 August from R and F sites (Table 3). NoV genome was detected in a single sample collected from the water entry of the resort involved in the outbreak (site R). Actually, only two out of 26 One-Step RT-PCR repetitions were positive for NoV viral capsid gene. The two ORF2-positive results were obtained at the University of Palermo Laboratory using both 1:10 diluted and undiluted extracted RNA. The sequence analysis of the partial ORF2 portion (289 bp) indicated the presence of the same GII.2 capsid genotype detected in the stool samples of the students involved in the outbreak, showing 98.9% nt id. No ORF1 NoV sequence could be obtained from the NoV-positive water sample. Follow-up analyses for NoV genome, conducted on water samples collected on 20 November both at R and F sites found no positive samples. By sequence alignment, the NoV sequences obtained from Mascali outbreak samples were phylogenetically closely related in both ORF1 and ORF2 (99 and 98.9% of nt id, respectively) to a GII.P2/GII.2 strain responsible for a sporadic paediatric gastroenteritis case occurred in Palermo in January 2016 (PA 36/16, accession numbers MG517464 and MG517460, for ORF1 and ORF2, respectively) and previously detected at the Enteric Viruses Laboratory, University of Palermo. Whereas, three GII.P2/GII.2 Italian strains (PA 280/11, PA 366/11 and PA 3/ 14, accession numbers MG517462, MG517465 and MG517463, for ORF1 and MG517457 to −59, for ORF2, respectively), detected in Palermo in 2011 and 2014, were more distantly associated with the outbreak strain both in RdRp and VP1 (90% to 97.4% nt id and 96.1% to 98.1% nt id, respectively, in ORF1 and ORF2). Discussion Epidemiological features, such as the presence of vomiting in more than half (88%) of symptomatic subjects, the 33.2 h average incubation period and the 38.4 h average duration of illness, suggested a NoV aetiology for the investigated outbreak [13]. The detection of NoV genome in all 25 faecal samples from symptomatic subjects fully complies with the requirement of detection of at least one positive sample out of four tested that is considered to be sufficient to ascertain the aetiology of a NoV outbreak [24]. Moreover, molecular typing indicated that a single viral strain, belonging to genotype GII.P2/GII.2, infected all symptomatic subjects involved in the outbreak. NoVs are highly infectious and can easily spread from person to person but also through contaminated food or water. Therefore, settings such as health care institutions, schools and touristic resorts are mainly affected, since close contact and shared food services facilitate transmission. In Mascali, two waves of transmission were observed over three days, being consistent with food/waterborne followed by person-to-person and/or airborne (vomit) transmission (Fig. 2). The first wave of cases presented on 20 May 2016 after a single shared meal, while the last case was detected on 22 May. Exposure to vomit may have been responsible for some of the secondary cases occurred after the first wave of illness. Vomiting produces aerosols that can be inhaled and swallowed. The extremely low infectious dose of NoV makes the inhalation of vomit aerosols an effective transmission route in settings where there is close contact, like on a bus trip [25]. All food consumed in the only shared lunch on 19 May at the resort was associated with the risk of disease (Table 2), but the association was inconclusive since all cases shared the same dishes included in the lunch menu on Day 1. Although many foods were potentially involved, our attention focused on the tap water. In fact, trivial information from local health authorities reported that local population avoided tap water consumption and relied on bottled water for drinking and cooking. Many small outbreaks of gastroenteritis involving restaurants and family groups had been reported in the same area, especially at the beginning of the summer season, when vacation houses and recreational activities were re-opened after a period of low water demand. Illness caused by NoV is common and if the level of water contamination is low, the number of cases remains limited and water can persist as unrecognised source of infection. If polluted water was served for drinking and used for cooking and washing dishes in the Mascali resort at Day 1 a generalised contamination of food and surfaces could have occurred. The potential role of the foodhandlers as a source of contamination could not be analysed in this study but should also be taken into account. If water contamination has already been present in the area for a long time, infections may have occurred among the local food-handlers. The analysis of water samples taken immediately after the outbreak revealed the presence of bacterial indicators of faecal contamination in the water used in the resort, with the only exception of a sample of filter purified water used both for cooking and as drinking water. When extending the sampling to other areas of the municipal water network, many other sites, mostly on the seaside, showed faecal pollution, which demonstrates a poor protection of the water network from faecal leakage (Fig. 1). NoV contamination of waters generally originates from sewage since these enteric pathogens, as other enteric viruses affecting humans, are mostly species-specific and are shed at high concentrations in the stools of patients. When the analyses on the stools of the students made it clear that the outbreak had been caused by NoV, the viral genome was also searched in appropriate volumes of water samples, but this happened when correction measures on the water network had already been taken. Nevertheless, a single water sample, obtained 2 months after the outbreak (1 August 2016), still contained genome fragments that could be correlated to the same NoV genotype causing the outbreak. Although the finding of NoV genome does not demonstrate that entire infecting viral particles were present in the water, it could indicate that a NoV belonging to the same genotype involved in the outbreak was still circulating in the local population several weeks after the outbreak and it was someway able to penetrate the water network, probably indicating continuing faecal contamination through leakage from sewage systems. NoV is particularly resistant in the environment and also to chlorine, therefore transmission through chlorinated water is not exceptional [9, 12, 26]. Drinking water from municipal supplies fed by lake or groundwater has already been identified in the past as the source of waterborne NoV outbreaks in Italy [27–29] and NoV is now considered one of the pathogens causing the largest number of affected consumers in drinking waterborne outbreaks [12]. NoV can persist in water longer than 15 days [30] and it is resistant to chlorine disinfection when free chlorine levels or pre-treatment processes are inadequate [31, 32]. Moreover, microfiltration filters labelled as certified by NSF Standards 53 or 58 (with a pore size ranging from 0.05 to 5 µm) are not effective in removing viruses and can even decrease chlorine concentration in filtered water [33]. Actually, the owner of the resort involved in our outbreak was encouraged to serve tap water as drinking water at the restaurant relying on the effectiveness of such a filter purifier. Unfortunately, we did not have a tap water sample from the restaurant of the resort taken before correction measures and whose volume was appropriate for virus detection. Therefore, we could not demonstrate the failure to eliminate viral pollution of the microfiltration system. In Mascali, correction measures on the water network consisted in the maintenance of the old chlorination device and in the addition of a new one at a key point located upstream and close to suspected illegal sewage dumping (Fig. 1), but as poor protection of the pipeline from sewage leakage risked to reproduce faecal pollution episodes, renovation of the municipal water system was also advised, as well as systematic control of illegal sewage dumping. The additional chlorination device was apparently the most effective intervention, since right after its activation faecal bacterial pollution indicators disappeared. However, NoV genome was detected in a water sample taken 6 weeks after the last detection of bacterial indicators. Unfortunately, detection of faecal contamination of drinking water is only based on the use of bacterial indicators despite no correlation between levels of enteric bacteria and enteric viruses has been found [34–36]. According to ISGEV enteric virus circulation surveillance data, in 2016 the circulation of GII.P2/GII.2 NoV strains in Palermo, home town of the travellers involved in the outbreak, was the highest ever observed from the starting of local NoV surveillance in 2004. GII.P2/GII.2 was the fourth most common genotype in 2016 (12.2% prevalence), following GII.P4/GII.4 Sydney 2012 (28.4%), GII.P16/GII.2 (18.9%) and GII.Pe/GII.4 Sydney 2012 (17.6%). GII.P2/GII.2 strains were mostly observed starting from July 2016, leaving space for the possibility that their increased circulation was linked to the introduction of the ‘Mascali clone’. However, a couple of GII.P2/GII.2 strains (including PA36/16) genetically close to the ‘Mascali clone’ had been already detected in January 2016 in Palermo and GII.P2/GII.2 NoV had sporadically been observed also in 2011 and 2014. Moreover, the circulation of GII.P2/GII.2 NoVs has been increasing worldwide over the last period [37]. Therefore, the unusually high circulation observed in Palermo might just reflect the global epidemiology of GII.P2/GII.2 strains. Noteworthy, the older GII.P2/GII.2 NoV strains detected in Palermo were closely related in the ORF2 (98.9% nt id) to the GII.P16/GII.2 Nashville strain (KY865307) which is supposed to have generated the recombinant GII.P16/GII.4 clone recently circulating epidemically worldwide [37, 38] and were more distantly associated (95.3% nt id) to the GII.2 capsid sequence (AB662868) of another putative GII.P16/GII.4 ancestor from Japan [39]. The increasing frequency of detection of waterborne NoV outbreaks calls for the introduction of routine monitoring for viruses of drinking waters and for the definition of affordable and reliable indicators for the presence of NoV in drinking waters. The use of ultrafiltration systems as a concentration method in order to recover viruses from environmental waters combined with RT-PCR to detect viral genomes has been demonstrated to grant a recovery rate of >50% of multiple viruses [40] and is at present one of the most sensitive approaches for the detection of NoVs, although the discrimination between viable and nonviable viruses remains a problem. The burden of NoV outbreaks, either food associated or waterborne, is still poorly defined due to the lack of specific diagnostic tools in many clinical laboratories. NoV epidemics should always be suspected when the main symptom is vomiting and the duration of disease is short (<48 h). Waterborne transmission is suggested by the rapid spread of disease and the high attack rate. Care should be taken to perform sensitive diagnostic tests on a sufficient number of stool samples in order to identify NoV as the causative agent of enteritis outbreaks, possibly preferring genome screening over antigen detection. Only the systematic detection of NoV outbreaks can guide in the choice of the safety procedures that could be most appropriate in containing such microbiological hazard.

What caused the event?

1543

Waterborne Norovirus outbreak at a seaside resort likely originating from municipal water distribution system failure

Abstract In May 2016 a Norovirus (NoV) gastroenteritis outbreak involved a high school class visiting a seaside resort near Taormina (Mascali, Sicily). Twenty-four students and a teacher were affected and 17 of them showed symptoms on the second day of the journey, while the others got ill within the following 2 days. Symptoms included vomiting, diarrhoea and fever, and 12 students required hospitalisation. Stool samples tested positive for NoV genome by Real-Time polymerase chain reaction assay in all 25 symptomatic subjects. The GII.P2/GII.2 NoV genotype was linked to the outbreak by ORF1/ORF2 sequence analysis. The epidemiological features of the outbreak were consistent with food/waterborne followed by person-to-person and/ or vomit transmission. Food consumed at a shared lunch on the first day of the trip was associated to illness and drinking un-bottled tap water was also considered as a risk factor. The analysis of water samples revealed the presence of bacterial indicators of faecal contamination in the water used in the resort as well as in other areas of the municipal water network, linking the NoV gastroenteritis outbreak to tap water pollution from sewage leakage. From a single water sample, an amplicon whose sequence corresponded to the capsid genotype recovered from patients could be obtained. Introduction Noroviruses (NoVs) are recognised as a leading cause of sporadic cases and outbreaks of acute gastroenteritis (AGE) across all age groups [1, 2]. They are transmitted primarily through the faecal-oral route either by direct person–person spread and with consumption of contaminated food or water [1, 3, 4]. NoVs are non-enveloped RNA viruses belonging to the Caliciviridae family and are classified into seven distinct genogroups (GI-GVII) on the basis of the major capsid protein VP1 sequence. GI, GII and GIV NoVs infect humans and they have been divided into more than 30 genotypes based on the sequences of polymerase (ORF1) and capsid proteins (ORF2) [5]. NoV GII.4 genotype has been associated with the vast majority of NoV related cases of AGE worldwide although many genotypes, belonging to both GI and GII genogroups are involved in foodborne and waterborne outbreaks [6, 7]. NoVs are highly contagious owing to their high stability in the environment and their ability to infect at low doses. In fact, just 10–100 virions of NoV may be sufficient to cause AGE [8, 9]. NoVs are considered the major cause of water-related disease in semi-closed institutions, including restaurants, schools, hospitals, nursing homes and cruise ships [10–12]. To confirm NoV as a cause of outbreaks, the presence of four epidemiological features of disease, vomiting in more than half of affected persons, a mean incubation period of 24/48 h, a mean duration of illness of 12–60 h and the absence of bacterial pathogens in stool culture, should be evaluated [13]. Waterborne NoV outbreaks can be caused by contaminated drinking water despite correct chlorine concentration and in the absence of coliform bacteria [10, 14–17]. The aim of this study was to investigate an outbreak of NoV gastroenteritis associated with drinking water involving a high school class travelling from Palermo (West Sicily) to Acireale (East Sicily) on an educational tour in May 2016. Stool samples were collected soon after the onset of AGE symptoms to ascertain the aetiology of illness. The source of infection was investigated collecting epidemiological data through an online questionnaire and, in order to explore a possible transmission route, laboratory tests on water from the municipal network were performed. Material and methods Epidemiological investigation From 20 to 22 May 2016 an AGE outbreak occurred in Mascali, a small town near Taormina, involving a school class of 28 students, 13 girls and 15 boys aged between 14 and 18 years and a 54-year-old female teacher, during a trip to participate in a theatre festival in Acireale. A total of 25/29 persons were affected by nausea, vomiting, abdominal cramps, diarrhoea and fever while they were staying at a seaside resort in Mascali, on the way to Acireale. To obtain data on symptoms and the correlated risk factors all participants filled an online questionnaire investigating personal data (age and sex), travel and accommodation details and food consumption over 48 h before symptoms presentation, time of onset and duration of symptoms, therapy and/or hospitalisation required and history of contact with people affected by diarrhoea and/or vomiting (Table 1). The severity of illness was analysed by the Vesikari severity scale allowing to distinguish non-severe (score <11) and severe (score ⩾11) AGE (Table 1). Vesikari score was calculated from maximum number of episodes/day and duration of both diarrhoea and vomiting (score 0–3 for each of the 4 parameters), temperature (score 0–3), level of dehydration (score 0–3) and treatment required (score 0–2) [18]. A case-control study was conducted to establish if food consumed before the onset of symptoms was associated with the disease. A case was defined as a student attending the tour and reporting diarrhoea and/or vomiting within 36 h from the first shared meal during the tour, while students who did not present any of the previous symptoms or those showing the first symptoms with a delay of ⩾48 h (secondary cases) were used as controls. Variables were studied using odds ratio (OR) and its 95% CI and statistical associations were determined using the χ2 test, with the level of statistical significance set at 0.05 (Table 2). NoV detection in stool samples A stool sample was obtained from each of the 29 travellers, soon after (3–5 days) the onset of the outbreak. No stool samples could be obtained from the food-handlers working at the resort. All samples were tested for NoV at the Enteric Viruses Laboratory of the University of Palermo, member of the Italian Study Group for Enteric Viruses (ISGEV), by a Real-Time polymerase chain reaction (RT-PCR) assay that allows to differentiate GI and GII NoVs [19]. Total RNA was extracted from stool samples using QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) and cDNA was generated by MMLV reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were considered positive for NoVs when cycle threshold (ct) of RT-PCR was ⩽30 and lowpositive when 30<ct⩽35. For all NoV-positive samples two genomic fragments were amplified by conventional RT-PCR using Superscript III One-Step RT-PCR system with Platinum Taq DNA polymerase (Invitrogen) and primers JV12a/JV13b, targeting 326 bps of the RdRp region A [20] and either COG1F/ GISKR or COG2F/GIISKR, amplifying a 315 bps 5′portion of GI or GII VP1 protein [19, 21]. Environmental investigation From May through August 2016, a total of 32 tap water samples were collected from the Mascali municipal network. In particular, seven samples were collected at the resort where the outbreak occurred and from its water entry (R), two in the city centre of Mascali (M), 15 in the village of Fondachello (F), on the seaside, three in Carrabba village (C), far from the coast, and two and three, respectively, from the water well ‘Carlino’ (W) and the source ‘Bufardo and Torregrossa’ (S) which feed the municipal system and provide drinking water to the Mascali area (Fig. 1, Table 3). One litre water samples were sent by the local health authorities to the SIALAB laboratory in Avola, Siracusa and 100 ml aliquots were screened for coliforms, E. coli and enterococci as faecal indicators, by standard bacteriological methods according to UNI EN ISO 9308-1:2014. Unfortunately, although the suspect of a water source for the viral outbreak was quickly communicated, only in August and November 2016, following the implementation of control measures, local authorities provided water samplings of appropriate volume to perform viral tests (10 l). Five water samples of 10 l each were collected from different sampling sites in F and R areas for viral screening (Table 3) and were concentrated according to the Italian National Institute of Health (Istituto Superiore di Sanità, ISS) protocols [22] through an ultra-filtration system (Sartoflow® Slice 200 Benchtop Crossflow System, Sartorius AG, Goettingen, Germany), using appropriate membranes (SG Hydrosart 10 kDa) pretreated with 300 ml of 3% Beef Extract (BE) at pH 7, by recirculating it for 10 min in the system, at a pressure not higher than 1.30 bar. Elution and recovery of viruses attached to the membranes were obtained by recirculating 200 ml of 3% BE at pH 9.5 for 3 h until a 10–12 ml volume of eluate was obtained. The pH of the eluate was then checked with litmus paper and if necessary brought to the value of 7 with HCl. Concentrated water samples were simultaneously tested for NoV in three laboratories: Enteric Viruses Laboratory, University of Palermo; Istituto Zooprofilattico Sperimentale della Sicilia (IZS), Palermo; and Istituto Superiore di Sanità (ISS), Rome. Total RNA was extracted from 500 and 560 µl of a concentrated water sample using BioMèrieux NucliSENS® Magnetic Extraction System (BioMèrieux) and QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) respectively, according to manufacturer’s instructions. Bonded RNA was eluted by 100 µl of Elution Buffer. Amplifications were performed at least in duplicate on whole and 1:10 diluted extracts. A Real-Time RT-PCR assay according to ISO/TS 15216- 2:2013, two in-house Real-Time RT-PCR protocols and the One-Step RT-PCR methods already used for amplifying ORF1 and ORF2 fragments in NoV-positive faecal samples were performed [19–21, 23]. Nucleotide sequences analysis and phylogenetic analysis PCR amplicons were purified by commercial columns (CleanSweep PCR Purification, Thermo Fisher Scientific, Vilnius, Lithuania) and sequenced by Sanger method and all nucleotide sequences were aligned using ClustalW. To identify NoV genotypes, RdRp and VP1 sequences were submitted to the RIVM NoV genotyping tool (http://www.rivm.nl/mpf/norovirus/typingtool). Phylogenetic trees were constructed using MEGA 7 and Neighbour-Joining method (bootstrap 1000 replicates). Sequences from stool and water samples are available at NCBI with the following accession numbers: MG517455 for the major capsid protein obtained from a water sample; MG517456 and MG517461, respectively, for the major capsid protein and RNA-dependent-RNA-polymerase, as representatives of the sequences obtained from the stool samples of those involved in the outbreak. Control measures After outbreak notification, on 21 May, the resort was closed for 2 days. On 25 May, following the first results of the bacteriological analyses on municipal water samples, the use of municipal water for human consumption was temporarily restricted (use of water was permitted only for cleaning purposes) in the Carrabba and Fondachello villages and in the area surrounding the resort. From 25 May the reservoir collecting the water from the Carlino well and its chlorination device were submitted to maintenance work and a new chlorination device was installed on the conduit bringing the water from the Bufardo source. Actually, the Bufardo source had been connected to the municipal network in 2003 but unfortunately a dedicated chlorination station had not been planned at that time. Therefore, unchlorinated water entered the system near Fondachello village on the seaside (Fig. 1), possibly diluting chlorine concentrations in water treated at the Carlino reservoir. Moreover, the exact path of the branch of the water network situated downstream the Bufardo source water connection was not marked in the maps of the municipality (Fig. 1, red line), indicating that this section had never been submitted to maintenance. The area surrounding the unmaintained section of the water pipeline was scattered with unauthorised cottages that were not connected to the sewerage systems. Likely, untreated waste from cesspools was unloaded directly into the soil, generating a risk of contamination of the municipal water system through leakage. No work had been made on the water pipelines before the outbreak but sections of the network had been temporarily closed during the winter season on the seaside (Fondachello). Local health authorities advised the connection of all households in the area to the sewerage system and the replacement of the water system section without maintenance. Health authorities also raised attention on AGE cases notification at all local Emergency Departments and Family Practitioners but there was no further evidence of AGE outbreaks in the months following the intervention. Results Data collected through the online questionnaire compiled 7–10 days after the outbreak revealed that none of the 28 students and the teacher taking part in the school trip showed gastroenteritis symptoms at the beginning of the journey and they had no contact with people who suffered from diarrhoea and/or vomiting in the previous days. Travellers had their first shared meal on 19 May (Day 1), 2016, at lunch on their arrival to the resort in Mascali. The list of foods included in the menu is shown in Table 2. Conversely, the dinner on Day 1 was not shared and the students went to dine in three different fast food restaurants in Acireale. On 20 May (Day 2), 17 students (58.6% of the travellers) presented with vomit and/or diarrhoea, falling ill 26–36 h after lunch on Day 1, strongly suggesting that exposure occurred on the first shared meal. Some students also vomited in the presence of the other travellers on the bus bringing them to Acireale. Five more students were symptomatic in the early hours of the following day (21 May, Day 3, from 0:30 am to 8:00 am), as a continuation of the wave of illnesses of the previous day. The teacher and one more student fell ill at 8:00 pm and 11:00 pm of Day 3, respectively, for a second wave of illnesses occurring >48 h from the first shared meal and <36 h from the first cases, with the last student falling sick on 22 May (Day 4) at 3:00 am (Fig. 2, Table 1). On the whole, the mean incubation period was 33.2 h. Based on the online survey, the severity of illness was evaluated by the Vesikari severity scale, grouping symptomatic travellers into severe (11/25 = 44%) and non-severe (14/25 = 56%) cases (Table 1). Vomiting and fever were the predominant symptoms, affecting 22 out of 25 (88%) symptomatic subjects, while diarrhoea was present in 14 out of 25 (56%). Hospitalisation for rehydration therapy was necessary for 13 students (91% [10/11] of severe and 21% [3/14] of non-severe cases). Emergency room (ER) or family practitioner visits were necessary for 21 students and 13 (62%) of them initiated an antibiotic therapy. Two students indicated that other family members had gastroenteric symptoms more than 24 h after their return home, on Day 3, 4 and 5. All those involved in the outbreak recovered within 4 days, with an average duration of illness of 38.4 h. A case-control study of the risk associated with eating foods at lunch on Day 1 or with attending different restaurants at dinner on Day 1 was attempted. A total of 19 subjects filled the section of the online survey concerning food consumption. Fourteen of the responders were among the 22 cases falling ill within 36 h from the first shared meal (first wave), while the remaining five responders were three travellers with supposed secondary infections (showing symptoms with a delay ⩾48 h; second wave) and two asymptomatic students that were used as controls. No statistically significant associations with any food items served at lunch on Day 1 was identified (P < 0.05; Table 2). However, all food consumed on that occasion was to some extent associated with the risk of disease, the highest OR being obtained with chicken (OR: 20.7143, P: 0.0678; Table 2). Actually, most of the cases ate all foods included in the menu, while one of the two asymptomatic controls did not participate in the meal and the other had a pasta dish only. No statistically significant association was found for attending a particular fast food restaurant at the dinner on Day 1 (OR: 2.5, P: 0.3635). The Real-Time RT-PCR assay allowed to detect GII genogroup NoV genome in all the 25 samples from symptomatic subjects, although two of them showed high Ct values, indicating lower shedding of viral genomes (Table 1). The samples from four asymptomatic students were negative for both NoV genogroups. By sequence analysis, a NoV genotype could be defined for 20/ 25 NoV-positive stool samples. All NoV strains from travellers involved in the outbreaks belonged to the same genotype, GII.P2/GII.2, showing >99% nucleotide identities (nt id) in both fragments (ORF1 and ORF2) analysed. Thirty-two water samples were collected and analysed for bacterial contamination (Table 3). Seven samples were taken from the resort (site R) and the remaining 25 from five other sites (C, F, M, S, W), representative of the entire municipal water network. The majority (10/12) of the water samples collected immediately after the onset of the outbreak, between 21 May and 16 June 2016, at the resort (site R) and at sites F and C, located close to the resort, on the seaside and in the interior, respectively, were positive for coliform bacteria. More than 1000 colony-forming units (CFU)/ 100 ml (up to 2700 CFU/100 ml) were detected in four samples, while their absence is required in 100 ml by the Standard of Water Quality in Italy and seven of 11 samples tested were also Escherichia coli-positive (up to 900 CFU/100 ml). Enterococci were searched in 11 samples and three of them were found positive for this faecal contamination indicator. From 7 July to 16 August 2016, after control measures on the water network were undertaken, 13 water samples were collected for follow-up laboratory examination of the water system at F, R and C sites. All of them were negative for bacterial indicators of faecal contamination. No faecal bacteria were identified in the seven samples collected from M, W and S sites, including the well and the source that fed the water network, at any sampling time, before and after implementing the control measures. The research of NoV genome was conducted on concentrated water samples collected on 1 August from R and F sites (Table 3). NoV genome was detected in a single sample collected from the water entry of the resort involved in the outbreak (site R). Actually, only two out of 26 One-Step RT-PCR repetitions were positive for NoV viral capsid gene. The two ORF2-positive results were obtained at the University of Palermo Laboratory using both 1:10 diluted and undiluted extracted RNA. The sequence analysis of the partial ORF2 portion (289 bp) indicated the presence of the same GII.2 capsid genotype detected in the stool samples of the students involved in the outbreak, showing 98.9% nt id. No ORF1 NoV sequence could be obtained from the NoV-positive water sample. Follow-up analyses for NoV genome, conducted on water samples collected on 20 November both at R and F sites found no positive samples. By sequence alignment, the NoV sequences obtained from Mascali outbreak samples were phylogenetically closely related in both ORF1 and ORF2 (99 and 98.9% of nt id, respectively) to a GII.P2/GII.2 strain responsible for a sporadic paediatric gastroenteritis case occurred in Palermo in January 2016 (PA 36/16, accession numbers MG517464 and MG517460, for ORF1 and ORF2, respectively) and previously detected at the Enteric Viruses Laboratory, University of Palermo. Whereas, three GII.P2/GII.2 Italian strains (PA 280/11, PA 366/11 and PA 3/ 14, accession numbers MG517462, MG517465 and MG517463, for ORF1 and MG517457 to −59, for ORF2, respectively), detected in Palermo in 2011 and 2014, were more distantly associated with the outbreak strain both in RdRp and VP1 (90% to 97.4% nt id and 96.1% to 98.1% nt id, respectively, in ORF1 and ORF2). Discussion Epidemiological features, such as the presence of vomiting in more than half (88%) of symptomatic subjects, the 33.2 h average incubation period and the 38.4 h average duration of illness, suggested a NoV aetiology for the investigated outbreak [13]. The detection of NoV genome in all 25 faecal samples from symptomatic subjects fully complies with the requirement of detection of at least one positive sample out of four tested that is considered to be sufficient to ascertain the aetiology of a NoV outbreak [24]. Moreover, molecular typing indicated that a single viral strain, belonging to genotype GII.P2/GII.2, infected all symptomatic subjects involved in the outbreak. NoVs are highly infectious and can easily spread from person to person but also through contaminated food or water. Therefore, settings such as health care institutions, schools and touristic resorts are mainly affected, since close contact and shared food services facilitate transmission. In Mascali, two waves of transmission were observed over three days, being consistent with food/waterborne followed by person-to-person and/or airborne (vomit) transmission (Fig. 2). The first wave of cases presented on 20 May 2016 after a single shared meal, while the last case was detected on 22 May. Exposure to vomit may have been responsible for some of the secondary cases occurred after the first wave of illness. Vomiting produces aerosols that can be inhaled and swallowed. The extremely low infectious dose of NoV makes the inhalation of vomit aerosols an effective transmission route in settings where there is close contact, like on a bus trip [25]. All food consumed in the only shared lunch on 19 May at the resort was associated with the risk of disease (Table 2), but the association was inconclusive since all cases shared the same dishes included in the lunch menu on Day 1. Although many foods were potentially involved, our attention focused on the tap water. In fact, trivial information from local health authorities reported that local population avoided tap water consumption and relied on bottled water for drinking and cooking. Many small outbreaks of gastroenteritis involving restaurants and family groups had been reported in the same area, especially at the beginning of the summer season, when vacation houses and recreational activities were re-opened after a period of low water demand. Illness caused by NoV is common and if the level of water contamination is low, the number of cases remains limited and water can persist as unrecognised source of infection. If polluted water was served for drinking and used for cooking and washing dishes in the Mascali resort at Day 1 a generalised contamination of food and surfaces could have occurred. The potential role of the foodhandlers as a source of contamination could not be analysed in this study but should also be taken into account. If water contamination has already been present in the area for a long time, infections may have occurred among the local food-handlers. The analysis of water samples taken immediately after the outbreak revealed the presence of bacterial indicators of faecal contamination in the water used in the resort, with the only exception of a sample of filter purified water used both for cooking and as drinking water. When extending the sampling to other areas of the municipal water network, many other sites, mostly on the seaside, showed faecal pollution, which demonstrates a poor protection of the water network from faecal leakage (Fig. 1). NoV contamination of waters generally originates from sewage since these enteric pathogens, as other enteric viruses affecting humans, are mostly species-specific and are shed at high concentrations in the stools of patients. When the analyses on the stools of the students made it clear that the outbreak had been caused by NoV, the viral genome was also searched in appropriate volumes of water samples, but this happened when correction measures on the water network had already been taken. Nevertheless, a single water sample, obtained 2 months after the outbreak (1 August 2016), still contained genome fragments that could be correlated to the same NoV genotype causing the outbreak. Although the finding of NoV genome does not demonstrate that entire infecting viral particles were present in the water, it could indicate that a NoV belonging to the same genotype involved in the outbreak was still circulating in the local population several weeks after the outbreak and it was someway able to penetrate the water network, probably indicating continuing faecal contamination through leakage from sewage systems. NoV is particularly resistant in the environment and also to chlorine, therefore transmission through chlorinated water is not exceptional [9, 12, 26]. Drinking water from municipal supplies fed by lake or groundwater has already been identified in the past as the source of waterborne NoV outbreaks in Italy [27–29] and NoV is now considered one of the pathogens causing the largest number of affected consumers in drinking waterborne outbreaks [12]. NoV can persist in water longer than 15 days [30] and it is resistant to chlorine disinfection when free chlorine levels or pre-treatment processes are inadequate [31, 32]. Moreover, microfiltration filters labelled as certified by NSF Standards 53 or 58 (with a pore size ranging from 0.05 to 5 µm) are not effective in removing viruses and can even decrease chlorine concentration in filtered water [33]. Actually, the owner of the resort involved in our outbreak was encouraged to serve tap water as drinking water at the restaurant relying on the effectiveness of such a filter purifier. Unfortunately, we did not have a tap water sample from the restaurant of the resort taken before correction measures and whose volume was appropriate for virus detection. Therefore, we could not demonstrate the failure to eliminate viral pollution of the microfiltration system. In Mascali, correction measures on the water network consisted in the maintenance of the old chlorination device and in the addition of a new one at a key point located upstream and close to suspected illegal sewage dumping (Fig. 1), but as poor protection of the pipeline from sewage leakage risked to reproduce faecal pollution episodes, renovation of the municipal water system was also advised, as well as systematic control of illegal sewage dumping. The additional chlorination device was apparently the most effective intervention, since right after its activation faecal bacterial pollution indicators disappeared. However, NoV genome was detected in a water sample taken 6 weeks after the last detection of bacterial indicators. Unfortunately, detection of faecal contamination of drinking water is only based on the use of bacterial indicators despite no correlation between levels of enteric bacteria and enteric viruses has been found [34–36]. According to ISGEV enteric virus circulation surveillance data, in 2016 the circulation of GII.P2/GII.2 NoV strains in Palermo, home town of the travellers involved in the outbreak, was the highest ever observed from the starting of local NoV surveillance in 2004. GII.P2/GII.2 was the fourth most common genotype in 2016 (12.2% prevalence), following GII.P4/GII.4 Sydney 2012 (28.4%), GII.P16/GII.2 (18.9%) and GII.Pe/GII.4 Sydney 2012 (17.6%). GII.P2/GII.2 strains were mostly observed starting from July 2016, leaving space for the possibility that their increased circulation was linked to the introduction of the ‘Mascali clone’. However, a couple of GII.P2/GII.2 strains (including PA36/16) genetically close to the ‘Mascali clone’ had been already detected in January 2016 in Palermo and GII.P2/GII.2 NoV had sporadically been observed also in 2011 and 2014. Moreover, the circulation of GII.P2/GII.2 NoVs has been increasing worldwide over the last period [37]. Therefore, the unusually high circulation observed in Palermo might just reflect the global epidemiology of GII.P2/GII.2 strains. Noteworthy, the older GII.P2/GII.2 NoV strains detected in Palermo were closely related in the ORF2 (98.9% nt id) to the GII.P16/GII.2 Nashville strain (KY865307) which is supposed to have generated the recombinant GII.P16/GII.4 clone recently circulating epidemically worldwide [37, 38] and were more distantly associated (95.3% nt id) to the GII.2 capsid sequence (AB662868) of another putative GII.P16/GII.4 ancestor from Japan [39]. The increasing frequency of detection of waterborne NoV outbreaks calls for the introduction of routine monitoring for viruses of drinking waters and for the definition of affordable and reliable indicators for the presence of NoV in drinking waters. The use of ultrafiltration systems as a concentration method in order to recover viruses from environmental waters combined with RT-PCR to detect viral genomes has been demonstrated to grant a recovery rate of >50% of multiple viruses [40] and is at present one of the most sensitive approaches for the detection of NoVs, although the discrimination between viable and nonviable viruses remains a problem. The burden of NoV outbreaks, either food associated or waterborne, is still poorly defined due to the lack of specific diagnostic tools in many clinical laboratories. NoV epidemics should always be suspected when the main symptom is vomiting and the duration of disease is short (<48 h). Waterborne transmission is suggested by the rapid spread of disease and the high attack rate. Care should be taken to perform sensitive diagnostic tests on a sufficient number of stool samples in order to identify NoV as the causative agent of enteritis outbreaks, possibly preferring genome screening over antigen detection. Only the systematic detection of NoV outbreaks can guide in the choice of the safety procedures that could be most appropriate in containing such microbiological hazard.

What was the cause of the event?

1544

Waterborne Norovirus outbreak at a seaside resort likely originating from municipal water distribution system failure

Abstract In May 2016 a Norovirus (NoV) gastroenteritis outbreak involved a high school class visiting a seaside resort near Taormina (Mascali, Sicily). Twenty-four students and a teacher were affected and 17 of them showed symptoms on the second day of the journey, while the others got ill within the following 2 days. Symptoms included vomiting, diarrhoea and fever, and 12 students required hospitalisation. Stool samples tested positive for NoV genome by Real-Time polymerase chain reaction assay in all 25 symptomatic subjects. The GII.P2/GII.2 NoV genotype was linked to the outbreak by ORF1/ORF2 sequence analysis. The epidemiological features of the outbreak were consistent with food/waterborne followed by person-to-person and/ or vomit transmission. Food consumed at a shared lunch on the first day of the trip was associated to illness and drinking un-bottled tap water was also considered as a risk factor. The analysis of water samples revealed the presence of bacterial indicators of faecal contamination in the water used in the resort as well as in other areas of the municipal water network, linking the NoV gastroenteritis outbreak to tap water pollution from sewage leakage. From a single water sample, an amplicon whose sequence corresponded to the capsid genotype recovered from patients could be obtained. Introduction Noroviruses (NoVs) are recognised as a leading cause of sporadic cases and outbreaks of acute gastroenteritis (AGE) across all age groups [1, 2]. They are transmitted primarily through the faecal-oral route either by direct person–person spread and with consumption of contaminated food or water [1, 3, 4]. NoVs are non-enveloped RNA viruses belonging to the Caliciviridae family and are classified into seven distinct genogroups (GI-GVII) on the basis of the major capsid protein VP1 sequence. GI, GII and GIV NoVs infect humans and they have been divided into more than 30 genotypes based on the sequences of polymerase (ORF1) and capsid proteins (ORF2) [5]. NoV GII.4 genotype has been associated with the vast majority of NoV related cases of AGE worldwide although many genotypes, belonging to both GI and GII genogroups are involved in foodborne and waterborne outbreaks [6, 7]. NoVs are highly contagious owing to their high stability in the environment and their ability to infect at low doses. In fact, just 10–100 virions of NoV may be sufficient to cause AGE [8, 9]. NoVs are considered the major cause of water-related disease in semi-closed institutions, including restaurants, schools, hospitals, nursing homes and cruise ships [10–12]. To confirm NoV as a cause of outbreaks, the presence of four epidemiological features of disease, vomiting in more than half of affected persons, a mean incubation period of 24/48 h, a mean duration of illness of 12–60 h and the absence of bacterial pathogens in stool culture, should be evaluated [13]. Waterborne NoV outbreaks can be caused by contaminated drinking water despite correct chlorine concentration and in the absence of coliform bacteria [10, 14–17]. The aim of this study was to investigate an outbreak of NoV gastroenteritis associated with drinking water involving a high school class travelling from Palermo (West Sicily) to Acireale (East Sicily) on an educational tour in May 2016. Stool samples were collected soon after the onset of AGE symptoms to ascertain the aetiology of illness. The source of infection was investigated collecting epidemiological data through an online questionnaire and, in order to explore a possible transmission route, laboratory tests on water from the municipal network were performed. Material and methods Epidemiological investigation From 20 to 22 May 2016 an AGE outbreak occurred in Mascali, a small town near Taormina, involving a school class of 28 students, 13 girls and 15 boys aged between 14 and 18 years and a 54-year-old female teacher, during a trip to participate in a theatre festival in Acireale. A total of 25/29 persons were affected by nausea, vomiting, abdominal cramps, diarrhoea and fever while they were staying at a seaside resort in Mascali, on the way to Acireale. To obtain data on symptoms and the correlated risk factors all participants filled an online questionnaire investigating personal data (age and sex), travel and accommodation details and food consumption over 48 h before symptoms presentation, time of onset and duration of symptoms, therapy and/or hospitalisation required and history of contact with people affected by diarrhoea and/or vomiting (Table 1). The severity of illness was analysed by the Vesikari severity scale allowing to distinguish non-severe (score <11) and severe (score ⩾11) AGE (Table 1). Vesikari score was calculated from maximum number of episodes/day and duration of both diarrhoea and vomiting (score 0–3 for each of the 4 parameters), temperature (score 0–3), level of dehydration (score 0–3) and treatment required (score 0–2) [18]. A case-control study was conducted to establish if food consumed before the onset of symptoms was associated with the disease. A case was defined as a student attending the tour and reporting diarrhoea and/or vomiting within 36 h from the first shared meal during the tour, while students who did not present any of the previous symptoms or those showing the first symptoms with a delay of ⩾48 h (secondary cases) were used as controls. Variables were studied using odds ratio (OR) and its 95% CI and statistical associations were determined using the χ2 test, with the level of statistical significance set at 0.05 (Table 2). NoV detection in stool samples A stool sample was obtained from each of the 29 travellers, soon after (3–5 days) the onset of the outbreak. No stool samples could be obtained from the food-handlers working at the resort. All samples were tested for NoV at the Enteric Viruses Laboratory of the University of Palermo, member of the Italian Study Group for Enteric Viruses (ISGEV), by a Real-Time polymerase chain reaction (RT-PCR) assay that allows to differentiate GI and GII NoVs [19]. Total RNA was extracted from stool samples using QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) and cDNA was generated by MMLV reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were considered positive for NoVs when cycle threshold (ct) of RT-PCR was ⩽30 and lowpositive when 30<ct⩽35. For all NoV-positive samples two genomic fragments were amplified by conventional RT-PCR using Superscript III One-Step RT-PCR system with Platinum Taq DNA polymerase (Invitrogen) and primers JV12a/JV13b, targeting 326 bps of the RdRp region A [20] and either COG1F/ GISKR or COG2F/GIISKR, amplifying a 315 bps 5′portion of GI or GII VP1 protein [19, 21]. Environmental investigation From May through August 2016, a total of 32 tap water samples were collected from the Mascali municipal network. In particular, seven samples were collected at the resort where the outbreak occurred and from its water entry (R), two in the city centre of Mascali (M), 15 in the village of Fondachello (F), on the seaside, three in Carrabba village (C), far from the coast, and two and three, respectively, from the water well ‘Carlino’ (W) and the source ‘Bufardo and Torregrossa’ (S) which feed the municipal system and provide drinking water to the Mascali area (Fig. 1, Table 3). One litre water samples were sent by the local health authorities to the SIALAB laboratory in Avola, Siracusa and 100 ml aliquots were screened for coliforms, E. coli and enterococci as faecal indicators, by standard bacteriological methods according to UNI EN ISO 9308-1:2014. Unfortunately, although the suspect of a water source for the viral outbreak was quickly communicated, only in August and November 2016, following the implementation of control measures, local authorities provided water samplings of appropriate volume to perform viral tests (10 l). Five water samples of 10 l each were collected from different sampling sites in F and R areas for viral screening (Table 3) and were concentrated according to the Italian National Institute of Health (Istituto Superiore di Sanità, ISS) protocols [22] through an ultra-filtration system (Sartoflow® Slice 200 Benchtop Crossflow System, Sartorius AG, Goettingen, Germany), using appropriate membranes (SG Hydrosart 10 kDa) pretreated with 300 ml of 3% Beef Extract (BE) at pH 7, by recirculating it for 10 min in the system, at a pressure not higher than 1.30 bar. Elution and recovery of viruses attached to the membranes were obtained by recirculating 200 ml of 3% BE at pH 9.5 for 3 h until a 10–12 ml volume of eluate was obtained. The pH of the eluate was then checked with litmus paper and if necessary brought to the value of 7 with HCl. Concentrated water samples were simultaneously tested for NoV in three laboratories: Enteric Viruses Laboratory, University of Palermo; Istituto Zooprofilattico Sperimentale della Sicilia (IZS), Palermo; and Istituto Superiore di Sanità (ISS), Rome. Total RNA was extracted from 500 and 560 µl of a concentrated water sample using BioMèrieux NucliSENS® Magnetic Extraction System (BioMèrieux) and QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) respectively, according to manufacturer’s instructions. Bonded RNA was eluted by 100 µl of Elution Buffer. Amplifications were performed at least in duplicate on whole and 1:10 diluted extracts. A Real-Time RT-PCR assay according to ISO/TS 15216- 2:2013, two in-house Real-Time RT-PCR protocols and the One-Step RT-PCR methods already used for amplifying ORF1 and ORF2 fragments in NoV-positive faecal samples were performed [19–21, 23]. Nucleotide sequences analysis and phylogenetic analysis PCR amplicons were purified by commercial columns (CleanSweep PCR Purification, Thermo Fisher Scientific, Vilnius, Lithuania) and sequenced by Sanger method and all nucleotide sequences were aligned using ClustalW. To identify NoV genotypes, RdRp and VP1 sequences were submitted to the RIVM NoV genotyping tool (http://www.rivm.nl/mpf/norovirus/typingtool). Phylogenetic trees were constructed using MEGA 7 and Neighbour-Joining method (bootstrap 1000 replicates). Sequences from stool and water samples are available at NCBI with the following accession numbers: MG517455 for the major capsid protein obtained from a water sample; MG517456 and MG517461, respectively, for the major capsid protein and RNA-dependent-RNA-polymerase, as representatives of the sequences obtained from the stool samples of those involved in the outbreak. Control measures After outbreak notification, on 21 May, the resort was closed for 2 days. On 25 May, following the first results of the bacteriological analyses on municipal water samples, the use of municipal water for human consumption was temporarily restricted (use of water was permitted only for cleaning purposes) in the Carrabba and Fondachello villages and in the area surrounding the resort. From 25 May the reservoir collecting the water from the Carlino well and its chlorination device were submitted to maintenance work and a new chlorination device was installed on the conduit bringing the water from the Bufardo source. Actually, the Bufardo source had been connected to the municipal network in 2003 but unfortunately a dedicated chlorination station had not been planned at that time. Therefore, unchlorinated water entered the system near Fondachello village on the seaside (Fig. 1), possibly diluting chlorine concentrations in water treated at the Carlino reservoir. Moreover, the exact path of the branch of the water network situated downstream the Bufardo source water connection was not marked in the maps of the municipality (Fig. 1, red line), indicating that this section had never been submitted to maintenance. The area surrounding the unmaintained section of the water pipeline was scattered with unauthorised cottages that were not connected to the sewerage systems. Likely, untreated waste from cesspools was unloaded directly into the soil, generating a risk of contamination of the municipal water system through leakage. No work had been made on the water pipelines before the outbreak but sections of the network had been temporarily closed during the winter season on the seaside (Fondachello). Local health authorities advised the connection of all households in the area to the sewerage system and the replacement of the water system section without maintenance. Health authorities also raised attention on AGE cases notification at all local Emergency Departments and Family Practitioners but there was no further evidence of AGE outbreaks in the months following the intervention. Results Data collected through the online questionnaire compiled 7–10 days after the outbreak revealed that none of the 28 students and the teacher taking part in the school trip showed gastroenteritis symptoms at the beginning of the journey and they had no contact with people who suffered from diarrhoea and/or vomiting in the previous days. Travellers had their first shared meal on 19 May (Day 1), 2016, at lunch on their arrival to the resort in Mascali. The list of foods included in the menu is shown in Table 2. Conversely, the dinner on Day 1 was not shared and the students went to dine in three different fast food restaurants in Acireale. On 20 May (Day 2), 17 students (58.6% of the travellers) presented with vomit and/or diarrhoea, falling ill 26–36 h after lunch on Day 1, strongly suggesting that exposure occurred on the first shared meal. Some students also vomited in the presence of the other travellers on the bus bringing them to Acireale. Five more students were symptomatic in the early hours of the following day (21 May, Day 3, from 0:30 am to 8:00 am), as a continuation of the wave of illnesses of the previous day. The teacher and one more student fell ill at 8:00 pm and 11:00 pm of Day 3, respectively, for a second wave of illnesses occurring >48 h from the first shared meal and <36 h from the first cases, with the last student falling sick on 22 May (Day 4) at 3:00 am (Fig. 2, Table 1). On the whole, the mean incubation period was 33.2 h. Based on the online survey, the severity of illness was evaluated by the Vesikari severity scale, grouping symptomatic travellers into severe (11/25 = 44%) and non-severe (14/25 = 56%) cases (Table 1). Vomiting and fever were the predominant symptoms, affecting 22 out of 25 (88%) symptomatic subjects, while diarrhoea was present in 14 out of 25 (56%). Hospitalisation for rehydration therapy was necessary for 13 students (91% [10/11] of severe and 21% [3/14] of non-severe cases). Emergency room (ER) or family practitioner visits were necessary for 21 students and 13 (62%) of them initiated an antibiotic therapy. Two students indicated that other family members had gastroenteric symptoms more than 24 h after their return home, on Day 3, 4 and 5. All those involved in the outbreak recovered within 4 days, with an average duration of illness of 38.4 h. A case-control study of the risk associated with eating foods at lunch on Day 1 or with attending different restaurants at dinner on Day 1 was attempted. A total of 19 subjects filled the section of the online survey concerning food consumption. Fourteen of the responders were among the 22 cases falling ill within 36 h from the first shared meal (first wave), while the remaining five responders were three travellers with supposed secondary infections (showing symptoms with a delay ⩾48 h; second wave) and two asymptomatic students that were used as controls. No statistically significant associations with any food items served at lunch on Day 1 was identified (P < 0.05; Table 2). However, all food consumed on that occasion was to some extent associated with the risk of disease, the highest OR being obtained with chicken (OR: 20.7143, P: 0.0678; Table 2). Actually, most of the cases ate all foods included in the menu, while one of the two asymptomatic controls did not participate in the meal and the other had a pasta dish only. No statistically significant association was found for attending a particular fast food restaurant at the dinner on Day 1 (OR: 2.5, P: 0.3635). The Real-Time RT-PCR assay allowed to detect GII genogroup NoV genome in all the 25 samples from symptomatic subjects, although two of them showed high Ct values, indicating lower shedding of viral genomes (Table 1). The samples from four asymptomatic students were negative for both NoV genogroups. By sequence analysis, a NoV genotype could be defined for 20/ 25 NoV-positive stool samples. All NoV strains from travellers involved in the outbreaks belonged to the same genotype, GII.P2/GII.2, showing >99% nucleotide identities (nt id) in both fragments (ORF1 and ORF2) analysed. Thirty-two water samples were collected and analysed for bacterial contamination (Table 3). Seven samples were taken from the resort (site R) and the remaining 25 from five other sites (C, F, M, S, W), representative of the entire municipal water network. The majority (10/12) of the water samples collected immediately after the onset of the outbreak, between 21 May and 16 June 2016, at the resort (site R) and at sites F and C, located close to the resort, on the seaside and in the interior, respectively, were positive for coliform bacteria. More than 1000 colony-forming units (CFU)/ 100 ml (up to 2700 CFU/100 ml) were detected in four samples, while their absence is required in 100 ml by the Standard of Water Quality in Italy and seven of 11 samples tested were also Escherichia coli-positive (up to 900 CFU/100 ml). Enterococci were searched in 11 samples and three of them were found positive for this faecal contamination indicator. From 7 July to 16 August 2016, after control measures on the water network were undertaken, 13 water samples were collected for follow-up laboratory examination of the water system at F, R and C sites. All of them were negative for bacterial indicators of faecal contamination. No faecal bacteria were identified in the seven samples collected from M, W and S sites, including the well and the source that fed the water network, at any sampling time, before and after implementing the control measures. The research of NoV genome was conducted on concentrated water samples collected on 1 August from R and F sites (Table 3). NoV genome was detected in a single sample collected from the water entry of the resort involved in the outbreak (site R). Actually, only two out of 26 One-Step RT-PCR repetitions were positive for NoV viral capsid gene. The two ORF2-positive results were obtained at the University of Palermo Laboratory using both 1:10 diluted and undiluted extracted RNA. The sequence analysis of the partial ORF2 portion (289 bp) indicated the presence of the same GII.2 capsid genotype detected in the stool samples of the students involved in the outbreak, showing 98.9% nt id. No ORF1 NoV sequence could be obtained from the NoV-positive water sample. Follow-up analyses for NoV genome, conducted on water samples collected on 20 November both at R and F sites found no positive samples. By sequence alignment, the NoV sequences obtained from Mascali outbreak samples were phylogenetically closely related in both ORF1 and ORF2 (99 and 98.9% of nt id, respectively) to a GII.P2/GII.2 strain responsible for a sporadic paediatric gastroenteritis case occurred in Palermo in January 2016 (PA 36/16, accession numbers MG517464 and MG517460, for ORF1 and ORF2, respectively) and previously detected at the Enteric Viruses Laboratory, University of Palermo. Whereas, three GII.P2/GII.2 Italian strains (PA 280/11, PA 366/11 and PA 3/ 14, accession numbers MG517462, MG517465 and MG517463, for ORF1 and MG517457 to −59, for ORF2, respectively), detected in Palermo in 2011 and 2014, were more distantly associated with the outbreak strain both in RdRp and VP1 (90% to 97.4% nt id and 96.1% to 98.1% nt id, respectively, in ORF1 and ORF2). Discussion Epidemiological features, such as the presence of vomiting in more than half (88%) of symptomatic subjects, the 33.2 h average incubation period and the 38.4 h average duration of illness, suggested a NoV aetiology for the investigated outbreak [13]. The detection of NoV genome in all 25 faecal samples from symptomatic subjects fully complies with the requirement of detection of at least one positive sample out of four tested that is considered to be sufficient to ascertain the aetiology of a NoV outbreak [24]. Moreover, molecular typing indicated that a single viral strain, belonging to genotype GII.P2/GII.2, infected all symptomatic subjects involved in the outbreak. NoVs are highly infectious and can easily spread from person to person but also through contaminated food or water. Therefore, settings such as health care institutions, schools and touristic resorts are mainly affected, since close contact and shared food services facilitate transmission. In Mascali, two waves of transmission were observed over three days, being consistent with food/waterborne followed by person-to-person and/or airborne (vomit) transmission (Fig. 2). The first wave of cases presented on 20 May 2016 after a single shared meal, while the last case was detected on 22 May. Exposure to vomit may have been responsible for some of the secondary cases occurred after the first wave of illness. Vomiting produces aerosols that can be inhaled and swallowed. The extremely low infectious dose of NoV makes the inhalation of vomit aerosols an effective transmission route in settings where there is close contact, like on a bus trip [25]. All food consumed in the only shared lunch on 19 May at the resort was associated with the risk of disease (Table 2), but the association was inconclusive since all cases shared the same dishes included in the lunch menu on Day 1. Although many foods were potentially involved, our attention focused on the tap water. In fact, trivial information from local health authorities reported that local population avoided tap water consumption and relied on bottled water for drinking and cooking. Many small outbreaks of gastroenteritis involving restaurants and family groups had been reported in the same area, especially at the beginning of the summer season, when vacation houses and recreational activities were re-opened after a period of low water demand. Illness caused by NoV is common and if the level of water contamination is low, the number of cases remains limited and water can persist as unrecognised source of infection. If polluted water was served for drinking and used for cooking and washing dishes in the Mascali resort at Day 1 a generalised contamination of food and surfaces could have occurred. The potential role of the foodhandlers as a source of contamination could not be analysed in this study but should also be taken into account. If water contamination has already been present in the area for a long time, infections may have occurred among the local food-handlers. The analysis of water samples taken immediately after the outbreak revealed the presence of bacterial indicators of faecal contamination in the water used in the resort, with the only exception of a sample of filter purified water used both for cooking and as drinking water. When extending the sampling to other areas of the municipal water network, many other sites, mostly on the seaside, showed faecal pollution, which demonstrates a poor protection of the water network from faecal leakage (Fig. 1). NoV contamination of waters generally originates from sewage since these enteric pathogens, as other enteric viruses affecting humans, are mostly species-specific and are shed at high concentrations in the stools of patients. When the analyses on the stools of the students made it clear that the outbreak had been caused by NoV, the viral genome was also searched in appropriate volumes of water samples, but this happened when correction measures on the water network had already been taken. Nevertheless, a single water sample, obtained 2 months after the outbreak (1 August 2016), still contained genome fragments that could be correlated to the same NoV genotype causing the outbreak. Although the finding of NoV genome does not demonstrate that entire infecting viral particles were present in the water, it could indicate that a NoV belonging to the same genotype involved in the outbreak was still circulating in the local population several weeks after the outbreak and it was someway able to penetrate the water network, probably indicating continuing faecal contamination through leakage from sewage systems. NoV is particularly resistant in the environment and also to chlorine, therefore transmission through chlorinated water is not exceptional [9, 12, 26]. Drinking water from municipal supplies fed by lake or groundwater has already been identified in the past as the source of waterborne NoV outbreaks in Italy [27–29] and NoV is now considered one of the pathogens causing the largest number of affected consumers in drinking waterborne outbreaks [12]. NoV can persist in water longer than 15 days [30] and it is resistant to chlorine disinfection when free chlorine levels or pre-treatment processes are inadequate [31, 32]. Moreover, microfiltration filters labelled as certified by NSF Standards 53 or 58 (with a pore size ranging from 0.05 to 5 µm) are not effective in removing viruses and can even decrease chlorine concentration in filtered water [33]. Actually, the owner of the resort involved in our outbreak was encouraged to serve tap water as drinking water at the restaurant relying on the effectiveness of such a filter purifier. Unfortunately, we did not have a tap water sample from the restaurant of the resort taken before correction measures and whose volume was appropriate for virus detection. Therefore, we could not demonstrate the failure to eliminate viral pollution of the microfiltration system. In Mascali, correction measures on the water network consisted in the maintenance of the old chlorination device and in the addition of a new one at a key point located upstream and close to suspected illegal sewage dumping (Fig. 1), but as poor protection of the pipeline from sewage leakage risked to reproduce faecal pollution episodes, renovation of the municipal water system was also advised, as well as systematic control of illegal sewage dumping. The additional chlorination device was apparently the most effective intervention, since right after its activation faecal bacterial pollution indicators disappeared. However, NoV genome was detected in a water sample taken 6 weeks after the last detection of bacterial indicators. Unfortunately, detection of faecal contamination of drinking water is only based on the use of bacterial indicators despite no correlation between levels of enteric bacteria and enteric viruses has been found [34–36]. According to ISGEV enteric virus circulation surveillance data, in 2016 the circulation of GII.P2/GII.2 NoV strains in Palermo, home town of the travellers involved in the outbreak, was the highest ever observed from the starting of local NoV surveillance in 2004. GII.P2/GII.2 was the fourth most common genotype in 2016 (12.2% prevalence), following GII.P4/GII.4 Sydney 2012 (28.4%), GII.P16/GII.2 (18.9%) and GII.Pe/GII.4 Sydney 2012 (17.6%). GII.P2/GII.2 strains were mostly observed starting from July 2016, leaving space for the possibility that their increased circulation was linked to the introduction of the ‘Mascali clone’. However, a couple of GII.P2/GII.2 strains (including PA36/16) genetically close to the ‘Mascali clone’ had been already detected in January 2016 in Palermo and GII.P2/GII.2 NoV had sporadically been observed also in 2011 and 2014. Moreover, the circulation of GII.P2/GII.2 NoVs has been increasing worldwide over the last period [37]. Therefore, the unusually high circulation observed in Palermo might just reflect the global epidemiology of GII.P2/GII.2 strains. Noteworthy, the older GII.P2/GII.2 NoV strains detected in Palermo were closely related in the ORF2 (98.9% nt id) to the GII.P16/GII.2 Nashville strain (KY865307) which is supposed to have generated the recombinant GII.P16/GII.4 clone recently circulating epidemically worldwide [37, 38] and were more distantly associated (95.3% nt id) to the GII.2 capsid sequence (AB662868) of another putative GII.P16/GII.4 ancestor from Japan [39]. The increasing frequency of detection of waterborne NoV outbreaks calls for the introduction of routine monitoring for viruses of drinking waters and for the definition of affordable and reliable indicators for the presence of NoV in drinking waters. The use of ultrafiltration systems as a concentration method in order to recover viruses from environmental waters combined with RT-PCR to detect viral genomes has been demonstrated to grant a recovery rate of >50% of multiple viruses [40] and is at present one of the most sensitive approaches for the detection of NoVs, although the discrimination between viable and nonviable viruses remains a problem. The burden of NoV outbreaks, either food associated or waterborne, is still poorly defined due to the lack of specific diagnostic tools in many clinical laboratories. NoV epidemics should always be suspected when the main symptom is vomiting and the duration of disease is short (<48 h). Waterborne transmission is suggested by the rapid spread of disease and the high attack rate. Care should be taken to perform sensitive diagnostic tests on a sufficient number of stool samples in order to identify NoV as the causative agent of enteritis outbreaks, possibly preferring genome screening over antigen detection. Only the systematic detection of NoV outbreaks can guide in the choice of the safety procedures that could be most appropriate in containing such microbiological hazard.

What source started the event?

1545

Waterborne Norovirus outbreak at a seaside resort likely originating from municipal water distribution system failure

Abstract In May 2016 a Norovirus (NoV) gastroenteritis outbreak involved a high school class visiting a seaside resort near Taormina (Mascali, Sicily). Twenty-four students and a teacher were affected and 17 of them showed symptoms on the second day of the journey, while the others got ill within the following 2 days. Symptoms included vomiting, diarrhoea and fever, and 12 students required hospitalisation. Stool samples tested positive for NoV genome by Real-Time polymerase chain reaction assay in all 25 symptomatic subjects. The GII.P2/GII.2 NoV genotype was linked to the outbreak by ORF1/ORF2 sequence analysis. The epidemiological features of the outbreak were consistent with food/waterborne followed by person-to-person and/ or vomit transmission. Food consumed at a shared lunch on the first day of the trip was associated to illness and drinking un-bottled tap water was also considered as a risk factor. The analysis of water samples revealed the presence of bacterial indicators of faecal contamination in the water used in the resort as well as in other areas of the municipal water network, linking the NoV gastroenteritis outbreak to tap water pollution from sewage leakage. From a single water sample, an amplicon whose sequence corresponded to the capsid genotype recovered from patients could be obtained. Introduction Noroviruses (NoVs) are recognised as a leading cause of sporadic cases and outbreaks of acute gastroenteritis (AGE) across all age groups [1, 2]. They are transmitted primarily through the faecal-oral route either by direct person–person spread and with consumption of contaminated food or water [1, 3, 4]. NoVs are non-enveloped RNA viruses belonging to the Caliciviridae family and are classified into seven distinct genogroups (GI-GVII) on the basis of the major capsid protein VP1 sequence. GI, GII and GIV NoVs infect humans and they have been divided into more than 30 genotypes based on the sequences of polymerase (ORF1) and capsid proteins (ORF2) [5]. NoV GII.4 genotype has been associated with the vast majority of NoV related cases of AGE worldwide although many genotypes, belonging to both GI and GII genogroups are involved in foodborne and waterborne outbreaks [6, 7]. NoVs are highly contagious owing to their high stability in the environment and their ability to infect at low doses. In fact, just 10–100 virions of NoV may be sufficient to cause AGE [8, 9]. NoVs are considered the major cause of water-related disease in semi-closed institutions, including restaurants, schools, hospitals, nursing homes and cruise ships [10–12]. To confirm NoV as a cause of outbreaks, the presence of four epidemiological features of disease, vomiting in more than half of affected persons, a mean incubation period of 24/48 h, a mean duration of illness of 12–60 h and the absence of bacterial pathogens in stool culture, should be evaluated [13]. Waterborne NoV outbreaks can be caused by contaminated drinking water despite correct chlorine concentration and in the absence of coliform bacteria [10, 14–17]. The aim of this study was to investigate an outbreak of NoV gastroenteritis associated with drinking water involving a high school class travelling from Palermo (West Sicily) to Acireale (East Sicily) on an educational tour in May 2016. Stool samples were collected soon after the onset of AGE symptoms to ascertain the aetiology of illness. The source of infection was investigated collecting epidemiological data through an online questionnaire and, in order to explore a possible transmission route, laboratory tests on water from the municipal network were performed. Material and methods Epidemiological investigation From 20 to 22 May 2016 an AGE outbreak occurred in Mascali, a small town near Taormina, involving a school class of 28 students, 13 girls and 15 boys aged between 14 and 18 years and a 54-year-old female teacher, during a trip to participate in a theatre festival in Acireale. A total of 25/29 persons were affected by nausea, vomiting, abdominal cramps, diarrhoea and fever while they were staying at a seaside resort in Mascali, on the way to Acireale. To obtain data on symptoms and the correlated risk factors all participants filled an online questionnaire investigating personal data (age and sex), travel and accommodation details and food consumption over 48 h before symptoms presentation, time of onset and duration of symptoms, therapy and/or hospitalisation required and history of contact with people affected by diarrhoea and/or vomiting (Table 1). The severity of illness was analysed by the Vesikari severity scale allowing to distinguish non-severe (score <11) and severe (score ⩾11) AGE (Table 1). Vesikari score was calculated from maximum number of episodes/day and duration of both diarrhoea and vomiting (score 0–3 for each of the 4 parameters), temperature (score 0–3), level of dehydration (score 0–3) and treatment required (score 0–2) [18]. A case-control study was conducted to establish if food consumed before the onset of symptoms was associated with the disease. A case was defined as a student attending the tour and reporting diarrhoea and/or vomiting within 36 h from the first shared meal during the tour, while students who did not present any of the previous symptoms or those showing the first symptoms with a delay of ⩾48 h (secondary cases) were used as controls. Variables were studied using odds ratio (OR) and its 95% CI and statistical associations were determined using the χ2 test, with the level of statistical significance set at 0.05 (Table 2). NoV detection in stool samples A stool sample was obtained from each of the 29 travellers, soon after (3–5 days) the onset of the outbreak. No stool samples could be obtained from the food-handlers working at the resort. All samples were tested for NoV at the Enteric Viruses Laboratory of the University of Palermo, member of the Italian Study Group for Enteric Viruses (ISGEV), by a Real-Time polymerase chain reaction (RT-PCR) assay that allows to differentiate GI and GII NoVs [19]. Total RNA was extracted from stool samples using QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) and cDNA was generated by MMLV reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were considered positive for NoVs when cycle threshold (ct) of RT-PCR was ⩽30 and lowpositive when 30<ct⩽35. For all NoV-positive samples two genomic fragments were amplified by conventional RT-PCR using Superscript III One-Step RT-PCR system with Platinum Taq DNA polymerase (Invitrogen) and primers JV12a/JV13b, targeting 326 bps of the RdRp region A [20] and either COG1F/ GISKR or COG2F/GIISKR, amplifying a 315 bps 5′portion of GI or GII VP1 protein [19, 21]. Environmental investigation From May through August 2016, a total of 32 tap water samples were collected from the Mascali municipal network. In particular, seven samples were collected at the resort where the outbreak occurred and from its water entry (R), two in the city centre of Mascali (M), 15 in the village of Fondachello (F), on the seaside, three in Carrabba village (C), far from the coast, and two and three, respectively, from the water well ‘Carlino’ (W) and the source ‘Bufardo and Torregrossa’ (S) which feed the municipal system and provide drinking water to the Mascali area (Fig. 1, Table 3). One litre water samples were sent by the local health authorities to the SIALAB laboratory in Avola, Siracusa and 100 ml aliquots were screened for coliforms, E. coli and enterococci as faecal indicators, by standard bacteriological methods according to UNI EN ISO 9308-1:2014. Unfortunately, although the suspect of a water source for the viral outbreak was quickly communicated, only in August and November 2016, following the implementation of control measures, local authorities provided water samplings of appropriate volume to perform viral tests (10 l). Five water samples of 10 l each were collected from different sampling sites in F and R areas for viral screening (Table 3) and were concentrated according to the Italian National Institute of Health (Istituto Superiore di Sanità, ISS) protocols [22] through an ultra-filtration system (Sartoflow® Slice 200 Benchtop Crossflow System, Sartorius AG, Goettingen, Germany), using appropriate membranes (SG Hydrosart 10 kDa) pretreated with 300 ml of 3% Beef Extract (BE) at pH 7, by recirculating it for 10 min in the system, at a pressure not higher than 1.30 bar. Elution and recovery of viruses attached to the membranes were obtained by recirculating 200 ml of 3% BE at pH 9.5 for 3 h until a 10–12 ml volume of eluate was obtained. The pH of the eluate was then checked with litmus paper and if necessary brought to the value of 7 with HCl. Concentrated water samples were simultaneously tested for NoV in three laboratories: Enteric Viruses Laboratory, University of Palermo; Istituto Zooprofilattico Sperimentale della Sicilia (IZS), Palermo; and Istituto Superiore di Sanità (ISS), Rome. Total RNA was extracted from 500 and 560 µl of a concentrated water sample using BioMèrieux NucliSENS® Magnetic Extraction System (BioMèrieux) and QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) respectively, according to manufacturer’s instructions. Bonded RNA was eluted by 100 µl of Elution Buffer. Amplifications were performed at least in duplicate on whole and 1:10 diluted extracts. A Real-Time RT-PCR assay according to ISO/TS 15216- 2:2013, two in-house Real-Time RT-PCR protocols and the One-Step RT-PCR methods already used for amplifying ORF1 and ORF2 fragments in NoV-positive faecal samples were performed [19–21, 23]. Nucleotide sequences analysis and phylogenetic analysis PCR amplicons were purified by commercial columns (CleanSweep PCR Purification, Thermo Fisher Scientific, Vilnius, Lithuania) and sequenced by Sanger method and all nucleotide sequences were aligned using ClustalW. To identify NoV genotypes, RdRp and VP1 sequences were submitted to the RIVM NoV genotyping tool (http://www.rivm.nl/mpf/norovirus/typingtool). Phylogenetic trees were constructed using MEGA 7 and Neighbour-Joining method (bootstrap 1000 replicates). Sequences from stool and water samples are available at NCBI with the following accession numbers: MG517455 for the major capsid protein obtained from a water sample; MG517456 and MG517461, respectively, for the major capsid protein and RNA-dependent-RNA-polymerase, as representatives of the sequences obtained from the stool samples of those involved in the outbreak. Control measures After outbreak notification, on 21 May, the resort was closed for 2 days. On 25 May, following the first results of the bacteriological analyses on municipal water samples, the use of municipal water for human consumption was temporarily restricted (use of water was permitted only for cleaning purposes) in the Carrabba and Fondachello villages and in the area surrounding the resort. From 25 May the reservoir collecting the water from the Carlino well and its chlorination device were submitted to maintenance work and a new chlorination device was installed on the conduit bringing the water from the Bufardo source. Actually, the Bufardo source had been connected to the municipal network in 2003 but unfortunately a dedicated chlorination station had not been planned at that time. Therefore, unchlorinated water entered the system near Fondachello village on the seaside (Fig. 1), possibly diluting chlorine concentrations in water treated at the Carlino reservoir. Moreover, the exact path of the branch of the water network situated downstream the Bufardo source water connection was not marked in the maps of the municipality (Fig. 1, red line), indicating that this section had never been submitted to maintenance. The area surrounding the unmaintained section of the water pipeline was scattered with unauthorised cottages that were not connected to the sewerage systems. Likely, untreated waste from cesspools was unloaded directly into the soil, generating a risk of contamination of the municipal water system through leakage. No work had been made on the water pipelines before the outbreak but sections of the network had been temporarily closed during the winter season on the seaside (Fondachello). Local health authorities advised the connection of all households in the area to the sewerage system and the replacement of the water system section without maintenance. Health authorities also raised attention on AGE cases notification at all local Emergency Departments and Family Practitioners but there was no further evidence of AGE outbreaks in the months following the intervention. Results Data collected through the online questionnaire compiled 7–10 days after the outbreak revealed that none of the 28 students and the teacher taking part in the school trip showed gastroenteritis symptoms at the beginning of the journey and they had no contact with people who suffered from diarrhoea and/or vomiting in the previous days. Travellers had their first shared meal on 19 May (Day 1), 2016, at lunch on their arrival to the resort in Mascali. The list of foods included in the menu is shown in Table 2. Conversely, the dinner on Day 1 was not shared and the students went to dine in three different fast food restaurants in Acireale. On 20 May (Day 2), 17 students (58.6% of the travellers) presented with vomit and/or diarrhoea, falling ill 26–36 h after lunch on Day 1, strongly suggesting that exposure occurred on the first shared meal. Some students also vomited in the presence of the other travellers on the bus bringing them to Acireale. Five more students were symptomatic in the early hours of the following day (21 May, Day 3, from 0:30 am to 8:00 am), as a continuation of the wave of illnesses of the previous day. The teacher and one more student fell ill at 8:00 pm and 11:00 pm of Day 3, respectively, for a second wave of illnesses occurring >48 h from the first shared meal and <36 h from the first cases, with the last student falling sick on 22 May (Day 4) at 3:00 am (Fig. 2, Table 1). On the whole, the mean incubation period was 33.2 h. Based on the online survey, the severity of illness was evaluated by the Vesikari severity scale, grouping symptomatic travellers into severe (11/25 = 44%) and non-severe (14/25 = 56%) cases (Table 1). Vomiting and fever were the predominant symptoms, affecting 22 out of 25 (88%) symptomatic subjects, while diarrhoea was present in 14 out of 25 (56%). Hospitalisation for rehydration therapy was necessary for 13 students (91% [10/11] of severe and 21% [3/14] of non-severe cases). Emergency room (ER) or family practitioner visits were necessary for 21 students and 13 (62%) of them initiated an antibiotic therapy. Two students indicated that other family members had gastroenteric symptoms more than 24 h after their return home, on Day 3, 4 and 5. All those involved in the outbreak recovered within 4 days, with an average duration of illness of 38.4 h. A case-control study of the risk associated with eating foods at lunch on Day 1 or with attending different restaurants at dinner on Day 1 was attempted. A total of 19 subjects filled the section of the online survey concerning food consumption. Fourteen of the responders were among the 22 cases falling ill within 36 h from the first shared meal (first wave), while the remaining five responders were three travellers with supposed secondary infections (showing symptoms with a delay ⩾48 h; second wave) and two asymptomatic students that were used as controls. No statistically significant associations with any food items served at lunch on Day 1 was identified (P < 0.05; Table 2). However, all food consumed on that occasion was to some extent associated with the risk of disease, the highest OR being obtained with chicken (OR: 20.7143, P: 0.0678; Table 2). Actually, most of the cases ate all foods included in the menu, while one of the two asymptomatic controls did not participate in the meal and the other had a pasta dish only. No statistically significant association was found for attending a particular fast food restaurant at the dinner on Day 1 (OR: 2.5, P: 0.3635). The Real-Time RT-PCR assay allowed to detect GII genogroup NoV genome in all the 25 samples from symptomatic subjects, although two of them showed high Ct values, indicating lower shedding of viral genomes (Table 1). The samples from four asymptomatic students were negative for both NoV genogroups. By sequence analysis, a NoV genotype could be defined for 20/ 25 NoV-positive stool samples. All NoV strains from travellers involved in the outbreaks belonged to the same genotype, GII.P2/GII.2, showing >99% nucleotide identities (nt id) in both fragments (ORF1 and ORF2) analysed. Thirty-two water samples were collected and analysed for bacterial contamination (Table 3). Seven samples were taken from the resort (site R) and the remaining 25 from five other sites (C, F, M, S, W), representative of the entire municipal water network. The majority (10/12) of the water samples collected immediately after the onset of the outbreak, between 21 May and 16 June 2016, at the resort (site R) and at sites F and C, located close to the resort, on the seaside and in the interior, respectively, were positive for coliform bacteria. More than 1000 colony-forming units (CFU)/ 100 ml (up to 2700 CFU/100 ml) were detected in four samples, while their absence is required in 100 ml by the Standard of Water Quality in Italy and seven of 11 samples tested were also Escherichia coli-positive (up to 900 CFU/100 ml). Enterococci were searched in 11 samples and three of them were found positive for this faecal contamination indicator. From 7 July to 16 August 2016, after control measures on the water network were undertaken, 13 water samples were collected for follow-up laboratory examination of the water system at F, R and C sites. All of them were negative for bacterial indicators of faecal contamination. No faecal bacteria were identified in the seven samples collected from M, W and S sites, including the well and the source that fed the water network, at any sampling time, before and after implementing the control measures. The research of NoV genome was conducted on concentrated water samples collected on 1 August from R and F sites (Table 3). NoV genome was detected in a single sample collected from the water entry of the resort involved in the outbreak (site R). Actually, only two out of 26 One-Step RT-PCR repetitions were positive for NoV viral capsid gene. The two ORF2-positive results were obtained at the University of Palermo Laboratory using both 1:10 diluted and undiluted extracted RNA. The sequence analysis of the partial ORF2 portion (289 bp) indicated the presence of the same GII.2 capsid genotype detected in the stool samples of the students involved in the outbreak, showing 98.9% nt id. No ORF1 NoV sequence could be obtained from the NoV-positive water sample. Follow-up analyses for NoV genome, conducted on water samples collected on 20 November both at R and F sites found no positive samples. By sequence alignment, the NoV sequences obtained from Mascali outbreak samples were phylogenetically closely related in both ORF1 and ORF2 (99 and 98.9% of nt id, respectively) to a GII.P2/GII.2 strain responsible for a sporadic paediatric gastroenteritis case occurred in Palermo in January 2016 (PA 36/16, accession numbers MG517464 and MG517460, for ORF1 and ORF2, respectively) and previously detected at the Enteric Viruses Laboratory, University of Palermo. Whereas, three GII.P2/GII.2 Italian strains (PA 280/11, PA 366/11 and PA 3/ 14, accession numbers MG517462, MG517465 and MG517463, for ORF1 and MG517457 to −59, for ORF2, respectively), detected in Palermo in 2011 and 2014, were more distantly associated with the outbreak strain both in RdRp and VP1 (90% to 97.4% nt id and 96.1% to 98.1% nt id, respectively, in ORF1 and ORF2). Discussion Epidemiological features, such as the presence of vomiting in more than half (88%) of symptomatic subjects, the 33.2 h average incubation period and the 38.4 h average duration of illness, suggested a NoV aetiology for the investigated outbreak [13]. The detection of NoV genome in all 25 faecal samples from symptomatic subjects fully complies with the requirement of detection of at least one positive sample out of four tested that is considered to be sufficient to ascertain the aetiology of a NoV outbreak [24]. Moreover, molecular typing indicated that a single viral strain, belonging to genotype GII.P2/GII.2, infected all symptomatic subjects involved in the outbreak. NoVs are highly infectious and can easily spread from person to person but also through contaminated food or water. Therefore, settings such as health care institutions, schools and touristic resorts are mainly affected, since close contact and shared food services facilitate transmission. In Mascali, two waves of transmission were observed over three days, being consistent with food/waterborne followed by person-to-person and/or airborne (vomit) transmission (Fig. 2). The first wave of cases presented on 20 May 2016 after a single shared meal, while the last case was detected on 22 May. Exposure to vomit may have been responsible for some of the secondary cases occurred after the first wave of illness. Vomiting produces aerosols that can be inhaled and swallowed. The extremely low infectious dose of NoV makes the inhalation of vomit aerosols an effective transmission route in settings where there is close contact, like on a bus trip [25]. All food consumed in the only shared lunch on 19 May at the resort was associated with the risk of disease (Table 2), but the association was inconclusive since all cases shared the same dishes included in the lunch menu on Day 1. Although many foods were potentially involved, our attention focused on the tap water. In fact, trivial information from local health authorities reported that local population avoided tap water consumption and relied on bottled water for drinking and cooking. Many small outbreaks of gastroenteritis involving restaurants and family groups had been reported in the same area, especially at the beginning of the summer season, when vacation houses and recreational activities were re-opened after a period of low water demand. Illness caused by NoV is common and if the level of water contamination is low, the number of cases remains limited and water can persist as unrecognised source of infection. If polluted water was served for drinking and used for cooking and washing dishes in the Mascali resort at Day 1 a generalised contamination of food and surfaces could have occurred. The potential role of the foodhandlers as a source of contamination could not be analysed in this study but should also be taken into account. If water contamination has already been present in the area for a long time, infections may have occurred among the local food-handlers. The analysis of water samples taken immediately after the outbreak revealed the presence of bacterial indicators of faecal contamination in the water used in the resort, with the only exception of a sample of filter purified water used both for cooking and as drinking water. When extending the sampling to other areas of the municipal water network, many other sites, mostly on the seaside, showed faecal pollution, which demonstrates a poor protection of the water network from faecal leakage (Fig. 1). NoV contamination of waters generally originates from sewage since these enteric pathogens, as other enteric viruses affecting humans, are mostly species-specific and are shed at high concentrations in the stools of patients. When the analyses on the stools of the students made it clear that the outbreak had been caused by NoV, the viral genome was also searched in appropriate volumes of water samples, but this happened when correction measures on the water network had already been taken. Nevertheless, a single water sample, obtained 2 months after the outbreak (1 August 2016), still contained genome fragments that could be correlated to the same NoV genotype causing the outbreak. Although the finding of NoV genome does not demonstrate that entire infecting viral particles were present in the water, it could indicate that a NoV belonging to the same genotype involved in the outbreak was still circulating in the local population several weeks after the outbreak and it was someway able to penetrate the water network, probably indicating continuing faecal contamination through leakage from sewage systems. NoV is particularly resistant in the environment and also to chlorine, therefore transmission through chlorinated water is not exceptional [9, 12, 26]. Drinking water from municipal supplies fed by lake or groundwater has already been identified in the past as the source of waterborne NoV outbreaks in Italy [27–29] and NoV is now considered one of the pathogens causing the largest number of affected consumers in drinking waterborne outbreaks [12]. NoV can persist in water longer than 15 days [30] and it is resistant to chlorine disinfection when free chlorine levels or pre-treatment processes are inadequate [31, 32]. Moreover, microfiltration filters labelled as certified by NSF Standards 53 or 58 (with a pore size ranging from 0.05 to 5 µm) are not effective in removing viruses and can even decrease chlorine concentration in filtered water [33]. Actually, the owner of the resort involved in our outbreak was encouraged to serve tap water as drinking water at the restaurant relying on the effectiveness of such a filter purifier. Unfortunately, we did not have a tap water sample from the restaurant of the resort taken before correction measures and whose volume was appropriate for virus detection. Therefore, we could not demonstrate the failure to eliminate viral pollution of the microfiltration system. In Mascali, correction measures on the water network consisted in the maintenance of the old chlorination device and in the addition of a new one at a key point located upstream and close to suspected illegal sewage dumping (Fig. 1), but as poor protection of the pipeline from sewage leakage risked to reproduce faecal pollution episodes, renovation of the municipal water system was also advised, as well as systematic control of illegal sewage dumping. The additional chlorination device was apparently the most effective intervention, since right after its activation faecal bacterial pollution indicators disappeared. However, NoV genome was detected in a water sample taken 6 weeks after the last detection of bacterial indicators. Unfortunately, detection of faecal contamination of drinking water is only based on the use of bacterial indicators despite no correlation between levels of enteric bacteria and enteric viruses has been found [34–36]. According to ISGEV enteric virus circulation surveillance data, in 2016 the circulation of GII.P2/GII.2 NoV strains in Palermo, home town of the travellers involved in the outbreak, was the highest ever observed from the starting of local NoV surveillance in 2004. GII.P2/GII.2 was the fourth most common genotype in 2016 (12.2% prevalence), following GII.P4/GII.4 Sydney 2012 (28.4%), GII.P16/GII.2 (18.9%) and GII.Pe/GII.4 Sydney 2012 (17.6%). GII.P2/GII.2 strains were mostly observed starting from July 2016, leaving space for the possibility that their increased circulation was linked to the introduction of the ‘Mascali clone’. However, a couple of GII.P2/GII.2 strains (including PA36/16) genetically close to the ‘Mascali clone’ had been already detected in January 2016 in Palermo and GII.P2/GII.2 NoV had sporadically been observed also in 2011 and 2014. Moreover, the circulation of GII.P2/GII.2 NoVs has been increasing worldwide over the last period [37]. Therefore, the unusually high circulation observed in Palermo might just reflect the global epidemiology of GII.P2/GII.2 strains. Noteworthy, the older GII.P2/GII.2 NoV strains detected in Palermo were closely related in the ORF2 (98.9% nt id) to the GII.P16/GII.2 Nashville strain (KY865307) which is supposed to have generated the recombinant GII.P16/GII.4 clone recently circulating epidemically worldwide [37, 38] and were more distantly associated (95.3% nt id) to the GII.2 capsid sequence (AB662868) of another putative GII.P16/GII.4 ancestor from Japan [39]. The increasing frequency of detection of waterborne NoV outbreaks calls for the introduction of routine monitoring for viruses of drinking waters and for the definition of affordable and reliable indicators for the presence of NoV in drinking waters. The use of ultrafiltration systems as a concentration method in order to recover viruses from environmental waters combined with RT-PCR to detect viral genomes has been demonstrated to grant a recovery rate of >50% of multiple viruses [40] and is at present one of the most sensitive approaches for the detection of NoVs, although the discrimination between viable and nonviable viruses remains a problem. The burden of NoV outbreaks, either food associated or waterborne, is still poorly defined due to the lack of specific diagnostic tools in many clinical laboratories. NoV epidemics should always be suspected when the main symptom is vomiting and the duration of disease is short (<48 h). Waterborne transmission is suggested by the rapid spread of disease and the high attack rate. Care should be taken to perform sensitive diagnostic tests on a sufficient number of stool samples in order to identify NoV as the causative agent of enteritis outbreaks, possibly preferring genome screening over antigen detection. Only the systematic detection of NoV outbreaks can guide in the choice of the safety procedures that could be most appropriate in containing such microbiological hazard.

How was the event first detected?

1546

Waterborne Norovirus outbreak at a seaside resort likely originating from municipal water distribution system failure

Abstract In May 2016 a Norovirus (NoV) gastroenteritis outbreak involved a high school class visiting a seaside resort near Taormina (Mascali, Sicily). Twenty-four students and a teacher were affected and 17 of them showed symptoms on the second day of the journey, while the others got ill within the following 2 days. Symptoms included vomiting, diarrhoea and fever, and 12 students required hospitalisation. Stool samples tested positive for NoV genome by Real-Time polymerase chain reaction assay in all 25 symptomatic subjects. The GII.P2/GII.2 NoV genotype was linked to the outbreak by ORF1/ORF2 sequence analysis. The epidemiological features of the outbreak were consistent with food/waterborne followed by person-to-person and/ or vomit transmission. Food consumed at a shared lunch on the first day of the trip was associated to illness and drinking un-bottled tap water was also considered as a risk factor. The analysis of water samples revealed the presence of bacterial indicators of faecal contamination in the water used in the resort as well as in other areas of the municipal water network, linking the NoV gastroenteritis outbreak to tap water pollution from sewage leakage. From a single water sample, an amplicon whose sequence corresponded to the capsid genotype recovered from patients could be obtained. Introduction Noroviruses (NoVs) are recognised as a leading cause of sporadic cases and outbreaks of acute gastroenteritis (AGE) across all age groups [1, 2]. They are transmitted primarily through the faecal-oral route either by direct person–person spread and with consumption of contaminated food or water [1, 3, 4]. NoVs are non-enveloped RNA viruses belonging to the Caliciviridae family and are classified into seven distinct genogroups (GI-GVII) on the basis of the major capsid protein VP1 sequence. GI, GII and GIV NoVs infect humans and they have been divided into more than 30 genotypes based on the sequences of polymerase (ORF1) and capsid proteins (ORF2) [5]. NoV GII.4 genotype has been associated with the vast majority of NoV related cases of AGE worldwide although many genotypes, belonging to both GI and GII genogroups are involved in foodborne and waterborne outbreaks [6, 7]. NoVs are highly contagious owing to their high stability in the environment and their ability to infect at low doses. In fact, just 10–100 virions of NoV may be sufficient to cause AGE [8, 9]. NoVs are considered the major cause of water-related disease in semi-closed institutions, including restaurants, schools, hospitals, nursing homes and cruise ships [10–12]. To confirm NoV as a cause of outbreaks, the presence of four epidemiological features of disease, vomiting in more than half of affected persons, a mean incubation period of 24/48 h, a mean duration of illness of 12–60 h and the absence of bacterial pathogens in stool culture, should be evaluated [13]. Waterborne NoV outbreaks can be caused by contaminated drinking water despite correct chlorine concentration and in the absence of coliform bacteria [10, 14–17]. The aim of this study was to investigate an outbreak of NoV gastroenteritis associated with drinking water involving a high school class travelling from Palermo (West Sicily) to Acireale (East Sicily) on an educational tour in May 2016. Stool samples were collected soon after the onset of AGE symptoms to ascertain the aetiology of illness. The source of infection was investigated collecting epidemiological data through an online questionnaire and, in order to explore a possible transmission route, laboratory tests on water from the municipal network were performed. Material and methods Epidemiological investigation From 20 to 22 May 2016 an AGE outbreak occurred in Mascali, a small town near Taormina, involving a school class of 28 students, 13 girls and 15 boys aged between 14 and 18 years and a 54-year-old female teacher, during a trip to participate in a theatre festival in Acireale. A total of 25/29 persons were affected by nausea, vomiting, abdominal cramps, diarrhoea and fever while they were staying at a seaside resort in Mascali, on the way to Acireale. To obtain data on symptoms and the correlated risk factors all participants filled an online questionnaire investigating personal data (age and sex), travel and accommodation details and food consumption over 48 h before symptoms presentation, time of onset and duration of symptoms, therapy and/or hospitalisation required and history of contact with people affected by diarrhoea and/or vomiting (Table 1). The severity of illness was analysed by the Vesikari severity scale allowing to distinguish non-severe (score <11) and severe (score ⩾11) AGE (Table 1). Vesikari score was calculated from maximum number of episodes/day and duration of both diarrhoea and vomiting (score 0–3 for each of the 4 parameters), temperature (score 0–3), level of dehydration (score 0–3) and treatment required (score 0–2) [18]. A case-control study was conducted to establish if food consumed before the onset of symptoms was associated with the disease. A case was defined as a student attending the tour and reporting diarrhoea and/or vomiting within 36 h from the first shared meal during the tour, while students who did not present any of the previous symptoms or those showing the first symptoms with a delay of ⩾48 h (secondary cases) were used as controls. Variables were studied using odds ratio (OR) and its 95% CI and statistical associations were determined using the χ2 test, with the level of statistical significance set at 0.05 (Table 2). NoV detection in stool samples A stool sample was obtained from each of the 29 travellers, soon after (3–5 days) the onset of the outbreak. No stool samples could be obtained from the food-handlers working at the resort. All samples were tested for NoV at the Enteric Viruses Laboratory of the University of Palermo, member of the Italian Study Group for Enteric Viruses (ISGEV), by a Real-Time polymerase chain reaction (RT-PCR) assay that allows to differentiate GI and GII NoVs [19]. Total RNA was extracted from stool samples using QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) and cDNA was generated by MMLV reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were considered positive for NoVs when cycle threshold (ct) of RT-PCR was ⩽30 and lowpositive when 30<ct⩽35. For all NoV-positive samples two genomic fragments were amplified by conventional RT-PCR using Superscript III One-Step RT-PCR system with Platinum Taq DNA polymerase (Invitrogen) and primers JV12a/JV13b, targeting 326 bps of the RdRp region A [20] and either COG1F/ GISKR or COG2F/GIISKR, amplifying a 315 bps 5′portion of GI or GII VP1 protein [19, 21]. Environmental investigation From May through August 2016, a total of 32 tap water samples were collected from the Mascali municipal network. In particular, seven samples were collected at the resort where the outbreak occurred and from its water entry (R), two in the city centre of Mascali (M), 15 in the village of Fondachello (F), on the seaside, three in Carrabba village (C), far from the coast, and two and three, respectively, from the water well ‘Carlino’ (W) and the source ‘Bufardo and Torregrossa’ (S) which feed the municipal system and provide drinking water to the Mascali area (Fig. 1, Table 3). One litre water samples were sent by the local health authorities to the SIALAB laboratory in Avola, Siracusa and 100 ml aliquots were screened for coliforms, E. coli and enterococci as faecal indicators, by standard bacteriological methods according to UNI EN ISO 9308-1:2014. Unfortunately, although the suspect of a water source for the viral outbreak was quickly communicated, only in August and November 2016, following the implementation of control measures, local authorities provided water samplings of appropriate volume to perform viral tests (10 l). Five water samples of 10 l each were collected from different sampling sites in F and R areas for viral screening (Table 3) and were concentrated according to the Italian National Institute of Health (Istituto Superiore di Sanità, ISS) protocols [22] through an ultra-filtration system (Sartoflow® Slice 200 Benchtop Crossflow System, Sartorius AG, Goettingen, Germany), using appropriate membranes (SG Hydrosart 10 kDa) pretreated with 300 ml of 3% Beef Extract (BE) at pH 7, by recirculating it for 10 min in the system, at a pressure not higher than 1.30 bar. Elution and recovery of viruses attached to the membranes were obtained by recirculating 200 ml of 3% BE at pH 9.5 for 3 h until a 10–12 ml volume of eluate was obtained. The pH of the eluate was then checked with litmus paper and if necessary brought to the value of 7 with HCl. Concentrated water samples were simultaneously tested for NoV in three laboratories: Enteric Viruses Laboratory, University of Palermo; Istituto Zooprofilattico Sperimentale della Sicilia (IZS), Palermo; and Istituto Superiore di Sanità (ISS), Rome. Total RNA was extracted from 500 and 560 µl of a concentrated water sample using BioMèrieux NucliSENS® Magnetic Extraction System (BioMèrieux) and QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) respectively, according to manufacturer’s instructions. Bonded RNA was eluted by 100 µl of Elution Buffer. Amplifications were performed at least in duplicate on whole and 1:10 diluted extracts. A Real-Time RT-PCR assay according to ISO/TS 15216- 2:2013, two in-house Real-Time RT-PCR protocols and the One-Step RT-PCR methods already used for amplifying ORF1 and ORF2 fragments in NoV-positive faecal samples were performed [19–21, 23]. Nucleotide sequences analysis and phylogenetic analysis PCR amplicons were purified by commercial columns (CleanSweep PCR Purification, Thermo Fisher Scientific, Vilnius, Lithuania) and sequenced by Sanger method and all nucleotide sequences were aligned using ClustalW. To identify NoV genotypes, RdRp and VP1 sequences were submitted to the RIVM NoV genotyping tool (http://www.rivm.nl/mpf/norovirus/typingtool). Phylogenetic trees were constructed using MEGA 7 and Neighbour-Joining method (bootstrap 1000 replicates). Sequences from stool and water samples are available at NCBI with the following accession numbers: MG517455 for the major capsid protein obtained from a water sample; MG517456 and MG517461, respectively, for the major capsid protein and RNA-dependent-RNA-polymerase, as representatives of the sequences obtained from the stool samples of those involved in the outbreak. Control measures After outbreak notification, on 21 May, the resort was closed for 2 days. On 25 May, following the first results of the bacteriological analyses on municipal water samples, the use of municipal water for human consumption was temporarily restricted (use of water was permitted only for cleaning purposes) in the Carrabba and Fondachello villages and in the area surrounding the resort. From 25 May the reservoir collecting the water from the Carlino well and its chlorination device were submitted to maintenance work and a new chlorination device was installed on the conduit bringing the water from the Bufardo source. Actually, the Bufardo source had been connected to the municipal network in 2003 but unfortunately a dedicated chlorination station had not been planned at that time. Therefore, unchlorinated water entered the system near Fondachello village on the seaside (Fig. 1), possibly diluting chlorine concentrations in water treated at the Carlino reservoir. Moreover, the exact path of the branch of the water network situated downstream the Bufardo source water connection was not marked in the maps of the municipality (Fig. 1, red line), indicating that this section had never been submitted to maintenance. The area surrounding the unmaintained section of the water pipeline was scattered with unauthorised cottages that were not connected to the sewerage systems. Likely, untreated waste from cesspools was unloaded directly into the soil, generating a risk of contamination of the municipal water system through leakage. No work had been made on the water pipelines before the outbreak but sections of the network had been temporarily closed during the winter season on the seaside (Fondachello). Local health authorities advised the connection of all households in the area to the sewerage system and the replacement of the water system section without maintenance. Health authorities also raised attention on AGE cases notification at all local Emergency Departments and Family Practitioners but there was no further evidence of AGE outbreaks in the months following the intervention. Results Data collected through the online questionnaire compiled 7–10 days after the outbreak revealed that none of the 28 students and the teacher taking part in the school trip showed gastroenteritis symptoms at the beginning of the journey and they had no contact with people who suffered from diarrhoea and/or vomiting in the previous days. Travellers had their first shared meal on 19 May (Day 1), 2016, at lunch on their arrival to the resort in Mascali. The list of foods included in the menu is shown in Table 2. Conversely, the dinner on Day 1 was not shared and the students went to dine in three different fast food restaurants in Acireale. On 20 May (Day 2), 17 students (58.6% of the travellers) presented with vomit and/or diarrhoea, falling ill 26–36 h after lunch on Day 1, strongly suggesting that exposure occurred on the first shared meal. Some students also vomited in the presence of the other travellers on the bus bringing them to Acireale. Five more students were symptomatic in the early hours of the following day (21 May, Day 3, from 0:30 am to 8:00 am), as a continuation of the wave of illnesses of the previous day. The teacher and one more student fell ill at 8:00 pm and 11:00 pm of Day 3, respectively, for a second wave of illnesses occurring >48 h from the first shared meal and <36 h from the first cases, with the last student falling sick on 22 May (Day 4) at 3:00 am (Fig. 2, Table 1). On the whole, the mean incubation period was 33.2 h. Based on the online survey, the severity of illness was evaluated by the Vesikari severity scale, grouping symptomatic travellers into severe (11/25 = 44%) and non-severe (14/25 = 56%) cases (Table 1). Vomiting and fever were the predominant symptoms, affecting 22 out of 25 (88%) symptomatic subjects, while diarrhoea was present in 14 out of 25 (56%). Hospitalisation for rehydration therapy was necessary for 13 students (91% [10/11] of severe and 21% [3/14] of non-severe cases). Emergency room (ER) or family practitioner visits were necessary for 21 students and 13 (62%) of them initiated an antibiotic therapy. Two students indicated that other family members had gastroenteric symptoms more than 24 h after their return home, on Day 3, 4 and 5. All those involved in the outbreak recovered within 4 days, with an average duration of illness of 38.4 h. A case-control study of the risk associated with eating foods at lunch on Day 1 or with attending different restaurants at dinner on Day 1 was attempted. A total of 19 subjects filled the section of the online survey concerning food consumption. Fourteen of the responders were among the 22 cases falling ill within 36 h from the first shared meal (first wave), while the remaining five responders were three travellers with supposed secondary infections (showing symptoms with a delay ⩾48 h; second wave) and two asymptomatic students that were used as controls. No statistically significant associations with any food items served at lunch on Day 1 was identified (P < 0.05; Table 2). However, all food consumed on that occasion was to some extent associated with the risk of disease, the highest OR being obtained with chicken (OR: 20.7143, P: 0.0678; Table 2). Actually, most of the cases ate all foods included in the menu, while one of the two asymptomatic controls did not participate in the meal and the other had a pasta dish only. No statistically significant association was found for attending a particular fast food restaurant at the dinner on Day 1 (OR: 2.5, P: 0.3635). The Real-Time RT-PCR assay allowed to detect GII genogroup NoV genome in all the 25 samples from symptomatic subjects, although two of them showed high Ct values, indicating lower shedding of viral genomes (Table 1). The samples from four asymptomatic students were negative for both NoV genogroups. By sequence analysis, a NoV genotype could be defined for 20/ 25 NoV-positive stool samples. All NoV strains from travellers involved in the outbreaks belonged to the same genotype, GII.P2/GII.2, showing >99% nucleotide identities (nt id) in both fragments (ORF1 and ORF2) analysed. Thirty-two water samples were collected and analysed for bacterial contamination (Table 3). Seven samples were taken from the resort (site R) and the remaining 25 from five other sites (C, F, M, S, W), representative of the entire municipal water network. The majority (10/12) of the water samples collected immediately after the onset of the outbreak, between 21 May and 16 June 2016, at the resort (site R) and at sites F and C, located close to the resort, on the seaside and in the interior, respectively, were positive for coliform bacteria. More than 1000 colony-forming units (CFU)/ 100 ml (up to 2700 CFU/100 ml) were detected in four samples, while their absence is required in 100 ml by the Standard of Water Quality in Italy and seven of 11 samples tested were also Escherichia coli-positive (up to 900 CFU/100 ml). Enterococci were searched in 11 samples and three of them were found positive for this faecal contamination indicator. From 7 July to 16 August 2016, after control measures on the water network were undertaken, 13 water samples were collected for follow-up laboratory examination of the water system at F, R and C sites. All of them were negative for bacterial indicators of faecal contamination. No faecal bacteria were identified in the seven samples collected from M, W and S sites, including the well and the source that fed the water network, at any sampling time, before and after implementing the control measures. The research of NoV genome was conducted on concentrated water samples collected on 1 August from R and F sites (Table 3). NoV genome was detected in a single sample collected from the water entry of the resort involved in the outbreak (site R). Actually, only two out of 26 One-Step RT-PCR repetitions were positive for NoV viral capsid gene. The two ORF2-positive results were obtained at the University of Palermo Laboratory using both 1:10 diluted and undiluted extracted RNA. The sequence analysis of the partial ORF2 portion (289 bp) indicated the presence of the same GII.2 capsid genotype detected in the stool samples of the students involved in the outbreak, showing 98.9% nt id. No ORF1 NoV sequence could be obtained from the NoV-positive water sample. Follow-up analyses for NoV genome, conducted on water samples collected on 20 November both at R and F sites found no positive samples. By sequence alignment, the NoV sequences obtained from Mascali outbreak samples were phylogenetically closely related in both ORF1 and ORF2 (99 and 98.9% of nt id, respectively) to a GII.P2/GII.2 strain responsible for a sporadic paediatric gastroenteritis case occurred in Palermo in January 2016 (PA 36/16, accession numbers MG517464 and MG517460, for ORF1 and ORF2, respectively) and previously detected at the Enteric Viruses Laboratory, University of Palermo. Whereas, three GII.P2/GII.2 Italian strains (PA 280/11, PA 366/11 and PA 3/ 14, accession numbers MG517462, MG517465 and MG517463, for ORF1 and MG517457 to −59, for ORF2, respectively), detected in Palermo in 2011 and 2014, were more distantly associated with the outbreak strain both in RdRp and VP1 (90% to 97.4% nt id and 96.1% to 98.1% nt id, respectively, in ORF1 and ORF2). Discussion Epidemiological features, such as the presence of vomiting in more than half (88%) of symptomatic subjects, the 33.2 h average incubation period and the 38.4 h average duration of illness, suggested a NoV aetiology for the investigated outbreak [13]. The detection of NoV genome in all 25 faecal samples from symptomatic subjects fully complies with the requirement of detection of at least one positive sample out of four tested that is considered to be sufficient to ascertain the aetiology of a NoV outbreak [24]. Moreover, molecular typing indicated that a single viral strain, belonging to genotype GII.P2/GII.2, infected all symptomatic subjects involved in the outbreak. NoVs are highly infectious and can easily spread from person to person but also through contaminated food or water. Therefore, settings such as health care institutions, schools and touristic resorts are mainly affected, since close contact and shared food services facilitate transmission. In Mascali, two waves of transmission were observed over three days, being consistent with food/waterborne followed by person-to-person and/or airborne (vomit) transmission (Fig. 2). The first wave of cases presented on 20 May 2016 after a single shared meal, while the last case was detected on 22 May. Exposure to vomit may have been responsible for some of the secondary cases occurred after the first wave of illness. Vomiting produces aerosols that can be inhaled and swallowed. The extremely low infectious dose of NoV makes the inhalation of vomit aerosols an effective transmission route in settings where there is close contact, like on a bus trip [25]. All food consumed in the only shared lunch on 19 May at the resort was associated with the risk of disease (Table 2), but the association was inconclusive since all cases shared the same dishes included in the lunch menu on Day 1. Although many foods were potentially involved, our attention focused on the tap water. In fact, trivial information from local health authorities reported that local population avoided tap water consumption and relied on bottled water for drinking and cooking. Many small outbreaks of gastroenteritis involving restaurants and family groups had been reported in the same area, especially at the beginning of the summer season, when vacation houses and recreational activities were re-opened after a period of low water demand. Illness caused by NoV is common and if the level of water contamination is low, the number of cases remains limited and water can persist as unrecognised source of infection. If polluted water was served for drinking and used for cooking and washing dishes in the Mascali resort at Day 1 a generalised contamination of food and surfaces could have occurred. The potential role of the foodhandlers as a source of contamination could not be analysed in this study but should also be taken into account. If water contamination has already been present in the area for a long time, infections may have occurred among the local food-handlers. The analysis of water samples taken immediately after the outbreak revealed the presence of bacterial indicators of faecal contamination in the water used in the resort, with the only exception of a sample of filter purified water used both for cooking and as drinking water. When extending the sampling to other areas of the municipal water network, many other sites, mostly on the seaside, showed faecal pollution, which demonstrates a poor protection of the water network from faecal leakage (Fig. 1). NoV contamination of waters generally originates from sewage since these enteric pathogens, as other enteric viruses affecting humans, are mostly species-specific and are shed at high concentrations in the stools of patients. When the analyses on the stools of the students made it clear that the outbreak had been caused by NoV, the viral genome was also searched in appropriate volumes of water samples, but this happened when correction measures on the water network had already been taken. Nevertheless, a single water sample, obtained 2 months after the outbreak (1 August 2016), still contained genome fragments that could be correlated to the same NoV genotype causing the outbreak. Although the finding of NoV genome does not demonstrate that entire infecting viral particles were present in the water, it could indicate that a NoV belonging to the same genotype involved in the outbreak was still circulating in the local population several weeks after the outbreak and it was someway able to penetrate the water network, probably indicating continuing faecal contamination through leakage from sewage systems. NoV is particularly resistant in the environment and also to chlorine, therefore transmission through chlorinated water is not exceptional [9, 12, 26]. Drinking water from municipal supplies fed by lake or groundwater has already been identified in the past as the source of waterborne NoV outbreaks in Italy [27–29] and NoV is now considered one of the pathogens causing the largest number of affected consumers in drinking waterborne outbreaks [12]. NoV can persist in water longer than 15 days [30] and it is resistant to chlorine disinfection when free chlorine levels or pre-treatment processes are inadequate [31, 32]. Moreover, microfiltration filters labelled as certified by NSF Standards 53 or 58 (with a pore size ranging from 0.05 to 5 µm) are not effective in removing viruses and can even decrease chlorine concentration in filtered water [33]. Actually, the owner of the resort involved in our outbreak was encouraged to serve tap water as drinking water at the restaurant relying on the effectiveness of such a filter purifier. Unfortunately, we did not have a tap water sample from the restaurant of the resort taken before correction measures and whose volume was appropriate for virus detection. Therefore, we could not demonstrate the failure to eliminate viral pollution of the microfiltration system. In Mascali, correction measures on the water network consisted in the maintenance of the old chlorination device and in the addition of a new one at a key point located upstream and close to suspected illegal sewage dumping (Fig. 1), but as poor protection of the pipeline from sewage leakage risked to reproduce faecal pollution episodes, renovation of the municipal water system was also advised, as well as systematic control of illegal sewage dumping. The additional chlorination device was apparently the most effective intervention, since right after its activation faecal bacterial pollution indicators disappeared. However, NoV genome was detected in a water sample taken 6 weeks after the last detection of bacterial indicators. Unfortunately, detection of faecal contamination of drinking water is only based on the use of bacterial indicators despite no correlation between levels of enteric bacteria and enteric viruses has been found [34–36]. According to ISGEV enteric virus circulation surveillance data, in 2016 the circulation of GII.P2/GII.2 NoV strains in Palermo, home town of the travellers involved in the outbreak, was the highest ever observed from the starting of local NoV surveillance in 2004. GII.P2/GII.2 was the fourth most common genotype in 2016 (12.2% prevalence), following GII.P4/GII.4 Sydney 2012 (28.4%), GII.P16/GII.2 (18.9%) and GII.Pe/GII.4 Sydney 2012 (17.6%). GII.P2/GII.2 strains were mostly observed starting from July 2016, leaving space for the possibility that their increased circulation was linked to the introduction of the ‘Mascali clone’. However, a couple of GII.P2/GII.2 strains (including PA36/16) genetically close to the ‘Mascali clone’ had been already detected in January 2016 in Palermo and GII.P2/GII.2 NoV had sporadically been observed also in 2011 and 2014. Moreover, the circulation of GII.P2/GII.2 NoVs has been increasing worldwide over the last period [37]. Therefore, the unusually high circulation observed in Palermo might just reflect the global epidemiology of GII.P2/GII.2 strains. Noteworthy, the older GII.P2/GII.2 NoV strains detected in Palermo were closely related in the ORF2 (98.9% nt id) to the GII.P16/GII.2 Nashville strain (KY865307) which is supposed to have generated the recombinant GII.P16/GII.4 clone recently circulating epidemically worldwide [37, 38] and were more distantly associated (95.3% nt id) to the GII.2 capsid sequence (AB662868) of another putative GII.P16/GII.4 ancestor from Japan [39]. The increasing frequency of detection of waterborne NoV outbreaks calls for the introduction of routine monitoring for viruses of drinking waters and for the definition of affordable and reliable indicators for the presence of NoV in drinking waters. The use of ultrafiltration systems as a concentration method in order to recover viruses from environmental waters combined with RT-PCR to detect viral genomes has been demonstrated to grant a recovery rate of >50% of multiple viruses [40] and is at present one of the most sensitive approaches for the detection of NoVs, although the discrimination between viable and nonviable viruses remains a problem. The burden of NoV outbreaks, either food associated or waterborne, is still poorly defined due to the lack of specific diagnostic tools in many clinical laboratories. NoV epidemics should always be suspected when the main symptom is vomiting and the duration of disease is short (<48 h). Waterborne transmission is suggested by the rapid spread of disease and the high attack rate. Care should be taken to perform sensitive diagnostic tests on a sufficient number of stool samples in order to identify NoV as the causative agent of enteritis outbreaks, possibly preferring genome screening over antigen detection. Only the systematic detection of NoV outbreaks can guide in the choice of the safety procedures that could be most appropriate in containing such microbiological hazard.

How many people were ill?

1547

Waterborne Norovirus outbreak at a seaside resort likely originating from municipal water distribution system failure

Abstract In May 2016 a Norovirus (NoV) gastroenteritis outbreak involved a high school class visiting a seaside resort near Taormina (Mascali, Sicily). Twenty-four students and a teacher were affected and 17 of them showed symptoms on the second day of the journey, while the others got ill within the following 2 days. Symptoms included vomiting, diarrhoea and fever, and 12 students required hospitalisation. Stool samples tested positive for NoV genome by Real-Time polymerase chain reaction assay in all 25 symptomatic subjects. The GII.P2/GII.2 NoV genotype was linked to the outbreak by ORF1/ORF2 sequence analysis. The epidemiological features of the outbreak were consistent with food/waterborne followed by person-to-person and/ or vomit transmission. Food consumed at a shared lunch on the first day of the trip was associated to illness and drinking un-bottled tap water was also considered as a risk factor. The analysis of water samples revealed the presence of bacterial indicators of faecal contamination in the water used in the resort as well as in other areas of the municipal water network, linking the NoV gastroenteritis outbreak to tap water pollution from sewage leakage. From a single water sample, an amplicon whose sequence corresponded to the capsid genotype recovered from patients could be obtained. Introduction Noroviruses (NoVs) are recognised as a leading cause of sporadic cases and outbreaks of acute gastroenteritis (AGE) across all age groups [1, 2]. They are transmitted primarily through the faecal-oral route either by direct person–person spread and with consumption of contaminated food or water [1, 3, 4]. NoVs are non-enveloped RNA viruses belonging to the Caliciviridae family and are classified into seven distinct genogroups (GI-GVII) on the basis of the major capsid protein VP1 sequence. GI, GII and GIV NoVs infect humans and they have been divided into more than 30 genotypes based on the sequences of polymerase (ORF1) and capsid proteins (ORF2) [5]. NoV GII.4 genotype has been associated with the vast majority of NoV related cases of AGE worldwide although many genotypes, belonging to both GI and GII genogroups are involved in foodborne and waterborne outbreaks [6, 7]. NoVs are highly contagious owing to their high stability in the environment and their ability to infect at low doses. In fact, just 10–100 virions of NoV may be sufficient to cause AGE [8, 9]. NoVs are considered the major cause of water-related disease in semi-closed institutions, including restaurants, schools, hospitals, nursing homes and cruise ships [10–12]. To confirm NoV as a cause of outbreaks, the presence of four epidemiological features of disease, vomiting in more than half of affected persons, a mean incubation period of 24/48 h, a mean duration of illness of 12–60 h and the absence of bacterial pathogens in stool culture, should be evaluated [13]. Waterborne NoV outbreaks can be caused by contaminated drinking water despite correct chlorine concentration and in the absence of coliform bacteria [10, 14–17]. The aim of this study was to investigate an outbreak of NoV gastroenteritis associated with drinking water involving a high school class travelling from Palermo (West Sicily) to Acireale (East Sicily) on an educational tour in May 2016. Stool samples were collected soon after the onset of AGE symptoms to ascertain the aetiology of illness. The source of infection was investigated collecting epidemiological data through an online questionnaire and, in order to explore a possible transmission route, laboratory tests on water from the municipal network were performed. Material and methods Epidemiological investigation From 20 to 22 May 2016 an AGE outbreak occurred in Mascali, a small town near Taormina, involving a school class of 28 students, 13 girls and 15 boys aged between 14 and 18 years and a 54-year-old female teacher, during a trip to participate in a theatre festival in Acireale. A total of 25/29 persons were affected by nausea, vomiting, abdominal cramps, diarrhoea and fever while they were staying at a seaside resort in Mascali, on the way to Acireale. To obtain data on symptoms and the correlated risk factors all participants filled an online questionnaire investigating personal data (age and sex), travel and accommodation details and food consumption over 48 h before symptoms presentation, time of onset and duration of symptoms, therapy and/or hospitalisation required and history of contact with people affected by diarrhoea and/or vomiting (Table 1). The severity of illness was analysed by the Vesikari severity scale allowing to distinguish non-severe (score <11) and severe (score ⩾11) AGE (Table 1). Vesikari score was calculated from maximum number of episodes/day and duration of both diarrhoea and vomiting (score 0–3 for each of the 4 parameters), temperature (score 0–3), level of dehydration (score 0–3) and treatment required (score 0–2) [18]. A case-control study was conducted to establish if food consumed before the onset of symptoms was associated with the disease. A case was defined as a student attending the tour and reporting diarrhoea and/or vomiting within 36 h from the first shared meal during the tour, while students who did not present any of the previous symptoms or those showing the first symptoms with a delay of ⩾48 h (secondary cases) were used as controls. Variables were studied using odds ratio (OR) and its 95% CI and statistical associations were determined using the χ2 test, with the level of statistical significance set at 0.05 (Table 2). NoV detection in stool samples A stool sample was obtained from each of the 29 travellers, soon after (3–5 days) the onset of the outbreak. No stool samples could be obtained from the food-handlers working at the resort. All samples were tested for NoV at the Enteric Viruses Laboratory of the University of Palermo, member of the Italian Study Group for Enteric Viruses (ISGEV), by a Real-Time polymerase chain reaction (RT-PCR) assay that allows to differentiate GI and GII NoVs [19]. Total RNA was extracted from stool samples using QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) and cDNA was generated by MMLV reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were considered positive for NoVs when cycle threshold (ct) of RT-PCR was ⩽30 and lowpositive when 30<ct⩽35. For all NoV-positive samples two genomic fragments were amplified by conventional RT-PCR using Superscript III One-Step RT-PCR system with Platinum Taq DNA polymerase (Invitrogen) and primers JV12a/JV13b, targeting 326 bps of the RdRp region A [20] and either COG1F/ GISKR or COG2F/GIISKR, amplifying a 315 bps 5′portion of GI or GII VP1 protein [19, 21]. Environmental investigation From May through August 2016, a total of 32 tap water samples were collected from the Mascali municipal network. In particular, seven samples were collected at the resort where the outbreak occurred and from its water entry (R), two in the city centre of Mascali (M), 15 in the village of Fondachello (F), on the seaside, three in Carrabba village (C), far from the coast, and two and three, respectively, from the water well ‘Carlino’ (W) and the source ‘Bufardo and Torregrossa’ (S) which feed the municipal system and provide drinking water to the Mascali area (Fig. 1, Table 3). One litre water samples were sent by the local health authorities to the SIALAB laboratory in Avola, Siracusa and 100 ml aliquots were screened for coliforms, E. coli and enterococci as faecal indicators, by standard bacteriological methods according to UNI EN ISO 9308-1:2014. Unfortunately, although the suspect of a water source for the viral outbreak was quickly communicated, only in August and November 2016, following the implementation of control measures, local authorities provided water samplings of appropriate volume to perform viral tests (10 l). Five water samples of 10 l each were collected from different sampling sites in F and R areas for viral screening (Table 3) and were concentrated according to the Italian National Institute of Health (Istituto Superiore di Sanità, ISS) protocols [22] through an ultra-filtration system (Sartoflow® Slice 200 Benchtop Crossflow System, Sartorius AG, Goettingen, Germany), using appropriate membranes (SG Hydrosart 10 kDa) pretreated with 300 ml of 3% Beef Extract (BE) at pH 7, by recirculating it for 10 min in the system, at a pressure not higher than 1.30 bar. Elution and recovery of viruses attached to the membranes were obtained by recirculating 200 ml of 3% BE at pH 9.5 for 3 h until a 10–12 ml volume of eluate was obtained. The pH of the eluate was then checked with litmus paper and if necessary brought to the value of 7 with HCl. Concentrated water samples were simultaneously tested for NoV in three laboratories: Enteric Viruses Laboratory, University of Palermo; Istituto Zooprofilattico Sperimentale della Sicilia (IZS), Palermo; and Istituto Superiore di Sanità (ISS), Rome. Total RNA was extracted from 500 and 560 µl of a concentrated water sample using BioMèrieux NucliSENS® Magnetic Extraction System (BioMèrieux) and QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) respectively, according to manufacturer’s instructions. Bonded RNA was eluted by 100 µl of Elution Buffer. Amplifications were performed at least in duplicate on whole and 1:10 diluted extracts. A Real-Time RT-PCR assay according to ISO/TS 15216- 2:2013, two in-house Real-Time RT-PCR protocols and the One-Step RT-PCR methods already used for amplifying ORF1 and ORF2 fragments in NoV-positive faecal samples were performed [19–21, 23]. Nucleotide sequences analysis and phylogenetic analysis PCR amplicons were purified by commercial columns (CleanSweep PCR Purification, Thermo Fisher Scientific, Vilnius, Lithuania) and sequenced by Sanger method and all nucleotide sequences were aligned using ClustalW. To identify NoV genotypes, RdRp and VP1 sequences were submitted to the RIVM NoV genotyping tool (http://www.rivm.nl/mpf/norovirus/typingtool). Phylogenetic trees were constructed using MEGA 7 and Neighbour-Joining method (bootstrap 1000 replicates). Sequences from stool and water samples are available at NCBI with the following accession numbers: MG517455 for the major capsid protein obtained from a water sample; MG517456 and MG517461, respectively, for the major capsid protein and RNA-dependent-RNA-polymerase, as representatives of the sequences obtained from the stool samples of those involved in the outbreak. Control measures After outbreak notification, on 21 May, the resort was closed for 2 days. On 25 May, following the first results of the bacteriological analyses on municipal water samples, the use of municipal water for human consumption was temporarily restricted (use of water was permitted only for cleaning purposes) in the Carrabba and Fondachello villages and in the area surrounding the resort. From 25 May the reservoir collecting the water from the Carlino well and its chlorination device were submitted to maintenance work and a new chlorination device was installed on the conduit bringing the water from the Bufardo source. Actually, the Bufardo source had been connected to the municipal network in 2003 but unfortunately a dedicated chlorination station had not been planned at that time. Therefore, unchlorinated water entered the system near Fondachello village on the seaside (Fig. 1), possibly diluting chlorine concentrations in water treated at the Carlino reservoir. Moreover, the exact path of the branch of the water network situated downstream the Bufardo source water connection was not marked in the maps of the municipality (Fig. 1, red line), indicating that this section had never been submitted to maintenance. The area surrounding the unmaintained section of the water pipeline was scattered with unauthorised cottages that were not connected to the sewerage systems. Likely, untreated waste from cesspools was unloaded directly into the soil, generating a risk of contamination of the municipal water system through leakage. No work had been made on the water pipelines before the outbreak but sections of the network had been temporarily closed during the winter season on the seaside (Fondachello). Local health authorities advised the connection of all households in the area to the sewerage system and the replacement of the water system section without maintenance. Health authorities also raised attention on AGE cases notification at all local Emergency Departments and Family Practitioners but there was no further evidence of AGE outbreaks in the months following the intervention. Results Data collected through the online questionnaire compiled 7–10 days after the outbreak revealed that none of the 28 students and the teacher taking part in the school trip showed gastroenteritis symptoms at the beginning of the journey and they had no contact with people who suffered from diarrhoea and/or vomiting in the previous days. Travellers had their first shared meal on 19 May (Day 1), 2016, at lunch on their arrival to the resort in Mascali. The list of foods included in the menu is shown in Table 2. Conversely, the dinner on Day 1 was not shared and the students went to dine in three different fast food restaurants in Acireale. On 20 May (Day 2), 17 students (58.6% of the travellers) presented with vomit and/or diarrhoea, falling ill 26–36 h after lunch on Day 1, strongly suggesting that exposure occurred on the first shared meal. Some students also vomited in the presence of the other travellers on the bus bringing them to Acireale. Five more students were symptomatic in the early hours of the following day (21 May, Day 3, from 0:30 am to 8:00 am), as a continuation of the wave of illnesses of the previous day. The teacher and one more student fell ill at 8:00 pm and 11:00 pm of Day 3, respectively, for a second wave of illnesses occurring >48 h from the first shared meal and <36 h from the first cases, with the last student falling sick on 22 May (Day 4) at 3:00 am (Fig. 2, Table 1). On the whole, the mean incubation period was 33.2 h. Based on the online survey, the severity of illness was evaluated by the Vesikari severity scale, grouping symptomatic travellers into severe (11/25 = 44%) and non-severe (14/25 = 56%) cases (Table 1). Vomiting and fever were the predominant symptoms, affecting 22 out of 25 (88%) symptomatic subjects, while diarrhoea was present in 14 out of 25 (56%). Hospitalisation for rehydration therapy was necessary for 13 students (91% [10/11] of severe and 21% [3/14] of non-severe cases). Emergency room (ER) or family practitioner visits were necessary for 21 students and 13 (62%) of them initiated an antibiotic therapy. Two students indicated that other family members had gastroenteric symptoms more than 24 h after their return home, on Day 3, 4 and 5. All those involved in the outbreak recovered within 4 days, with an average duration of illness of 38.4 h. A case-control study of the risk associated with eating foods at lunch on Day 1 or with attending different restaurants at dinner on Day 1 was attempted. A total of 19 subjects filled the section of the online survey concerning food consumption. Fourteen of the responders were among the 22 cases falling ill within 36 h from the first shared meal (first wave), while the remaining five responders were three travellers with supposed secondary infections (showing symptoms with a delay ⩾48 h; second wave) and two asymptomatic students that were used as controls. No statistically significant associations with any food items served at lunch on Day 1 was identified (P < 0.05; Table 2). However, all food consumed on that occasion was to some extent associated with the risk of disease, the highest OR being obtained with chicken (OR: 20.7143, P: 0.0678; Table 2). Actually, most of the cases ate all foods included in the menu, while one of the two asymptomatic controls did not participate in the meal and the other had a pasta dish only. No statistically significant association was found for attending a particular fast food restaurant at the dinner on Day 1 (OR: 2.5, P: 0.3635). The Real-Time RT-PCR assay allowed to detect GII genogroup NoV genome in all the 25 samples from symptomatic subjects, although two of them showed high Ct values, indicating lower shedding of viral genomes (Table 1). The samples from four asymptomatic students were negative for both NoV genogroups. By sequence analysis, a NoV genotype could be defined for 20/ 25 NoV-positive stool samples. All NoV strains from travellers involved in the outbreaks belonged to the same genotype, GII.P2/GII.2, showing >99% nucleotide identities (nt id) in both fragments (ORF1 and ORF2) analysed. Thirty-two water samples were collected and analysed for bacterial contamination (Table 3). Seven samples were taken from the resort (site R) and the remaining 25 from five other sites (C, F, M, S, W), representative of the entire municipal water network. The majority (10/12) of the water samples collected immediately after the onset of the outbreak, between 21 May and 16 June 2016, at the resort (site R) and at sites F and C, located close to the resort, on the seaside and in the interior, respectively, were positive for coliform bacteria. More than 1000 colony-forming units (CFU)/ 100 ml (up to 2700 CFU/100 ml) were detected in four samples, while their absence is required in 100 ml by the Standard of Water Quality in Italy and seven of 11 samples tested were also Escherichia coli-positive (up to 900 CFU/100 ml). Enterococci were searched in 11 samples and three of them were found positive for this faecal contamination indicator. From 7 July to 16 August 2016, after control measures on the water network were undertaken, 13 water samples were collected for follow-up laboratory examination of the water system at F, R and C sites. All of them were negative for bacterial indicators of faecal contamination. No faecal bacteria were identified in the seven samples collected from M, W and S sites, including the well and the source that fed the water network, at any sampling time, before and after implementing the control measures. The research of NoV genome was conducted on concentrated water samples collected on 1 August from R and F sites (Table 3). NoV genome was detected in a single sample collected from the water entry of the resort involved in the outbreak (site R). Actually, only two out of 26 One-Step RT-PCR repetitions were positive for NoV viral capsid gene. The two ORF2-positive results were obtained at the University of Palermo Laboratory using both 1:10 diluted and undiluted extracted RNA. The sequence analysis of the partial ORF2 portion (289 bp) indicated the presence of the same GII.2 capsid genotype detected in the stool samples of the students involved in the outbreak, showing 98.9% nt id. No ORF1 NoV sequence could be obtained from the NoV-positive water sample. Follow-up analyses for NoV genome, conducted on water samples collected on 20 November both at R and F sites found no positive samples. By sequence alignment, the NoV sequences obtained from Mascali outbreak samples were phylogenetically closely related in both ORF1 and ORF2 (99 and 98.9% of nt id, respectively) to a GII.P2/GII.2 strain responsible for a sporadic paediatric gastroenteritis case occurred in Palermo in January 2016 (PA 36/16, accession numbers MG517464 and MG517460, for ORF1 and ORF2, respectively) and previously detected at the Enteric Viruses Laboratory, University of Palermo. Whereas, three GII.P2/GII.2 Italian strains (PA 280/11, PA 366/11 and PA 3/ 14, accession numbers MG517462, MG517465 and MG517463, for ORF1 and MG517457 to −59, for ORF2, respectively), detected in Palermo in 2011 and 2014, were more distantly associated with the outbreak strain both in RdRp and VP1 (90% to 97.4% nt id and 96.1% to 98.1% nt id, respectively, in ORF1 and ORF2). Discussion Epidemiological features, such as the presence of vomiting in more than half (88%) of symptomatic subjects, the 33.2 h average incubation period and the 38.4 h average duration of illness, suggested a NoV aetiology for the investigated outbreak [13]. The detection of NoV genome in all 25 faecal samples from symptomatic subjects fully complies with the requirement of detection of at least one positive sample out of four tested that is considered to be sufficient to ascertain the aetiology of a NoV outbreak [24]. Moreover, molecular typing indicated that a single viral strain, belonging to genotype GII.P2/GII.2, infected all symptomatic subjects involved in the outbreak. NoVs are highly infectious and can easily spread from person to person but also through contaminated food or water. Therefore, settings such as health care institutions, schools and touristic resorts are mainly affected, since close contact and shared food services facilitate transmission. In Mascali, two waves of transmission were observed over three days, being consistent with food/waterborne followed by person-to-person and/or airborne (vomit) transmission (Fig. 2). The first wave of cases presented on 20 May 2016 after a single shared meal, while the last case was detected on 22 May. Exposure to vomit may have been responsible for some of the secondary cases occurred after the first wave of illness. Vomiting produces aerosols that can be inhaled and swallowed. The extremely low infectious dose of NoV makes the inhalation of vomit aerosols an effective transmission route in settings where there is close contact, like on a bus trip [25]. All food consumed in the only shared lunch on 19 May at the resort was associated with the risk of disease (Table 2), but the association was inconclusive since all cases shared the same dishes included in the lunch menu on Day 1. Although many foods were potentially involved, our attention focused on the tap water. In fact, trivial information from local health authorities reported that local population avoided tap water consumption and relied on bottled water for drinking and cooking. Many small outbreaks of gastroenteritis involving restaurants and family groups had been reported in the same area, especially at the beginning of the summer season, when vacation houses and recreational activities were re-opened after a period of low water demand. Illness caused by NoV is common and if the level of water contamination is low, the number of cases remains limited and water can persist as unrecognised source of infection. If polluted water was served for drinking and used for cooking and washing dishes in the Mascali resort at Day 1 a generalised contamination of food and surfaces could have occurred. The potential role of the foodhandlers as a source of contamination could not be analysed in this study but should also be taken into account. If water contamination has already been present in the area for a long time, infections may have occurred among the local food-handlers. The analysis of water samples taken immediately after the outbreak revealed the presence of bacterial indicators of faecal contamination in the water used in the resort, with the only exception of a sample of filter purified water used both for cooking and as drinking water. When extending the sampling to other areas of the municipal water network, many other sites, mostly on the seaside, showed faecal pollution, which demonstrates a poor protection of the water network from faecal leakage (Fig. 1). NoV contamination of waters generally originates from sewage since these enteric pathogens, as other enteric viruses affecting humans, are mostly species-specific and are shed at high concentrations in the stools of patients. When the analyses on the stools of the students made it clear that the outbreak had been caused by NoV, the viral genome was also searched in appropriate volumes of water samples, but this happened when correction measures on the water network had already been taken. Nevertheless, a single water sample, obtained 2 months after the outbreak (1 August 2016), still contained genome fragments that could be correlated to the same NoV genotype causing the outbreak. Although the finding of NoV genome does not demonstrate that entire infecting viral particles were present in the water, it could indicate that a NoV belonging to the same genotype involved in the outbreak was still circulating in the local population several weeks after the outbreak and it was someway able to penetrate the water network, probably indicating continuing faecal contamination through leakage from sewage systems. NoV is particularly resistant in the environment and also to chlorine, therefore transmission through chlorinated water is not exceptional [9, 12, 26]. Drinking water from municipal supplies fed by lake or groundwater has already been identified in the past as the source of waterborne NoV outbreaks in Italy [27–29] and NoV is now considered one of the pathogens causing the largest number of affected consumers in drinking waterborne outbreaks [12]. NoV can persist in water longer than 15 days [30] and it is resistant to chlorine disinfection when free chlorine levels or pre-treatment processes are inadequate [31, 32]. Moreover, microfiltration filters labelled as certified by NSF Standards 53 or 58 (with a pore size ranging from 0.05 to 5 µm) are not effective in removing viruses and can even decrease chlorine concentration in filtered water [33]. Actually, the owner of the resort involved in our outbreak was encouraged to serve tap water as drinking water at the restaurant relying on the effectiveness of such a filter purifier. Unfortunately, we did not have a tap water sample from the restaurant of the resort taken before correction measures and whose volume was appropriate for virus detection. Therefore, we could not demonstrate the failure to eliminate viral pollution of the microfiltration system. In Mascali, correction measures on the water network consisted in the maintenance of the old chlorination device and in the addition of a new one at a key point located upstream and close to suspected illegal sewage dumping (Fig. 1), but as poor protection of the pipeline from sewage leakage risked to reproduce faecal pollution episodes, renovation of the municipal water system was also advised, as well as systematic control of illegal sewage dumping. The additional chlorination device was apparently the most effective intervention, since right after its activation faecal bacterial pollution indicators disappeared. However, NoV genome was detected in a water sample taken 6 weeks after the last detection of bacterial indicators. Unfortunately, detection of faecal contamination of drinking water is only based on the use of bacterial indicators despite no correlation between levels of enteric bacteria and enteric viruses has been found [34–36]. According to ISGEV enteric virus circulation surveillance data, in 2016 the circulation of GII.P2/GII.2 NoV strains in Palermo, home town of the travellers involved in the outbreak, was the highest ever observed from the starting of local NoV surveillance in 2004. GII.P2/GII.2 was the fourth most common genotype in 2016 (12.2% prevalence), following GII.P4/GII.4 Sydney 2012 (28.4%), GII.P16/GII.2 (18.9%) and GII.Pe/GII.4 Sydney 2012 (17.6%). GII.P2/GII.2 strains were mostly observed starting from July 2016, leaving space for the possibility that their increased circulation was linked to the introduction of the ‘Mascali clone’. However, a couple of GII.P2/GII.2 strains (including PA36/16) genetically close to the ‘Mascali clone’ had been already detected in January 2016 in Palermo and GII.P2/GII.2 NoV had sporadically been observed also in 2011 and 2014. Moreover, the circulation of GII.P2/GII.2 NoVs has been increasing worldwide over the last period [37]. Therefore, the unusually high circulation observed in Palermo might just reflect the global epidemiology of GII.P2/GII.2 strains. Noteworthy, the older GII.P2/GII.2 NoV strains detected in Palermo were closely related in the ORF2 (98.9% nt id) to the GII.P16/GII.2 Nashville strain (KY865307) which is supposed to have generated the recombinant GII.P16/GII.4 clone recently circulating epidemically worldwide [37, 38] and were more distantly associated (95.3% nt id) to the GII.2 capsid sequence (AB662868) of another putative GII.P16/GII.4 ancestor from Japan [39]. The increasing frequency of detection of waterborne NoV outbreaks calls for the introduction of routine monitoring for viruses of drinking waters and for the definition of affordable and reliable indicators for the presence of NoV in drinking waters. The use of ultrafiltration systems as a concentration method in order to recover viruses from environmental waters combined with RT-PCR to detect viral genomes has been demonstrated to grant a recovery rate of >50% of multiple viruses [40] and is at present one of the most sensitive approaches for the detection of NoVs, although the discrimination between viable and nonviable viruses remains a problem. The burden of NoV outbreaks, either food associated or waterborne, is still poorly defined due to the lack of specific diagnostic tools in many clinical laboratories. NoV epidemics should always be suspected when the main symptom is vomiting and the duration of disease is short (<48 h). Waterborne transmission is suggested by the rapid spread of disease and the high attack rate. Care should be taken to perform sensitive diagnostic tests on a sufficient number of stool samples in order to identify NoV as the causative agent of enteritis outbreaks, possibly preferring genome screening over antigen detection. Only the systematic detection of NoV outbreaks can guide in the choice of the safety procedures that could be most appropriate in containing such microbiological hazard.

How many people were hospitalized?

1548

Waterborne Norovirus outbreak at a seaside resort likely originating from municipal water distribution system failure

Abstract In May 2016 a Norovirus (NoV) gastroenteritis outbreak involved a high school class visiting a seaside resort near Taormina (Mascali, Sicily). Twenty-four students and a teacher were affected and 17 of them showed symptoms on the second day of the journey, while the others got ill within the following 2 days. Symptoms included vomiting, diarrhoea and fever, and 12 students required hospitalisation. Stool samples tested positive for NoV genome by Real-Time polymerase chain reaction assay in all 25 symptomatic subjects. The GII.P2/GII.2 NoV genotype was linked to the outbreak by ORF1/ORF2 sequence analysis. The epidemiological features of the outbreak were consistent with food/waterborne followed by person-to-person and/ or vomit transmission. Food consumed at a shared lunch on the first day of the trip was associated to illness and drinking un-bottled tap water was also considered as a risk factor. The analysis of water samples revealed the presence of bacterial indicators of faecal contamination in the water used in the resort as well as in other areas of the municipal water network, linking the NoV gastroenteritis outbreak to tap water pollution from sewage leakage. From a single water sample, an amplicon whose sequence corresponded to the capsid genotype recovered from patients could be obtained. Introduction Noroviruses (NoVs) are recognised as a leading cause of sporadic cases and outbreaks of acute gastroenteritis (AGE) across all age groups [1, 2]. They are transmitted primarily through the faecal-oral route either by direct person–person spread and with consumption of contaminated food or water [1, 3, 4]. NoVs are non-enveloped RNA viruses belonging to the Caliciviridae family and are classified into seven distinct genogroups (GI-GVII) on the basis of the major capsid protein VP1 sequence. GI, GII and GIV NoVs infect humans and they have been divided into more than 30 genotypes based on the sequences of polymerase (ORF1) and capsid proteins (ORF2) [5]. NoV GII.4 genotype has been associated with the vast majority of NoV related cases of AGE worldwide although many genotypes, belonging to both GI and GII genogroups are involved in foodborne and waterborne outbreaks [6, 7]. NoVs are highly contagious owing to their high stability in the environment and their ability to infect at low doses. In fact, just 10–100 virions of NoV may be sufficient to cause AGE [8, 9]. NoVs are considered the major cause of water-related disease in semi-closed institutions, including restaurants, schools, hospitals, nursing homes and cruise ships [10–12]. To confirm NoV as a cause of outbreaks, the presence of four epidemiological features of disease, vomiting in more than half of affected persons, a mean incubation period of 24/48 h, a mean duration of illness of 12–60 h and the absence of bacterial pathogens in stool culture, should be evaluated [13]. Waterborne NoV outbreaks can be caused by contaminated drinking water despite correct chlorine concentration and in the absence of coliform bacteria [10, 14–17]. The aim of this study was to investigate an outbreak of NoV gastroenteritis associated with drinking water involving a high school class travelling from Palermo (West Sicily) to Acireale (East Sicily) on an educational tour in May 2016. Stool samples were collected soon after the onset of AGE symptoms to ascertain the aetiology of illness. The source of infection was investigated collecting epidemiological data through an online questionnaire and, in order to explore a possible transmission route, laboratory tests on water from the municipal network were performed. Material and methods Epidemiological investigation From 20 to 22 May 2016 an AGE outbreak occurred in Mascali, a small town near Taormina, involving a school class of 28 students, 13 girls and 15 boys aged between 14 and 18 years and a 54-year-old female teacher, during a trip to participate in a theatre festival in Acireale. A total of 25/29 persons were affected by nausea, vomiting, abdominal cramps, diarrhoea and fever while they were staying at a seaside resort in Mascali, on the way to Acireale. To obtain data on symptoms and the correlated risk factors all participants filled an online questionnaire investigating personal data (age and sex), travel and accommodation details and food consumption over 48 h before symptoms presentation, time of onset and duration of symptoms, therapy and/or hospitalisation required and history of contact with people affected by diarrhoea and/or vomiting (Table 1). The severity of illness was analysed by the Vesikari severity scale allowing to distinguish non-severe (score <11) and severe (score ⩾11) AGE (Table 1). Vesikari score was calculated from maximum number of episodes/day and duration of both diarrhoea and vomiting (score 0–3 for each of the 4 parameters), temperature (score 0–3), level of dehydration (score 0–3) and treatment required (score 0–2) [18]. A case-control study was conducted to establish if food consumed before the onset of symptoms was associated with the disease. A case was defined as a student attending the tour and reporting diarrhoea and/or vomiting within 36 h from the first shared meal during the tour, while students who did not present any of the previous symptoms or those showing the first symptoms with a delay of ⩾48 h (secondary cases) were used as controls. Variables were studied using odds ratio (OR) and its 95% CI and statistical associations were determined using the χ2 test, with the level of statistical significance set at 0.05 (Table 2). NoV detection in stool samples A stool sample was obtained from each of the 29 travellers, soon after (3–5 days) the onset of the outbreak. No stool samples could be obtained from the food-handlers working at the resort. All samples were tested for NoV at the Enteric Viruses Laboratory of the University of Palermo, member of the Italian Study Group for Enteric Viruses (ISGEV), by a Real-Time polymerase chain reaction (RT-PCR) assay that allows to differentiate GI and GII NoVs [19]. Total RNA was extracted from stool samples using QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) and cDNA was generated by MMLV reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were considered positive for NoVs when cycle threshold (ct) of RT-PCR was ⩽30 and lowpositive when 30<ct⩽35. For all NoV-positive samples two genomic fragments were amplified by conventional RT-PCR using Superscript III One-Step RT-PCR system with Platinum Taq DNA polymerase (Invitrogen) and primers JV12a/JV13b, targeting 326 bps of the RdRp region A [20] and either COG1F/ GISKR or COG2F/GIISKR, amplifying a 315 bps 5′portion of GI or GII VP1 protein [19, 21]. Environmental investigation From May through August 2016, a total of 32 tap water samples were collected from the Mascali municipal network. In particular, seven samples were collected at the resort where the outbreak occurred and from its water entry (R), two in the city centre of Mascali (M), 15 in the village of Fondachello (F), on the seaside, three in Carrabba village (C), far from the coast, and two and three, respectively, from the water well ‘Carlino’ (W) and the source ‘Bufardo and Torregrossa’ (S) which feed the municipal system and provide drinking water to the Mascali area (Fig. 1, Table 3). One litre water samples were sent by the local health authorities to the SIALAB laboratory in Avola, Siracusa and 100 ml aliquots were screened for coliforms, E. coli and enterococci as faecal indicators, by standard bacteriological methods according to UNI EN ISO 9308-1:2014. Unfortunately, although the suspect of a water source for the viral outbreak was quickly communicated, only in August and November 2016, following the implementation of control measures, local authorities provided water samplings of appropriate volume to perform viral tests (10 l). Five water samples of 10 l each were collected from different sampling sites in F and R areas for viral screening (Table 3) and were concentrated according to the Italian National Institute of Health (Istituto Superiore di Sanità, ISS) protocols [22] through an ultra-filtration system (Sartoflow® Slice 200 Benchtop Crossflow System, Sartorius AG, Goettingen, Germany), using appropriate membranes (SG Hydrosart 10 kDa) pretreated with 300 ml of 3% Beef Extract (BE) at pH 7, by recirculating it for 10 min in the system, at a pressure not higher than 1.30 bar. Elution and recovery of viruses attached to the membranes were obtained by recirculating 200 ml of 3% BE at pH 9.5 for 3 h until a 10–12 ml volume of eluate was obtained. The pH of the eluate was then checked with litmus paper and if necessary brought to the value of 7 with HCl. Concentrated water samples were simultaneously tested for NoV in three laboratories: Enteric Viruses Laboratory, University of Palermo; Istituto Zooprofilattico Sperimentale della Sicilia (IZS), Palermo; and Istituto Superiore di Sanità (ISS), Rome. Total RNA was extracted from 500 and 560 µl of a concentrated water sample using BioMèrieux NucliSENS® Magnetic Extraction System (BioMèrieux) and QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) respectively, according to manufacturer’s instructions. Bonded RNA was eluted by 100 µl of Elution Buffer. Amplifications were performed at least in duplicate on whole and 1:10 diluted extracts. A Real-Time RT-PCR assay according to ISO/TS 15216- 2:2013, two in-house Real-Time RT-PCR protocols and the One-Step RT-PCR methods already used for amplifying ORF1 and ORF2 fragments in NoV-positive faecal samples were performed [19–21, 23]. Nucleotide sequences analysis and phylogenetic analysis PCR amplicons were purified by commercial columns (CleanSweep PCR Purification, Thermo Fisher Scientific, Vilnius, Lithuania) and sequenced by Sanger method and all nucleotide sequences were aligned using ClustalW. To identify NoV genotypes, RdRp and VP1 sequences were submitted to the RIVM NoV genotyping tool (http://www.rivm.nl/mpf/norovirus/typingtool). Phylogenetic trees were constructed using MEGA 7 and Neighbour-Joining method (bootstrap 1000 replicates). Sequences from stool and water samples are available at NCBI with the following accession numbers: MG517455 for the major capsid protein obtained from a water sample; MG517456 and MG517461, respectively, for the major capsid protein and RNA-dependent-RNA-polymerase, as representatives of the sequences obtained from the stool samples of those involved in the outbreak. Control measures After outbreak notification, on 21 May, the resort was closed for 2 days. On 25 May, following the first results of the bacteriological analyses on municipal water samples, the use of municipal water for human consumption was temporarily restricted (use of water was permitted only for cleaning purposes) in the Carrabba and Fondachello villages and in the area surrounding the resort. From 25 May the reservoir collecting the water from the Carlino well and its chlorination device were submitted to maintenance work and a new chlorination device was installed on the conduit bringing the water from the Bufardo source. Actually, the Bufardo source had been connected to the municipal network in 2003 but unfortunately a dedicated chlorination station had not been planned at that time. Therefore, unchlorinated water entered the system near Fondachello village on the seaside (Fig. 1), possibly diluting chlorine concentrations in water treated at the Carlino reservoir. Moreover, the exact path of the branch of the water network situated downstream the Bufardo source water connection was not marked in the maps of the municipality (Fig. 1, red line), indicating that this section had never been submitted to maintenance. The area surrounding the unmaintained section of the water pipeline was scattered with unauthorised cottages that were not connected to the sewerage systems. Likely, untreated waste from cesspools was unloaded directly into the soil, generating a risk of contamination of the municipal water system through leakage. No work had been made on the water pipelines before the outbreak but sections of the network had been temporarily closed during the winter season on the seaside (Fondachello). Local health authorities advised the connection of all households in the area to the sewerage system and the replacement of the water system section without maintenance. Health authorities also raised attention on AGE cases notification at all local Emergency Departments and Family Practitioners but there was no further evidence of AGE outbreaks in the months following the intervention. Results Data collected through the online questionnaire compiled 7–10 days after the outbreak revealed that none of the 28 students and the teacher taking part in the school trip showed gastroenteritis symptoms at the beginning of the journey and they had no contact with people who suffered from diarrhoea and/or vomiting in the previous days. Travellers had their first shared meal on 19 May (Day 1), 2016, at lunch on their arrival to the resort in Mascali. The list of foods included in the menu is shown in Table 2. Conversely, the dinner on Day 1 was not shared and the students went to dine in three different fast food restaurants in Acireale. On 20 May (Day 2), 17 students (58.6% of the travellers) presented with vomit and/or diarrhoea, falling ill 26–36 h after lunch on Day 1, strongly suggesting that exposure occurred on the first shared meal. Some students also vomited in the presence of the other travellers on the bus bringing them to Acireale. Five more students were symptomatic in the early hours of the following day (21 May, Day 3, from 0:30 am to 8:00 am), as a continuation of the wave of illnesses of the previous day. The teacher and one more student fell ill at 8:00 pm and 11:00 pm of Day 3, respectively, for a second wave of illnesses occurring >48 h from the first shared meal and <36 h from the first cases, with the last student falling sick on 22 May (Day 4) at 3:00 am (Fig. 2, Table 1). On the whole, the mean incubation period was 33.2 h. Based on the online survey, the severity of illness was evaluated by the Vesikari severity scale, grouping symptomatic travellers into severe (11/25 = 44%) and non-severe (14/25 = 56%) cases (Table 1). Vomiting and fever were the predominant symptoms, affecting 22 out of 25 (88%) symptomatic subjects, while diarrhoea was present in 14 out of 25 (56%). Hospitalisation for rehydration therapy was necessary for 13 students (91% [10/11] of severe and 21% [3/14] of non-severe cases). Emergency room (ER) or family practitioner visits were necessary for 21 students and 13 (62%) of them initiated an antibiotic therapy. Two students indicated that other family members had gastroenteric symptoms more than 24 h after their return home, on Day 3, 4 and 5. All those involved in the outbreak recovered within 4 days, with an average duration of illness of 38.4 h. A case-control study of the risk associated with eating foods at lunch on Day 1 or with attending different restaurants at dinner on Day 1 was attempted. A total of 19 subjects filled the section of the online survey concerning food consumption. Fourteen of the responders were among the 22 cases falling ill within 36 h from the first shared meal (first wave), while the remaining five responders were three travellers with supposed secondary infections (showing symptoms with a delay ⩾48 h; second wave) and two asymptomatic students that were used as controls. No statistically significant associations with any food items served at lunch on Day 1 was identified (P < 0.05; Table 2). However, all food consumed on that occasion was to some extent associated with the risk of disease, the highest OR being obtained with chicken (OR: 20.7143, P: 0.0678; Table 2). Actually, most of the cases ate all foods included in the menu, while one of the two asymptomatic controls did not participate in the meal and the other had a pasta dish only. No statistically significant association was found for attending a particular fast food restaurant at the dinner on Day 1 (OR: 2.5, P: 0.3635). The Real-Time RT-PCR assay allowed to detect GII genogroup NoV genome in all the 25 samples from symptomatic subjects, although two of them showed high Ct values, indicating lower shedding of viral genomes (Table 1). The samples from four asymptomatic students were negative for both NoV genogroups. By sequence analysis, a NoV genotype could be defined for 20/ 25 NoV-positive stool samples. All NoV strains from travellers involved in the outbreaks belonged to the same genotype, GII.P2/GII.2, showing >99% nucleotide identities (nt id) in both fragments (ORF1 and ORF2) analysed. Thirty-two water samples were collected and analysed for bacterial contamination (Table 3). Seven samples were taken from the resort (site R) and the remaining 25 from five other sites (C, F, M, S, W), representative of the entire municipal water network. The majority (10/12) of the water samples collected immediately after the onset of the outbreak, between 21 May and 16 June 2016, at the resort (site R) and at sites F and C, located close to the resort, on the seaside and in the interior, respectively, were positive for coliform bacteria. More than 1000 colony-forming units (CFU)/ 100 ml (up to 2700 CFU/100 ml) were detected in four samples, while their absence is required in 100 ml by the Standard of Water Quality in Italy and seven of 11 samples tested were also Escherichia coli-positive (up to 900 CFU/100 ml). Enterococci were searched in 11 samples and three of them were found positive for this faecal contamination indicator. From 7 July to 16 August 2016, after control measures on the water network were undertaken, 13 water samples were collected for follow-up laboratory examination of the water system at F, R and C sites. All of them were negative for bacterial indicators of faecal contamination. No faecal bacteria were identified in the seven samples collected from M, W and S sites, including the well and the source that fed the water network, at any sampling time, before and after implementing the control measures. The research of NoV genome was conducted on concentrated water samples collected on 1 August from R and F sites (Table 3). NoV genome was detected in a single sample collected from the water entry of the resort involved in the outbreak (site R). Actually, only two out of 26 One-Step RT-PCR repetitions were positive for NoV viral capsid gene. The two ORF2-positive results were obtained at the University of Palermo Laboratory using both 1:10 diluted and undiluted extracted RNA. The sequence analysis of the partial ORF2 portion (289 bp) indicated the presence of the same GII.2 capsid genotype detected in the stool samples of the students involved in the outbreak, showing 98.9% nt id. No ORF1 NoV sequence could be obtained from the NoV-positive water sample. Follow-up analyses for NoV genome, conducted on water samples collected on 20 November both at R and F sites found no positive samples. By sequence alignment, the NoV sequences obtained from Mascali outbreak samples were phylogenetically closely related in both ORF1 and ORF2 (99 and 98.9% of nt id, respectively) to a GII.P2/GII.2 strain responsible for a sporadic paediatric gastroenteritis case occurred in Palermo in January 2016 (PA 36/16, accession numbers MG517464 and MG517460, for ORF1 and ORF2, respectively) and previously detected at the Enteric Viruses Laboratory, University of Palermo. Whereas, three GII.P2/GII.2 Italian strains (PA 280/11, PA 366/11 and PA 3/ 14, accession numbers MG517462, MG517465 and MG517463, for ORF1 and MG517457 to −59, for ORF2, respectively), detected in Palermo in 2011 and 2014, were more distantly associated with the outbreak strain both in RdRp and VP1 (90% to 97.4% nt id and 96.1% to 98.1% nt id, respectively, in ORF1 and ORF2). Discussion Epidemiological features, such as the presence of vomiting in more than half (88%) of symptomatic subjects, the 33.2 h average incubation period and the 38.4 h average duration of illness, suggested a NoV aetiology for the investigated outbreak [13]. The detection of NoV genome in all 25 faecal samples from symptomatic subjects fully complies with the requirement of detection of at least one positive sample out of four tested that is considered to be sufficient to ascertain the aetiology of a NoV outbreak [24]. Moreover, molecular typing indicated that a single viral strain, belonging to genotype GII.P2/GII.2, infected all symptomatic subjects involved in the outbreak. NoVs are highly infectious and can easily spread from person to person but also through contaminated food or water. Therefore, settings such as health care institutions, schools and touristic resorts are mainly affected, since close contact and shared food services facilitate transmission. In Mascali, two waves of transmission were observed over three days, being consistent with food/waterborne followed by person-to-person and/or airborne (vomit) transmission (Fig. 2). The first wave of cases presented on 20 May 2016 after a single shared meal, while the last case was detected on 22 May. Exposure to vomit may have been responsible for some of the secondary cases occurred after the first wave of illness. Vomiting produces aerosols that can be inhaled and swallowed. The extremely low infectious dose of NoV makes the inhalation of vomit aerosols an effective transmission route in settings where there is close contact, like on a bus trip [25]. All food consumed in the only shared lunch on 19 May at the resort was associated with the risk of disease (Table 2), but the association was inconclusive since all cases shared the same dishes included in the lunch menu on Day 1. Although many foods were potentially involved, our attention focused on the tap water. In fact, trivial information from local health authorities reported that local population avoided tap water consumption and relied on bottled water for drinking and cooking. Many small outbreaks of gastroenteritis involving restaurants and family groups had been reported in the same area, especially at the beginning of the summer season, when vacation houses and recreational activities were re-opened after a period of low water demand. Illness caused by NoV is common and if the level of water contamination is low, the number of cases remains limited and water can persist as unrecognised source of infection. If polluted water was served for drinking and used for cooking and washing dishes in the Mascali resort at Day 1 a generalised contamination of food and surfaces could have occurred. The potential role of the foodhandlers as a source of contamination could not be analysed in this study but should also be taken into account. If water contamination has already been present in the area for a long time, infections may have occurred among the local food-handlers. The analysis of water samples taken immediately after the outbreak revealed the presence of bacterial indicators of faecal contamination in the water used in the resort, with the only exception of a sample of filter purified water used both for cooking and as drinking water. When extending the sampling to other areas of the municipal water network, many other sites, mostly on the seaside, showed faecal pollution, which demonstrates a poor protection of the water network from faecal leakage (Fig. 1). NoV contamination of waters generally originates from sewage since these enteric pathogens, as other enteric viruses affecting humans, are mostly species-specific and are shed at high concentrations in the stools of patients. When the analyses on the stools of the students made it clear that the outbreak had been caused by NoV, the viral genome was also searched in appropriate volumes of water samples, but this happened when correction measures on the water network had already been taken. Nevertheless, a single water sample, obtained 2 months after the outbreak (1 August 2016), still contained genome fragments that could be correlated to the same NoV genotype causing the outbreak. Although the finding of NoV genome does not demonstrate that entire infecting viral particles were present in the water, it could indicate that a NoV belonging to the same genotype involved in the outbreak was still circulating in the local population several weeks after the outbreak and it was someway able to penetrate the water network, probably indicating continuing faecal contamination through leakage from sewage systems. NoV is particularly resistant in the environment and also to chlorine, therefore transmission through chlorinated water is not exceptional [9, 12, 26]. Drinking water from municipal supplies fed by lake or groundwater has already been identified in the past as the source of waterborne NoV outbreaks in Italy [27–29] and NoV is now considered one of the pathogens causing the largest number of affected consumers in drinking waterborne outbreaks [12]. NoV can persist in water longer than 15 days [30] and it is resistant to chlorine disinfection when free chlorine levels or pre-treatment processes are inadequate [31, 32]. Moreover, microfiltration filters labelled as certified by NSF Standards 53 or 58 (with a pore size ranging from 0.05 to 5 µm) are not effective in removing viruses and can even decrease chlorine concentration in filtered water [33]. Actually, the owner of the resort involved in our outbreak was encouraged to serve tap water as drinking water at the restaurant relying on the effectiveness of such a filter purifier. Unfortunately, we did not have a tap water sample from the restaurant of the resort taken before correction measures and whose volume was appropriate for virus detection. Therefore, we could not demonstrate the failure to eliminate viral pollution of the microfiltration system. In Mascali, correction measures on the water network consisted in the maintenance of the old chlorination device and in the addition of a new one at a key point located upstream and close to suspected illegal sewage dumping (Fig. 1), but as poor protection of the pipeline from sewage leakage risked to reproduce faecal pollution episodes, renovation of the municipal water system was also advised, as well as systematic control of illegal sewage dumping. The additional chlorination device was apparently the most effective intervention, since right after its activation faecal bacterial pollution indicators disappeared. However, NoV genome was detected in a water sample taken 6 weeks after the last detection of bacterial indicators. Unfortunately, detection of faecal contamination of drinking water is only based on the use of bacterial indicators despite no correlation between levels of enteric bacteria and enteric viruses has been found [34–36]. According to ISGEV enteric virus circulation surveillance data, in 2016 the circulation of GII.P2/GII.2 NoV strains in Palermo, home town of the travellers involved in the outbreak, was the highest ever observed from the starting of local NoV surveillance in 2004. GII.P2/GII.2 was the fourth most common genotype in 2016 (12.2% prevalence), following GII.P4/GII.4 Sydney 2012 (28.4%), GII.P16/GII.2 (18.9%) and GII.Pe/GII.4 Sydney 2012 (17.6%). GII.P2/GII.2 strains were mostly observed starting from July 2016, leaving space for the possibility that their increased circulation was linked to the introduction of the ‘Mascali clone’. However, a couple of GII.P2/GII.2 strains (including PA36/16) genetically close to the ‘Mascali clone’ had been already detected in January 2016 in Palermo and GII.P2/GII.2 NoV had sporadically been observed also in 2011 and 2014. Moreover, the circulation of GII.P2/GII.2 NoVs has been increasing worldwide over the last period [37]. Therefore, the unusually high circulation observed in Palermo might just reflect the global epidemiology of GII.P2/GII.2 strains. Noteworthy, the older GII.P2/GII.2 NoV strains detected in Palermo were closely related in the ORF2 (98.9% nt id) to the GII.P16/GII.2 Nashville strain (KY865307) which is supposed to have generated the recombinant GII.P16/GII.4 clone recently circulating epidemically worldwide [37, 38] and were more distantly associated (95.3% nt id) to the GII.2 capsid sequence (AB662868) of another putative GII.P16/GII.4 ancestor from Japan [39]. The increasing frequency of detection of waterborne NoV outbreaks calls for the introduction of routine monitoring for viruses of drinking waters and for the definition of affordable and reliable indicators for the presence of NoV in drinking waters. The use of ultrafiltration systems as a concentration method in order to recover viruses from environmental waters combined with RT-PCR to detect viral genomes has been demonstrated to grant a recovery rate of >50% of multiple viruses [40] and is at present one of the most sensitive approaches for the detection of NoVs, although the discrimination between viable and nonviable viruses remains a problem. The burden of NoV outbreaks, either food associated or waterborne, is still poorly defined due to the lack of specific diagnostic tools in many clinical laboratories. NoV epidemics should always be suspected when the main symptom is vomiting and the duration of disease is short (<48 h). Waterborne transmission is suggested by the rapid spread of disease and the high attack rate. Care should be taken to perform sensitive diagnostic tests on a sufficient number of stool samples in order to identify NoV as the causative agent of enteritis outbreaks, possibly preferring genome screening over antigen detection. Only the systematic detection of NoV outbreaks can guide in the choice of the safety procedures that could be most appropriate in containing such microbiological hazard.

How many people were dead?

1549

Waterborne Norovirus outbreak at a seaside resort likely originating from municipal water distribution system failure

Abstract In May 2016 a Norovirus (NoV) gastroenteritis outbreak involved a high school class visiting a seaside resort near Taormina (Mascali, Sicily). Twenty-four students and a teacher were affected and 17 of them showed symptoms on the second day of the journey, while the others got ill within the following 2 days. Symptoms included vomiting, diarrhoea and fever, and 12 students required hospitalisation. Stool samples tested positive for NoV genome by Real-Time polymerase chain reaction assay in all 25 symptomatic subjects. The GII.P2/GII.2 NoV genotype was linked to the outbreak by ORF1/ORF2 sequence analysis. The epidemiological features of the outbreak were consistent with food/waterborne followed by person-to-person and/ or vomit transmission. Food consumed at a shared lunch on the first day of the trip was associated to illness and drinking un-bottled tap water was also considered as a risk factor. The analysis of water samples revealed the presence of bacterial indicators of faecal contamination in the water used in the resort as well as in other areas of the municipal water network, linking the NoV gastroenteritis outbreak to tap water pollution from sewage leakage. From a single water sample, an amplicon whose sequence corresponded to the capsid genotype recovered from patients could be obtained. Introduction Noroviruses (NoVs) are recognised as a leading cause of sporadic cases and outbreaks of acute gastroenteritis (AGE) across all age groups [1, 2]. They are transmitted primarily through the faecal-oral route either by direct person–person spread and with consumption of contaminated food or water [1, 3, 4]. NoVs are non-enveloped RNA viruses belonging to the Caliciviridae family and are classified into seven distinct genogroups (GI-GVII) on the basis of the major capsid protein VP1 sequence. GI, GII and GIV NoVs infect humans and they have been divided into more than 30 genotypes based on the sequences of polymerase (ORF1) and capsid proteins (ORF2) [5]. NoV GII.4 genotype has been associated with the vast majority of NoV related cases of AGE worldwide although many genotypes, belonging to both GI and GII genogroups are involved in foodborne and waterborne outbreaks [6, 7]. NoVs are highly contagious owing to their high stability in the environment and their ability to infect at low doses. In fact, just 10–100 virions of NoV may be sufficient to cause AGE [8, 9]. NoVs are considered the major cause of water-related disease in semi-closed institutions, including restaurants, schools, hospitals, nursing homes and cruise ships [10–12]. To confirm NoV as a cause of outbreaks, the presence of four epidemiological features of disease, vomiting in more than half of affected persons, a mean incubation period of 24/48 h, a mean duration of illness of 12–60 h and the absence of bacterial pathogens in stool culture, should be evaluated [13]. Waterborne NoV outbreaks can be caused by contaminated drinking water despite correct chlorine concentration and in the absence of coliform bacteria [10, 14–17]. The aim of this study was to investigate an outbreak of NoV gastroenteritis associated with drinking water involving a high school class travelling from Palermo (West Sicily) to Acireale (East Sicily) on an educational tour in May 2016. Stool samples were collected soon after the onset of AGE symptoms to ascertain the aetiology of illness. The source of infection was investigated collecting epidemiological data through an online questionnaire and, in order to explore a possible transmission route, laboratory tests on water from the municipal network were performed. Material and methods Epidemiological investigation From 20 to 22 May 2016 an AGE outbreak occurred in Mascali, a small town near Taormina, involving a school class of 28 students, 13 girls and 15 boys aged between 14 and 18 years and a 54-year-old female teacher, during a trip to participate in a theatre festival in Acireale. A total of 25/29 persons were affected by nausea, vomiting, abdominal cramps, diarrhoea and fever while they were staying at a seaside resort in Mascali, on the way to Acireale. To obtain data on symptoms and the correlated risk factors all participants filled an online questionnaire investigating personal data (age and sex), travel and accommodation details and food consumption over 48 h before symptoms presentation, time of onset and duration of symptoms, therapy and/or hospitalisation required and history of contact with people affected by diarrhoea and/or vomiting (Table 1). The severity of illness was analysed by the Vesikari severity scale allowing to distinguish non-severe (score <11) and severe (score ⩾11) AGE (Table 1). Vesikari score was calculated from maximum number of episodes/day and duration of both diarrhoea and vomiting (score 0–3 for each of the 4 parameters), temperature (score 0–3), level of dehydration (score 0–3) and treatment required (score 0–2) [18]. A case-control study was conducted to establish if food consumed before the onset of symptoms was associated with the disease. A case was defined as a student attending the tour and reporting diarrhoea and/or vomiting within 36 h from the first shared meal during the tour, while students who did not present any of the previous symptoms or those showing the first symptoms with a delay of ⩾48 h (secondary cases) were used as controls. Variables were studied using odds ratio (OR) and its 95% CI and statistical associations were determined using the χ2 test, with the level of statistical significance set at 0.05 (Table 2). NoV detection in stool samples A stool sample was obtained from each of the 29 travellers, soon after (3–5 days) the onset of the outbreak. No stool samples could be obtained from the food-handlers working at the resort. All samples were tested for NoV at the Enteric Viruses Laboratory of the University of Palermo, member of the Italian Study Group for Enteric Viruses (ISGEV), by a Real-Time polymerase chain reaction (RT-PCR) assay that allows to differentiate GI and GII NoVs [19]. Total RNA was extracted from stool samples using QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) and cDNA was generated by MMLV reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were considered positive for NoVs when cycle threshold (ct) of RT-PCR was ⩽30 and lowpositive when 30<ct⩽35. For all NoV-positive samples two genomic fragments were amplified by conventional RT-PCR using Superscript III One-Step RT-PCR system with Platinum Taq DNA polymerase (Invitrogen) and primers JV12a/JV13b, targeting 326 bps of the RdRp region A [20] and either COG1F/ GISKR or COG2F/GIISKR, amplifying a 315 bps 5′portion of GI or GII VP1 protein [19, 21]. Environmental investigation From May through August 2016, a total of 32 tap water samples were collected from the Mascali municipal network. In particular, seven samples were collected at the resort where the outbreak occurred and from its water entry (R), two in the city centre of Mascali (M), 15 in the village of Fondachello (F), on the seaside, three in Carrabba village (C), far from the coast, and two and three, respectively, from the water well ‘Carlino’ (W) and the source ‘Bufardo and Torregrossa’ (S) which feed the municipal system and provide drinking water to the Mascali area (Fig. 1, Table 3). One litre water samples were sent by the local health authorities to the SIALAB laboratory in Avola, Siracusa and 100 ml aliquots were screened for coliforms, E. coli and enterococci as faecal indicators, by standard bacteriological methods according to UNI EN ISO 9308-1:2014. Unfortunately, although the suspect of a water source for the viral outbreak was quickly communicated, only in August and November 2016, following the implementation of control measures, local authorities provided water samplings of appropriate volume to perform viral tests (10 l). Five water samples of 10 l each were collected from different sampling sites in F and R areas for viral screening (Table 3) and were concentrated according to the Italian National Institute of Health (Istituto Superiore di Sanità, ISS) protocols [22] through an ultra-filtration system (Sartoflow® Slice 200 Benchtop Crossflow System, Sartorius AG, Goettingen, Germany), using appropriate membranes (SG Hydrosart 10 kDa) pretreated with 300 ml of 3% Beef Extract (BE) at pH 7, by recirculating it for 10 min in the system, at a pressure not higher than 1.30 bar. Elution and recovery of viruses attached to the membranes were obtained by recirculating 200 ml of 3% BE at pH 9.5 for 3 h until a 10–12 ml volume of eluate was obtained. The pH of the eluate was then checked with litmus paper and if necessary brought to the value of 7 with HCl. Concentrated water samples were simultaneously tested for NoV in three laboratories: Enteric Viruses Laboratory, University of Palermo; Istituto Zooprofilattico Sperimentale della Sicilia (IZS), Palermo; and Istituto Superiore di Sanità (ISS), Rome. Total RNA was extracted from 500 and 560 µl of a concentrated water sample using BioMèrieux NucliSENS® Magnetic Extraction System (BioMèrieux) and QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) respectively, according to manufacturer’s instructions. Bonded RNA was eluted by 100 µl of Elution Buffer. Amplifications were performed at least in duplicate on whole and 1:10 diluted extracts. A Real-Time RT-PCR assay according to ISO/TS 15216- 2:2013, two in-house Real-Time RT-PCR protocols and the One-Step RT-PCR methods already used for amplifying ORF1 and ORF2 fragments in NoV-positive faecal samples were performed [19–21, 23]. Nucleotide sequences analysis and phylogenetic analysis PCR amplicons were purified by commercial columns (CleanSweep PCR Purification, Thermo Fisher Scientific, Vilnius, Lithuania) and sequenced by Sanger method and all nucleotide sequences were aligned using ClustalW. To identify NoV genotypes, RdRp and VP1 sequences were submitted to the RIVM NoV genotyping tool (http://www.rivm.nl/mpf/norovirus/typingtool). Phylogenetic trees were constructed using MEGA 7 and Neighbour-Joining method (bootstrap 1000 replicates). Sequences from stool and water samples are available at NCBI with the following accession numbers: MG517455 for the major capsid protein obtained from a water sample; MG517456 and MG517461, respectively, for the major capsid protein and RNA-dependent-RNA-polymerase, as representatives of the sequences obtained from the stool samples of those involved in the outbreak. Control measures After outbreak notification, on 21 May, the resort was closed for 2 days. On 25 May, following the first results of the bacteriological analyses on municipal water samples, the use of municipal water for human consumption was temporarily restricted (use of water was permitted only for cleaning purposes) in the Carrabba and Fondachello villages and in the area surrounding the resort. From 25 May the reservoir collecting the water from the Carlino well and its chlorination device were submitted to maintenance work and a new chlorination device was installed on the conduit bringing the water from the Bufardo source. Actually, the Bufardo source had been connected to the municipal network in 2003 but unfortunately a dedicated chlorination station had not been planned at that time. Therefore, unchlorinated water entered the system near Fondachello village on the seaside (Fig. 1), possibly diluting chlorine concentrations in water treated at the Carlino reservoir. Moreover, the exact path of the branch of the water network situated downstream the Bufardo source water connection was not marked in the maps of the municipality (Fig. 1, red line), indicating that this section had never been submitted to maintenance. The area surrounding the unmaintained section of the water pipeline was scattered with unauthorised cottages that were not connected to the sewerage systems. Likely, untreated waste from cesspools was unloaded directly into the soil, generating a risk of contamination of the municipal water system through leakage. No work had been made on the water pipelines before the outbreak but sections of the network had been temporarily closed during the winter season on the seaside (Fondachello). Local health authorities advised the connection of all households in the area to the sewerage system and the replacement of the water system section without maintenance. Health authorities also raised attention on AGE cases notification at all local Emergency Departments and Family Practitioners but there was no further evidence of AGE outbreaks in the months following the intervention. Results Data collected through the online questionnaire compiled 7–10 days after the outbreak revealed that none of the 28 students and the teacher taking part in the school trip showed gastroenteritis symptoms at the beginning of the journey and they had no contact with people who suffered from diarrhoea and/or vomiting in the previous days. Travellers had their first shared meal on 19 May (Day 1), 2016, at lunch on their arrival to the resort in Mascali. The list of foods included in the menu is shown in Table 2. Conversely, the dinner on Day 1 was not shared and the students went to dine in three different fast food restaurants in Acireale. On 20 May (Day 2), 17 students (58.6% of the travellers) presented with vomit and/or diarrhoea, falling ill 26–36 h after lunch on Day 1, strongly suggesting that exposure occurred on the first shared meal. Some students also vomited in the presence of the other travellers on the bus bringing them to Acireale. Five more students were symptomatic in the early hours of the following day (21 May, Day 3, from 0:30 am to 8:00 am), as a continuation of the wave of illnesses of the previous day. The teacher and one more student fell ill at 8:00 pm and 11:00 pm of Day 3, respectively, for a second wave of illnesses occurring >48 h from the first shared meal and <36 h from the first cases, with the last student falling sick on 22 May (Day 4) at 3:00 am (Fig. 2, Table 1). On the whole, the mean incubation period was 33.2 h. Based on the online survey, the severity of illness was evaluated by the Vesikari severity scale, grouping symptomatic travellers into severe (11/25 = 44%) and non-severe (14/25 = 56%) cases (Table 1). Vomiting and fever were the predominant symptoms, affecting 22 out of 25 (88%) symptomatic subjects, while diarrhoea was present in 14 out of 25 (56%). Hospitalisation for rehydration therapy was necessary for 13 students (91% [10/11] of severe and 21% [3/14] of non-severe cases). Emergency room (ER) or family practitioner visits were necessary for 21 students and 13 (62%) of them initiated an antibiotic therapy. Two students indicated that other family members had gastroenteric symptoms more than 24 h after their return home, on Day 3, 4 and 5. All those involved in the outbreak recovered within 4 days, with an average duration of illness of 38.4 h. A case-control study of the risk associated with eating foods at lunch on Day 1 or with attending different restaurants at dinner on Day 1 was attempted. A total of 19 subjects filled the section of the online survey concerning food consumption. Fourteen of the responders were among the 22 cases falling ill within 36 h from the first shared meal (first wave), while the remaining five responders were three travellers with supposed secondary infections (showing symptoms with a delay ⩾48 h; second wave) and two asymptomatic students that were used as controls. No statistically significant associations with any food items served at lunch on Day 1 was identified (P < 0.05; Table 2). However, all food consumed on that occasion was to some extent associated with the risk of disease, the highest OR being obtained with chicken (OR: 20.7143, P: 0.0678; Table 2). Actually, most of the cases ate all foods included in the menu, while one of the two asymptomatic controls did not participate in the meal and the other had a pasta dish only. No statistically significant association was found for attending a particular fast food restaurant at the dinner on Day 1 (OR: 2.5, P: 0.3635). The Real-Time RT-PCR assay allowed to detect GII genogroup NoV genome in all the 25 samples from symptomatic subjects, although two of them showed high Ct values, indicating lower shedding of viral genomes (Table 1). The samples from four asymptomatic students were negative for both NoV genogroups. By sequence analysis, a NoV genotype could be defined for 20/ 25 NoV-positive stool samples. All NoV strains from travellers involved in the outbreaks belonged to the same genotype, GII.P2/GII.2, showing >99% nucleotide identities (nt id) in both fragments (ORF1 and ORF2) analysed. Thirty-two water samples were collected and analysed for bacterial contamination (Table 3). Seven samples were taken from the resort (site R) and the remaining 25 from five other sites (C, F, M, S, W), representative of the entire municipal water network. The majority (10/12) of the water samples collected immediately after the onset of the outbreak, between 21 May and 16 June 2016, at the resort (site R) and at sites F and C, located close to the resort, on the seaside and in the interior, respectively, were positive for coliform bacteria. More than 1000 colony-forming units (CFU)/ 100 ml (up to 2700 CFU/100 ml) were detected in four samples, while their absence is required in 100 ml by the Standard of Water Quality in Italy and seven of 11 samples tested were also Escherichia coli-positive (up to 900 CFU/100 ml). Enterococci were searched in 11 samples and three of them were found positive for this faecal contamination indicator. From 7 July to 16 August 2016, after control measures on the water network were undertaken, 13 water samples were collected for follow-up laboratory examination of the water system at F, R and C sites. All of them were negative for bacterial indicators of faecal contamination. No faecal bacteria were identified in the seven samples collected from M, W and S sites, including the well and the source that fed the water network, at any sampling time, before and after implementing the control measures. The research of NoV genome was conducted on concentrated water samples collected on 1 August from R and F sites (Table 3). NoV genome was detected in a single sample collected from the water entry of the resort involved in the outbreak (site R). Actually, only two out of 26 One-Step RT-PCR repetitions were positive for NoV viral capsid gene. The two ORF2-positive results were obtained at the University of Palermo Laboratory using both 1:10 diluted and undiluted extracted RNA. The sequence analysis of the partial ORF2 portion (289 bp) indicated the presence of the same GII.2 capsid genotype detected in the stool samples of the students involved in the outbreak, showing 98.9% nt id. No ORF1 NoV sequence could be obtained from the NoV-positive water sample. Follow-up analyses for NoV genome, conducted on water samples collected on 20 November both at R and F sites found no positive samples. By sequence alignment, the NoV sequences obtained from Mascali outbreak samples were phylogenetically closely related in both ORF1 and ORF2 (99 and 98.9% of nt id, respectively) to a GII.P2/GII.2 strain responsible for a sporadic paediatric gastroenteritis case occurred in Palermo in January 2016 (PA 36/16, accession numbers MG517464 and MG517460, for ORF1 and ORF2, respectively) and previously detected at the Enteric Viruses Laboratory, University of Palermo. Whereas, three GII.P2/GII.2 Italian strains (PA 280/11, PA 366/11 and PA 3/ 14, accession numbers MG517462, MG517465 and MG517463, for ORF1 and MG517457 to −59, for ORF2, respectively), detected in Palermo in 2011 and 2014, were more distantly associated with the outbreak strain both in RdRp and VP1 (90% to 97.4% nt id and 96.1% to 98.1% nt id, respectively, in ORF1 and ORF2). Discussion Epidemiological features, such as the presence of vomiting in more than half (88%) of symptomatic subjects, the 33.2 h average incubation period and the 38.4 h average duration of illness, suggested a NoV aetiology for the investigated outbreak [13]. The detection of NoV genome in all 25 faecal samples from symptomatic subjects fully complies with the requirement of detection of at least one positive sample out of four tested that is considered to be sufficient to ascertain the aetiology of a NoV outbreak [24]. Moreover, molecular typing indicated that a single viral strain, belonging to genotype GII.P2/GII.2, infected all symptomatic subjects involved in the outbreak. NoVs are highly infectious and can easily spread from person to person but also through contaminated food or water. Therefore, settings such as health care institutions, schools and touristic resorts are mainly affected, since close contact and shared food services facilitate transmission. In Mascali, two waves of transmission were observed over three days, being consistent with food/waterborne followed by person-to-person and/or airborne (vomit) transmission (Fig. 2). The first wave of cases presented on 20 May 2016 after a single shared meal, while the last case was detected on 22 May. Exposure to vomit may have been responsible for some of the secondary cases occurred after the first wave of illness. Vomiting produces aerosols that can be inhaled and swallowed. The extremely low infectious dose of NoV makes the inhalation of vomit aerosols an effective transmission route in settings where there is close contact, like on a bus trip [25]. All food consumed in the only shared lunch on 19 May at the resort was associated with the risk of disease (Table 2), but the association was inconclusive since all cases shared the same dishes included in the lunch menu on Day 1. Although many foods were potentially involved, our attention focused on the tap water. In fact, trivial information from local health authorities reported that local population avoided tap water consumption and relied on bottled water for drinking and cooking. Many small outbreaks of gastroenteritis involving restaurants and family groups had been reported in the same area, especially at the beginning of the summer season, when vacation houses and recreational activities were re-opened after a period of low water demand. Illness caused by NoV is common and if the level of water contamination is low, the number of cases remains limited and water can persist as unrecognised source of infection. If polluted water was served for drinking and used for cooking and washing dishes in the Mascali resort at Day 1 a generalised contamination of food and surfaces could have occurred. The potential role of the foodhandlers as a source of contamination could not be analysed in this study but should also be taken into account. If water contamination has already been present in the area for a long time, infections may have occurred among the local food-handlers. The analysis of water samples taken immediately after the outbreak revealed the presence of bacterial indicators of faecal contamination in the water used in the resort, with the only exception of a sample of filter purified water used both for cooking and as drinking water. When extending the sampling to other areas of the municipal water network, many other sites, mostly on the seaside, showed faecal pollution, which demonstrates a poor protection of the water network from faecal leakage (Fig. 1). NoV contamination of waters generally originates from sewage since these enteric pathogens, as other enteric viruses affecting humans, are mostly species-specific and are shed at high concentrations in the stools of patients. When the analyses on the stools of the students made it clear that the outbreak had been caused by NoV, the viral genome was also searched in appropriate volumes of water samples, but this happened when correction measures on the water network had already been taken. Nevertheless, a single water sample, obtained 2 months after the outbreak (1 August 2016), still contained genome fragments that could be correlated to the same NoV genotype causing the outbreak. Although the finding of NoV genome does not demonstrate that entire infecting viral particles were present in the water, it could indicate that a NoV belonging to the same genotype involved in the outbreak was still circulating in the local population several weeks after the outbreak and it was someway able to penetrate the water network, probably indicating continuing faecal contamination through leakage from sewage systems. NoV is particularly resistant in the environment and also to chlorine, therefore transmission through chlorinated water is not exceptional [9, 12, 26]. Drinking water from municipal supplies fed by lake or groundwater has already been identified in the past as the source of waterborne NoV outbreaks in Italy [27–29] and NoV is now considered one of the pathogens causing the largest number of affected consumers in drinking waterborne outbreaks [12]. NoV can persist in water longer than 15 days [30] and it is resistant to chlorine disinfection when free chlorine levels or pre-treatment processes are inadequate [31, 32]. Moreover, microfiltration filters labelled as certified by NSF Standards 53 or 58 (with a pore size ranging from 0.05 to 5 µm) are not effective in removing viruses and can even decrease chlorine concentration in filtered water [33]. Actually, the owner of the resort involved in our outbreak was encouraged to serve tap water as drinking water at the restaurant relying on the effectiveness of such a filter purifier. Unfortunately, we did not have a tap water sample from the restaurant of the resort taken before correction measures and whose volume was appropriate for virus detection. Therefore, we could not demonstrate the failure to eliminate viral pollution of the microfiltration system. In Mascali, correction measures on the water network consisted in the maintenance of the old chlorination device and in the addition of a new one at a key point located upstream and close to suspected illegal sewage dumping (Fig. 1), but as poor protection of the pipeline from sewage leakage risked to reproduce faecal pollution episodes, renovation of the municipal water system was also advised, as well as systematic control of illegal sewage dumping. The additional chlorination device was apparently the most effective intervention, since right after its activation faecal bacterial pollution indicators disappeared. However, NoV genome was detected in a water sample taken 6 weeks after the last detection of bacterial indicators. Unfortunately, detection of faecal contamination of drinking water is only based on the use of bacterial indicators despite no correlation between levels of enteric bacteria and enteric viruses has been found [34–36]. According to ISGEV enteric virus circulation surveillance data, in 2016 the circulation of GII.P2/GII.2 NoV strains in Palermo, home town of the travellers involved in the outbreak, was the highest ever observed from the starting of local NoV surveillance in 2004. GII.P2/GII.2 was the fourth most common genotype in 2016 (12.2% prevalence), following GII.P4/GII.4 Sydney 2012 (28.4%), GII.P16/GII.2 (18.9%) and GII.Pe/GII.4 Sydney 2012 (17.6%). GII.P2/GII.2 strains were mostly observed starting from July 2016, leaving space for the possibility that their increased circulation was linked to the introduction of the ‘Mascali clone’. However, a couple of GII.P2/GII.2 strains (including PA36/16) genetically close to the ‘Mascali clone’ had been already detected in January 2016 in Palermo and GII.P2/GII.2 NoV had sporadically been observed also in 2011 and 2014. Moreover, the circulation of GII.P2/GII.2 NoVs has been increasing worldwide over the last period [37]. Therefore, the unusually high circulation observed in Palermo might just reflect the global epidemiology of GII.P2/GII.2 strains. Noteworthy, the older GII.P2/GII.2 NoV strains detected in Palermo were closely related in the ORF2 (98.9% nt id) to the GII.P16/GII.2 Nashville strain (KY865307) which is supposed to have generated the recombinant GII.P16/GII.4 clone recently circulating epidemically worldwide [37, 38] and were more distantly associated (95.3% nt id) to the GII.2 capsid sequence (AB662868) of another putative GII.P16/GII.4 ancestor from Japan [39]. The increasing frequency of detection of waterborne NoV outbreaks calls for the introduction of routine monitoring for viruses of drinking waters and for the definition of affordable and reliable indicators for the presence of NoV in drinking waters. The use of ultrafiltration systems as a concentration method in order to recover viruses from environmental waters combined with RT-PCR to detect viral genomes has been demonstrated to grant a recovery rate of >50% of multiple viruses [40] and is at present one of the most sensitive approaches for the detection of NoVs, although the discrimination between viable and nonviable viruses remains a problem. The burden of NoV outbreaks, either food associated or waterborne, is still poorly defined due to the lack of specific diagnostic tools in many clinical laboratories. NoV epidemics should always be suspected when the main symptom is vomiting and the duration of disease is short (<48 h). Waterborne transmission is suggested by the rapid spread of disease and the high attack rate. Care should be taken to perform sensitive diagnostic tests on a sufficient number of stool samples in order to identify NoV as the causative agent of enteritis outbreaks, possibly preferring genome screening over antigen detection. Only the systematic detection of NoV outbreaks can guide in the choice of the safety procedures that could be most appropriate in containing such microbiological hazard.

Which contaminants or viruses or bacteria were found?

1550

Waterborne Norovirus outbreak at a seaside resort likely originating from municipal water distribution system failure

Abstract In May 2016 a Norovirus (NoV) gastroenteritis outbreak involved a high school class visiting a seaside resort near Taormina (Mascali, Sicily). Twenty-four students and a teacher were affected and 17 of them showed symptoms on the second day of the journey, while the others got ill within the following 2 days. Symptoms included vomiting, diarrhoea and fever, and 12 students required hospitalisation. Stool samples tested positive for NoV genome by Real-Time polymerase chain reaction assay in all 25 symptomatic subjects. The GII.P2/GII.2 NoV genotype was linked to the outbreak by ORF1/ORF2 sequence analysis. The epidemiological features of the outbreak were consistent with food/waterborne followed by person-to-person and/ or vomit transmission. Food consumed at a shared lunch on the first day of the trip was associated to illness and drinking un-bottled tap water was also considered as a risk factor. The analysis of water samples revealed the presence of bacterial indicators of faecal contamination in the water used in the resort as well as in other areas of the municipal water network, linking the NoV gastroenteritis outbreak to tap water pollution from sewage leakage. From a single water sample, an amplicon whose sequence corresponded to the capsid genotype recovered from patients could be obtained. Introduction Noroviruses (NoVs) are recognised as a leading cause of sporadic cases and outbreaks of acute gastroenteritis (AGE) across all age groups [1, 2]. They are transmitted primarily through the faecal-oral route either by direct person–person spread and with consumption of contaminated food or water [1, 3, 4]. NoVs are non-enveloped RNA viruses belonging to the Caliciviridae family and are classified into seven distinct genogroups (GI-GVII) on the basis of the major capsid protein VP1 sequence. GI, GII and GIV NoVs infect humans and they have been divided into more than 30 genotypes based on the sequences of polymerase (ORF1) and capsid proteins (ORF2) [5]. NoV GII.4 genotype has been associated with the vast majority of NoV related cases of AGE worldwide although many genotypes, belonging to both GI and GII genogroups are involved in foodborne and waterborne outbreaks [6, 7]. NoVs are highly contagious owing to their high stability in the environment and their ability to infect at low doses. In fact, just 10–100 virions of NoV may be sufficient to cause AGE [8, 9]. NoVs are considered the major cause of water-related disease in semi-closed institutions, including restaurants, schools, hospitals, nursing homes and cruise ships [10–12]. To confirm NoV as a cause of outbreaks, the presence of four epidemiological features of disease, vomiting in more than half of affected persons, a mean incubation period of 24/48 h, a mean duration of illness of 12–60 h and the absence of bacterial pathogens in stool culture, should be evaluated [13]. Waterborne NoV outbreaks can be caused by contaminated drinking water despite correct chlorine concentration and in the absence of coliform bacteria [10, 14–17]. The aim of this study was to investigate an outbreak of NoV gastroenteritis associated with drinking water involving a high school class travelling from Palermo (West Sicily) to Acireale (East Sicily) on an educational tour in May 2016. Stool samples were collected soon after the onset of AGE symptoms to ascertain the aetiology of illness. The source of infection was investigated collecting epidemiological data through an online questionnaire and, in order to explore a possible transmission route, laboratory tests on water from the municipal network were performed. Material and methods Epidemiological investigation From 20 to 22 May 2016 an AGE outbreak occurred in Mascali, a small town near Taormina, involving a school class of 28 students, 13 girls and 15 boys aged between 14 and 18 years and a 54-year-old female teacher, during a trip to participate in a theatre festival in Acireale. A total of 25/29 persons were affected by nausea, vomiting, abdominal cramps, diarrhoea and fever while they were staying at a seaside resort in Mascali, on the way to Acireale. To obtain data on symptoms and the correlated risk factors all participants filled an online questionnaire investigating personal data (age and sex), travel and accommodation details and food consumption over 48 h before symptoms presentation, time of onset and duration of symptoms, therapy and/or hospitalisation required and history of contact with people affected by diarrhoea and/or vomiting (Table 1). The severity of illness was analysed by the Vesikari severity scale allowing to distinguish non-severe (score <11) and severe (score ⩾11) AGE (Table 1). Vesikari score was calculated from maximum number of episodes/day and duration of both diarrhoea and vomiting (score 0–3 for each of the 4 parameters), temperature (score 0–3), level of dehydration (score 0–3) and treatment required (score 0–2) [18]. A case-control study was conducted to establish if food consumed before the onset of symptoms was associated with the disease. A case was defined as a student attending the tour and reporting diarrhoea and/or vomiting within 36 h from the first shared meal during the tour, while students who did not present any of the previous symptoms or those showing the first symptoms with a delay of ⩾48 h (secondary cases) were used as controls. Variables were studied using odds ratio (OR) and its 95% CI and statistical associations were determined using the χ2 test, with the level of statistical significance set at 0.05 (Table 2). NoV detection in stool samples A stool sample was obtained from each of the 29 travellers, soon after (3–5 days) the onset of the outbreak. No stool samples could be obtained from the food-handlers working at the resort. All samples were tested for NoV at the Enteric Viruses Laboratory of the University of Palermo, member of the Italian Study Group for Enteric Viruses (ISGEV), by a Real-Time polymerase chain reaction (RT-PCR) assay that allows to differentiate GI and GII NoVs [19]. Total RNA was extracted from stool samples using QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) and cDNA was generated by MMLV reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were considered positive for NoVs when cycle threshold (ct) of RT-PCR was ⩽30 and lowpositive when 30<ct⩽35. For all NoV-positive samples two genomic fragments were amplified by conventional RT-PCR using Superscript III One-Step RT-PCR system with Platinum Taq DNA polymerase (Invitrogen) and primers JV12a/JV13b, targeting 326 bps of the RdRp region A [20] and either COG1F/ GISKR or COG2F/GIISKR, amplifying a 315 bps 5′portion of GI or GII VP1 protein [19, 21]. Environmental investigation From May through August 2016, a total of 32 tap water samples were collected from the Mascali municipal network. In particular, seven samples were collected at the resort where the outbreak occurred and from its water entry (R), two in the city centre of Mascali (M), 15 in the village of Fondachello (F), on the seaside, three in Carrabba village (C), far from the coast, and two and three, respectively, from the water well ‘Carlino’ (W) and the source ‘Bufardo and Torregrossa’ (S) which feed the municipal system and provide drinking water to the Mascali area (Fig. 1, Table 3). One litre water samples were sent by the local health authorities to the SIALAB laboratory in Avola, Siracusa and 100 ml aliquots were screened for coliforms, E. coli and enterococci as faecal indicators, by standard bacteriological methods according to UNI EN ISO 9308-1:2014. Unfortunately, although the suspect of a water source for the viral outbreak was quickly communicated, only in August and November 2016, following the implementation of control measures, local authorities provided water samplings of appropriate volume to perform viral tests (10 l). Five water samples of 10 l each were collected from different sampling sites in F and R areas for viral screening (Table 3) and were concentrated according to the Italian National Institute of Health (Istituto Superiore di Sanità, ISS) protocols [22] through an ultra-filtration system (Sartoflow® Slice 200 Benchtop Crossflow System, Sartorius AG, Goettingen, Germany), using appropriate membranes (SG Hydrosart 10 kDa) pretreated with 300 ml of 3% Beef Extract (BE) at pH 7, by recirculating it for 10 min in the system, at a pressure not higher than 1.30 bar. Elution and recovery of viruses attached to the membranes were obtained by recirculating 200 ml of 3% BE at pH 9.5 for 3 h until a 10–12 ml volume of eluate was obtained. The pH of the eluate was then checked with litmus paper and if necessary brought to the value of 7 with HCl. Concentrated water samples were simultaneously tested for NoV in three laboratories: Enteric Viruses Laboratory, University of Palermo; Istituto Zooprofilattico Sperimentale della Sicilia (IZS), Palermo; and Istituto Superiore di Sanità (ISS), Rome. Total RNA was extracted from 500 and 560 µl of a concentrated water sample using BioMèrieux NucliSENS® Magnetic Extraction System (BioMèrieux) and QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) respectively, according to manufacturer’s instructions. Bonded RNA was eluted by 100 µl of Elution Buffer. Amplifications were performed at least in duplicate on whole and 1:10 diluted extracts. A Real-Time RT-PCR assay according to ISO/TS 15216- 2:2013, two in-house Real-Time RT-PCR protocols and the One-Step RT-PCR methods already used for amplifying ORF1 and ORF2 fragments in NoV-positive faecal samples were performed [19–21, 23]. Nucleotide sequences analysis and phylogenetic analysis PCR amplicons were purified by commercial columns (CleanSweep PCR Purification, Thermo Fisher Scientific, Vilnius, Lithuania) and sequenced by Sanger method and all nucleotide sequences were aligned using ClustalW. To identify NoV genotypes, RdRp and VP1 sequences were submitted to the RIVM NoV genotyping tool (http://www.rivm.nl/mpf/norovirus/typingtool). Phylogenetic trees were constructed using MEGA 7 and Neighbour-Joining method (bootstrap 1000 replicates). Sequences from stool and water samples are available at NCBI with the following accession numbers: MG517455 for the major capsid protein obtained from a water sample; MG517456 and MG517461, respectively, for the major capsid protein and RNA-dependent-RNA-polymerase, as representatives of the sequences obtained from the stool samples of those involved in the outbreak. Control measures After outbreak notification, on 21 May, the resort was closed for 2 days. On 25 May, following the first results of the bacteriological analyses on municipal water samples, the use of municipal water for human consumption was temporarily restricted (use of water was permitted only for cleaning purposes) in the Carrabba and Fondachello villages and in the area surrounding the resort. From 25 May the reservoir collecting the water from the Carlino well and its chlorination device were submitted to maintenance work and a new chlorination device was installed on the conduit bringing the water from the Bufardo source. Actually, the Bufardo source had been connected to the municipal network in 2003 but unfortunately a dedicated chlorination station had not been planned at that time. Therefore, unchlorinated water entered the system near Fondachello village on the seaside (Fig. 1), possibly diluting chlorine concentrations in water treated at the Carlino reservoir. Moreover, the exact path of the branch of the water network situated downstream the Bufardo source water connection was not marked in the maps of the municipality (Fig. 1, red line), indicating that this section had never been submitted to maintenance. The area surrounding the unmaintained section of the water pipeline was scattered with unauthorised cottages that were not connected to the sewerage systems. Likely, untreated waste from cesspools was unloaded directly into the soil, generating a risk of contamination of the municipal water system through leakage. No work had been made on the water pipelines before the outbreak but sections of the network had been temporarily closed during the winter season on the seaside (Fondachello). Local health authorities advised the connection of all households in the area to the sewerage system and the replacement of the water system section without maintenance. Health authorities also raised attention on AGE cases notification at all local Emergency Departments and Family Practitioners but there was no further evidence of AGE outbreaks in the months following the intervention. Results Data collected through the online questionnaire compiled 7–10 days after the outbreak revealed that none of the 28 students and the teacher taking part in the school trip showed gastroenteritis symptoms at the beginning of the journey and they had no contact with people who suffered from diarrhoea and/or vomiting in the previous days. Travellers had their first shared meal on 19 May (Day 1), 2016, at lunch on their arrival to the resort in Mascali. The list of foods included in the menu is shown in Table 2. Conversely, the dinner on Day 1 was not shared and the students went to dine in three different fast food restaurants in Acireale. On 20 May (Day 2), 17 students (58.6% of the travellers) presented with vomit and/or diarrhoea, falling ill 26–36 h after lunch on Day 1, strongly suggesting that exposure occurred on the first shared meal. Some students also vomited in the presence of the other travellers on the bus bringing them to Acireale. Five more students were symptomatic in the early hours of the following day (21 May, Day 3, from 0:30 am to 8:00 am), as a continuation of the wave of illnesses of the previous day. The teacher and one more student fell ill at 8:00 pm and 11:00 pm of Day 3, respectively, for a second wave of illnesses occurring >48 h from the first shared meal and <36 h from the first cases, with the last student falling sick on 22 May (Day 4) at 3:00 am (Fig. 2, Table 1). On the whole, the mean incubation period was 33.2 h. Based on the online survey, the severity of illness was evaluated by the Vesikari severity scale, grouping symptomatic travellers into severe (11/25 = 44%) and non-severe (14/25 = 56%) cases (Table 1). Vomiting and fever were the predominant symptoms, affecting 22 out of 25 (88%) symptomatic subjects, while diarrhoea was present in 14 out of 25 (56%). Hospitalisation for rehydration therapy was necessary for 13 students (91% [10/11] of severe and 21% [3/14] of non-severe cases). Emergency room (ER) or family practitioner visits were necessary for 21 students and 13 (62%) of them initiated an antibiotic therapy. Two students indicated that other family members had gastroenteric symptoms more than 24 h after their return home, on Day 3, 4 and 5. All those involved in the outbreak recovered within 4 days, with an average duration of illness of 38.4 h. A case-control study of the risk associated with eating foods at lunch on Day 1 or with attending different restaurants at dinner on Day 1 was attempted. A total of 19 subjects filled the section of the online survey concerning food consumption. Fourteen of the responders were among the 22 cases falling ill within 36 h from the first shared meal (first wave), while the remaining five responders were three travellers with supposed secondary infections (showing symptoms with a delay ⩾48 h; second wave) and two asymptomatic students that were used as controls. No statistically significant associations with any food items served at lunch on Day 1 was identified (P < 0.05; Table 2). However, all food consumed on that occasion was to some extent associated with the risk of disease, the highest OR being obtained with chicken (OR: 20.7143, P: 0.0678; Table 2). Actually, most of the cases ate all foods included in the menu, while one of the two asymptomatic controls did not participate in the meal and the other had a pasta dish only. No statistically significant association was found for attending a particular fast food restaurant at the dinner on Day 1 (OR: 2.5, P: 0.3635). The Real-Time RT-PCR assay allowed to detect GII genogroup NoV genome in all the 25 samples from symptomatic subjects, although two of them showed high Ct values, indicating lower shedding of viral genomes (Table 1). The samples from four asymptomatic students were negative for both NoV genogroups. By sequence analysis, a NoV genotype could be defined for 20/ 25 NoV-positive stool samples. All NoV strains from travellers involved in the outbreaks belonged to the same genotype, GII.P2/GII.2, showing >99% nucleotide identities (nt id) in both fragments (ORF1 and ORF2) analysed. Thirty-two water samples were collected and analysed for bacterial contamination (Table 3). Seven samples were taken from the resort (site R) and the remaining 25 from five other sites (C, F, M, S, W), representative of the entire municipal water network. The majority (10/12) of the water samples collected immediately after the onset of the outbreak, between 21 May and 16 June 2016, at the resort (site R) and at sites F and C, located close to the resort, on the seaside and in the interior, respectively, were positive for coliform bacteria. More than 1000 colony-forming units (CFU)/ 100 ml (up to 2700 CFU/100 ml) were detected in four samples, while their absence is required in 100 ml by the Standard of Water Quality in Italy and seven of 11 samples tested were also Escherichia coli-positive (up to 900 CFU/100 ml). Enterococci were searched in 11 samples and three of them were found positive for this faecal contamination indicator. From 7 July to 16 August 2016, after control measures on the water network were undertaken, 13 water samples were collected for follow-up laboratory examination of the water system at F, R and C sites. All of them were negative for bacterial indicators of faecal contamination. No faecal bacteria were identified in the seven samples collected from M, W and S sites, including the well and the source that fed the water network, at any sampling time, before and after implementing the control measures. The research of NoV genome was conducted on concentrated water samples collected on 1 August from R and F sites (Table 3). NoV genome was detected in a single sample collected from the water entry of the resort involved in the outbreak (site R). Actually, only two out of 26 One-Step RT-PCR repetitions were positive for NoV viral capsid gene. The two ORF2-positive results were obtained at the University of Palermo Laboratory using both 1:10 diluted and undiluted extracted RNA. The sequence analysis of the partial ORF2 portion (289 bp) indicated the presence of the same GII.2 capsid genotype detected in the stool samples of the students involved in the outbreak, showing 98.9% nt id. No ORF1 NoV sequence could be obtained from the NoV-positive water sample. Follow-up analyses for NoV genome, conducted on water samples collected on 20 November both at R and F sites found no positive samples. By sequence alignment, the NoV sequences obtained from Mascali outbreak samples were phylogenetically closely related in both ORF1 and ORF2 (99 and 98.9% of nt id, respectively) to a GII.P2/GII.2 strain responsible for a sporadic paediatric gastroenteritis case occurred in Palermo in January 2016 (PA 36/16, accession numbers MG517464 and MG517460, for ORF1 and ORF2, respectively) and previously detected at the Enteric Viruses Laboratory, University of Palermo. Whereas, three GII.P2/GII.2 Italian strains (PA 280/11, PA 366/11 and PA 3/ 14, accession numbers MG517462, MG517465 and MG517463, for ORF1 and MG517457 to −59, for ORF2, respectively), detected in Palermo in 2011 and 2014, were more distantly associated with the outbreak strain both in RdRp and VP1 (90% to 97.4% nt id and 96.1% to 98.1% nt id, respectively, in ORF1 and ORF2). Discussion Epidemiological features, such as the presence of vomiting in more than half (88%) of symptomatic subjects, the 33.2 h average incubation period and the 38.4 h average duration of illness, suggested a NoV aetiology for the investigated outbreak [13]. The detection of NoV genome in all 25 faecal samples from symptomatic subjects fully complies with the requirement of detection of at least one positive sample out of four tested that is considered to be sufficient to ascertain the aetiology of a NoV outbreak [24]. Moreover, molecular typing indicated that a single viral strain, belonging to genotype GII.P2/GII.2, infected all symptomatic subjects involved in the outbreak. NoVs are highly infectious and can easily spread from person to person but also through contaminated food or water. Therefore, settings such as health care institutions, schools and touristic resorts are mainly affected, since close contact and shared food services facilitate transmission. In Mascali, two waves of transmission were observed over three days, being consistent with food/waterborne followed by person-to-person and/or airborne (vomit) transmission (Fig. 2). The first wave of cases presented on 20 May 2016 after a single shared meal, while the last case was detected on 22 May. Exposure to vomit may have been responsible for some of the secondary cases occurred after the first wave of illness. Vomiting produces aerosols that can be inhaled and swallowed. The extremely low infectious dose of NoV makes the inhalation of vomit aerosols an effective transmission route in settings where there is close contact, like on a bus trip [25]. All food consumed in the only shared lunch on 19 May at the resort was associated with the risk of disease (Table 2), but the association was inconclusive since all cases shared the same dishes included in the lunch menu on Day 1. Although many foods were potentially involved, our attention focused on the tap water. In fact, trivial information from local health authorities reported that local population avoided tap water consumption and relied on bottled water for drinking and cooking. Many small outbreaks of gastroenteritis involving restaurants and family groups had been reported in the same area, especially at the beginning of the summer season, when vacation houses and recreational activities were re-opened after a period of low water demand. Illness caused by NoV is common and if the level of water contamination is low, the number of cases remains limited and water can persist as unrecognised source of infection. If polluted water was served for drinking and used for cooking and washing dishes in the Mascali resort at Day 1 a generalised contamination of food and surfaces could have occurred. The potential role of the foodhandlers as a source of contamination could not be analysed in this study but should also be taken into account. If water contamination has already been present in the area for a long time, infections may have occurred among the local food-handlers. The analysis of water samples taken immediately after the outbreak revealed the presence of bacterial indicators of faecal contamination in the water used in the resort, with the only exception of a sample of filter purified water used both for cooking and as drinking water. When extending the sampling to other areas of the municipal water network, many other sites, mostly on the seaside, showed faecal pollution, which demonstrates a poor protection of the water network from faecal leakage (Fig. 1). NoV contamination of waters generally originates from sewage since these enteric pathogens, as other enteric viruses affecting humans, are mostly species-specific and are shed at high concentrations in the stools of patients. When the analyses on the stools of the students made it clear that the outbreak had been caused by NoV, the viral genome was also searched in appropriate volumes of water samples, but this happened when correction measures on the water network had already been taken. Nevertheless, a single water sample, obtained 2 months after the outbreak (1 August 2016), still contained genome fragments that could be correlated to the same NoV genotype causing the outbreak. Although the finding of NoV genome does not demonstrate that entire infecting viral particles were present in the water, it could indicate that a NoV belonging to the same genotype involved in the outbreak was still circulating in the local population several weeks after the outbreak and it was someway able to penetrate the water network, probably indicating continuing faecal contamination through leakage from sewage systems. NoV is particularly resistant in the environment and also to chlorine, therefore transmission through chlorinated water is not exceptional [9, 12, 26]. Drinking water from municipal supplies fed by lake or groundwater has already been identified in the past as the source of waterborne NoV outbreaks in Italy [27–29] and NoV is now considered one of the pathogens causing the largest number of affected consumers in drinking waterborne outbreaks [12]. NoV can persist in water longer than 15 days [30] and it is resistant to chlorine disinfection when free chlorine levels or pre-treatment processes are inadequate [31, 32]. Moreover, microfiltration filters labelled as certified by NSF Standards 53 or 58 (with a pore size ranging from 0.05 to 5 µm) are not effective in removing viruses and can even decrease chlorine concentration in filtered water [33]. Actually, the owner of the resort involved in our outbreak was encouraged to serve tap water as drinking water at the restaurant relying on the effectiveness of such a filter purifier. Unfortunately, we did not have a tap water sample from the restaurant of the resort taken before correction measures and whose volume was appropriate for virus detection. Therefore, we could not demonstrate the failure to eliminate viral pollution of the microfiltration system. In Mascali, correction measures on the water network consisted in the maintenance of the old chlorination device and in the addition of a new one at a key point located upstream and close to suspected illegal sewage dumping (Fig. 1), but as poor protection of the pipeline from sewage leakage risked to reproduce faecal pollution episodes, renovation of the municipal water system was also advised, as well as systematic control of illegal sewage dumping. The additional chlorination device was apparently the most effective intervention, since right after its activation faecal bacterial pollution indicators disappeared. However, NoV genome was detected in a water sample taken 6 weeks after the last detection of bacterial indicators. Unfortunately, detection of faecal contamination of drinking water is only based on the use of bacterial indicators despite no correlation between levels of enteric bacteria and enteric viruses has been found [34–36]. According to ISGEV enteric virus circulation surveillance data, in 2016 the circulation of GII.P2/GII.2 NoV strains in Palermo, home town of the travellers involved in the outbreak, was the highest ever observed from the starting of local NoV surveillance in 2004. GII.P2/GII.2 was the fourth most common genotype in 2016 (12.2% prevalence), following GII.P4/GII.4 Sydney 2012 (28.4%), GII.P16/GII.2 (18.9%) and GII.Pe/GII.4 Sydney 2012 (17.6%). GII.P2/GII.2 strains were mostly observed starting from July 2016, leaving space for the possibility that their increased circulation was linked to the introduction of the ‘Mascali clone’. However, a couple of GII.P2/GII.2 strains (including PA36/16) genetically close to the ‘Mascali clone’ had been already detected in January 2016 in Palermo and GII.P2/GII.2 NoV had sporadically been observed also in 2011 and 2014. Moreover, the circulation of GII.P2/GII.2 NoVs has been increasing worldwide over the last period [37]. Therefore, the unusually high circulation observed in Palermo might just reflect the global epidemiology of GII.P2/GII.2 strains. Noteworthy, the older GII.P2/GII.2 NoV strains detected in Palermo were closely related in the ORF2 (98.9% nt id) to the GII.P16/GII.2 Nashville strain (KY865307) which is supposed to have generated the recombinant GII.P16/GII.4 clone recently circulating epidemically worldwide [37, 38] and were more distantly associated (95.3% nt id) to the GII.2 capsid sequence (AB662868) of another putative GII.P16/GII.4 ancestor from Japan [39]. The increasing frequency of detection of waterborne NoV outbreaks calls for the introduction of routine monitoring for viruses of drinking waters and for the definition of affordable and reliable indicators for the presence of NoV in drinking waters. The use of ultrafiltration systems as a concentration method in order to recover viruses from environmental waters combined with RT-PCR to detect viral genomes has been demonstrated to grant a recovery rate of >50% of multiple viruses [40] and is at present one of the most sensitive approaches for the detection of NoVs, although the discrimination between viable and nonviable viruses remains a problem. The burden of NoV outbreaks, either food associated or waterborne, is still poorly defined due to the lack of specific diagnostic tools in many clinical laboratories. NoV epidemics should always be suspected when the main symptom is vomiting and the duration of disease is short (<48 h). Waterborne transmission is suggested by the rapid spread of disease and the high attack rate. Care should be taken to perform sensitive diagnostic tests on a sufficient number of stool samples in order to identify NoV as the causative agent of enteritis outbreaks, possibly preferring genome screening over antigen detection. Only the systematic detection of NoV outbreaks can guide in the choice of the safety procedures that could be most appropriate in containing such microbiological hazard.

Which were the symptoms?

1551

Waterborne Norovirus outbreak at a seaside resort likely originating from municipal water distribution system failure

Abstract In May 2016 a Norovirus (NoV) gastroenteritis outbreak involved a high school class visiting a seaside resort near Taormina (Mascali, Sicily). Twenty-four students and a teacher were affected and 17 of them showed symptoms on the second day of the journey, while the others got ill within the following 2 days. Symptoms included vomiting, diarrhoea and fever, and 12 students required hospitalisation. Stool samples tested positive for NoV genome by Real-Time polymerase chain reaction assay in all 25 symptomatic subjects. The GII.P2/GII.2 NoV genotype was linked to the outbreak by ORF1/ORF2 sequence analysis. The epidemiological features of the outbreak were consistent with food/waterborne followed by person-to-person and/ or vomit transmission. Food consumed at a shared lunch on the first day of the trip was associated to illness and drinking un-bottled tap water was also considered as a risk factor. The analysis of water samples revealed the presence of bacterial indicators of faecal contamination in the water used in the resort as well as in other areas of the municipal water network, linking the NoV gastroenteritis outbreak to tap water pollution from sewage leakage. From a single water sample, an amplicon whose sequence corresponded to the capsid genotype recovered from patients could be obtained. Introduction Noroviruses (NoVs) are recognised as a leading cause of sporadic cases and outbreaks of acute gastroenteritis (AGE) across all age groups [1, 2]. They are transmitted primarily through the faecal-oral route either by direct person–person spread and with consumption of contaminated food or water [1, 3, 4]. NoVs are non-enveloped RNA viruses belonging to the Caliciviridae family and are classified into seven distinct genogroups (GI-GVII) on the basis of the major capsid protein VP1 sequence. GI, GII and GIV NoVs infect humans and they have been divided into more than 30 genotypes based on the sequences of polymerase (ORF1) and capsid proteins (ORF2) [5]. NoV GII.4 genotype has been associated with the vast majority of NoV related cases of AGE worldwide although many genotypes, belonging to both GI and GII genogroups are involved in foodborne and waterborne outbreaks [6, 7]. NoVs are highly contagious owing to their high stability in the environment and their ability to infect at low doses. In fact, just 10–100 virions of NoV may be sufficient to cause AGE [8, 9]. NoVs are considered the major cause of water-related disease in semi-closed institutions, including restaurants, schools, hospitals, nursing homes and cruise ships [10–12]. To confirm NoV as a cause of outbreaks, the presence of four epidemiological features of disease, vomiting in more than half of affected persons, a mean incubation period of 24/48 h, a mean duration of illness of 12–60 h and the absence of bacterial pathogens in stool culture, should be evaluated [13]. Waterborne NoV outbreaks can be caused by contaminated drinking water despite correct chlorine concentration and in the absence of coliform bacteria [10, 14–17]. The aim of this study was to investigate an outbreak of NoV gastroenteritis associated with drinking water involving a high school class travelling from Palermo (West Sicily) to Acireale (East Sicily) on an educational tour in May 2016. Stool samples were collected soon after the onset of AGE symptoms to ascertain the aetiology of illness. The source of infection was investigated collecting epidemiological data through an online questionnaire and, in order to explore a possible transmission route, laboratory tests on water from the municipal network were performed. Material and methods Epidemiological investigation From 20 to 22 May 2016 an AGE outbreak occurred in Mascali, a small town near Taormina, involving a school class of 28 students, 13 girls and 15 boys aged between 14 and 18 years and a 54-year-old female teacher, during a trip to participate in a theatre festival in Acireale. A total of 25/29 persons were affected by nausea, vomiting, abdominal cramps, diarrhoea and fever while they were staying at a seaside resort in Mascali, on the way to Acireale. To obtain data on symptoms and the correlated risk factors all participants filled an online questionnaire investigating personal data (age and sex), travel and accommodation details and food consumption over 48 h before symptoms presentation, time of onset and duration of symptoms, therapy and/or hospitalisation required and history of contact with people affected by diarrhoea and/or vomiting (Table 1). The severity of illness was analysed by the Vesikari severity scale allowing to distinguish non-severe (score <11) and severe (score ⩾11) AGE (Table 1). Vesikari score was calculated from maximum number of episodes/day and duration of both diarrhoea and vomiting (score 0–3 for each of the 4 parameters), temperature (score 0–3), level of dehydration (score 0–3) and treatment required (score 0–2) [18]. A case-control study was conducted to establish if food consumed before the onset of symptoms was associated with the disease. A case was defined as a student attending the tour and reporting diarrhoea and/or vomiting within 36 h from the first shared meal during the tour, while students who did not present any of the previous symptoms or those showing the first symptoms with a delay of ⩾48 h (secondary cases) were used as controls. Variables were studied using odds ratio (OR) and its 95% CI and statistical associations were determined using the χ2 test, with the level of statistical significance set at 0.05 (Table 2). NoV detection in stool samples A stool sample was obtained from each of the 29 travellers, soon after (3–5 days) the onset of the outbreak. No stool samples could be obtained from the food-handlers working at the resort. All samples were tested for NoV at the Enteric Viruses Laboratory of the University of Palermo, member of the Italian Study Group for Enteric Viruses (ISGEV), by a Real-Time polymerase chain reaction (RT-PCR) assay that allows to differentiate GI and GII NoVs [19]. Total RNA was extracted from stool samples using QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) and cDNA was generated by MMLV reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were considered positive for NoVs when cycle threshold (ct) of RT-PCR was ⩽30 and lowpositive when 30<ct⩽35. For all NoV-positive samples two genomic fragments were amplified by conventional RT-PCR using Superscript III One-Step RT-PCR system with Platinum Taq DNA polymerase (Invitrogen) and primers JV12a/JV13b, targeting 326 bps of the RdRp region A [20] and either COG1F/ GISKR or COG2F/GIISKR, amplifying a 315 bps 5′portion of GI or GII VP1 protein [19, 21]. Environmental investigation From May through August 2016, a total of 32 tap water samples were collected from the Mascali municipal network. In particular, seven samples were collected at the resort where the outbreak occurred and from its water entry (R), two in the city centre of Mascali (M), 15 in the village of Fondachello (F), on the seaside, three in Carrabba village (C), far from the coast, and two and three, respectively, from the water well ‘Carlino’ (W) and the source ‘Bufardo and Torregrossa’ (S) which feed the municipal system and provide drinking water to the Mascali area (Fig. 1, Table 3). One litre water samples were sent by the local health authorities to the SIALAB laboratory in Avola, Siracusa and 100 ml aliquots were screened for coliforms, E. coli and enterococci as faecal indicators, by standard bacteriological methods according to UNI EN ISO 9308-1:2014. Unfortunately, although the suspect of a water source for the viral outbreak was quickly communicated, only in August and November 2016, following the implementation of control measures, local authorities provided water samplings of appropriate volume to perform viral tests (10 l). Five water samples of 10 l each were collected from different sampling sites in F and R areas for viral screening (Table 3) and were concentrated according to the Italian National Institute of Health (Istituto Superiore di Sanità, ISS) protocols [22] through an ultra-filtration system (Sartoflow® Slice 200 Benchtop Crossflow System, Sartorius AG, Goettingen, Germany), using appropriate membranes (SG Hydrosart 10 kDa) pretreated with 300 ml of 3% Beef Extract (BE) at pH 7, by recirculating it for 10 min in the system, at a pressure not higher than 1.30 bar. Elution and recovery of viruses attached to the membranes were obtained by recirculating 200 ml of 3% BE at pH 9.5 for 3 h until a 10–12 ml volume of eluate was obtained. The pH of the eluate was then checked with litmus paper and if necessary brought to the value of 7 with HCl. Concentrated water samples were simultaneously tested for NoV in three laboratories: Enteric Viruses Laboratory, University of Palermo; Istituto Zooprofilattico Sperimentale della Sicilia (IZS), Palermo; and Istituto Superiore di Sanità (ISS), Rome. Total RNA was extracted from 500 and 560 µl of a concentrated water sample using BioMèrieux NucliSENS® Magnetic Extraction System (BioMèrieux) and QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) respectively, according to manufacturer’s instructions. Bonded RNA was eluted by 100 µl of Elution Buffer. Amplifications were performed at least in duplicate on whole and 1:10 diluted extracts. A Real-Time RT-PCR assay according to ISO/TS 15216- 2:2013, two in-house Real-Time RT-PCR protocols and the One-Step RT-PCR methods already used for amplifying ORF1 and ORF2 fragments in NoV-positive faecal samples were performed [19–21, 23]. Nucleotide sequences analysis and phylogenetic analysis PCR amplicons were purified by commercial columns (CleanSweep PCR Purification, Thermo Fisher Scientific, Vilnius, Lithuania) and sequenced by Sanger method and all nucleotide sequences were aligned using ClustalW. To identify NoV genotypes, RdRp and VP1 sequences were submitted to the RIVM NoV genotyping tool (http://www.rivm.nl/mpf/norovirus/typingtool). Phylogenetic trees were constructed using MEGA 7 and Neighbour-Joining method (bootstrap 1000 replicates). Sequences from stool and water samples are available at NCBI with the following accession numbers: MG517455 for the major capsid protein obtained from a water sample; MG517456 and MG517461, respectively, for the major capsid protein and RNA-dependent-RNA-polymerase, as representatives of the sequences obtained from the stool samples of those involved in the outbreak. Control measures After outbreak notification, on 21 May, the resort was closed for 2 days. On 25 May, following the first results of the bacteriological analyses on municipal water samples, the use of municipal water for human consumption was temporarily restricted (use of water was permitted only for cleaning purposes) in the Carrabba and Fondachello villages and in the area surrounding the resort. From 25 May the reservoir collecting the water from the Carlino well and its chlorination device were submitted to maintenance work and a new chlorination device was installed on the conduit bringing the water from the Bufardo source. Actually, the Bufardo source had been connected to the municipal network in 2003 but unfortunately a dedicated chlorination station had not been planned at that time. Therefore, unchlorinated water entered the system near Fondachello village on the seaside (Fig. 1), possibly diluting chlorine concentrations in water treated at the Carlino reservoir. Moreover, the exact path of the branch of the water network situated downstream the Bufardo source water connection was not marked in the maps of the municipality (Fig. 1, red line), indicating that this section had never been submitted to maintenance. The area surrounding the unmaintained section of the water pipeline was scattered with unauthorised cottages that were not connected to the sewerage systems. Likely, untreated waste from cesspools was unloaded directly into the soil, generating a risk of contamination of the municipal water system through leakage. No work had been made on the water pipelines before the outbreak but sections of the network had been temporarily closed during the winter season on the seaside (Fondachello). Local health authorities advised the connection of all households in the area to the sewerage system and the replacement of the water system section without maintenance. Health authorities also raised attention on AGE cases notification at all local Emergency Departments and Family Practitioners but there was no further evidence of AGE outbreaks in the months following the intervention. Results Data collected through the online questionnaire compiled 7–10 days after the outbreak revealed that none of the 28 students and the teacher taking part in the school trip showed gastroenteritis symptoms at the beginning of the journey and they had no contact with people who suffered from diarrhoea and/or vomiting in the previous days. Travellers had their first shared meal on 19 May (Day 1), 2016, at lunch on their arrival to the resort in Mascali. The list of foods included in the menu is shown in Table 2. Conversely, the dinner on Day 1 was not shared and the students went to dine in three different fast food restaurants in Acireale. On 20 May (Day 2), 17 students (58.6% of the travellers) presented with vomit and/or diarrhoea, falling ill 26–36 h after lunch on Day 1, strongly suggesting that exposure occurred on the first shared meal. Some students also vomited in the presence of the other travellers on the bus bringing them to Acireale. Five more students were symptomatic in the early hours of the following day (21 May, Day 3, from 0:30 am to 8:00 am), as a continuation of the wave of illnesses of the previous day. The teacher and one more student fell ill at 8:00 pm and 11:00 pm of Day 3, respectively, for a second wave of illnesses occurring >48 h from the first shared meal and <36 h from the first cases, with the last student falling sick on 22 May (Day 4) at 3:00 am (Fig. 2, Table 1). On the whole, the mean incubation period was 33.2 h. Based on the online survey, the severity of illness was evaluated by the Vesikari severity scale, grouping symptomatic travellers into severe (11/25 = 44%) and non-severe (14/25 = 56%) cases (Table 1). Vomiting and fever were the predominant symptoms, affecting 22 out of 25 (88%) symptomatic subjects, while diarrhoea was present in 14 out of 25 (56%). Hospitalisation for rehydration therapy was necessary for 13 students (91% [10/11] of severe and 21% [3/14] of non-severe cases). Emergency room (ER) or family practitioner visits were necessary for 21 students and 13 (62%) of them initiated an antibiotic therapy. Two students indicated that other family members had gastroenteric symptoms more than 24 h after their return home, on Day 3, 4 and 5. All those involved in the outbreak recovered within 4 days, with an average duration of illness of 38.4 h. A case-control study of the risk associated with eating foods at lunch on Day 1 or with attending different restaurants at dinner on Day 1 was attempted. A total of 19 subjects filled the section of the online survey concerning food consumption. Fourteen of the responders were among the 22 cases falling ill within 36 h from the first shared meal (first wave), while the remaining five responders were three travellers with supposed secondary infections (showing symptoms with a delay ⩾48 h; second wave) and two asymptomatic students that were used as controls. No statistically significant associations with any food items served at lunch on Day 1 was identified (P < 0.05; Table 2). However, all food consumed on that occasion was to some extent associated with the risk of disease, the highest OR being obtained with chicken (OR: 20.7143, P: 0.0678; Table 2). Actually, most of the cases ate all foods included in the menu, while one of the two asymptomatic controls did not participate in the meal and the other had a pasta dish only. No statistically significant association was found for attending a particular fast food restaurant at the dinner on Day 1 (OR: 2.5, P: 0.3635). The Real-Time RT-PCR assay allowed to detect GII genogroup NoV genome in all the 25 samples from symptomatic subjects, although two of them showed high Ct values, indicating lower shedding of viral genomes (Table 1). The samples from four asymptomatic students were negative for both NoV genogroups. By sequence analysis, a NoV genotype could be defined for 20/ 25 NoV-positive stool samples. All NoV strains from travellers involved in the outbreaks belonged to the same genotype, GII.P2/GII.2, showing >99% nucleotide identities (nt id) in both fragments (ORF1 and ORF2) analysed. Thirty-two water samples were collected and analysed for bacterial contamination (Table 3). Seven samples were taken from the resort (site R) and the remaining 25 from five other sites (C, F, M, S, W), representative of the entire municipal water network. The majority (10/12) of the water samples collected immediately after the onset of the outbreak, between 21 May and 16 June 2016, at the resort (site R) and at sites F and C, located close to the resort, on the seaside and in the interior, respectively, were positive for coliform bacteria. More than 1000 colony-forming units (CFU)/ 100 ml (up to 2700 CFU/100 ml) were detected in four samples, while their absence is required in 100 ml by the Standard of Water Quality in Italy and seven of 11 samples tested were also Escherichia coli-positive (up to 900 CFU/100 ml). Enterococci were searched in 11 samples and three of them were found positive for this faecal contamination indicator. From 7 July to 16 August 2016, after control measures on the water network were undertaken, 13 water samples were collected for follow-up laboratory examination of the water system at F, R and C sites. All of them were negative for bacterial indicators of faecal contamination. No faecal bacteria were identified in the seven samples collected from M, W and S sites, including the well and the source that fed the water network, at any sampling time, before and after implementing the control measures. The research of NoV genome was conducted on concentrated water samples collected on 1 August from R and F sites (Table 3). NoV genome was detected in a single sample collected from the water entry of the resort involved in the outbreak (site R). Actually, only two out of 26 One-Step RT-PCR repetitions were positive for NoV viral capsid gene. The two ORF2-positive results were obtained at the University of Palermo Laboratory using both 1:10 diluted and undiluted extracted RNA. The sequence analysis of the partial ORF2 portion (289 bp) indicated the presence of the same GII.2 capsid genotype detected in the stool samples of the students involved in the outbreak, showing 98.9% nt id. No ORF1 NoV sequence could be obtained from the NoV-positive water sample. Follow-up analyses for NoV genome, conducted on water samples collected on 20 November both at R and F sites found no positive samples. By sequence alignment, the NoV sequences obtained from Mascali outbreak samples were phylogenetically closely related in both ORF1 and ORF2 (99 and 98.9% of nt id, respectively) to a GII.P2/GII.2 strain responsible for a sporadic paediatric gastroenteritis case occurred in Palermo in January 2016 (PA 36/16, accession numbers MG517464 and MG517460, for ORF1 and ORF2, respectively) and previously detected at the Enteric Viruses Laboratory, University of Palermo. Whereas, three GII.P2/GII.2 Italian strains (PA 280/11, PA 366/11 and PA 3/ 14, accession numbers MG517462, MG517465 and MG517463, for ORF1 and MG517457 to −59, for ORF2, respectively), detected in Palermo in 2011 and 2014, were more distantly associated with the outbreak strain both in RdRp and VP1 (90% to 97.4% nt id and 96.1% to 98.1% nt id, respectively, in ORF1 and ORF2). Discussion Epidemiological features, such as the presence of vomiting in more than half (88%) of symptomatic subjects, the 33.2 h average incubation period and the 38.4 h average duration of illness, suggested a NoV aetiology for the investigated outbreak [13]. The detection of NoV genome in all 25 faecal samples from symptomatic subjects fully complies with the requirement of detection of at least one positive sample out of four tested that is considered to be sufficient to ascertain the aetiology of a NoV outbreak [24]. Moreover, molecular typing indicated that a single viral strain, belonging to genotype GII.P2/GII.2, infected all symptomatic subjects involved in the outbreak. NoVs are highly infectious and can easily spread from person to person but also through contaminated food or water. Therefore, settings such as health care institutions, schools and touristic resorts are mainly affected, since close contact and shared food services facilitate transmission. In Mascali, two waves of transmission were observed over three days, being consistent with food/waterborne followed by person-to-person and/or airborne (vomit) transmission (Fig. 2). The first wave of cases presented on 20 May 2016 after a single shared meal, while the last case was detected on 22 May. Exposure to vomit may have been responsible for some of the secondary cases occurred after the first wave of illness. Vomiting produces aerosols that can be inhaled and swallowed. The extremely low infectious dose of NoV makes the inhalation of vomit aerosols an effective transmission route in settings where there is close contact, like on a bus trip [25]. All food consumed in the only shared lunch on 19 May at the resort was associated with the risk of disease (Table 2), but the association was inconclusive since all cases shared the same dishes included in the lunch menu on Day 1. Although many foods were potentially involved, our attention focused on the tap water. In fact, trivial information from local health authorities reported that local population avoided tap water consumption and relied on bottled water for drinking and cooking. Many small outbreaks of gastroenteritis involving restaurants and family groups had been reported in the same area, especially at the beginning of the summer season, when vacation houses and recreational activities were re-opened after a period of low water demand. Illness caused by NoV is common and if the level of water contamination is low, the number of cases remains limited and water can persist as unrecognised source of infection. If polluted water was served for drinking and used for cooking and washing dishes in the Mascali resort at Day 1 a generalised contamination of food and surfaces could have occurred. The potential role of the foodhandlers as a source of contamination could not be analysed in this study but should also be taken into account. If water contamination has already been present in the area for a long time, infections may have occurred among the local food-handlers. The analysis of water samples taken immediately after the outbreak revealed the presence of bacterial indicators of faecal contamination in the water used in the resort, with the only exception of a sample of filter purified water used both for cooking and as drinking water. When extending the sampling to other areas of the municipal water network, many other sites, mostly on the seaside, showed faecal pollution, which demonstrates a poor protection of the water network from faecal leakage (Fig. 1). NoV contamination of waters generally originates from sewage since these enteric pathogens, as other enteric viruses affecting humans, are mostly species-specific and are shed at high concentrations in the stools of patients. When the analyses on the stools of the students made it clear that the outbreak had been caused by NoV, the viral genome was also searched in appropriate volumes of water samples, but this happened when correction measures on the water network had already been taken. Nevertheless, a single water sample, obtained 2 months after the outbreak (1 August 2016), still contained genome fragments that could be correlated to the same NoV genotype causing the outbreak. Although the finding of NoV genome does not demonstrate that entire infecting viral particles were present in the water, it could indicate that a NoV belonging to the same genotype involved in the outbreak was still circulating in the local population several weeks after the outbreak and it was someway able to penetrate the water network, probably indicating continuing faecal contamination through leakage from sewage systems. NoV is particularly resistant in the environment and also to chlorine, therefore transmission through chlorinated water is not exceptional [9, 12, 26]. Drinking water from municipal supplies fed by lake or groundwater has already been identified in the past as the source of waterborne NoV outbreaks in Italy [27–29] and NoV is now considered one of the pathogens causing the largest number of affected consumers in drinking waterborne outbreaks [12]. NoV can persist in water longer than 15 days [30] and it is resistant to chlorine disinfection when free chlorine levels or pre-treatment processes are inadequate [31, 32]. Moreover, microfiltration filters labelled as certified by NSF Standards 53 or 58 (with a pore size ranging from 0.05 to 5 µm) are not effective in removing viruses and can even decrease chlorine concentration in filtered water [33]. Actually, the owner of the resort involved in our outbreak was encouraged to serve tap water as drinking water at the restaurant relying on the effectiveness of such a filter purifier. Unfortunately, we did not have a tap water sample from the restaurant of the resort taken before correction measures and whose volume was appropriate for virus detection. Therefore, we could not demonstrate the failure to eliminate viral pollution of the microfiltration system. In Mascali, correction measures on the water network consisted in the maintenance of the old chlorination device and in the addition of a new one at a key point located upstream and close to suspected illegal sewage dumping (Fig. 1), but as poor protection of the pipeline from sewage leakage risked to reproduce faecal pollution episodes, renovation of the municipal water system was also advised, as well as systematic control of illegal sewage dumping. The additional chlorination device was apparently the most effective intervention, since right after its activation faecal bacterial pollution indicators disappeared. However, NoV genome was detected in a water sample taken 6 weeks after the last detection of bacterial indicators. Unfortunately, detection of faecal contamination of drinking water is only based on the use of bacterial indicators despite no correlation between levels of enteric bacteria and enteric viruses has been found [34–36]. According to ISGEV enteric virus circulation surveillance data, in 2016 the circulation of GII.P2/GII.2 NoV strains in Palermo, home town of the travellers involved in the outbreak, was the highest ever observed from the starting of local NoV surveillance in 2004. GII.P2/GII.2 was the fourth most common genotype in 2016 (12.2% prevalence), following GII.P4/GII.4 Sydney 2012 (28.4%), GII.P16/GII.2 (18.9%) and GII.Pe/GII.4 Sydney 2012 (17.6%). GII.P2/GII.2 strains were mostly observed starting from July 2016, leaving space for the possibility that their increased circulation was linked to the introduction of the ‘Mascali clone’. However, a couple of GII.P2/GII.2 strains (including PA36/16) genetically close to the ‘Mascali clone’ had been already detected in January 2016 in Palermo and GII.P2/GII.2 NoV had sporadically been observed also in 2011 and 2014. Moreover, the circulation of GII.P2/GII.2 NoVs has been increasing worldwide over the last period [37]. Therefore, the unusually high circulation observed in Palermo might just reflect the global epidemiology of GII.P2/GII.2 strains. Noteworthy, the older GII.P2/GII.2 NoV strains detected in Palermo were closely related in the ORF2 (98.9% nt id) to the GII.P16/GII.2 Nashville strain (KY865307) which is supposed to have generated the recombinant GII.P16/GII.4 clone recently circulating epidemically worldwide [37, 38] and were more distantly associated (95.3% nt id) to the GII.2 capsid sequence (AB662868) of another putative GII.P16/GII.4 ancestor from Japan [39]. The increasing frequency of detection of waterborne NoV outbreaks calls for the introduction of routine monitoring for viruses of drinking waters and for the definition of affordable and reliable indicators for the presence of NoV in drinking waters. The use of ultrafiltration systems as a concentration method in order to recover viruses from environmental waters combined with RT-PCR to detect viral genomes has been demonstrated to grant a recovery rate of >50% of multiple viruses [40] and is at present one of the most sensitive approaches for the detection of NoVs, although the discrimination between viable and nonviable viruses remains a problem. The burden of NoV outbreaks, either food associated or waterborne, is still poorly defined due to the lack of specific diagnostic tools in many clinical laboratories. NoV epidemics should always be suspected when the main symptom is vomiting and the duration of disease is short (<48 h). Waterborne transmission is suggested by the rapid spread of disease and the high attack rate. Care should be taken to perform sensitive diagnostic tests on a sufficient number of stool samples in order to identify NoV as the causative agent of enteritis outbreaks, possibly preferring genome screening over antigen detection. Only the systematic detection of NoV outbreaks can guide in the choice of the safety procedures that could be most appropriate in containing such microbiological hazard.

What did the patients have?

1552

Waterborne Norovirus outbreak at a seaside resort likely originating from municipal water distribution system failure

Abstract In May 2016 a Norovirus (NoV) gastroenteritis outbreak involved a high school class visiting a seaside resort near Taormina (Mascali, Sicily). Twenty-four students and a teacher were affected and 17 of them showed symptoms on the second day of the journey, while the others got ill within the following 2 days. Symptoms included vomiting, diarrhoea and fever, and 12 students required hospitalisation. Stool samples tested positive for NoV genome by Real-Time polymerase chain reaction assay in all 25 symptomatic subjects. The GII.P2/GII.2 NoV genotype was linked to the outbreak by ORF1/ORF2 sequence analysis. The epidemiological features of the outbreak were consistent with food/waterborne followed by person-to-person and/ or vomit transmission. Food consumed at a shared lunch on the first day of the trip was associated to illness and drinking un-bottled tap water was also considered as a risk factor. The analysis of water samples revealed the presence of bacterial indicators of faecal contamination in the water used in the resort as well as in other areas of the municipal water network, linking the NoV gastroenteritis outbreak to tap water pollution from sewage leakage. From a single water sample, an amplicon whose sequence corresponded to the capsid genotype recovered from patients could be obtained. Introduction Noroviruses (NoVs) are recognised as a leading cause of sporadic cases and outbreaks of acute gastroenteritis (AGE) across all age groups [1, 2]. They are transmitted primarily through the faecal-oral route either by direct person–person spread and with consumption of contaminated food or water [1, 3, 4]. NoVs are non-enveloped RNA viruses belonging to the Caliciviridae family and are classified into seven distinct genogroups (GI-GVII) on the basis of the major capsid protein VP1 sequence. GI, GII and GIV NoVs infect humans and they have been divided into more than 30 genotypes based on the sequences of polymerase (ORF1) and capsid proteins (ORF2) [5]. NoV GII.4 genotype has been associated with the vast majority of NoV related cases of AGE worldwide although many genotypes, belonging to both GI and GII genogroups are involved in foodborne and waterborne outbreaks [6, 7]. NoVs are highly contagious owing to their high stability in the environment and their ability to infect at low doses. In fact, just 10–100 virions of NoV may be sufficient to cause AGE [8, 9]. NoVs are considered the major cause of water-related disease in semi-closed institutions, including restaurants, schools, hospitals, nursing homes and cruise ships [10–12]. To confirm NoV as a cause of outbreaks, the presence of four epidemiological features of disease, vomiting in more than half of affected persons, a mean incubation period of 24/48 h, a mean duration of illness of 12–60 h and the absence of bacterial pathogens in stool culture, should be evaluated [13]. Waterborne NoV outbreaks can be caused by contaminated drinking water despite correct chlorine concentration and in the absence of coliform bacteria [10, 14–17]. The aim of this study was to investigate an outbreak of NoV gastroenteritis associated with drinking water involving a high school class travelling from Palermo (West Sicily) to Acireale (East Sicily) on an educational tour in May 2016. Stool samples were collected soon after the onset of AGE symptoms to ascertain the aetiology of illness. The source of infection was investigated collecting epidemiological data through an online questionnaire and, in order to explore a possible transmission route, laboratory tests on water from the municipal network were performed. Material and methods Epidemiological investigation From 20 to 22 May 2016 an AGE outbreak occurred in Mascali, a small town near Taormina, involving a school class of 28 students, 13 girls and 15 boys aged between 14 and 18 years and a 54-year-old female teacher, during a trip to participate in a theatre festival in Acireale. A total of 25/29 persons were affected by nausea, vomiting, abdominal cramps, diarrhoea and fever while they were staying at a seaside resort in Mascali, on the way to Acireale. To obtain data on symptoms and the correlated risk factors all participants filled an online questionnaire investigating personal data (age and sex), travel and accommodation details and food consumption over 48 h before symptoms presentation, time of onset and duration of symptoms, therapy and/or hospitalisation required and history of contact with people affected by diarrhoea and/or vomiting (Table 1). The severity of illness was analysed by the Vesikari severity scale allowing to distinguish non-severe (score <11) and severe (score ⩾11) AGE (Table 1). Vesikari score was calculated from maximum number of episodes/day and duration of both diarrhoea and vomiting (score 0–3 for each of the 4 parameters), temperature (score 0–3), level of dehydration (score 0–3) and treatment required (score 0–2) [18]. A case-control study was conducted to establish if food consumed before the onset of symptoms was associated with the disease. A case was defined as a student attending the tour and reporting diarrhoea and/or vomiting within 36 h from the first shared meal during the tour, while students who did not present any of the previous symptoms or those showing the first symptoms with a delay of ⩾48 h (secondary cases) were used as controls. Variables were studied using odds ratio (OR) and its 95% CI and statistical associations were determined using the χ2 test, with the level of statistical significance set at 0.05 (Table 2). NoV detection in stool samples A stool sample was obtained from each of the 29 travellers, soon after (3–5 days) the onset of the outbreak. No stool samples could be obtained from the food-handlers working at the resort. All samples were tested for NoV at the Enteric Viruses Laboratory of the University of Palermo, member of the Italian Study Group for Enteric Viruses (ISGEV), by a Real-Time polymerase chain reaction (RT-PCR) assay that allows to differentiate GI and GII NoVs [19]. Total RNA was extracted from stool samples using QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) and cDNA was generated by MMLV reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were considered positive for NoVs when cycle threshold (ct) of RT-PCR was ⩽30 and lowpositive when 30<ct⩽35. For all NoV-positive samples two genomic fragments were amplified by conventional RT-PCR using Superscript III One-Step RT-PCR system with Platinum Taq DNA polymerase (Invitrogen) and primers JV12a/JV13b, targeting 326 bps of the RdRp region A [20] and either COG1F/ GISKR or COG2F/GIISKR, amplifying a 315 bps 5′portion of GI or GII VP1 protein [19, 21]. Environmental investigation From May through August 2016, a total of 32 tap water samples were collected from the Mascali municipal network. In particular, seven samples were collected at the resort where the outbreak occurred and from its water entry (R), two in the city centre of Mascali (M), 15 in the village of Fondachello (F), on the seaside, three in Carrabba village (C), far from the coast, and two and three, respectively, from the water well ‘Carlino’ (W) and the source ‘Bufardo and Torregrossa’ (S) which feed the municipal system and provide drinking water to the Mascali area (Fig. 1, Table 3). One litre water samples were sent by the local health authorities to the SIALAB laboratory in Avola, Siracusa and 100 ml aliquots were screened for coliforms, E. coli and enterococci as faecal indicators, by standard bacteriological methods according to UNI EN ISO 9308-1:2014. Unfortunately, although the suspect of a water source for the viral outbreak was quickly communicated, only in August and November 2016, following the implementation of control measures, local authorities provided water samplings of appropriate volume to perform viral tests (10 l). Five water samples of 10 l each were collected from different sampling sites in F and R areas for viral screening (Table 3) and were concentrated according to the Italian National Institute of Health (Istituto Superiore di Sanità, ISS) protocols [22] through an ultra-filtration system (Sartoflow® Slice 200 Benchtop Crossflow System, Sartorius AG, Goettingen, Germany), using appropriate membranes (SG Hydrosart 10 kDa) pretreated with 300 ml of 3% Beef Extract (BE) at pH 7, by recirculating it for 10 min in the system, at a pressure not higher than 1.30 bar. Elution and recovery of viruses attached to the membranes were obtained by recirculating 200 ml of 3% BE at pH 9.5 for 3 h until a 10–12 ml volume of eluate was obtained. The pH of the eluate was then checked with litmus paper and if necessary brought to the value of 7 with HCl. Concentrated water samples were simultaneously tested for NoV in three laboratories: Enteric Viruses Laboratory, University of Palermo; Istituto Zooprofilattico Sperimentale della Sicilia (IZS), Palermo; and Istituto Superiore di Sanità (ISS), Rome. Total RNA was extracted from 500 and 560 µl of a concentrated water sample using BioMèrieux NucliSENS® Magnetic Extraction System (BioMèrieux) and QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) respectively, according to manufacturer’s instructions. Bonded RNA was eluted by 100 µl of Elution Buffer. Amplifications were performed at least in duplicate on whole and 1:10 diluted extracts. A Real-Time RT-PCR assay according to ISO/TS 15216- 2:2013, two in-house Real-Time RT-PCR protocols and the One-Step RT-PCR methods already used for amplifying ORF1 and ORF2 fragments in NoV-positive faecal samples were performed [19–21, 23]. Nucleotide sequences analysis and phylogenetic analysis PCR amplicons were purified by commercial columns (CleanSweep PCR Purification, Thermo Fisher Scientific, Vilnius, Lithuania) and sequenced by Sanger method and all nucleotide sequences were aligned using ClustalW. To identify NoV genotypes, RdRp and VP1 sequences were submitted to the RIVM NoV genotyping tool (http://www.rivm.nl/mpf/norovirus/typingtool). Phylogenetic trees were constructed using MEGA 7 and Neighbour-Joining method (bootstrap 1000 replicates). Sequences from stool and water samples are available at NCBI with the following accession numbers: MG517455 for the major capsid protein obtained from a water sample; MG517456 and MG517461, respectively, for the major capsid protein and RNA-dependent-RNA-polymerase, as representatives of the sequences obtained from the stool samples of those involved in the outbreak. Control measures After outbreak notification, on 21 May, the resort was closed for 2 days. On 25 May, following the first results of the bacteriological analyses on municipal water samples, the use of municipal water for human consumption was temporarily restricted (use of water was permitted only for cleaning purposes) in the Carrabba and Fondachello villages and in the area surrounding the resort. From 25 May the reservoir collecting the water from the Carlino well and its chlorination device were submitted to maintenance work and a new chlorination device was installed on the conduit bringing the water from the Bufardo source. Actually, the Bufardo source had been connected to the municipal network in 2003 but unfortunately a dedicated chlorination station had not been planned at that time. Therefore, unchlorinated water entered the system near Fondachello village on the seaside (Fig. 1), possibly diluting chlorine concentrations in water treated at the Carlino reservoir. Moreover, the exact path of the branch of the water network situated downstream the Bufardo source water connection was not marked in the maps of the municipality (Fig. 1, red line), indicating that this section had never been submitted to maintenance. The area surrounding the unmaintained section of the water pipeline was scattered with unauthorised cottages that were not connected to the sewerage systems. Likely, untreated waste from cesspools was unloaded directly into the soil, generating a risk of contamination of the municipal water system through leakage. No work had been made on the water pipelines before the outbreak but sections of the network had been temporarily closed during the winter season on the seaside (Fondachello). Local health authorities advised the connection of all households in the area to the sewerage system and the replacement of the water system section without maintenance. Health authorities also raised attention on AGE cases notification at all local Emergency Departments and Family Practitioners but there was no further evidence of AGE outbreaks in the months following the intervention. Results Data collected through the online questionnaire compiled 7–10 days after the outbreak revealed that none of the 28 students and the teacher taking part in the school trip showed gastroenteritis symptoms at the beginning of the journey and they had no contact with people who suffered from diarrhoea and/or vomiting in the previous days. Travellers had their first shared meal on 19 May (Day 1), 2016, at lunch on their arrival to the resort in Mascali. The list of foods included in the menu is shown in Table 2. Conversely, the dinner on Day 1 was not shared and the students went to dine in three different fast food restaurants in Acireale. On 20 May (Day 2), 17 students (58.6% of the travellers) presented with vomit and/or diarrhoea, falling ill 26–36 h after lunch on Day 1, strongly suggesting that exposure occurred on the first shared meal. Some students also vomited in the presence of the other travellers on the bus bringing them to Acireale. Five more students were symptomatic in the early hours of the following day (21 May, Day 3, from 0:30 am to 8:00 am), as a continuation of the wave of illnesses of the previous day. The teacher and one more student fell ill at 8:00 pm and 11:00 pm of Day 3, respectively, for a second wave of illnesses occurring >48 h from the first shared meal and <36 h from the first cases, with the last student falling sick on 22 May (Day 4) at 3:00 am (Fig. 2, Table 1). On the whole, the mean incubation period was 33.2 h. Based on the online survey, the severity of illness was evaluated by the Vesikari severity scale, grouping symptomatic travellers into severe (11/25 = 44%) and non-severe (14/25 = 56%) cases (Table 1). Vomiting and fever were the predominant symptoms, affecting 22 out of 25 (88%) symptomatic subjects, while diarrhoea was present in 14 out of 25 (56%). Hospitalisation for rehydration therapy was necessary for 13 students (91% [10/11] of severe and 21% [3/14] of non-severe cases). Emergency room (ER) or family practitioner visits were necessary for 21 students and 13 (62%) of them initiated an antibiotic therapy. Two students indicated that other family members had gastroenteric symptoms more than 24 h after their return home, on Day 3, 4 and 5. All those involved in the outbreak recovered within 4 days, with an average duration of illness of 38.4 h. A case-control study of the risk associated with eating foods at lunch on Day 1 or with attending different restaurants at dinner on Day 1 was attempted. A total of 19 subjects filled the section of the online survey concerning food consumption. Fourteen of the responders were among the 22 cases falling ill within 36 h from the first shared meal (first wave), while the remaining five responders were three travellers with supposed secondary infections (showing symptoms with a delay ⩾48 h; second wave) and two asymptomatic students that were used as controls. No statistically significant associations with any food items served at lunch on Day 1 was identified (P < 0.05; Table 2). However, all food consumed on that occasion was to some extent associated with the risk of disease, the highest OR being obtained with chicken (OR: 20.7143, P: 0.0678; Table 2). Actually, most of the cases ate all foods included in the menu, while one of the two asymptomatic controls did not participate in the meal and the other had a pasta dish only. No statistically significant association was found for attending a particular fast food restaurant at the dinner on Day 1 (OR: 2.5, P: 0.3635). The Real-Time RT-PCR assay allowed to detect GII genogroup NoV genome in all the 25 samples from symptomatic subjects, although two of them showed high Ct values, indicating lower shedding of viral genomes (Table 1). The samples from four asymptomatic students were negative for both NoV genogroups. By sequence analysis, a NoV genotype could be defined for 20/ 25 NoV-positive stool samples. All NoV strains from travellers involved in the outbreaks belonged to the same genotype, GII.P2/GII.2, showing >99% nucleotide identities (nt id) in both fragments (ORF1 and ORF2) analysed. Thirty-two water samples were collected and analysed for bacterial contamination (Table 3). Seven samples were taken from the resort (site R) and the remaining 25 from five other sites (C, F, M, S, W), representative of the entire municipal water network. The majority (10/12) of the water samples collected immediately after the onset of the outbreak, between 21 May and 16 June 2016, at the resort (site R) and at sites F and C, located close to the resort, on the seaside and in the interior, respectively, were positive for coliform bacteria. More than 1000 colony-forming units (CFU)/ 100 ml (up to 2700 CFU/100 ml) were detected in four samples, while their absence is required in 100 ml by the Standard of Water Quality in Italy and seven of 11 samples tested were also Escherichia coli-positive (up to 900 CFU/100 ml). Enterococci were searched in 11 samples and three of them were found positive for this faecal contamination indicator. From 7 July to 16 August 2016, after control measures on the water network were undertaken, 13 water samples were collected for follow-up laboratory examination of the water system at F, R and C sites. All of them were negative for bacterial indicators of faecal contamination. No faecal bacteria were identified in the seven samples collected from M, W and S sites, including the well and the source that fed the water network, at any sampling time, before and after implementing the control measures. The research of NoV genome was conducted on concentrated water samples collected on 1 August from R and F sites (Table 3). NoV genome was detected in a single sample collected from the water entry of the resort involved in the outbreak (site R). Actually, only two out of 26 One-Step RT-PCR repetitions were positive for NoV viral capsid gene. The two ORF2-positive results were obtained at the University of Palermo Laboratory using both 1:10 diluted and undiluted extracted RNA. The sequence analysis of the partial ORF2 portion (289 bp) indicated the presence of the same GII.2 capsid genotype detected in the stool samples of the students involved in the outbreak, showing 98.9% nt id. No ORF1 NoV sequence could be obtained from the NoV-positive water sample. Follow-up analyses for NoV genome, conducted on water samples collected on 20 November both at R and F sites found no positive samples. By sequence alignment, the NoV sequences obtained from Mascali outbreak samples were phylogenetically closely related in both ORF1 and ORF2 (99 and 98.9% of nt id, respectively) to a GII.P2/GII.2 strain responsible for a sporadic paediatric gastroenteritis case occurred in Palermo in January 2016 (PA 36/16, accession numbers MG517464 and MG517460, for ORF1 and ORF2, respectively) and previously detected at the Enteric Viruses Laboratory, University of Palermo. Whereas, three GII.P2/GII.2 Italian strains (PA 280/11, PA 366/11 and PA 3/ 14, accession numbers MG517462, MG517465 and MG517463, for ORF1 and MG517457 to −59, for ORF2, respectively), detected in Palermo in 2011 and 2014, were more distantly associated with the outbreak strain both in RdRp and VP1 (90% to 97.4% nt id and 96.1% to 98.1% nt id, respectively, in ORF1 and ORF2). Discussion Epidemiological features, such as the presence of vomiting in more than half (88%) of symptomatic subjects, the 33.2 h average incubation period and the 38.4 h average duration of illness, suggested a NoV aetiology for the investigated outbreak [13]. The detection of NoV genome in all 25 faecal samples from symptomatic subjects fully complies with the requirement of detection of at least one positive sample out of four tested that is considered to be sufficient to ascertain the aetiology of a NoV outbreak [24]. Moreover, molecular typing indicated that a single viral strain, belonging to genotype GII.P2/GII.2, infected all symptomatic subjects involved in the outbreak. NoVs are highly infectious and can easily spread from person to person but also through contaminated food or water. Therefore, settings such as health care institutions, schools and touristic resorts are mainly affected, since close contact and shared food services facilitate transmission. In Mascali, two waves of transmission were observed over three days, being consistent with food/waterborne followed by person-to-person and/or airborne (vomit) transmission (Fig. 2). The first wave of cases presented on 20 May 2016 after a single shared meal, while the last case was detected on 22 May. Exposure to vomit may have been responsible for some of the secondary cases occurred after the first wave of illness. Vomiting produces aerosols that can be inhaled and swallowed. The extremely low infectious dose of NoV makes the inhalation of vomit aerosols an effective transmission route in settings where there is close contact, like on a bus trip [25]. All food consumed in the only shared lunch on 19 May at the resort was associated with the risk of disease (Table 2), but the association was inconclusive since all cases shared the same dishes included in the lunch menu on Day 1. Although many foods were potentially involved, our attention focused on the tap water. In fact, trivial information from local health authorities reported that local population avoided tap water consumption and relied on bottled water for drinking and cooking. Many small outbreaks of gastroenteritis involving restaurants and family groups had been reported in the same area, especially at the beginning of the summer season, when vacation houses and recreational activities were re-opened after a period of low water demand. Illness caused by NoV is common and if the level of water contamination is low, the number of cases remains limited and water can persist as unrecognised source of infection. If polluted water was served for drinking and used for cooking and washing dishes in the Mascali resort at Day 1 a generalised contamination of food and surfaces could have occurred. The potential role of the foodhandlers as a source of contamination could not be analysed in this study but should also be taken into account. If water contamination has already been present in the area for a long time, infections may have occurred among the local food-handlers. The analysis of water samples taken immediately after the outbreak revealed the presence of bacterial indicators of faecal contamination in the water used in the resort, with the only exception of a sample of filter purified water used both for cooking and as drinking water. When extending the sampling to other areas of the municipal water network, many other sites, mostly on the seaside, showed faecal pollution, which demonstrates a poor protection of the water network from faecal leakage (Fig. 1). NoV contamination of waters generally originates from sewage since these enteric pathogens, as other enteric viruses affecting humans, are mostly species-specific and are shed at high concentrations in the stools of patients. When the analyses on the stools of the students made it clear that the outbreak had been caused by NoV, the viral genome was also searched in appropriate volumes of water samples, but this happened when correction measures on the water network had already been taken. Nevertheless, a single water sample, obtained 2 months after the outbreak (1 August 2016), still contained genome fragments that could be correlated to the same NoV genotype causing the outbreak. Although the finding of NoV genome does not demonstrate that entire infecting viral particles were present in the water, it could indicate that a NoV belonging to the same genotype involved in the outbreak was still circulating in the local population several weeks after the outbreak and it was someway able to penetrate the water network, probably indicating continuing faecal contamination through leakage from sewage systems. NoV is particularly resistant in the environment and also to chlorine, therefore transmission through chlorinated water is not exceptional [9, 12, 26]. Drinking water from municipal supplies fed by lake or groundwater has already been identified in the past as the source of waterborne NoV outbreaks in Italy [27–29] and NoV is now considered one of the pathogens causing the largest number of affected consumers in drinking waterborne outbreaks [12]. NoV can persist in water longer than 15 days [30] and it is resistant to chlorine disinfection when free chlorine levels or pre-treatment processes are inadequate [31, 32]. Moreover, microfiltration filters labelled as certified by NSF Standards 53 or 58 (with a pore size ranging from 0.05 to 5 µm) are not effective in removing viruses and can even decrease chlorine concentration in filtered water [33]. Actually, the owner of the resort involved in our outbreak was encouraged to serve tap water as drinking water at the restaurant relying on the effectiveness of such a filter purifier. Unfortunately, we did not have a tap water sample from the restaurant of the resort taken before correction measures and whose volume was appropriate for virus detection. Therefore, we could not demonstrate the failure to eliminate viral pollution of the microfiltration system. In Mascali, correction measures on the water network consisted in the maintenance of the old chlorination device and in the addition of a new one at a key point located upstream and close to suspected illegal sewage dumping (Fig. 1), but as poor protection of the pipeline from sewage leakage risked to reproduce faecal pollution episodes, renovation of the municipal water system was also advised, as well as systematic control of illegal sewage dumping. The additional chlorination device was apparently the most effective intervention, since right after its activation faecal bacterial pollution indicators disappeared. However, NoV genome was detected in a water sample taken 6 weeks after the last detection of bacterial indicators. Unfortunately, detection of faecal contamination of drinking water is only based on the use of bacterial indicators despite no correlation between levels of enteric bacteria and enteric viruses has been found [34–36]. According to ISGEV enteric virus circulation surveillance data, in 2016 the circulation of GII.P2/GII.2 NoV strains in Palermo, home town of the travellers involved in the outbreak, was the highest ever observed from the starting of local NoV surveillance in 2004. GII.P2/GII.2 was the fourth most common genotype in 2016 (12.2% prevalence), following GII.P4/GII.4 Sydney 2012 (28.4%), GII.P16/GII.2 (18.9%) and GII.Pe/GII.4 Sydney 2012 (17.6%). GII.P2/GII.2 strains were mostly observed starting from July 2016, leaving space for the possibility that their increased circulation was linked to the introduction of the ‘Mascali clone’. However, a couple of GII.P2/GII.2 strains (including PA36/16) genetically close to the ‘Mascali clone’ had been already detected in January 2016 in Palermo and GII.P2/GII.2 NoV had sporadically been observed also in 2011 and 2014. Moreover, the circulation of GII.P2/GII.2 NoVs has been increasing worldwide over the last period [37]. Therefore, the unusually high circulation observed in Palermo might just reflect the global epidemiology of GII.P2/GII.2 strains. Noteworthy, the older GII.P2/GII.2 NoV strains detected in Palermo were closely related in the ORF2 (98.9% nt id) to the GII.P16/GII.2 Nashville strain (KY865307) which is supposed to have generated the recombinant GII.P16/GII.4 clone recently circulating epidemically worldwide [37, 38] and were more distantly associated (95.3% nt id) to the GII.2 capsid sequence (AB662868) of another putative GII.P16/GII.4 ancestor from Japan [39]. The increasing frequency of detection of waterborne NoV outbreaks calls for the introduction of routine monitoring for viruses of drinking waters and for the definition of affordable and reliable indicators for the presence of NoV in drinking waters. The use of ultrafiltration systems as a concentration method in order to recover viruses from environmental waters combined with RT-PCR to detect viral genomes has been demonstrated to grant a recovery rate of >50% of multiple viruses [40] and is at present one of the most sensitive approaches for the detection of NoVs, although the discrimination between viable and nonviable viruses remains a problem. The burden of NoV outbreaks, either food associated or waterborne, is still poorly defined due to the lack of specific diagnostic tools in many clinical laboratories. NoV epidemics should always be suspected when the main symptom is vomiting and the duration of disease is short (<48 h). Waterborne transmission is suggested by the rapid spread of disease and the high attack rate. Care should be taken to perform sensitive diagnostic tests on a sufficient number of stool samples in order to identify NoV as the causative agent of enteritis outbreaks, possibly preferring genome screening over antigen detection. Only the systematic detection of NoV outbreaks can guide in the choice of the safety procedures that could be most appropriate in containing such microbiological hazard.

What were the first steps?

1553

Waterborne Norovirus outbreak at a seaside resort likely originating from municipal water distribution system failure

Abstract In May 2016 a Norovirus (NoV) gastroenteritis outbreak involved a high school class visiting a seaside resort near Taormina (Mascali, Sicily). Twenty-four students and a teacher were affected and 17 of them showed symptoms on the second day of the journey, while the others got ill within the following 2 days. Symptoms included vomiting, diarrhoea and fever, and 12 students required hospitalisation. Stool samples tested positive for NoV genome by Real-Time polymerase chain reaction assay in all 25 symptomatic subjects. The GII.P2/GII.2 NoV genotype was linked to the outbreak by ORF1/ORF2 sequence analysis. The epidemiological features of the outbreak were consistent with food/waterborne followed by person-to-person and/ or vomit transmission. Food consumed at a shared lunch on the first day of the trip was associated to illness and drinking un-bottled tap water was also considered as a risk factor. The analysis of water samples revealed the presence of bacterial indicators of faecal contamination in the water used in the resort as well as in other areas of the municipal water network, linking the NoV gastroenteritis outbreak to tap water pollution from sewage leakage. From a single water sample, an amplicon whose sequence corresponded to the capsid genotype recovered from patients could be obtained. Introduction Noroviruses (NoVs) are recognised as a leading cause of sporadic cases and outbreaks of acute gastroenteritis (AGE) across all age groups [1, 2]. They are transmitted primarily through the faecal-oral route either by direct person–person spread and with consumption of contaminated food or water [1, 3, 4]. NoVs are non-enveloped RNA viruses belonging to the Caliciviridae family and are classified into seven distinct genogroups (GI-GVII) on the basis of the major capsid protein VP1 sequence. GI, GII and GIV NoVs infect humans and they have been divided into more than 30 genotypes based on the sequences of polymerase (ORF1) and capsid proteins (ORF2) [5]. NoV GII.4 genotype has been associated with the vast majority of NoV related cases of AGE worldwide although many genotypes, belonging to both GI and GII genogroups are involved in foodborne and waterborne outbreaks [6, 7]. NoVs are highly contagious owing to their high stability in the environment and their ability to infect at low doses. In fact, just 10–100 virions of NoV may be sufficient to cause AGE [8, 9]. NoVs are considered the major cause of water-related disease in semi-closed institutions, including restaurants, schools, hospitals, nursing homes and cruise ships [10–12]. To confirm NoV as a cause of outbreaks, the presence of four epidemiological features of disease, vomiting in more than half of affected persons, a mean incubation period of 24/48 h, a mean duration of illness of 12–60 h and the absence of bacterial pathogens in stool culture, should be evaluated [13]. Waterborne NoV outbreaks can be caused by contaminated drinking water despite correct chlorine concentration and in the absence of coliform bacteria [10, 14–17]. The aim of this study was to investigate an outbreak of NoV gastroenteritis associated with drinking water involving a high school class travelling from Palermo (West Sicily) to Acireale (East Sicily) on an educational tour in May 2016. Stool samples were collected soon after the onset of AGE symptoms to ascertain the aetiology of illness. The source of infection was investigated collecting epidemiological data through an online questionnaire and, in order to explore a possible transmission route, laboratory tests on water from the municipal network were performed. Material and methods Epidemiological investigation From 20 to 22 May 2016 an AGE outbreak occurred in Mascali, a small town near Taormina, involving a school class of 28 students, 13 girls and 15 boys aged between 14 and 18 years and a 54-year-old female teacher, during a trip to participate in a theatre festival in Acireale. A total of 25/29 persons were affected by nausea, vomiting, abdominal cramps, diarrhoea and fever while they were staying at a seaside resort in Mascali, on the way to Acireale. To obtain data on symptoms and the correlated risk factors all participants filled an online questionnaire investigating personal data (age and sex), travel and accommodation details and food consumption over 48 h before symptoms presentation, time of onset and duration of symptoms, therapy and/or hospitalisation required and history of contact with people affected by diarrhoea and/or vomiting (Table 1). The severity of illness was analysed by the Vesikari severity scale allowing to distinguish non-severe (score <11) and severe (score ⩾11) AGE (Table 1). Vesikari score was calculated from maximum number of episodes/day and duration of both diarrhoea and vomiting (score 0–3 for each of the 4 parameters), temperature (score 0–3), level of dehydration (score 0–3) and treatment required (score 0–2) [18]. A case-control study was conducted to establish if food consumed before the onset of symptoms was associated with the disease. A case was defined as a student attending the tour and reporting diarrhoea and/or vomiting within 36 h from the first shared meal during the tour, while students who did not present any of the previous symptoms or those showing the first symptoms with a delay of ⩾48 h (secondary cases) were used as controls. Variables were studied using odds ratio (OR) and its 95% CI and statistical associations were determined using the χ2 test, with the level of statistical significance set at 0.05 (Table 2). NoV detection in stool samples A stool sample was obtained from each of the 29 travellers, soon after (3–5 days) the onset of the outbreak. No stool samples could be obtained from the food-handlers working at the resort. All samples were tested for NoV at the Enteric Viruses Laboratory of the University of Palermo, member of the Italian Study Group for Enteric Viruses (ISGEV), by a Real-Time polymerase chain reaction (RT-PCR) assay that allows to differentiate GI and GII NoVs [19]. Total RNA was extracted from stool samples using QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) and cDNA was generated by MMLV reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were considered positive for NoVs when cycle threshold (ct) of RT-PCR was ⩽30 and lowpositive when 30<ct⩽35. For all NoV-positive samples two genomic fragments were amplified by conventional RT-PCR using Superscript III One-Step RT-PCR system with Platinum Taq DNA polymerase (Invitrogen) and primers JV12a/JV13b, targeting 326 bps of the RdRp region A [20] and either COG1F/ GISKR or COG2F/GIISKR, amplifying a 315 bps 5′portion of GI or GII VP1 protein [19, 21]. Environmental investigation From May through August 2016, a total of 32 tap water samples were collected from the Mascali municipal network. In particular, seven samples were collected at the resort where the outbreak occurred and from its water entry (R), two in the city centre of Mascali (M), 15 in the village of Fondachello (F), on the seaside, three in Carrabba village (C), far from the coast, and two and three, respectively, from the water well ‘Carlino’ (W) and the source ‘Bufardo and Torregrossa’ (S) which feed the municipal system and provide drinking water to the Mascali area (Fig. 1, Table 3). One litre water samples were sent by the local health authorities to the SIALAB laboratory in Avola, Siracusa and 100 ml aliquots were screened for coliforms, E. coli and enterococci as faecal indicators, by standard bacteriological methods according to UNI EN ISO 9308-1:2014. Unfortunately, although the suspect of a water source for the viral outbreak was quickly communicated, only in August and November 2016, following the implementation of control measures, local authorities provided water samplings of appropriate volume to perform viral tests (10 l). Five water samples of 10 l each were collected from different sampling sites in F and R areas for viral screening (Table 3) and were concentrated according to the Italian National Institute of Health (Istituto Superiore di Sanità, ISS) protocols [22] through an ultra-filtration system (Sartoflow® Slice 200 Benchtop Crossflow System, Sartorius AG, Goettingen, Germany), using appropriate membranes (SG Hydrosart 10 kDa) pretreated with 300 ml of 3% Beef Extract (BE) at pH 7, by recirculating it for 10 min in the system, at a pressure not higher than 1.30 bar. Elution and recovery of viruses attached to the membranes were obtained by recirculating 200 ml of 3% BE at pH 9.5 for 3 h until a 10–12 ml volume of eluate was obtained. The pH of the eluate was then checked with litmus paper and if necessary brought to the value of 7 with HCl. Concentrated water samples were simultaneously tested for NoV in three laboratories: Enteric Viruses Laboratory, University of Palermo; Istituto Zooprofilattico Sperimentale della Sicilia (IZS), Palermo; and Istituto Superiore di Sanità (ISS), Rome. Total RNA was extracted from 500 and 560 µl of a concentrated water sample using BioMèrieux NucliSENS® Magnetic Extraction System (BioMèrieux) and QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) respectively, according to manufacturer’s instructions. Bonded RNA was eluted by 100 µl of Elution Buffer. Amplifications were performed at least in duplicate on whole and 1:10 diluted extracts. A Real-Time RT-PCR assay according to ISO/TS 15216- 2:2013, two in-house Real-Time RT-PCR protocols and the One-Step RT-PCR methods already used for amplifying ORF1 and ORF2 fragments in NoV-positive faecal samples were performed [19–21, 23]. Nucleotide sequences analysis and phylogenetic analysis PCR amplicons were purified by commercial columns (CleanSweep PCR Purification, Thermo Fisher Scientific, Vilnius, Lithuania) and sequenced by Sanger method and all nucleotide sequences were aligned using ClustalW. To identify NoV genotypes, RdRp and VP1 sequences were submitted to the RIVM NoV genotyping tool (http://www.rivm.nl/mpf/norovirus/typingtool). Phylogenetic trees were constructed using MEGA 7 and Neighbour-Joining method (bootstrap 1000 replicates). Sequences from stool and water samples are available at NCBI with the following accession numbers: MG517455 for the major capsid protein obtained from a water sample; MG517456 and MG517461, respectively, for the major capsid protein and RNA-dependent-RNA-polymerase, as representatives of the sequences obtained from the stool samples of those involved in the outbreak. Control measures After outbreak notification, on 21 May, the resort was closed for 2 days. On 25 May, following the first results of the bacteriological analyses on municipal water samples, the use of municipal water for human consumption was temporarily restricted (use of water was permitted only for cleaning purposes) in the Carrabba and Fondachello villages and in the area surrounding the resort. From 25 May the reservoir collecting the water from the Carlino well and its chlorination device were submitted to maintenance work and a new chlorination device was installed on the conduit bringing the water from the Bufardo source. Actually, the Bufardo source had been connected to the municipal network in 2003 but unfortunately a dedicated chlorination station had not been planned at that time. Therefore, unchlorinated water entered the system near Fondachello village on the seaside (Fig. 1), possibly diluting chlorine concentrations in water treated at the Carlino reservoir. Moreover, the exact path of the branch of the water network situated downstream the Bufardo source water connection was not marked in the maps of the municipality (Fig. 1, red line), indicating that this section had never been submitted to maintenance. The area surrounding the unmaintained section of the water pipeline was scattered with unauthorised cottages that were not connected to the sewerage systems. Likely, untreated waste from cesspools was unloaded directly into the soil, generating a risk of contamination of the municipal water system through leakage. No work had been made on the water pipelines before the outbreak but sections of the network had been temporarily closed during the winter season on the seaside (Fondachello). Local health authorities advised the connection of all households in the area to the sewerage system and the replacement of the water system section without maintenance. Health authorities also raised attention on AGE cases notification at all local Emergency Departments and Family Practitioners but there was no further evidence of AGE outbreaks in the months following the intervention. Results Data collected through the online questionnaire compiled 7–10 days after the outbreak revealed that none of the 28 students and the teacher taking part in the school trip showed gastroenteritis symptoms at the beginning of the journey and they had no contact with people who suffered from diarrhoea and/or vomiting in the previous days. Travellers had their first shared meal on 19 May (Day 1), 2016, at lunch on their arrival to the resort in Mascali. The list of foods included in the menu is shown in Table 2. Conversely, the dinner on Day 1 was not shared and the students went to dine in three different fast food restaurants in Acireale. On 20 May (Day 2), 17 students (58.6% of the travellers) presented with vomit and/or diarrhoea, falling ill 26–36 h after lunch on Day 1, strongly suggesting that exposure occurred on the first shared meal. Some students also vomited in the presence of the other travellers on the bus bringing them to Acireale. Five more students were symptomatic in the early hours of the following day (21 May, Day 3, from 0:30 am to 8:00 am), as a continuation of the wave of illnesses of the previous day. The teacher and one more student fell ill at 8:00 pm and 11:00 pm of Day 3, respectively, for a second wave of illnesses occurring >48 h from the first shared meal and <36 h from the first cases, with the last student falling sick on 22 May (Day 4) at 3:00 am (Fig. 2, Table 1). On the whole, the mean incubation period was 33.2 h. Based on the online survey, the severity of illness was evaluated by the Vesikari severity scale, grouping symptomatic travellers into severe (11/25 = 44%) and non-severe (14/25 = 56%) cases (Table 1). Vomiting and fever were the predominant symptoms, affecting 22 out of 25 (88%) symptomatic subjects, while diarrhoea was present in 14 out of 25 (56%). Hospitalisation for rehydration therapy was necessary for 13 students (91% [10/11] of severe and 21% [3/14] of non-severe cases). Emergency room (ER) or family practitioner visits were necessary for 21 students and 13 (62%) of them initiated an antibiotic therapy. Two students indicated that other family members had gastroenteric symptoms more than 24 h after their return home, on Day 3, 4 and 5. All those involved in the outbreak recovered within 4 days, with an average duration of illness of 38.4 h. A case-control study of the risk associated with eating foods at lunch on Day 1 or with attending different restaurants at dinner on Day 1 was attempted. A total of 19 subjects filled the section of the online survey concerning food consumption. Fourteen of the responders were among the 22 cases falling ill within 36 h from the first shared meal (first wave), while the remaining five responders were three travellers with supposed secondary infections (showing symptoms with a delay ⩾48 h; second wave) and two asymptomatic students that were used as controls. No statistically significant associations with any food items served at lunch on Day 1 was identified (P < 0.05; Table 2). However, all food consumed on that occasion was to some extent associated with the risk of disease, the highest OR being obtained with chicken (OR: 20.7143, P: 0.0678; Table 2). Actually, most of the cases ate all foods included in the menu, while one of the two asymptomatic controls did not participate in the meal and the other had a pasta dish only. No statistically significant association was found for attending a particular fast food restaurant at the dinner on Day 1 (OR: 2.5, P: 0.3635). The Real-Time RT-PCR assay allowed to detect GII genogroup NoV genome in all the 25 samples from symptomatic subjects, although two of them showed high Ct values, indicating lower shedding of viral genomes (Table 1). The samples from four asymptomatic students were negative for both NoV genogroups. By sequence analysis, a NoV genotype could be defined for 20/ 25 NoV-positive stool samples. All NoV strains from travellers involved in the outbreaks belonged to the same genotype, GII.P2/GII.2, showing >99% nucleotide identities (nt id) in both fragments (ORF1 and ORF2) analysed. Thirty-two water samples were collected and analysed for bacterial contamination (Table 3). Seven samples were taken from the resort (site R) and the remaining 25 from five other sites (C, F, M, S, W), representative of the entire municipal water network. The majority (10/12) of the water samples collected immediately after the onset of the outbreak, between 21 May and 16 June 2016, at the resort (site R) and at sites F and C, located close to the resort, on the seaside and in the interior, respectively, were positive for coliform bacteria. More than 1000 colony-forming units (CFU)/ 100 ml (up to 2700 CFU/100 ml) were detected in four samples, while their absence is required in 100 ml by the Standard of Water Quality in Italy and seven of 11 samples tested were also Escherichia coli-positive (up to 900 CFU/100 ml). Enterococci were searched in 11 samples and three of them were found positive for this faecal contamination indicator. From 7 July to 16 August 2016, after control measures on the water network were undertaken, 13 water samples were collected for follow-up laboratory examination of the water system at F, R and C sites. All of them were negative for bacterial indicators of faecal contamination. No faecal bacteria were identified in the seven samples collected from M, W and S sites, including the well and the source that fed the water network, at any sampling time, before and after implementing the control measures. The research of NoV genome was conducted on concentrated water samples collected on 1 August from R and F sites (Table 3). NoV genome was detected in a single sample collected from the water entry of the resort involved in the outbreak (site R). Actually, only two out of 26 One-Step RT-PCR repetitions were positive for NoV viral capsid gene. The two ORF2-positive results were obtained at the University of Palermo Laboratory using both 1:10 diluted and undiluted extracted RNA. The sequence analysis of the partial ORF2 portion (289 bp) indicated the presence of the same GII.2 capsid genotype detected in the stool samples of the students involved in the outbreak, showing 98.9% nt id. No ORF1 NoV sequence could be obtained from the NoV-positive water sample. Follow-up analyses for NoV genome, conducted on water samples collected on 20 November both at R and F sites found no positive samples. By sequence alignment, the NoV sequences obtained from Mascali outbreak samples were phylogenetically closely related in both ORF1 and ORF2 (99 and 98.9% of nt id, respectively) to a GII.P2/GII.2 strain responsible for a sporadic paediatric gastroenteritis case occurred in Palermo in January 2016 (PA 36/16, accession numbers MG517464 and MG517460, for ORF1 and ORF2, respectively) and previously detected at the Enteric Viruses Laboratory, University of Palermo. Whereas, three GII.P2/GII.2 Italian strains (PA 280/11, PA 366/11 and PA 3/ 14, accession numbers MG517462, MG517465 and MG517463, for ORF1 and MG517457 to −59, for ORF2, respectively), detected in Palermo in 2011 and 2014, were more distantly associated with the outbreak strain both in RdRp and VP1 (90% to 97.4% nt id and 96.1% to 98.1% nt id, respectively, in ORF1 and ORF2). Discussion Epidemiological features, such as the presence of vomiting in more than half (88%) of symptomatic subjects, the 33.2 h average incubation period and the 38.4 h average duration of illness, suggested a NoV aetiology for the investigated outbreak [13]. The detection of NoV genome in all 25 faecal samples from symptomatic subjects fully complies with the requirement of detection of at least one positive sample out of four tested that is considered to be sufficient to ascertain the aetiology of a NoV outbreak [24]. Moreover, molecular typing indicated that a single viral strain, belonging to genotype GII.P2/GII.2, infected all symptomatic subjects involved in the outbreak. NoVs are highly infectious and can easily spread from person to person but also through contaminated food or water. Therefore, settings such as health care institutions, schools and touristic resorts are mainly affected, since close contact and shared food services facilitate transmission. In Mascali, two waves of transmission were observed over three days, being consistent with food/waterborne followed by person-to-person and/or airborne (vomit) transmission (Fig. 2). The first wave of cases presented on 20 May 2016 after a single shared meal, while the last case was detected on 22 May. Exposure to vomit may have been responsible for some of the secondary cases occurred after the first wave of illness. Vomiting produces aerosols that can be inhaled and swallowed. The extremely low infectious dose of NoV makes the inhalation of vomit aerosols an effective transmission route in settings where there is close contact, like on a bus trip [25]. All food consumed in the only shared lunch on 19 May at the resort was associated with the risk of disease (Table 2), but the association was inconclusive since all cases shared the same dishes included in the lunch menu on Day 1. Although many foods were potentially involved, our attention focused on the tap water. In fact, trivial information from local health authorities reported that local population avoided tap water consumption and relied on bottled water for drinking and cooking. Many small outbreaks of gastroenteritis involving restaurants and family groups had been reported in the same area, especially at the beginning of the summer season, when vacation houses and recreational activities were re-opened after a period of low water demand. Illness caused by NoV is common and if the level of water contamination is low, the number of cases remains limited and water can persist as unrecognised source of infection. If polluted water was served for drinking and used for cooking and washing dishes in the Mascali resort at Day 1 a generalised contamination of food and surfaces could have occurred. The potential role of the foodhandlers as a source of contamination could not be analysed in this study but should also be taken into account. If water contamination has already been present in the area for a long time, infections may have occurred among the local food-handlers. The analysis of water samples taken immediately after the outbreak revealed the presence of bacterial indicators of faecal contamination in the water used in the resort, with the only exception of a sample of filter purified water used both for cooking and as drinking water. When extending the sampling to other areas of the municipal water network, many other sites, mostly on the seaside, showed faecal pollution, which demonstrates a poor protection of the water network from faecal leakage (Fig. 1). NoV contamination of waters generally originates from sewage since these enteric pathogens, as other enteric viruses affecting humans, are mostly species-specific and are shed at high concentrations in the stools of patients. When the analyses on the stools of the students made it clear that the outbreak had been caused by NoV, the viral genome was also searched in appropriate volumes of water samples, but this happened when correction measures on the water network had already been taken. Nevertheless, a single water sample, obtained 2 months after the outbreak (1 August 2016), still contained genome fragments that could be correlated to the same NoV genotype causing the outbreak. Although the finding of NoV genome does not demonstrate that entire infecting viral particles were present in the water, it could indicate that a NoV belonging to the same genotype involved in the outbreak was still circulating in the local population several weeks after the outbreak and it was someway able to penetrate the water network, probably indicating continuing faecal contamination through leakage from sewage systems. NoV is particularly resistant in the environment and also to chlorine, therefore transmission through chlorinated water is not exceptional [9, 12, 26]. Drinking water from municipal supplies fed by lake or groundwater has already been identified in the past as the source of waterborne NoV outbreaks in Italy [27–29] and NoV is now considered one of the pathogens causing the largest number of affected consumers in drinking waterborne outbreaks [12]. NoV can persist in water longer than 15 days [30] and it is resistant to chlorine disinfection when free chlorine levels or pre-treatment processes are inadequate [31, 32]. Moreover, microfiltration filters labelled as certified by NSF Standards 53 or 58 (with a pore size ranging from 0.05 to 5 µm) are not effective in removing viruses and can even decrease chlorine concentration in filtered water [33]. Actually, the owner of the resort involved in our outbreak was encouraged to serve tap water as drinking water at the restaurant relying on the effectiveness of such a filter purifier. Unfortunately, we did not have a tap water sample from the restaurant of the resort taken before correction measures and whose volume was appropriate for virus detection. Therefore, we could not demonstrate the failure to eliminate viral pollution of the microfiltration system. In Mascali, correction measures on the water network consisted in the maintenance of the old chlorination device and in the addition of a new one at a key point located upstream and close to suspected illegal sewage dumping (Fig. 1), but as poor protection of the pipeline from sewage leakage risked to reproduce faecal pollution episodes, renovation of the municipal water system was also advised, as well as systematic control of illegal sewage dumping. The additional chlorination device was apparently the most effective intervention, since right after its activation faecal bacterial pollution indicators disappeared. However, NoV genome was detected in a water sample taken 6 weeks after the last detection of bacterial indicators. Unfortunately, detection of faecal contamination of drinking water is only based on the use of bacterial indicators despite no correlation between levels of enteric bacteria and enteric viruses has been found [34–36]. According to ISGEV enteric virus circulation surveillance data, in 2016 the circulation of GII.P2/GII.2 NoV strains in Palermo, home town of the travellers involved in the outbreak, was the highest ever observed from the starting of local NoV surveillance in 2004. GII.P2/GII.2 was the fourth most common genotype in 2016 (12.2% prevalence), following GII.P4/GII.4 Sydney 2012 (28.4%), GII.P16/GII.2 (18.9%) and GII.Pe/GII.4 Sydney 2012 (17.6%). GII.P2/GII.2 strains were mostly observed starting from July 2016, leaving space for the possibility that their increased circulation was linked to the introduction of the ‘Mascali clone’. However, a couple of GII.P2/GII.2 strains (including PA36/16) genetically close to the ‘Mascali clone’ had been already detected in January 2016 in Palermo and GII.P2/GII.2 NoV had sporadically been observed also in 2011 and 2014. Moreover, the circulation of GII.P2/GII.2 NoVs has been increasing worldwide over the last period [37]. Therefore, the unusually high circulation observed in Palermo might just reflect the global epidemiology of GII.P2/GII.2 strains. Noteworthy, the older GII.P2/GII.2 NoV strains detected in Palermo were closely related in the ORF2 (98.9% nt id) to the GII.P16/GII.2 Nashville strain (KY865307) which is supposed to have generated the recombinant GII.P16/GII.4 clone recently circulating epidemically worldwide [37, 38] and were more distantly associated (95.3% nt id) to the GII.2 capsid sequence (AB662868) of another putative GII.P16/GII.4 ancestor from Japan [39]. The increasing frequency of detection of waterborne NoV outbreaks calls for the introduction of routine monitoring for viruses of drinking waters and for the definition of affordable and reliable indicators for the presence of NoV in drinking waters. The use of ultrafiltration systems as a concentration method in order to recover viruses from environmental waters combined with RT-PCR to detect viral genomes has been demonstrated to grant a recovery rate of >50% of multiple viruses [40] and is at present one of the most sensitive approaches for the detection of NoVs, although the discrimination between viable and nonviable viruses remains a problem. The burden of NoV outbreaks, either food associated or waterborne, is still poorly defined due to the lack of specific diagnostic tools in many clinical laboratories. NoV epidemics should always be suspected when the main symptom is vomiting and the duration of disease is short (<48 h). Waterborne transmission is suggested by the rapid spread of disease and the high attack rate. Care should be taken to perform sensitive diagnostic tests on a sufficient number of stool samples in order to identify NoV as the causative agent of enteritis outbreaks, possibly preferring genome screening over antigen detection. Only the systematic detection of NoV outbreaks can guide in the choice of the safety procedures that could be most appropriate in containing such microbiological hazard.

What did they do to control the problem?

1554

Waterborne Norovirus outbreak at a seaside resort likely originating from municipal water distribution system failure

Abstract In May 2016 a Norovirus (NoV) gastroenteritis outbreak involved a high school class visiting a seaside resort near Taormina (Mascali, Sicily). Twenty-four students and a teacher were affected and 17 of them showed symptoms on the second day of the journey, while the others got ill within the following 2 days. Symptoms included vomiting, diarrhoea and fever, and 12 students required hospitalisation. Stool samples tested positive for NoV genome by Real-Time polymerase chain reaction assay in all 25 symptomatic subjects. The GII.P2/GII.2 NoV genotype was linked to the outbreak by ORF1/ORF2 sequence analysis. The epidemiological features of the outbreak were consistent with food/waterborne followed by person-to-person and/ or vomit transmission. Food consumed at a shared lunch on the first day of the trip was associated to illness and drinking un-bottled tap water was also considered as a risk factor. The analysis of water samples revealed the presence of bacterial indicators of faecal contamination in the water used in the resort as well as in other areas of the municipal water network, linking the NoV gastroenteritis outbreak to tap water pollution from sewage leakage. From a single water sample, an amplicon whose sequence corresponded to the capsid genotype recovered from patients could be obtained. Introduction Noroviruses (NoVs) are recognised as a leading cause of sporadic cases and outbreaks of acute gastroenteritis (AGE) across all age groups [1, 2]. They are transmitted primarily through the faecal-oral route either by direct person–person spread and with consumption of contaminated food or water [1, 3, 4]. NoVs are non-enveloped RNA viruses belonging to the Caliciviridae family and are classified into seven distinct genogroups (GI-GVII) on the basis of the major capsid protein VP1 sequence. GI, GII and GIV NoVs infect humans and they have been divided into more than 30 genotypes based on the sequences of polymerase (ORF1) and capsid proteins (ORF2) [5]. NoV GII.4 genotype has been associated with the vast majority of NoV related cases of AGE worldwide although many genotypes, belonging to both GI and GII genogroups are involved in foodborne and waterborne outbreaks [6, 7]. NoVs are highly contagious owing to their high stability in the environment and their ability to infect at low doses. In fact, just 10–100 virions of NoV may be sufficient to cause AGE [8, 9]. NoVs are considered the major cause of water-related disease in semi-closed institutions, including restaurants, schools, hospitals, nursing homes and cruise ships [10–12]. To confirm NoV as a cause of outbreaks, the presence of four epidemiological features of disease, vomiting in more than half of affected persons, a mean incubation period of 24/48 h, a mean duration of illness of 12–60 h and the absence of bacterial pathogens in stool culture, should be evaluated [13]. Waterborne NoV outbreaks can be caused by contaminated drinking water despite correct chlorine concentration and in the absence of coliform bacteria [10, 14–17]. The aim of this study was to investigate an outbreak of NoV gastroenteritis associated with drinking water involving a high school class travelling from Palermo (West Sicily) to Acireale (East Sicily) on an educational tour in May 2016. Stool samples were collected soon after the onset of AGE symptoms to ascertain the aetiology of illness. The source of infection was investigated collecting epidemiological data through an online questionnaire and, in order to explore a possible transmission route, laboratory tests on water from the municipal network were performed. Material and methods Epidemiological investigation From 20 to 22 May 2016 an AGE outbreak occurred in Mascali, a small town near Taormina, involving a school class of 28 students, 13 girls and 15 boys aged between 14 and 18 years and a 54-year-old female teacher, during a trip to participate in a theatre festival in Acireale. A total of 25/29 persons were affected by nausea, vomiting, abdominal cramps, diarrhoea and fever while they were staying at a seaside resort in Mascali, on the way to Acireale. To obtain data on symptoms and the correlated risk factors all participants filled an online questionnaire investigating personal data (age and sex), travel and accommodation details and food consumption over 48 h before symptoms presentation, time of onset and duration of symptoms, therapy and/or hospitalisation required and history of contact with people affected by diarrhoea and/or vomiting (Table 1). The severity of illness was analysed by the Vesikari severity scale allowing to distinguish non-severe (score <11) and severe (score ⩾11) AGE (Table 1). Vesikari score was calculated from maximum number of episodes/day and duration of both diarrhoea and vomiting (score 0–3 for each of the 4 parameters), temperature (score 0–3), level of dehydration (score 0–3) and treatment required (score 0–2) [18]. A case-control study was conducted to establish if food consumed before the onset of symptoms was associated with the disease. A case was defined as a student attending the tour and reporting diarrhoea and/or vomiting within 36 h from the first shared meal during the tour, while students who did not present any of the previous symptoms or those showing the first symptoms with a delay of ⩾48 h (secondary cases) were used as controls. Variables were studied using odds ratio (OR) and its 95% CI and statistical associations were determined using the χ2 test, with the level of statistical significance set at 0.05 (Table 2). NoV detection in stool samples A stool sample was obtained from each of the 29 travellers, soon after (3–5 days) the onset of the outbreak. No stool samples could be obtained from the food-handlers working at the resort. All samples were tested for NoV at the Enteric Viruses Laboratory of the University of Palermo, member of the Italian Study Group for Enteric Viruses (ISGEV), by a Real-Time polymerase chain reaction (RT-PCR) assay that allows to differentiate GI and GII NoVs [19]. Total RNA was extracted from stool samples using QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) and cDNA was generated by MMLV reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were considered positive for NoVs when cycle threshold (ct) of RT-PCR was ⩽30 and lowpositive when 30<ct⩽35. For all NoV-positive samples two genomic fragments were amplified by conventional RT-PCR using Superscript III One-Step RT-PCR system with Platinum Taq DNA polymerase (Invitrogen) and primers JV12a/JV13b, targeting 326 bps of the RdRp region A [20] and either COG1F/ GISKR or COG2F/GIISKR, amplifying a 315 bps 5′portion of GI or GII VP1 protein [19, 21]. Environmental investigation From May through August 2016, a total of 32 tap water samples were collected from the Mascali municipal network. In particular, seven samples were collected at the resort where the outbreak occurred and from its water entry (R), two in the city centre of Mascali (M), 15 in the village of Fondachello (F), on the seaside, three in Carrabba village (C), far from the coast, and two and three, respectively, from the water well ‘Carlino’ (W) and the source ‘Bufardo and Torregrossa’ (S) which feed the municipal system and provide drinking water to the Mascali area (Fig. 1, Table 3). One litre water samples were sent by the local health authorities to the SIALAB laboratory in Avola, Siracusa and 100 ml aliquots were screened for coliforms, E. coli and enterococci as faecal indicators, by standard bacteriological methods according to UNI EN ISO 9308-1:2014. Unfortunately, although the suspect of a water source for the viral outbreak was quickly communicated, only in August and November 2016, following the implementation of control measures, local authorities provided water samplings of appropriate volume to perform viral tests (10 l). Five water samples of 10 l each were collected from different sampling sites in F and R areas for viral screening (Table 3) and were concentrated according to the Italian National Institute of Health (Istituto Superiore di Sanità, ISS) protocols [22] through an ultra-filtration system (Sartoflow® Slice 200 Benchtop Crossflow System, Sartorius AG, Goettingen, Germany), using appropriate membranes (SG Hydrosart 10 kDa) pretreated with 300 ml of 3% Beef Extract (BE) at pH 7, by recirculating it for 10 min in the system, at a pressure not higher than 1.30 bar. Elution and recovery of viruses attached to the membranes were obtained by recirculating 200 ml of 3% BE at pH 9.5 for 3 h until a 10–12 ml volume of eluate was obtained. The pH of the eluate was then checked with litmus paper and if necessary brought to the value of 7 with HCl. Concentrated water samples were simultaneously tested for NoV in three laboratories: Enteric Viruses Laboratory, University of Palermo; Istituto Zooprofilattico Sperimentale della Sicilia (IZS), Palermo; and Istituto Superiore di Sanità (ISS), Rome. Total RNA was extracted from 500 and 560 µl of a concentrated water sample using BioMèrieux NucliSENS® Magnetic Extraction System (BioMèrieux) and QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) respectively, according to manufacturer’s instructions. Bonded RNA was eluted by 100 µl of Elution Buffer. Amplifications were performed at least in duplicate on whole and 1:10 diluted extracts. A Real-Time RT-PCR assay according to ISO/TS 15216- 2:2013, two in-house Real-Time RT-PCR protocols and the One-Step RT-PCR methods already used for amplifying ORF1 and ORF2 fragments in NoV-positive faecal samples were performed [19–21, 23]. Nucleotide sequences analysis and phylogenetic analysis PCR amplicons were purified by commercial columns (CleanSweep PCR Purification, Thermo Fisher Scientific, Vilnius, Lithuania) and sequenced by Sanger method and all nucleotide sequences were aligned using ClustalW. To identify NoV genotypes, RdRp and VP1 sequences were submitted to the RIVM NoV genotyping tool (http://www.rivm.nl/mpf/norovirus/typingtool). Phylogenetic trees were constructed using MEGA 7 and Neighbour-Joining method (bootstrap 1000 replicates). Sequences from stool and water samples are available at NCBI with the following accession numbers: MG517455 for the major capsid protein obtained from a water sample; MG517456 and MG517461, respectively, for the major capsid protein and RNA-dependent-RNA-polymerase, as representatives of the sequences obtained from the stool samples of those involved in the outbreak. Control measures After outbreak notification, on 21 May, the resort was closed for 2 days. On 25 May, following the first results of the bacteriological analyses on municipal water samples, the use of municipal water for human consumption was temporarily restricted (use of water was permitted only for cleaning purposes) in the Carrabba and Fondachello villages and in the area surrounding the resort. From 25 May the reservoir collecting the water from the Carlino well and its chlorination device were submitted to maintenance work and a new chlorination device was installed on the conduit bringing the water from the Bufardo source. Actually, the Bufardo source had been connected to the municipal network in 2003 but unfortunately a dedicated chlorination station had not been planned at that time. Therefore, unchlorinated water entered the system near Fondachello village on the seaside (Fig. 1), possibly diluting chlorine concentrations in water treated at the Carlino reservoir. Moreover, the exact path of the branch of the water network situated downstream the Bufardo source water connection was not marked in the maps of the municipality (Fig. 1, red line), indicating that this section had never been submitted to maintenance. The area surrounding the unmaintained section of the water pipeline was scattered with unauthorised cottages that were not connected to the sewerage systems. Likely, untreated waste from cesspools was unloaded directly into the soil, generating a risk of contamination of the municipal water system through leakage. No work had been made on the water pipelines before the outbreak but sections of the network had been temporarily closed during the winter season on the seaside (Fondachello). Local health authorities advised the connection of all households in the area to the sewerage system and the replacement of the water system section without maintenance. Health authorities also raised attention on AGE cases notification at all local Emergency Departments and Family Practitioners but there was no further evidence of AGE outbreaks in the months following the intervention. Results Data collected through the online questionnaire compiled 7–10 days after the outbreak revealed that none of the 28 students and the teacher taking part in the school trip showed gastroenteritis symptoms at the beginning of the journey and they had no contact with people who suffered from diarrhoea and/or vomiting in the previous days. Travellers had their first shared meal on 19 May (Day 1), 2016, at lunch on their arrival to the resort in Mascali. The list of foods included in the menu is shown in Table 2. Conversely, the dinner on Day 1 was not shared and the students went to dine in three different fast food restaurants in Acireale. On 20 May (Day 2), 17 students (58.6% of the travellers) presented with vomit and/or diarrhoea, falling ill 26–36 h after lunch on Day 1, strongly suggesting that exposure occurred on the first shared meal. Some students also vomited in the presence of the other travellers on the bus bringing them to Acireale. Five more students were symptomatic in the early hours of the following day (21 May, Day 3, from 0:30 am to 8:00 am), as a continuation of the wave of illnesses of the previous day. The teacher and one more student fell ill at 8:00 pm and 11:00 pm of Day 3, respectively, for a second wave of illnesses occurring >48 h from the first shared meal and <36 h from the first cases, with the last student falling sick on 22 May (Day 4) at 3:00 am (Fig. 2, Table 1). On the whole, the mean incubation period was 33.2 h. Based on the online survey, the severity of illness was evaluated by the Vesikari severity scale, grouping symptomatic travellers into severe (11/25 = 44%) and non-severe (14/25 = 56%) cases (Table 1). Vomiting and fever were the predominant symptoms, affecting 22 out of 25 (88%) symptomatic subjects, while diarrhoea was present in 14 out of 25 (56%). Hospitalisation for rehydration therapy was necessary for 13 students (91% [10/11] of severe and 21% [3/14] of non-severe cases). Emergency room (ER) or family practitioner visits were necessary for 21 students and 13 (62%) of them initiated an antibiotic therapy. Two students indicated that other family members had gastroenteric symptoms more than 24 h after their return home, on Day 3, 4 and 5. All those involved in the outbreak recovered within 4 days, with an average duration of illness of 38.4 h. A case-control study of the risk associated with eating foods at lunch on Day 1 or with attending different restaurants at dinner on Day 1 was attempted. A total of 19 subjects filled the section of the online survey concerning food consumption. Fourteen of the responders were among the 22 cases falling ill within 36 h from the first shared meal (first wave), while the remaining five responders were three travellers with supposed secondary infections (showing symptoms with a delay ⩾48 h; second wave) and two asymptomatic students that were used as controls. No statistically significant associations with any food items served at lunch on Day 1 was identified (P < 0.05; Table 2). However, all food consumed on that occasion was to some extent associated with the risk of disease, the highest OR being obtained with chicken (OR: 20.7143, P: 0.0678; Table 2). Actually, most of the cases ate all foods included in the menu, while one of the two asymptomatic controls did not participate in the meal and the other had a pasta dish only. No statistically significant association was found for attending a particular fast food restaurant at the dinner on Day 1 (OR: 2.5, P: 0.3635). The Real-Time RT-PCR assay allowed to detect GII genogroup NoV genome in all the 25 samples from symptomatic subjects, although two of them showed high Ct values, indicating lower shedding of viral genomes (Table 1). The samples from four asymptomatic students were negative for both NoV genogroups. By sequence analysis, a NoV genotype could be defined for 20/ 25 NoV-positive stool samples. All NoV strains from travellers involved in the outbreaks belonged to the same genotype, GII.P2/GII.2, showing >99% nucleotide identities (nt id) in both fragments (ORF1 and ORF2) analysed. Thirty-two water samples were collected and analysed for bacterial contamination (Table 3). Seven samples were taken from the resort (site R) and the remaining 25 from five other sites (C, F, M, S, W), representative of the entire municipal water network. The majority (10/12) of the water samples collected immediately after the onset of the outbreak, between 21 May and 16 June 2016, at the resort (site R) and at sites F and C, located close to the resort, on the seaside and in the interior, respectively, were positive for coliform bacteria. More than 1000 colony-forming units (CFU)/ 100 ml (up to 2700 CFU/100 ml) were detected in four samples, while their absence is required in 100 ml by the Standard of Water Quality in Italy and seven of 11 samples tested were also Escherichia coli-positive (up to 900 CFU/100 ml). Enterococci were searched in 11 samples and three of them were found positive for this faecal contamination indicator. From 7 July to 16 August 2016, after control measures on the water network were undertaken, 13 water samples were collected for follow-up laboratory examination of the water system at F, R and C sites. All of them were negative for bacterial indicators of faecal contamination. No faecal bacteria were identified in the seven samples collected from M, W and S sites, including the well and the source that fed the water network, at any sampling time, before and after implementing the control measures. The research of NoV genome was conducted on concentrated water samples collected on 1 August from R and F sites (Table 3). NoV genome was detected in a single sample collected from the water entry of the resort involved in the outbreak (site R). Actually, only two out of 26 One-Step RT-PCR repetitions were positive for NoV viral capsid gene. The two ORF2-positive results were obtained at the University of Palermo Laboratory using both 1:10 diluted and undiluted extracted RNA. The sequence analysis of the partial ORF2 portion (289 bp) indicated the presence of the same GII.2 capsid genotype detected in the stool samples of the students involved in the outbreak, showing 98.9% nt id. No ORF1 NoV sequence could be obtained from the NoV-positive water sample. Follow-up analyses for NoV genome, conducted on water samples collected on 20 November both at R and F sites found no positive samples. By sequence alignment, the NoV sequences obtained from Mascali outbreak samples were phylogenetically closely related in both ORF1 and ORF2 (99 and 98.9% of nt id, respectively) to a GII.P2/GII.2 strain responsible for a sporadic paediatric gastroenteritis case occurred in Palermo in January 2016 (PA 36/16, accession numbers MG517464 and MG517460, for ORF1 and ORF2, respectively) and previously detected at the Enteric Viruses Laboratory, University of Palermo. Whereas, three GII.P2/GII.2 Italian strains (PA 280/11, PA 366/11 and PA 3/ 14, accession numbers MG517462, MG517465 and MG517463, for ORF1 and MG517457 to −59, for ORF2, respectively), detected in Palermo in 2011 and 2014, were more distantly associated with the outbreak strain both in RdRp and VP1 (90% to 97.4% nt id and 96.1% to 98.1% nt id, respectively, in ORF1 and ORF2). Discussion Epidemiological features, such as the presence of vomiting in more than half (88%) of symptomatic subjects, the 33.2 h average incubation period and the 38.4 h average duration of illness, suggested a NoV aetiology for the investigated outbreak [13]. The detection of NoV genome in all 25 faecal samples from symptomatic subjects fully complies with the requirement of detection of at least one positive sample out of four tested that is considered to be sufficient to ascertain the aetiology of a NoV outbreak [24]. Moreover, molecular typing indicated that a single viral strain, belonging to genotype GII.P2/GII.2, infected all symptomatic subjects involved in the outbreak. NoVs are highly infectious and can easily spread from person to person but also through contaminated food or water. Therefore, settings such as health care institutions, schools and touristic resorts are mainly affected, since close contact and shared food services facilitate transmission. In Mascali, two waves of transmission were observed over three days, being consistent with food/waterborne followed by person-to-person and/or airborne (vomit) transmission (Fig. 2). The first wave of cases presented on 20 May 2016 after a single shared meal, while the last case was detected on 22 May. Exposure to vomit may have been responsible for some of the secondary cases occurred after the first wave of illness. Vomiting produces aerosols that can be inhaled and swallowed. The extremely low infectious dose of NoV makes the inhalation of vomit aerosols an effective transmission route in settings where there is close contact, like on a bus trip [25]. All food consumed in the only shared lunch on 19 May at the resort was associated with the risk of disease (Table 2), but the association was inconclusive since all cases shared the same dishes included in the lunch menu on Day 1. Although many foods were potentially involved, our attention focused on the tap water. In fact, trivial information from local health authorities reported that local population avoided tap water consumption and relied on bottled water for drinking and cooking. Many small outbreaks of gastroenteritis involving restaurants and family groups had been reported in the same area, especially at the beginning of the summer season, when vacation houses and recreational activities were re-opened after a period of low water demand. Illness caused by NoV is common and if the level of water contamination is low, the number of cases remains limited and water can persist as unrecognised source of infection. If polluted water was served for drinking and used for cooking and washing dishes in the Mascali resort at Day 1 a generalised contamination of food and surfaces could have occurred. The potential role of the foodhandlers as a source of contamination could not be analysed in this study but should also be taken into account. If water contamination has already been present in the area for a long time, infections may have occurred among the local food-handlers. The analysis of water samples taken immediately after the outbreak revealed the presence of bacterial indicators of faecal contamination in the water used in the resort, with the only exception of a sample of filter purified water used both for cooking and as drinking water. When extending the sampling to other areas of the municipal water network, many other sites, mostly on the seaside, showed faecal pollution, which demonstrates a poor protection of the water network from faecal leakage (Fig. 1). NoV contamination of waters generally originates from sewage since these enteric pathogens, as other enteric viruses affecting humans, are mostly species-specific and are shed at high concentrations in the stools of patients. When the analyses on the stools of the students made it clear that the outbreak had been caused by NoV, the viral genome was also searched in appropriate volumes of water samples, but this happened when correction measures on the water network had already been taken. Nevertheless, a single water sample, obtained 2 months after the outbreak (1 August 2016), still contained genome fragments that could be correlated to the same NoV genotype causing the outbreak. Although the finding of NoV genome does not demonstrate that entire infecting viral particles were present in the water, it could indicate that a NoV belonging to the same genotype involved in the outbreak was still circulating in the local population several weeks after the outbreak and it was someway able to penetrate the water network, probably indicating continuing faecal contamination through leakage from sewage systems. NoV is particularly resistant in the environment and also to chlorine, therefore transmission through chlorinated water is not exceptional [9, 12, 26]. Drinking water from municipal supplies fed by lake or groundwater has already been identified in the past as the source of waterborne NoV outbreaks in Italy [27–29] and NoV is now considered one of the pathogens causing the largest number of affected consumers in drinking waterborne outbreaks [12]. NoV can persist in water longer than 15 days [30] and it is resistant to chlorine disinfection when free chlorine levels or pre-treatment processes are inadequate [31, 32]. Moreover, microfiltration filters labelled as certified by NSF Standards 53 or 58 (with a pore size ranging from 0.05 to 5 µm) are not effective in removing viruses and can even decrease chlorine concentration in filtered water [33]. Actually, the owner of the resort involved in our outbreak was encouraged to serve tap water as drinking water at the restaurant relying on the effectiveness of such a filter purifier. Unfortunately, we did not have a tap water sample from the restaurant of the resort taken before correction measures and whose volume was appropriate for virus detection. Therefore, we could not demonstrate the failure to eliminate viral pollution of the microfiltration system. In Mascali, correction measures on the water network consisted in the maintenance of the old chlorination device and in the addition of a new one at a key point located upstream and close to suspected illegal sewage dumping (Fig. 1), but as poor protection of the pipeline from sewage leakage risked to reproduce faecal pollution episodes, renovation of the municipal water system was also advised, as well as systematic control of illegal sewage dumping. The additional chlorination device was apparently the most effective intervention, since right after its activation faecal bacterial pollution indicators disappeared. However, NoV genome was detected in a water sample taken 6 weeks after the last detection of bacterial indicators. Unfortunately, detection of faecal contamination of drinking water is only based on the use of bacterial indicators despite no correlation between levels of enteric bacteria and enteric viruses has been found [34–36]. According to ISGEV enteric virus circulation surveillance data, in 2016 the circulation of GII.P2/GII.2 NoV strains in Palermo, home town of the travellers involved in the outbreak, was the highest ever observed from the starting of local NoV surveillance in 2004. GII.P2/GII.2 was the fourth most common genotype in 2016 (12.2% prevalence), following GII.P4/GII.4 Sydney 2012 (28.4%), GII.P16/GII.2 (18.9%) and GII.Pe/GII.4 Sydney 2012 (17.6%). GII.P2/GII.2 strains were mostly observed starting from July 2016, leaving space for the possibility that their increased circulation was linked to the introduction of the ‘Mascali clone’. However, a couple of GII.P2/GII.2 strains (including PA36/16) genetically close to the ‘Mascali clone’ had been already detected in January 2016 in Palermo and GII.P2/GII.2 NoV had sporadically been observed also in 2011 and 2014. Moreover, the circulation of GII.P2/GII.2 NoVs has been increasing worldwide over the last period [37]. Therefore, the unusually high circulation observed in Palermo might just reflect the global epidemiology of GII.P2/GII.2 strains. Noteworthy, the older GII.P2/GII.2 NoV strains detected in Palermo were closely related in the ORF2 (98.9% nt id) to the GII.P16/GII.2 Nashville strain (KY865307) which is supposed to have generated the recombinant GII.P16/GII.4 clone recently circulating epidemically worldwide [37, 38] and were more distantly associated (95.3% nt id) to the GII.2 capsid sequence (AB662868) of another putative GII.P16/GII.4 ancestor from Japan [39]. The increasing frequency of detection of waterborne NoV outbreaks calls for the introduction of routine monitoring for viruses of drinking waters and for the definition of affordable and reliable indicators for the presence of NoV in drinking waters. The use of ultrafiltration systems as a concentration method in order to recover viruses from environmental waters combined with RT-PCR to detect viral genomes has been demonstrated to grant a recovery rate of >50% of multiple viruses [40] and is at present one of the most sensitive approaches for the detection of NoVs, although the discrimination between viable and nonviable viruses remains a problem. The burden of NoV outbreaks, either food associated or waterborne, is still poorly defined due to the lack of specific diagnostic tools in many clinical laboratories. NoV epidemics should always be suspected when the main symptom is vomiting and the duration of disease is short (<48 h). Waterborne transmission is suggested by the rapid spread of disease and the high attack rate. Care should be taken to perform sensitive diagnostic tests on a sufficient number of stool samples in order to identify NoV as the causative agent of enteritis outbreaks, possibly preferring genome screening over antigen detection. Only the systematic detection of NoV outbreaks can guide in the choice of the safety procedures that could be most appropriate in containing such microbiological hazard.

What did the local authorities advise?

1555

Waterborne Norovirus outbreak at a seaside resort likely originating from municipal water distribution system failure

Abstract In May 2016 a Norovirus (NoV) gastroenteritis outbreak involved a high school class visiting a seaside resort near Taormina (Mascali, Sicily). Twenty-four students and a teacher were affected and 17 of them showed symptoms on the second day of the journey, while the others got ill within the following 2 days. Symptoms included vomiting, diarrhoea and fever, and 12 students required hospitalisation. Stool samples tested positive for NoV genome by Real-Time polymerase chain reaction assay in all 25 symptomatic subjects. The GII.P2/GII.2 NoV genotype was linked to the outbreak by ORF1/ORF2 sequence analysis. The epidemiological features of the outbreak were consistent with food/waterborne followed by person-to-person and/ or vomit transmission. Food consumed at a shared lunch on the first day of the trip was associated to illness and drinking un-bottled tap water was also considered as a risk factor. The analysis of water samples revealed the presence of bacterial indicators of faecal contamination in the water used in the resort as well as in other areas of the municipal water network, linking the NoV gastroenteritis outbreak to tap water pollution from sewage leakage. From a single water sample, an amplicon whose sequence corresponded to the capsid genotype recovered from patients could be obtained. Introduction Noroviruses (NoVs) are recognised as a leading cause of sporadic cases and outbreaks of acute gastroenteritis (AGE) across all age groups [1, 2]. They are transmitted primarily through the faecal-oral route either by direct person–person spread and with consumption of contaminated food or water [1, 3, 4]. NoVs are non-enveloped RNA viruses belonging to the Caliciviridae family and are classified into seven distinct genogroups (GI-GVII) on the basis of the major capsid protein VP1 sequence. GI, GII and GIV NoVs infect humans and they have been divided into more than 30 genotypes based on the sequences of polymerase (ORF1) and capsid proteins (ORF2) [5]. NoV GII.4 genotype has been associated with the vast majority of NoV related cases of AGE worldwide although many genotypes, belonging to both GI and GII genogroups are involved in foodborne and waterborne outbreaks [6, 7]. NoVs are highly contagious owing to their high stability in the environment and their ability to infect at low doses. In fact, just 10–100 virions of NoV may be sufficient to cause AGE [8, 9]. NoVs are considered the major cause of water-related disease in semi-closed institutions, including restaurants, schools, hospitals, nursing homes and cruise ships [10–12]. To confirm NoV as a cause of outbreaks, the presence of four epidemiological features of disease, vomiting in more than half of affected persons, a mean incubation period of 24/48 h, a mean duration of illness of 12–60 h and the absence of bacterial pathogens in stool culture, should be evaluated [13]. Waterborne NoV outbreaks can be caused by contaminated drinking water despite correct chlorine concentration and in the absence of coliform bacteria [10, 14–17]. The aim of this study was to investigate an outbreak of NoV gastroenteritis associated with drinking water involving a high school class travelling from Palermo (West Sicily) to Acireale (East Sicily) on an educational tour in May 2016. Stool samples were collected soon after the onset of AGE symptoms to ascertain the aetiology of illness. The source of infection was investigated collecting epidemiological data through an online questionnaire and, in order to explore a possible transmission route, laboratory tests on water from the municipal network were performed. Material and methods Epidemiological investigation From 20 to 22 May 2016 an AGE outbreak occurred in Mascali, a small town near Taormina, involving a school class of 28 students, 13 girls and 15 boys aged between 14 and 18 years and a 54-year-old female teacher, during a trip to participate in a theatre festival in Acireale. A total of 25/29 persons were affected by nausea, vomiting, abdominal cramps, diarrhoea and fever while they were staying at a seaside resort in Mascali, on the way to Acireale. To obtain data on symptoms and the correlated risk factors all participants filled an online questionnaire investigating personal data (age and sex), travel and accommodation details and food consumption over 48 h before symptoms presentation, time of onset and duration of symptoms, therapy and/or hospitalisation required and history of contact with people affected by diarrhoea and/or vomiting (Table 1). The severity of illness was analysed by the Vesikari severity scale allowing to distinguish non-severe (score <11) and severe (score ⩾11) AGE (Table 1). Vesikari score was calculated from maximum number of episodes/day and duration of both diarrhoea and vomiting (score 0–3 for each of the 4 parameters), temperature (score 0–3), level of dehydration (score 0–3) and treatment required (score 0–2) [18]. A case-control study was conducted to establish if food consumed before the onset of symptoms was associated with the disease. A case was defined as a student attending the tour and reporting diarrhoea and/or vomiting within 36 h from the first shared meal during the tour, while students who did not present any of the previous symptoms or those showing the first symptoms with a delay of ⩾48 h (secondary cases) were used as controls. Variables were studied using odds ratio (OR) and its 95% CI and statistical associations were determined using the χ2 test, with the level of statistical significance set at 0.05 (Table 2). NoV detection in stool samples A stool sample was obtained from each of the 29 travellers, soon after (3–5 days) the onset of the outbreak. No stool samples could be obtained from the food-handlers working at the resort. All samples were tested for NoV at the Enteric Viruses Laboratory of the University of Palermo, member of the Italian Study Group for Enteric Viruses (ISGEV), by a Real-Time polymerase chain reaction (RT-PCR) assay that allows to differentiate GI and GII NoVs [19]. Total RNA was extracted from stool samples using QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) and cDNA was generated by MMLV reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were considered positive for NoVs when cycle threshold (ct) of RT-PCR was ⩽30 and lowpositive when 30<ct⩽35. For all NoV-positive samples two genomic fragments were amplified by conventional RT-PCR using Superscript III One-Step RT-PCR system with Platinum Taq DNA polymerase (Invitrogen) and primers JV12a/JV13b, targeting 326 bps of the RdRp region A [20] and either COG1F/ GISKR or COG2F/GIISKR, amplifying a 315 bps 5′portion of GI or GII VP1 protein [19, 21]. Environmental investigation From May through August 2016, a total of 32 tap water samples were collected from the Mascali municipal network. In particular, seven samples were collected at the resort where the outbreak occurred and from its water entry (R), two in the city centre of Mascali (M), 15 in the village of Fondachello (F), on the seaside, three in Carrabba village (C), far from the coast, and two and three, respectively, from the water well ‘Carlino’ (W) and the source ‘Bufardo and Torregrossa’ (S) which feed the municipal system and provide drinking water to the Mascali area (Fig. 1, Table 3). One litre water samples were sent by the local health authorities to the SIALAB laboratory in Avola, Siracusa and 100 ml aliquots were screened for coliforms, E. coli and enterococci as faecal indicators, by standard bacteriological methods according to UNI EN ISO 9308-1:2014. Unfortunately, although the suspect of a water source for the viral outbreak was quickly communicated, only in August and November 2016, following the implementation of control measures, local authorities provided water samplings of appropriate volume to perform viral tests (10 l). Five water samples of 10 l each were collected from different sampling sites in F and R areas for viral screening (Table 3) and were concentrated according to the Italian National Institute of Health (Istituto Superiore di Sanità, ISS) protocols [22] through an ultra-filtration system (Sartoflow® Slice 200 Benchtop Crossflow System, Sartorius AG, Goettingen, Germany), using appropriate membranes (SG Hydrosart 10 kDa) pretreated with 300 ml of 3% Beef Extract (BE) at pH 7, by recirculating it for 10 min in the system, at a pressure not higher than 1.30 bar. Elution and recovery of viruses attached to the membranes were obtained by recirculating 200 ml of 3% BE at pH 9.5 for 3 h until a 10–12 ml volume of eluate was obtained. The pH of the eluate was then checked with litmus paper and if necessary brought to the value of 7 with HCl. Concentrated water samples were simultaneously tested for NoV in three laboratories: Enteric Viruses Laboratory, University of Palermo; Istituto Zooprofilattico Sperimentale della Sicilia (IZS), Palermo; and Istituto Superiore di Sanità (ISS), Rome. Total RNA was extracted from 500 and 560 µl of a concentrated water sample using BioMèrieux NucliSENS® Magnetic Extraction System (BioMèrieux) and QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) respectively, according to manufacturer’s instructions. Bonded RNA was eluted by 100 µl of Elution Buffer. Amplifications were performed at least in duplicate on whole and 1:10 diluted extracts. A Real-Time RT-PCR assay according to ISO/TS 15216- 2:2013, two in-house Real-Time RT-PCR protocols and the One-Step RT-PCR methods already used for amplifying ORF1 and ORF2 fragments in NoV-positive faecal samples were performed [19–21, 23]. Nucleotide sequences analysis and phylogenetic analysis PCR amplicons were purified by commercial columns (CleanSweep PCR Purification, Thermo Fisher Scientific, Vilnius, Lithuania) and sequenced by Sanger method and all nucleotide sequences were aligned using ClustalW. To identify NoV genotypes, RdRp and VP1 sequences were submitted to the RIVM NoV genotyping tool (http://www.rivm.nl/mpf/norovirus/typingtool). Phylogenetic trees were constructed using MEGA 7 and Neighbour-Joining method (bootstrap 1000 replicates). Sequences from stool and water samples are available at NCBI with the following accession numbers: MG517455 for the major capsid protein obtained from a water sample; MG517456 and MG517461, respectively, for the major capsid protein and RNA-dependent-RNA-polymerase, as representatives of the sequences obtained from the stool samples of those involved in the outbreak. Control measures After outbreak notification, on 21 May, the resort was closed for 2 days. On 25 May, following the first results of the bacteriological analyses on municipal water samples, the use of municipal water for human consumption was temporarily restricted (use of water was permitted only for cleaning purposes) in the Carrabba and Fondachello villages and in the area surrounding the resort. From 25 May the reservoir collecting the water from the Carlino well and its chlorination device were submitted to maintenance work and a new chlorination device was installed on the conduit bringing the water from the Bufardo source. Actually, the Bufardo source had been connected to the municipal network in 2003 but unfortunately a dedicated chlorination station had not been planned at that time. Therefore, unchlorinated water entered the system near Fondachello village on the seaside (Fig. 1), possibly diluting chlorine concentrations in water treated at the Carlino reservoir. Moreover, the exact path of the branch of the water network situated downstream the Bufardo source water connection was not marked in the maps of the municipality (Fig. 1, red line), indicating that this section had never been submitted to maintenance. The area surrounding the unmaintained section of the water pipeline was scattered with unauthorised cottages that were not connected to the sewerage systems. Likely, untreated waste from cesspools was unloaded directly into the soil, generating a risk of contamination of the municipal water system through leakage. No work had been made on the water pipelines before the outbreak but sections of the network had been temporarily closed during the winter season on the seaside (Fondachello). Local health authorities advised the connection of all households in the area to the sewerage system and the replacement of the water system section without maintenance. Health authorities also raised attention on AGE cases notification at all local Emergency Departments and Family Practitioners but there was no further evidence of AGE outbreaks in the months following the intervention. Results Data collected through the online questionnaire compiled 7–10 days after the outbreak revealed that none of the 28 students and the teacher taking part in the school trip showed gastroenteritis symptoms at the beginning of the journey and they had no contact with people who suffered from diarrhoea and/or vomiting in the previous days. Travellers had their first shared meal on 19 May (Day 1), 2016, at lunch on their arrival to the resort in Mascali. The list of foods included in the menu is shown in Table 2. Conversely, the dinner on Day 1 was not shared and the students went to dine in three different fast food restaurants in Acireale. On 20 May (Day 2), 17 students (58.6% of the travellers) presented with vomit and/or diarrhoea, falling ill 26–36 h after lunch on Day 1, strongly suggesting that exposure occurred on the first shared meal. Some students also vomited in the presence of the other travellers on the bus bringing them to Acireale. Five more students were symptomatic in the early hours of the following day (21 May, Day 3, from 0:30 am to 8:00 am), as a continuation of the wave of illnesses of the previous day. The teacher and one more student fell ill at 8:00 pm and 11:00 pm of Day 3, respectively, for a second wave of illnesses occurring >48 h from the first shared meal and <36 h from the first cases, with the last student falling sick on 22 May (Day 4) at 3:00 am (Fig. 2, Table 1). On the whole, the mean incubation period was 33.2 h. Based on the online survey, the severity of illness was evaluated by the Vesikari severity scale, grouping symptomatic travellers into severe (11/25 = 44%) and non-severe (14/25 = 56%) cases (Table 1). Vomiting and fever were the predominant symptoms, affecting 22 out of 25 (88%) symptomatic subjects, while diarrhoea was present in 14 out of 25 (56%). Hospitalisation for rehydration therapy was necessary for 13 students (91% [10/11] of severe and 21% [3/14] of non-severe cases). Emergency room (ER) or family practitioner visits were necessary for 21 students and 13 (62%) of them initiated an antibiotic therapy. Two students indicated that other family members had gastroenteric symptoms more than 24 h after their return home, on Day 3, 4 and 5. All those involved in the outbreak recovered within 4 days, with an average duration of illness of 38.4 h. A case-control study of the risk associated with eating foods at lunch on Day 1 or with attending different restaurants at dinner on Day 1 was attempted. A total of 19 subjects filled the section of the online survey concerning food consumption. Fourteen of the responders were among the 22 cases falling ill within 36 h from the first shared meal (first wave), while the remaining five responders were three travellers with supposed secondary infections (showing symptoms with a delay ⩾48 h; second wave) and two asymptomatic students that were used as controls. No statistically significant associations with any food items served at lunch on Day 1 was identified (P < 0.05; Table 2). However, all food consumed on that occasion was to some extent associated with the risk of disease, the highest OR being obtained with chicken (OR: 20.7143, P: 0.0678; Table 2). Actually, most of the cases ate all foods included in the menu, while one of the two asymptomatic controls did not participate in the meal and the other had a pasta dish only. No statistically significant association was found for attending a particular fast food restaurant at the dinner on Day 1 (OR: 2.5, P: 0.3635). The Real-Time RT-PCR assay allowed to detect GII genogroup NoV genome in all the 25 samples from symptomatic subjects, although two of them showed high Ct values, indicating lower shedding of viral genomes (Table 1). The samples from four asymptomatic students were negative for both NoV genogroups. By sequence analysis, a NoV genotype could be defined for 20/ 25 NoV-positive stool samples. All NoV strains from travellers involved in the outbreaks belonged to the same genotype, GII.P2/GII.2, showing >99% nucleotide identities (nt id) in both fragments (ORF1 and ORF2) analysed. Thirty-two water samples were collected and analysed for bacterial contamination (Table 3). Seven samples were taken from the resort (site R) and the remaining 25 from five other sites (C, F, M, S, W), representative of the entire municipal water network. The majority (10/12) of the water samples collected immediately after the onset of the outbreak, between 21 May and 16 June 2016, at the resort (site R) and at sites F and C, located close to the resort, on the seaside and in the interior, respectively, were positive for coliform bacteria. More than 1000 colony-forming units (CFU)/ 100 ml (up to 2700 CFU/100 ml) were detected in four samples, while their absence is required in 100 ml by the Standard of Water Quality in Italy and seven of 11 samples tested were also Escherichia coli-positive (up to 900 CFU/100 ml). Enterococci were searched in 11 samples and three of them were found positive for this faecal contamination indicator. From 7 July to 16 August 2016, after control measures on the water network were undertaken, 13 water samples were collected for follow-up laboratory examination of the water system at F, R and C sites. All of them were negative for bacterial indicators of faecal contamination. No faecal bacteria were identified in the seven samples collected from M, W and S sites, including the well and the source that fed the water network, at any sampling time, before and after implementing the control measures. The research of NoV genome was conducted on concentrated water samples collected on 1 August from R and F sites (Table 3). NoV genome was detected in a single sample collected from the water entry of the resort involved in the outbreak (site R). Actually, only two out of 26 One-Step RT-PCR repetitions were positive for NoV viral capsid gene. The two ORF2-positive results were obtained at the University of Palermo Laboratory using both 1:10 diluted and undiluted extracted RNA. The sequence analysis of the partial ORF2 portion (289 bp) indicated the presence of the same GII.2 capsid genotype detected in the stool samples of the students involved in the outbreak, showing 98.9% nt id. No ORF1 NoV sequence could be obtained from the NoV-positive water sample. Follow-up analyses for NoV genome, conducted on water samples collected on 20 November both at R and F sites found no positive samples. By sequence alignment, the NoV sequences obtained from Mascali outbreak samples were phylogenetically closely related in both ORF1 and ORF2 (99 and 98.9% of nt id, respectively) to a GII.P2/GII.2 strain responsible for a sporadic paediatric gastroenteritis case occurred in Palermo in January 2016 (PA 36/16, accession numbers MG517464 and MG517460, for ORF1 and ORF2, respectively) and previously detected at the Enteric Viruses Laboratory, University of Palermo. Whereas, three GII.P2/GII.2 Italian strains (PA 280/11, PA 366/11 and PA 3/ 14, accession numbers MG517462, MG517465 and MG517463, for ORF1 and MG517457 to −59, for ORF2, respectively), detected in Palermo in 2011 and 2014, were more distantly associated with the outbreak strain both in RdRp and VP1 (90% to 97.4% nt id and 96.1% to 98.1% nt id, respectively, in ORF1 and ORF2). Discussion Epidemiological features, such as the presence of vomiting in more than half (88%) of symptomatic subjects, the 33.2 h average incubation period and the 38.4 h average duration of illness, suggested a NoV aetiology for the investigated outbreak [13]. The detection of NoV genome in all 25 faecal samples from symptomatic subjects fully complies with the requirement of detection of at least one positive sample out of four tested that is considered to be sufficient to ascertain the aetiology of a NoV outbreak [24]. Moreover, molecular typing indicated that a single viral strain, belonging to genotype GII.P2/GII.2, infected all symptomatic subjects involved in the outbreak. NoVs are highly infectious and can easily spread from person to person but also through contaminated food or water. Therefore, settings such as health care institutions, schools and touristic resorts are mainly affected, since close contact and shared food services facilitate transmission. In Mascali, two waves of transmission were observed over three days, being consistent with food/waterborne followed by person-to-person and/or airborne (vomit) transmission (Fig. 2). The first wave of cases presented on 20 May 2016 after a single shared meal, while the last case was detected on 22 May. Exposure to vomit may have been responsible for some of the secondary cases occurred after the first wave of illness. Vomiting produces aerosols that can be inhaled and swallowed. The extremely low infectious dose of NoV makes the inhalation of vomit aerosols an effective transmission route in settings where there is close contact, like on a bus trip [25]. All food consumed in the only shared lunch on 19 May at the resort was associated with the risk of disease (Table 2), but the association was inconclusive since all cases shared the same dishes included in the lunch menu on Day 1. Although many foods were potentially involved, our attention focused on the tap water. In fact, trivial information from local health authorities reported that local population avoided tap water consumption and relied on bottled water for drinking and cooking. Many small outbreaks of gastroenteritis involving restaurants and family groups had been reported in the same area, especially at the beginning of the summer season, when vacation houses and recreational activities were re-opened after a period of low water demand. Illness caused by NoV is common and if the level of water contamination is low, the number of cases remains limited and water can persist as unrecognised source of infection. If polluted water was served for drinking and used for cooking and washing dishes in the Mascali resort at Day 1 a generalised contamination of food and surfaces could have occurred. The potential role of the foodhandlers as a source of contamination could not be analysed in this study but should also be taken into account. If water contamination has already been present in the area for a long time, infections may have occurred among the local food-handlers. The analysis of water samples taken immediately after the outbreak revealed the presence of bacterial indicators of faecal contamination in the water used in the resort, with the only exception of a sample of filter purified water used both for cooking and as drinking water. When extending the sampling to other areas of the municipal water network, many other sites, mostly on the seaside, showed faecal pollution, which demonstrates a poor protection of the water network from faecal leakage (Fig. 1). NoV contamination of waters generally originates from sewage since these enteric pathogens, as other enteric viruses affecting humans, are mostly species-specific and are shed at high concentrations in the stools of patients. When the analyses on the stools of the students made it clear that the outbreak had been caused by NoV, the viral genome was also searched in appropriate volumes of water samples, but this happened when correction measures on the water network had already been taken. Nevertheless, a single water sample, obtained 2 months after the outbreak (1 August 2016), still contained genome fragments that could be correlated to the same NoV genotype causing the outbreak. Although the finding of NoV genome does not demonstrate that entire infecting viral particles were present in the water, it could indicate that a NoV belonging to the same genotype involved in the outbreak was still circulating in the local population several weeks after the outbreak and it was someway able to penetrate the water network, probably indicating continuing faecal contamination through leakage from sewage systems. NoV is particularly resistant in the environment and also to chlorine, therefore transmission through chlorinated water is not exceptional [9, 12, 26]. Drinking water from municipal supplies fed by lake or groundwater has already been identified in the past as the source of waterborne NoV outbreaks in Italy [27–29] and NoV is now considered one of the pathogens causing the largest number of affected consumers in drinking waterborne outbreaks [12]. NoV can persist in water longer than 15 days [30] and it is resistant to chlorine disinfection when free chlorine levels or pre-treatment processes are inadequate [31, 32]. Moreover, microfiltration filters labelled as certified by NSF Standards 53 or 58 (with a pore size ranging from 0.05 to 5 µm) are not effective in removing viruses and can even decrease chlorine concentration in filtered water [33]. Actually, the owner of the resort involved in our outbreak was encouraged to serve tap water as drinking water at the restaurant relying on the effectiveness of such a filter purifier. Unfortunately, we did not have a tap water sample from the restaurant of the resort taken before correction measures and whose volume was appropriate for virus detection. Therefore, we could not demonstrate the failure to eliminate viral pollution of the microfiltration system. In Mascali, correction measures on the water network consisted in the maintenance of the old chlorination device and in the addition of a new one at a key point located upstream and close to suspected illegal sewage dumping (Fig. 1), but as poor protection of the pipeline from sewage leakage risked to reproduce faecal pollution episodes, renovation of the municipal water system was also advised, as well as systematic control of illegal sewage dumping. The additional chlorination device was apparently the most effective intervention, since right after its activation faecal bacterial pollution indicators disappeared. However, NoV genome was detected in a water sample taken 6 weeks after the last detection of bacterial indicators. Unfortunately, detection of faecal contamination of drinking water is only based on the use of bacterial indicators despite no correlation between levels of enteric bacteria and enteric viruses has been found [34–36]. According to ISGEV enteric virus circulation surveillance data, in 2016 the circulation of GII.P2/GII.2 NoV strains in Palermo, home town of the travellers involved in the outbreak, was the highest ever observed from the starting of local NoV surveillance in 2004. GII.P2/GII.2 was the fourth most common genotype in 2016 (12.2% prevalence), following GII.P4/GII.4 Sydney 2012 (28.4%), GII.P16/GII.2 (18.9%) and GII.Pe/GII.4 Sydney 2012 (17.6%). GII.P2/GII.2 strains were mostly observed starting from July 2016, leaving space for the possibility that their increased circulation was linked to the introduction of the ‘Mascali clone’. However, a couple of GII.P2/GII.2 strains (including PA36/16) genetically close to the ‘Mascali clone’ had been already detected in January 2016 in Palermo and GII.P2/GII.2 NoV had sporadically been observed also in 2011 and 2014. Moreover, the circulation of GII.P2/GII.2 NoVs has been increasing worldwide over the last period [37]. Therefore, the unusually high circulation observed in Palermo might just reflect the global epidemiology of GII.P2/GII.2 strains. Noteworthy, the older GII.P2/GII.2 NoV strains detected in Palermo were closely related in the ORF2 (98.9% nt id) to the GII.P16/GII.2 Nashville strain (KY865307) which is supposed to have generated the recombinant GII.P16/GII.4 clone recently circulating epidemically worldwide [37, 38] and were more distantly associated (95.3% nt id) to the GII.2 capsid sequence (AB662868) of another putative GII.P16/GII.4 ancestor from Japan [39]. The increasing frequency of detection of waterborne NoV outbreaks calls for the introduction of routine monitoring for viruses of drinking waters and for the definition of affordable and reliable indicators for the presence of NoV in drinking waters. The use of ultrafiltration systems as a concentration method in order to recover viruses from environmental waters combined with RT-PCR to detect viral genomes has been demonstrated to grant a recovery rate of >50% of multiple viruses [40] and is at present one of the most sensitive approaches for the detection of NoVs, although the discrimination between viable and nonviable viruses remains a problem. The burden of NoV outbreaks, either food associated or waterborne, is still poorly defined due to the lack of specific diagnostic tools in many clinical laboratories. NoV epidemics should always be suspected when the main symptom is vomiting and the duration of disease is short (<48 h). Waterborne transmission is suggested by the rapid spread of disease and the high attack rate. Care should be taken to perform sensitive diagnostic tests on a sufficient number of stool samples in order to identify NoV as the causative agent of enteritis outbreaks, possibly preferring genome screening over antigen detection. Only the systematic detection of NoV outbreaks can guide in the choice of the safety procedures that could be most appropriate in containing such microbiological hazard.

What were the control measures?

1556

Waterborne Norovirus outbreak at a seaside resort likely originating from municipal water distribution system failure

Abstract In May 2016 a Norovirus (NoV) gastroenteritis outbreak involved a high school class visiting a seaside resort near Taormina (Mascali, Sicily). Twenty-four students and a teacher were affected and 17 of them showed symptoms on the second day of the journey, while the others got ill within the following 2 days. Symptoms included vomiting, diarrhoea and fever, and 12 students required hospitalisation. Stool samples tested positive for NoV genome by Real-Time polymerase chain reaction assay in all 25 symptomatic subjects. The GII.P2/GII.2 NoV genotype was linked to the outbreak by ORF1/ORF2 sequence analysis. The epidemiological features of the outbreak were consistent with food/waterborne followed by person-to-person and/ or vomit transmission. Food consumed at a shared lunch on the first day of the trip was associated to illness and drinking un-bottled tap water was also considered as a risk factor. The analysis of water samples revealed the presence of bacterial indicators of faecal contamination in the water used in the resort as well as in other areas of the municipal water network, linking the NoV gastroenteritis outbreak to tap water pollution from sewage leakage. From a single water sample, an amplicon whose sequence corresponded to the capsid genotype recovered from patients could be obtained. Introduction Noroviruses (NoVs) are recognised as a leading cause of sporadic cases and outbreaks of acute gastroenteritis (AGE) across all age groups [1, 2]. They are transmitted primarily through the faecal-oral route either by direct person–person spread and with consumption of contaminated food or water [1, 3, 4]. NoVs are non-enveloped RNA viruses belonging to the Caliciviridae family and are classified into seven distinct genogroups (GI-GVII) on the basis of the major capsid protein VP1 sequence. GI, GII and GIV NoVs infect humans and they have been divided into more than 30 genotypes based on the sequences of polymerase (ORF1) and capsid proteins (ORF2) [5]. NoV GII.4 genotype has been associated with the vast majority of NoV related cases of AGE worldwide although many genotypes, belonging to both GI and GII genogroups are involved in foodborne and waterborne outbreaks [6, 7]. NoVs are highly contagious owing to their high stability in the environment and their ability to infect at low doses. In fact, just 10–100 virions of NoV may be sufficient to cause AGE [8, 9]. NoVs are considered the major cause of water-related disease in semi-closed institutions, including restaurants, schools, hospitals, nursing homes and cruise ships [10–12]. To confirm NoV as a cause of outbreaks, the presence of four epidemiological features of disease, vomiting in more than half of affected persons, a mean incubation period of 24/48 h, a mean duration of illness of 12–60 h and the absence of bacterial pathogens in stool culture, should be evaluated [13]. Waterborne NoV outbreaks can be caused by contaminated drinking water despite correct chlorine concentration and in the absence of coliform bacteria [10, 14–17]. The aim of this study was to investigate an outbreak of NoV gastroenteritis associated with drinking water involving a high school class travelling from Palermo (West Sicily) to Acireale (East Sicily) on an educational tour in May 2016. Stool samples were collected soon after the onset of AGE symptoms to ascertain the aetiology of illness. The source of infection was investigated collecting epidemiological data through an online questionnaire and, in order to explore a possible transmission route, laboratory tests on water from the municipal network were performed. Material and methods Epidemiological investigation From 20 to 22 May 2016 an AGE outbreak occurred in Mascali, a small town near Taormina, involving a school class of 28 students, 13 girls and 15 boys aged between 14 and 18 years and a 54-year-old female teacher, during a trip to participate in a theatre festival in Acireale. A total of 25/29 persons were affected by nausea, vomiting, abdominal cramps, diarrhoea and fever while they were staying at a seaside resort in Mascali, on the way to Acireale. To obtain data on symptoms and the correlated risk factors all participants filled an online questionnaire investigating personal data (age and sex), travel and accommodation details and food consumption over 48 h before symptoms presentation, time of onset and duration of symptoms, therapy and/or hospitalisation required and history of contact with people affected by diarrhoea and/or vomiting (Table 1). The severity of illness was analysed by the Vesikari severity scale allowing to distinguish non-severe (score <11) and severe (score ⩾11) AGE (Table 1). Vesikari score was calculated from maximum number of episodes/day and duration of both diarrhoea and vomiting (score 0–3 for each of the 4 parameters), temperature (score 0–3), level of dehydration (score 0–3) and treatment required (score 0–2) [18]. A case-control study was conducted to establish if food consumed before the onset of symptoms was associated with the disease. A case was defined as a student attending the tour and reporting diarrhoea and/or vomiting within 36 h from the first shared meal during the tour, while students who did not present any of the previous symptoms or those showing the first symptoms with a delay of ⩾48 h (secondary cases) were used as controls. Variables were studied using odds ratio (OR) and its 95% CI and statistical associations were determined using the χ2 test, with the level of statistical significance set at 0.05 (Table 2). NoV detection in stool samples A stool sample was obtained from each of the 29 travellers, soon after (3–5 days) the onset of the outbreak. No stool samples could be obtained from the food-handlers working at the resort. All samples were tested for NoV at the Enteric Viruses Laboratory of the University of Palermo, member of the Italian Study Group for Enteric Viruses (ISGEV), by a Real-Time polymerase chain reaction (RT-PCR) assay that allows to differentiate GI and GII NoVs [19]. Total RNA was extracted from stool samples using QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) and cDNA was generated by MMLV reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were considered positive for NoVs when cycle threshold (ct) of RT-PCR was ⩽30 and lowpositive when 30<ct⩽35. For all NoV-positive samples two genomic fragments were amplified by conventional RT-PCR using Superscript III One-Step RT-PCR system with Platinum Taq DNA polymerase (Invitrogen) and primers JV12a/JV13b, targeting 326 bps of the RdRp region A [20] and either COG1F/ GISKR or COG2F/GIISKR, amplifying a 315 bps 5′portion of GI or GII VP1 protein [19, 21]. Environmental investigation From May through August 2016, a total of 32 tap water samples were collected from the Mascali municipal network. In particular, seven samples were collected at the resort where the outbreak occurred and from its water entry (R), two in the city centre of Mascali (M), 15 in the village of Fondachello (F), on the seaside, three in Carrabba village (C), far from the coast, and two and three, respectively, from the water well ‘Carlino’ (W) and the source ‘Bufardo and Torregrossa’ (S) which feed the municipal system and provide drinking water to the Mascali area (Fig. 1, Table 3). One litre water samples were sent by the local health authorities to the SIALAB laboratory in Avola, Siracusa and 100 ml aliquots were screened for coliforms, E. coli and enterococci as faecal indicators, by standard bacteriological methods according to UNI EN ISO 9308-1:2014. Unfortunately, although the suspect of a water source for the viral outbreak was quickly communicated, only in August and November 2016, following the implementation of control measures, local authorities provided water samplings of appropriate volume to perform viral tests (10 l). Five water samples of 10 l each were collected from different sampling sites in F and R areas for viral screening (Table 3) and were concentrated according to the Italian National Institute of Health (Istituto Superiore di Sanità, ISS) protocols [22] through an ultra-filtration system (Sartoflow® Slice 200 Benchtop Crossflow System, Sartorius AG, Goettingen, Germany), using appropriate membranes (SG Hydrosart 10 kDa) pretreated with 300 ml of 3% Beef Extract (BE) at pH 7, by recirculating it for 10 min in the system, at a pressure not higher than 1.30 bar. Elution and recovery of viruses attached to the membranes were obtained by recirculating 200 ml of 3% BE at pH 9.5 for 3 h until a 10–12 ml volume of eluate was obtained. The pH of the eluate was then checked with litmus paper and if necessary brought to the value of 7 with HCl. Concentrated water samples were simultaneously tested for NoV in three laboratories: Enteric Viruses Laboratory, University of Palermo; Istituto Zooprofilattico Sperimentale della Sicilia (IZS), Palermo; and Istituto Superiore di Sanità (ISS), Rome. Total RNA was extracted from 500 and 560 µl of a concentrated water sample using BioMèrieux NucliSENS® Magnetic Extraction System (BioMèrieux) and QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) respectively, according to manufacturer’s instructions. Bonded RNA was eluted by 100 µl of Elution Buffer. Amplifications were performed at least in duplicate on whole and 1:10 diluted extracts. A Real-Time RT-PCR assay according to ISO/TS 15216- 2:2013, two in-house Real-Time RT-PCR protocols and the One-Step RT-PCR methods already used for amplifying ORF1 and ORF2 fragments in NoV-positive faecal samples were performed [19–21, 23]. Nucleotide sequences analysis and phylogenetic analysis PCR amplicons were purified by commercial columns (CleanSweep PCR Purification, Thermo Fisher Scientific, Vilnius, Lithuania) and sequenced by Sanger method and all nucleotide sequences were aligned using ClustalW. To identify NoV genotypes, RdRp and VP1 sequences were submitted to the RIVM NoV genotyping tool (http://www.rivm.nl/mpf/norovirus/typingtool). Phylogenetic trees were constructed using MEGA 7 and Neighbour-Joining method (bootstrap 1000 replicates). Sequences from stool and water samples are available at NCBI with the following accession numbers: MG517455 for the major capsid protein obtained from a water sample; MG517456 and MG517461, respectively, for the major capsid protein and RNA-dependent-RNA-polymerase, as representatives of the sequences obtained from the stool samples of those involved in the outbreak. Control measures After outbreak notification, on 21 May, the resort was closed for 2 days. On 25 May, following the first results of the bacteriological analyses on municipal water samples, the use of municipal water for human consumption was temporarily restricted (use of water was permitted only for cleaning purposes) in the Carrabba and Fondachello villages and in the area surrounding the resort. From 25 May the reservoir collecting the water from the Carlino well and its chlorination device were submitted to maintenance work and a new chlorination device was installed on the conduit bringing the water from the Bufardo source. Actually, the Bufardo source had been connected to the municipal network in 2003 but unfortunately a dedicated chlorination station had not been planned at that time. Therefore, unchlorinated water entered the system near Fondachello village on the seaside (Fig. 1), possibly diluting chlorine concentrations in water treated at the Carlino reservoir. Moreover, the exact path of the branch of the water network situated downstream the Bufardo source water connection was not marked in the maps of the municipality (Fig. 1, red line), indicating that this section had never been submitted to maintenance. The area surrounding the unmaintained section of the water pipeline was scattered with unauthorised cottages that were not connected to the sewerage systems. Likely, untreated waste from cesspools was unloaded directly into the soil, generating a risk of contamination of the municipal water system through leakage. No work had been made on the water pipelines before the outbreak but sections of the network had been temporarily closed during the winter season on the seaside (Fondachello). Local health authorities advised the connection of all households in the area to the sewerage system and the replacement of the water system section without maintenance. Health authorities also raised attention on AGE cases notification at all local Emergency Departments and Family Practitioners but there was no further evidence of AGE outbreaks in the months following the intervention. Results Data collected through the online questionnaire compiled 7–10 days after the outbreak revealed that none of the 28 students and the teacher taking part in the school trip showed gastroenteritis symptoms at the beginning of the journey and they had no contact with people who suffered from diarrhoea and/or vomiting in the previous days. Travellers had their first shared meal on 19 May (Day 1), 2016, at lunch on their arrival to the resort in Mascali. The list of foods included in the menu is shown in Table 2. Conversely, the dinner on Day 1 was not shared and the students went to dine in three different fast food restaurants in Acireale. On 20 May (Day 2), 17 students (58.6% of the travellers) presented with vomit and/or diarrhoea, falling ill 26–36 h after lunch on Day 1, strongly suggesting that exposure occurred on the first shared meal. Some students also vomited in the presence of the other travellers on the bus bringing them to Acireale. Five more students were symptomatic in the early hours of the following day (21 May, Day 3, from 0:30 am to 8:00 am), as a continuation of the wave of illnesses of the previous day. The teacher and one more student fell ill at 8:00 pm and 11:00 pm of Day 3, respectively, for a second wave of illnesses occurring >48 h from the first shared meal and <36 h from the first cases, with the last student falling sick on 22 May (Day 4) at 3:00 am (Fig. 2, Table 1). On the whole, the mean incubation period was 33.2 h. Based on the online survey, the severity of illness was evaluated by the Vesikari severity scale, grouping symptomatic travellers into severe (11/25 = 44%) and non-severe (14/25 = 56%) cases (Table 1). Vomiting and fever were the predominant symptoms, affecting 22 out of 25 (88%) symptomatic subjects, while diarrhoea was present in 14 out of 25 (56%). Hospitalisation for rehydration therapy was necessary for 13 students (91% [10/11] of severe and 21% [3/14] of non-severe cases). Emergency room (ER) or family practitioner visits were necessary for 21 students and 13 (62%) of them initiated an antibiotic therapy. Two students indicated that other family members had gastroenteric symptoms more than 24 h after their return home, on Day 3, 4 and 5. All those involved in the outbreak recovered within 4 days, with an average duration of illness of 38.4 h. A case-control study of the risk associated with eating foods at lunch on Day 1 or with attending different restaurants at dinner on Day 1 was attempted. A total of 19 subjects filled the section of the online survey concerning food consumption. Fourteen of the responders were among the 22 cases falling ill within 36 h from the first shared meal (first wave), while the remaining five responders were three travellers with supposed secondary infections (showing symptoms with a delay ⩾48 h; second wave) and two asymptomatic students that were used as controls. No statistically significant associations with any food items served at lunch on Day 1 was identified (P < 0.05; Table 2). However, all food consumed on that occasion was to some extent associated with the risk of disease, the highest OR being obtained with chicken (OR: 20.7143, P: 0.0678; Table 2). Actually, most of the cases ate all foods included in the menu, while one of the two asymptomatic controls did not participate in the meal and the other had a pasta dish only. No statistically significant association was found for attending a particular fast food restaurant at the dinner on Day 1 (OR: 2.5, P: 0.3635). The Real-Time RT-PCR assay allowed to detect GII genogroup NoV genome in all the 25 samples from symptomatic subjects, although two of them showed high Ct values, indicating lower shedding of viral genomes (Table 1). The samples from four asymptomatic students were negative for both NoV genogroups. By sequence analysis, a NoV genotype could be defined for 20/ 25 NoV-positive stool samples. All NoV strains from travellers involved in the outbreaks belonged to the same genotype, GII.P2/GII.2, showing >99% nucleotide identities (nt id) in both fragments (ORF1 and ORF2) analysed. Thirty-two water samples were collected and analysed for bacterial contamination (Table 3). Seven samples were taken from the resort (site R) and the remaining 25 from five other sites (C, F, M, S, W), representative of the entire municipal water network. The majority (10/12) of the water samples collected immediately after the onset of the outbreak, between 21 May and 16 June 2016, at the resort (site R) and at sites F and C, located close to the resort, on the seaside and in the interior, respectively, were positive for coliform bacteria. More than 1000 colony-forming units (CFU)/ 100 ml (up to 2700 CFU/100 ml) were detected in four samples, while their absence is required in 100 ml by the Standard of Water Quality in Italy and seven of 11 samples tested were also Escherichia coli-positive (up to 900 CFU/100 ml). Enterococci were searched in 11 samples and three of them were found positive for this faecal contamination indicator. From 7 July to 16 August 2016, after control measures on the water network were undertaken, 13 water samples were collected for follow-up laboratory examination of the water system at F, R and C sites. All of them were negative for bacterial indicators of faecal contamination. No faecal bacteria were identified in the seven samples collected from M, W and S sites, including the well and the source that fed the water network, at any sampling time, before and after implementing the control measures. The research of NoV genome was conducted on concentrated water samples collected on 1 August from R and F sites (Table 3). NoV genome was detected in a single sample collected from the water entry of the resort involved in the outbreak (site R). Actually, only two out of 26 One-Step RT-PCR repetitions were positive for NoV viral capsid gene. The two ORF2-positive results were obtained at the University of Palermo Laboratory using both 1:10 diluted and undiluted extracted RNA. The sequence analysis of the partial ORF2 portion (289 bp) indicated the presence of the same GII.2 capsid genotype detected in the stool samples of the students involved in the outbreak, showing 98.9% nt id. No ORF1 NoV sequence could be obtained from the NoV-positive water sample. Follow-up analyses for NoV genome, conducted on water samples collected on 20 November both at R and F sites found no positive samples. By sequence alignment, the NoV sequences obtained from Mascali outbreak samples were phylogenetically closely related in both ORF1 and ORF2 (99 and 98.9% of nt id, respectively) to a GII.P2/GII.2 strain responsible for a sporadic paediatric gastroenteritis case occurred in Palermo in January 2016 (PA 36/16, accession numbers MG517464 and MG517460, for ORF1 and ORF2, respectively) and previously detected at the Enteric Viruses Laboratory, University of Palermo. Whereas, three GII.P2/GII.2 Italian strains (PA 280/11, PA 366/11 and PA 3/ 14, accession numbers MG517462, MG517465 and MG517463, for ORF1 and MG517457 to −59, for ORF2, respectively), detected in Palermo in 2011 and 2014, were more distantly associated with the outbreak strain both in RdRp and VP1 (90% to 97.4% nt id and 96.1% to 98.1% nt id, respectively, in ORF1 and ORF2). Discussion Epidemiological features, such as the presence of vomiting in more than half (88%) of symptomatic subjects, the 33.2 h average incubation period and the 38.4 h average duration of illness, suggested a NoV aetiology for the investigated outbreak [13]. The detection of NoV genome in all 25 faecal samples from symptomatic subjects fully complies with the requirement of detection of at least one positive sample out of four tested that is considered to be sufficient to ascertain the aetiology of a NoV outbreak [24]. Moreover, molecular typing indicated that a single viral strain, belonging to genotype GII.P2/GII.2, infected all symptomatic subjects involved in the outbreak. NoVs are highly infectious and can easily spread from person to person but also through contaminated food or water. Therefore, settings such as health care institutions, schools and touristic resorts are mainly affected, since close contact and shared food services facilitate transmission. In Mascali, two waves of transmission were observed over three days, being consistent with food/waterborne followed by person-to-person and/or airborne (vomit) transmission (Fig. 2). The first wave of cases presented on 20 May 2016 after a single shared meal, while the last case was detected on 22 May. Exposure to vomit may have been responsible for some of the secondary cases occurred after the first wave of illness. Vomiting produces aerosols that can be inhaled and swallowed. The extremely low infectious dose of NoV makes the inhalation of vomit aerosols an effective transmission route in settings where there is close contact, like on a bus trip [25]. All food consumed in the only shared lunch on 19 May at the resort was associated with the risk of disease (Table 2), but the association was inconclusive since all cases shared the same dishes included in the lunch menu on Day 1. Although many foods were potentially involved, our attention focused on the tap water. In fact, trivial information from local health authorities reported that local population avoided tap water consumption and relied on bottled water for drinking and cooking. Many small outbreaks of gastroenteritis involving restaurants and family groups had been reported in the same area, especially at the beginning of the summer season, when vacation houses and recreational activities were re-opened after a period of low water demand. Illness caused by NoV is common and if the level of water contamination is low, the number of cases remains limited and water can persist as unrecognised source of infection. If polluted water was served for drinking and used for cooking and washing dishes in the Mascali resort at Day 1 a generalised contamination of food and surfaces could have occurred. The potential role of the foodhandlers as a source of contamination could not be analysed in this study but should also be taken into account. If water contamination has already been present in the area for a long time, infections may have occurred among the local food-handlers. The analysis of water samples taken immediately after the outbreak revealed the presence of bacterial indicators of faecal contamination in the water used in the resort, with the only exception of a sample of filter purified water used both for cooking and as drinking water. When extending the sampling to other areas of the municipal water network, many other sites, mostly on the seaside, showed faecal pollution, which demonstrates a poor protection of the water network from faecal leakage (Fig. 1). NoV contamination of waters generally originates from sewage since these enteric pathogens, as other enteric viruses affecting humans, are mostly species-specific and are shed at high concentrations in the stools of patients. When the analyses on the stools of the students made it clear that the outbreak had been caused by NoV, the viral genome was also searched in appropriate volumes of water samples, but this happened when correction measures on the water network had already been taken. Nevertheless, a single water sample, obtained 2 months after the outbreak (1 August 2016), still contained genome fragments that could be correlated to the same NoV genotype causing the outbreak. Although the finding of NoV genome does not demonstrate that entire infecting viral particles were present in the water, it could indicate that a NoV belonging to the same genotype involved in the outbreak was still circulating in the local population several weeks after the outbreak and it was someway able to penetrate the water network, probably indicating continuing faecal contamination through leakage from sewage systems. NoV is particularly resistant in the environment and also to chlorine, therefore transmission through chlorinated water is not exceptional [9, 12, 26]. Drinking water from municipal supplies fed by lake or groundwater has already been identified in the past as the source of waterborne NoV outbreaks in Italy [27–29] and NoV is now considered one of the pathogens causing the largest number of affected consumers in drinking waterborne outbreaks [12]. NoV can persist in water longer than 15 days [30] and it is resistant to chlorine disinfection when free chlorine levels or pre-treatment processes are inadequate [31, 32]. Moreover, microfiltration filters labelled as certified by NSF Standards 53 or 58 (with a pore size ranging from 0.05 to 5 µm) are not effective in removing viruses and can even decrease chlorine concentration in filtered water [33]. Actually, the owner of the resort involved in our outbreak was encouraged to serve tap water as drinking water at the restaurant relying on the effectiveness of such a filter purifier. Unfortunately, we did not have a tap water sample from the restaurant of the resort taken before correction measures and whose volume was appropriate for virus detection. Therefore, we could not demonstrate the failure to eliminate viral pollution of the microfiltration system. In Mascali, correction measures on the water network consisted in the maintenance of the old chlorination device and in the addition of a new one at a key point located upstream and close to suspected illegal sewage dumping (Fig. 1), but as poor protection of the pipeline from sewage leakage risked to reproduce faecal pollution episodes, renovation of the municipal water system was also advised, as well as systematic control of illegal sewage dumping. The additional chlorination device was apparently the most effective intervention, since right after its activation faecal bacterial pollution indicators disappeared. However, NoV genome was detected in a water sample taken 6 weeks after the last detection of bacterial indicators. Unfortunately, detection of faecal contamination of drinking water is only based on the use of bacterial indicators despite no correlation between levels of enteric bacteria and enteric viruses has been found [34–36]. According to ISGEV enteric virus circulation surveillance data, in 2016 the circulation of GII.P2/GII.2 NoV strains in Palermo, home town of the travellers involved in the outbreak, was the highest ever observed from the starting of local NoV surveillance in 2004. GII.P2/GII.2 was the fourth most common genotype in 2016 (12.2% prevalence), following GII.P4/GII.4 Sydney 2012 (28.4%), GII.P16/GII.2 (18.9%) and GII.Pe/GII.4 Sydney 2012 (17.6%). GII.P2/GII.2 strains were mostly observed starting from July 2016, leaving space for the possibility that their increased circulation was linked to the introduction of the ‘Mascali clone’. However, a couple of GII.P2/GII.2 strains (including PA36/16) genetically close to the ‘Mascali clone’ had been already detected in January 2016 in Palermo and GII.P2/GII.2 NoV had sporadically been observed also in 2011 and 2014. Moreover, the circulation of GII.P2/GII.2 NoVs has been increasing worldwide over the last period [37]. Therefore, the unusually high circulation observed in Palermo might just reflect the global epidemiology of GII.P2/GII.2 strains. Noteworthy, the older GII.P2/GII.2 NoV strains detected in Palermo were closely related in the ORF2 (98.9% nt id) to the GII.P16/GII.2 Nashville strain (KY865307) which is supposed to have generated the recombinant GII.P16/GII.4 clone recently circulating epidemically worldwide [37, 38] and were more distantly associated (95.3% nt id) to the GII.2 capsid sequence (AB662868) of another putative GII.P16/GII.4 ancestor from Japan [39]. The increasing frequency of detection of waterborne NoV outbreaks calls for the introduction of routine monitoring for viruses of drinking waters and for the definition of affordable and reliable indicators for the presence of NoV in drinking waters. The use of ultrafiltration systems as a concentration method in order to recover viruses from environmental waters combined with RT-PCR to detect viral genomes has been demonstrated to grant a recovery rate of >50% of multiple viruses [40] and is at present one of the most sensitive approaches for the detection of NoVs, although the discrimination between viable and nonviable viruses remains a problem. The burden of NoV outbreaks, either food associated or waterborne, is still poorly defined due to the lack of specific diagnostic tools in many clinical laboratories. NoV epidemics should always be suspected when the main symptom is vomiting and the duration of disease is short (<48 h). Waterborne transmission is suggested by the rapid spread of disease and the high attack rate. Care should be taken to perform sensitive diagnostic tests on a sufficient number of stool samples in order to identify NoV as the causative agent of enteritis outbreaks, possibly preferring genome screening over antigen detection. Only the systematic detection of NoV outbreaks can guide in the choice of the safety procedures that could be most appropriate in containing such microbiological hazard.

What type of samples were examined?

1557

Waterborne Norovirus outbreak at a seaside resort likely originating from municipal water distribution system failure

Abstract In May 2016 a Norovirus (NoV) gastroenteritis outbreak involved a high school class visiting a seaside resort near Taormina (Mascali, Sicily). Twenty-four students and a teacher were affected and 17 of them showed symptoms on the second day of the journey, while the others got ill within the following 2 days. Symptoms included vomiting, diarrhoea and fever, and 12 students required hospitalisation. Stool samples tested positive for NoV genome by Real-Time polymerase chain reaction assay in all 25 symptomatic subjects. The GII.P2/GII.2 NoV genotype was linked to the outbreak by ORF1/ORF2 sequence analysis. The epidemiological features of the outbreak were consistent with food/waterborne followed by person-to-person and/ or vomit transmission. Food consumed at a shared lunch on the first day of the trip was associated to illness and drinking un-bottled tap water was also considered as a risk factor. The analysis of water samples revealed the presence of bacterial indicators of faecal contamination in the water used in the resort as well as in other areas of the municipal water network, linking the NoV gastroenteritis outbreak to tap water pollution from sewage leakage. From a single water sample, an amplicon whose sequence corresponded to the capsid genotype recovered from patients could be obtained. Introduction Noroviruses (NoVs) are recognised as a leading cause of sporadic cases and outbreaks of acute gastroenteritis (AGE) across all age groups [1, 2]. They are transmitted primarily through the faecal-oral route either by direct person–person spread and with consumption of contaminated food or water [1, 3, 4]. NoVs are non-enveloped RNA viruses belonging to the Caliciviridae family and are classified into seven distinct genogroups (GI-GVII) on the basis of the major capsid protein VP1 sequence. GI, GII and GIV NoVs infect humans and they have been divided into more than 30 genotypes based on the sequences of polymerase (ORF1) and capsid proteins (ORF2) [5]. NoV GII.4 genotype has been associated with the vast majority of NoV related cases of AGE worldwide although many genotypes, belonging to both GI and GII genogroups are involved in foodborne and waterborne outbreaks [6, 7]. NoVs are highly contagious owing to their high stability in the environment and their ability to infect at low doses. In fact, just 10–100 virions of NoV may be sufficient to cause AGE [8, 9]. NoVs are considered the major cause of water-related disease in semi-closed institutions, including restaurants, schools, hospitals, nursing homes and cruise ships [10–12]. To confirm NoV as a cause of outbreaks, the presence of four epidemiological features of disease, vomiting in more than half of affected persons, a mean incubation period of 24/48 h, a mean duration of illness of 12–60 h and the absence of bacterial pathogens in stool culture, should be evaluated [13]. Waterborne NoV outbreaks can be caused by contaminated drinking water despite correct chlorine concentration and in the absence of coliform bacteria [10, 14–17]. The aim of this study was to investigate an outbreak of NoV gastroenteritis associated with drinking water involving a high school class travelling from Palermo (West Sicily) to Acireale (East Sicily) on an educational tour in May 2016. Stool samples were collected soon after the onset of AGE symptoms to ascertain the aetiology of illness. The source of infection was investigated collecting epidemiological data through an online questionnaire and, in order to explore a possible transmission route, laboratory tests on water from the municipal network were performed. Material and methods Epidemiological investigation From 20 to 22 May 2016 an AGE outbreak occurred in Mascali, a small town near Taormina, involving a school class of 28 students, 13 girls and 15 boys aged between 14 and 18 years and a 54-year-old female teacher, during a trip to participate in a theatre festival in Acireale. A total of 25/29 persons were affected by nausea, vomiting, abdominal cramps, diarrhoea and fever while they were staying at a seaside resort in Mascali, on the way to Acireale. To obtain data on symptoms and the correlated risk factors all participants filled an online questionnaire investigating personal data (age and sex), travel and accommodation details and food consumption over 48 h before symptoms presentation, time of onset and duration of symptoms, therapy and/or hospitalisation required and history of contact with people affected by diarrhoea and/or vomiting (Table 1). The severity of illness was analysed by the Vesikari severity scale allowing to distinguish non-severe (score <11) and severe (score ⩾11) AGE (Table 1). Vesikari score was calculated from maximum number of episodes/day and duration of both diarrhoea and vomiting (score 0–3 for each of the 4 parameters), temperature (score 0–3), level of dehydration (score 0–3) and treatment required (score 0–2) [18]. A case-control study was conducted to establish if food consumed before the onset of symptoms was associated with the disease. A case was defined as a student attending the tour and reporting diarrhoea and/or vomiting within 36 h from the first shared meal during the tour, while students who did not present any of the previous symptoms or those showing the first symptoms with a delay of ⩾48 h (secondary cases) were used as controls. Variables were studied using odds ratio (OR) and its 95% CI and statistical associations were determined using the χ2 test, with the level of statistical significance set at 0.05 (Table 2). NoV detection in stool samples A stool sample was obtained from each of the 29 travellers, soon after (3–5 days) the onset of the outbreak. No stool samples could be obtained from the food-handlers working at the resort. All samples were tested for NoV at the Enteric Viruses Laboratory of the University of Palermo, member of the Italian Study Group for Enteric Viruses (ISGEV), by a Real-Time polymerase chain reaction (RT-PCR) assay that allows to differentiate GI and GII NoVs [19]. Total RNA was extracted from stool samples using QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) and cDNA was generated by MMLV reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were considered positive for NoVs when cycle threshold (ct) of RT-PCR was ⩽30 and lowpositive when 30<ct⩽35. For all NoV-positive samples two genomic fragments were amplified by conventional RT-PCR using Superscript III One-Step RT-PCR system with Platinum Taq DNA polymerase (Invitrogen) and primers JV12a/JV13b, targeting 326 bps of the RdRp region A [20] and either COG1F/ GISKR or COG2F/GIISKR, amplifying a 315 bps 5′portion of GI or GII VP1 protein [19, 21]. Environmental investigation From May through August 2016, a total of 32 tap water samples were collected from the Mascali municipal network. In particular, seven samples were collected at the resort where the outbreak occurred and from its water entry (R), two in the city centre of Mascali (M), 15 in the village of Fondachello (F), on the seaside, three in Carrabba village (C), far from the coast, and two and three, respectively, from the water well ‘Carlino’ (W) and the source ‘Bufardo and Torregrossa’ (S) which feed the municipal system and provide drinking water to the Mascali area (Fig. 1, Table 3). One litre water samples were sent by the local health authorities to the SIALAB laboratory in Avola, Siracusa and 100 ml aliquots were screened for coliforms, E. coli and enterococci as faecal indicators, by standard bacteriological methods according to UNI EN ISO 9308-1:2014. Unfortunately, although the suspect of a water source for the viral outbreak was quickly communicated, only in August and November 2016, following the implementation of control measures, local authorities provided water samplings of appropriate volume to perform viral tests (10 l). Five water samples of 10 l each were collected from different sampling sites in F and R areas for viral screening (Table 3) and were concentrated according to the Italian National Institute of Health (Istituto Superiore di Sanità, ISS) protocols [22] through an ultra-filtration system (Sartoflow® Slice 200 Benchtop Crossflow System, Sartorius AG, Goettingen, Germany), using appropriate membranes (SG Hydrosart 10 kDa) pretreated with 300 ml of 3% Beef Extract (BE) at pH 7, by recirculating it for 10 min in the system, at a pressure not higher than 1.30 bar. Elution and recovery of viruses attached to the membranes were obtained by recirculating 200 ml of 3% BE at pH 9.5 for 3 h until a 10–12 ml volume of eluate was obtained. The pH of the eluate was then checked with litmus paper and if necessary brought to the value of 7 with HCl. Concentrated water samples were simultaneously tested for NoV in three laboratories: Enteric Viruses Laboratory, University of Palermo; Istituto Zooprofilattico Sperimentale della Sicilia (IZS), Palermo; and Istituto Superiore di Sanità (ISS), Rome. Total RNA was extracted from 500 and 560 µl of a concentrated water sample using BioMèrieux NucliSENS® Magnetic Extraction System (BioMèrieux) and QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) respectively, according to manufacturer’s instructions. Bonded RNA was eluted by 100 µl of Elution Buffer. Amplifications were performed at least in duplicate on whole and 1:10 diluted extracts. A Real-Time RT-PCR assay according to ISO/TS 15216- 2:2013, two in-house Real-Time RT-PCR protocols and the One-Step RT-PCR methods already used for amplifying ORF1 and ORF2 fragments in NoV-positive faecal samples were performed [19–21, 23]. Nucleotide sequences analysis and phylogenetic analysis PCR amplicons were purified by commercial columns (CleanSweep PCR Purification, Thermo Fisher Scientific, Vilnius, Lithuania) and sequenced by Sanger method and all nucleotide sequences were aligned using ClustalW. To identify NoV genotypes, RdRp and VP1 sequences were submitted to the RIVM NoV genotyping tool (http://www.rivm.nl/mpf/norovirus/typingtool). Phylogenetic trees were constructed using MEGA 7 and Neighbour-Joining method (bootstrap 1000 replicates). Sequences from stool and water samples are available at NCBI with the following accession numbers: MG517455 for the major capsid protein obtained from a water sample; MG517456 and MG517461, respectively, for the major capsid protein and RNA-dependent-RNA-polymerase, as representatives of the sequences obtained from the stool samples of those involved in the outbreak. Control measures After outbreak notification, on 21 May, the resort was closed for 2 days. On 25 May, following the first results of the bacteriological analyses on municipal water samples, the use of municipal water for human consumption was temporarily restricted (use of water was permitted only for cleaning purposes) in the Carrabba and Fondachello villages and in the area surrounding the resort. From 25 May the reservoir collecting the water from the Carlino well and its chlorination device were submitted to maintenance work and a new chlorination device was installed on the conduit bringing the water from the Bufardo source. Actually, the Bufardo source had been connected to the municipal network in 2003 but unfortunately a dedicated chlorination station had not been planned at that time. Therefore, unchlorinated water entered the system near Fondachello village on the seaside (Fig. 1), possibly diluting chlorine concentrations in water treated at the Carlino reservoir. Moreover, the exact path of the branch of the water network situated downstream the Bufardo source water connection was not marked in the maps of the municipality (Fig. 1, red line), indicating that this section had never been submitted to maintenance. The area surrounding the unmaintained section of the water pipeline was scattered with unauthorised cottages that were not connected to the sewerage systems. Likely, untreated waste from cesspools was unloaded directly into the soil, generating a risk of contamination of the municipal water system through leakage. No work had been made on the water pipelines before the outbreak but sections of the network had been temporarily closed during the winter season on the seaside (Fondachello). Local health authorities advised the connection of all households in the area to the sewerage system and the replacement of the water system section without maintenance. Health authorities also raised attention on AGE cases notification at all local Emergency Departments and Family Practitioners but there was no further evidence of AGE outbreaks in the months following the intervention. Results Data collected through the online questionnaire compiled 7–10 days after the outbreak revealed that none of the 28 students and the teacher taking part in the school trip showed gastroenteritis symptoms at the beginning of the journey and they had no contact with people who suffered from diarrhoea and/or vomiting in the previous days. Travellers had their first shared meal on 19 May (Day 1), 2016, at lunch on their arrival to the resort in Mascali. The list of foods included in the menu is shown in Table 2. Conversely, the dinner on Day 1 was not shared and the students went to dine in three different fast food restaurants in Acireale. On 20 May (Day 2), 17 students (58.6% of the travellers) presented with vomit and/or diarrhoea, falling ill 26–36 h after lunch on Day 1, strongly suggesting that exposure occurred on the first shared meal. Some students also vomited in the presence of the other travellers on the bus bringing them to Acireale. Five more students were symptomatic in the early hours of the following day (21 May, Day 3, from 0:30 am to 8:00 am), as a continuation of the wave of illnesses of the previous day. The teacher and one more student fell ill at 8:00 pm and 11:00 pm of Day 3, respectively, for a second wave of illnesses occurring >48 h from the first shared meal and <36 h from the first cases, with the last student falling sick on 22 May (Day 4) at 3:00 am (Fig. 2, Table 1). On the whole, the mean incubation period was 33.2 h. Based on the online survey, the severity of illness was evaluated by the Vesikari severity scale, grouping symptomatic travellers into severe (11/25 = 44%) and non-severe (14/25 = 56%) cases (Table 1). Vomiting and fever were the predominant symptoms, affecting 22 out of 25 (88%) symptomatic subjects, while diarrhoea was present in 14 out of 25 (56%). Hospitalisation for rehydration therapy was necessary for 13 students (91% [10/11] of severe and 21% [3/14] of non-severe cases). Emergency room (ER) or family practitioner visits were necessary for 21 students and 13 (62%) of them initiated an antibiotic therapy. Two students indicated that other family members had gastroenteric symptoms more than 24 h after their return home, on Day 3, 4 and 5. All those involved in the outbreak recovered within 4 days, with an average duration of illness of 38.4 h. A case-control study of the risk associated with eating foods at lunch on Day 1 or with attending different restaurants at dinner on Day 1 was attempted. A total of 19 subjects filled the section of the online survey concerning food consumption. Fourteen of the responders were among the 22 cases falling ill within 36 h from the first shared meal (first wave), while the remaining five responders were three travellers with supposed secondary infections (showing symptoms with a delay ⩾48 h; second wave) and two asymptomatic students that were used as controls. No statistically significant associations with any food items served at lunch on Day 1 was identified (P < 0.05; Table 2). However, all food consumed on that occasion was to some extent associated with the risk of disease, the highest OR being obtained with chicken (OR: 20.7143, P: 0.0678; Table 2). Actually, most of the cases ate all foods included in the menu, while one of the two asymptomatic controls did not participate in the meal and the other had a pasta dish only. No statistically significant association was found for attending a particular fast food restaurant at the dinner on Day 1 (OR: 2.5, P: 0.3635). The Real-Time RT-PCR assay allowed to detect GII genogroup NoV genome in all the 25 samples from symptomatic subjects, although two of them showed high Ct values, indicating lower shedding of viral genomes (Table 1). The samples from four asymptomatic students were negative for both NoV genogroups. By sequence analysis, a NoV genotype could be defined for 20/ 25 NoV-positive stool samples. All NoV strains from travellers involved in the outbreaks belonged to the same genotype, GII.P2/GII.2, showing >99% nucleotide identities (nt id) in both fragments (ORF1 and ORF2) analysed. Thirty-two water samples were collected and analysed for bacterial contamination (Table 3). Seven samples were taken from the resort (site R) and the remaining 25 from five other sites (C, F, M, S, W), representative of the entire municipal water network. The majority (10/12) of the water samples collected immediately after the onset of the outbreak, between 21 May and 16 June 2016, at the resort (site R) and at sites F and C, located close to the resort, on the seaside and in the interior, respectively, were positive for coliform bacteria. More than 1000 colony-forming units (CFU)/ 100 ml (up to 2700 CFU/100 ml) were detected in four samples, while their absence is required in 100 ml by the Standard of Water Quality in Italy and seven of 11 samples tested were also Escherichia coli-positive (up to 900 CFU/100 ml). Enterococci were searched in 11 samples and three of them were found positive for this faecal contamination indicator. From 7 July to 16 August 2016, after control measures on the water network were undertaken, 13 water samples were collected for follow-up laboratory examination of the water system at F, R and C sites. All of them were negative for bacterial indicators of faecal contamination. No faecal bacteria were identified in the seven samples collected from M, W and S sites, including the well and the source that fed the water network, at any sampling time, before and after implementing the control measures. The research of NoV genome was conducted on concentrated water samples collected on 1 August from R and F sites (Table 3). NoV genome was detected in a single sample collected from the water entry of the resort involved in the outbreak (site R). Actually, only two out of 26 One-Step RT-PCR repetitions were positive for NoV viral capsid gene. The two ORF2-positive results were obtained at the University of Palermo Laboratory using both 1:10 diluted and undiluted extracted RNA. The sequence analysis of the partial ORF2 portion (289 bp) indicated the presence of the same GII.2 capsid genotype detected in the stool samples of the students involved in the outbreak, showing 98.9% nt id. No ORF1 NoV sequence could be obtained from the NoV-positive water sample. Follow-up analyses for NoV genome, conducted on water samples collected on 20 November both at R and F sites found no positive samples. By sequence alignment, the NoV sequences obtained from Mascali outbreak samples were phylogenetically closely related in both ORF1 and ORF2 (99 and 98.9% of nt id, respectively) to a GII.P2/GII.2 strain responsible for a sporadic paediatric gastroenteritis case occurred in Palermo in January 2016 (PA 36/16, accession numbers MG517464 and MG517460, for ORF1 and ORF2, respectively) and previously detected at the Enteric Viruses Laboratory, University of Palermo. Whereas, three GII.P2/GII.2 Italian strains (PA 280/11, PA 366/11 and PA 3/ 14, accession numbers MG517462, MG517465 and MG517463, for ORF1 and MG517457 to −59, for ORF2, respectively), detected in Palermo in 2011 and 2014, were more distantly associated with the outbreak strain both in RdRp and VP1 (90% to 97.4% nt id and 96.1% to 98.1% nt id, respectively, in ORF1 and ORF2). Discussion Epidemiological features, such as the presence of vomiting in more than half (88%) of symptomatic subjects, the 33.2 h average incubation period and the 38.4 h average duration of illness, suggested a NoV aetiology for the investigated outbreak [13]. The detection of NoV genome in all 25 faecal samples from symptomatic subjects fully complies with the requirement of detection of at least one positive sample out of four tested that is considered to be sufficient to ascertain the aetiology of a NoV outbreak [24]. Moreover, molecular typing indicated that a single viral strain, belonging to genotype GII.P2/GII.2, infected all symptomatic subjects involved in the outbreak. NoVs are highly infectious and can easily spread from person to person but also through contaminated food or water. Therefore, settings such as health care institutions, schools and touristic resorts are mainly affected, since close contact and shared food services facilitate transmission. In Mascali, two waves of transmission were observed over three days, being consistent with food/waterborne followed by person-to-person and/or airborne (vomit) transmission (Fig. 2). The first wave of cases presented on 20 May 2016 after a single shared meal, while the last case was detected on 22 May. Exposure to vomit may have been responsible for some of the secondary cases occurred after the first wave of illness. Vomiting produces aerosols that can be inhaled and swallowed. The extremely low infectious dose of NoV makes the inhalation of vomit aerosols an effective transmission route in settings where there is close contact, like on a bus trip [25]. All food consumed in the only shared lunch on 19 May at the resort was associated with the risk of disease (Table 2), but the association was inconclusive since all cases shared the same dishes included in the lunch menu on Day 1. Although many foods were potentially involved, our attention focused on the tap water. In fact, trivial information from local health authorities reported that local population avoided tap water consumption and relied on bottled water for drinking and cooking. Many small outbreaks of gastroenteritis involving restaurants and family groups had been reported in the same area, especially at the beginning of the summer season, when vacation houses and recreational activities were re-opened after a period of low water demand. Illness caused by NoV is common and if the level of water contamination is low, the number of cases remains limited and water can persist as unrecognised source of infection. If polluted water was served for drinking and used for cooking and washing dishes in the Mascali resort at Day 1 a generalised contamination of food and surfaces could have occurred. The potential role of the foodhandlers as a source of contamination could not be analysed in this study but should also be taken into account. If water contamination has already been present in the area for a long time, infections may have occurred among the local food-handlers. The analysis of water samples taken immediately after the outbreak revealed the presence of bacterial indicators of faecal contamination in the water used in the resort, with the only exception of a sample of filter purified water used both for cooking and as drinking water. When extending the sampling to other areas of the municipal water network, many other sites, mostly on the seaside, showed faecal pollution, which demonstrates a poor protection of the water network from faecal leakage (Fig. 1). NoV contamination of waters generally originates from sewage since these enteric pathogens, as other enteric viruses affecting humans, are mostly species-specific and are shed at high concentrations in the stools of patients. When the analyses on the stools of the students made it clear that the outbreak had been caused by NoV, the viral genome was also searched in appropriate volumes of water samples, but this happened when correction measures on the water network had already been taken. Nevertheless, a single water sample, obtained 2 months after the outbreak (1 August 2016), still contained genome fragments that could be correlated to the same NoV genotype causing the outbreak. Although the finding of NoV genome does not demonstrate that entire infecting viral particles were present in the water, it could indicate that a NoV belonging to the same genotype involved in the outbreak was still circulating in the local population several weeks after the outbreak and it was someway able to penetrate the water network, probably indicating continuing faecal contamination through leakage from sewage systems. NoV is particularly resistant in the environment and also to chlorine, therefore transmission through chlorinated water is not exceptional [9, 12, 26]. Drinking water from municipal supplies fed by lake or groundwater has already been identified in the past as the source of waterborne NoV outbreaks in Italy [27–29] and NoV is now considered one of the pathogens causing the largest number of affected consumers in drinking waterborne outbreaks [12]. NoV can persist in water longer than 15 days [30] and it is resistant to chlorine disinfection when free chlorine levels or pre-treatment processes are inadequate [31, 32]. Moreover, microfiltration filters labelled as certified by NSF Standards 53 or 58 (with a pore size ranging from 0.05 to 5 µm) are not effective in removing viruses and can even decrease chlorine concentration in filtered water [33]. Actually, the owner of the resort involved in our outbreak was encouraged to serve tap water as drinking water at the restaurant relying on the effectiveness of such a filter purifier. Unfortunately, we did not have a tap water sample from the restaurant of the resort taken before correction measures and whose volume was appropriate for virus detection. Therefore, we could not demonstrate the failure to eliminate viral pollution of the microfiltration system. In Mascali, correction measures on the water network consisted in the maintenance of the old chlorination device and in the addition of a new one at a key point located upstream and close to suspected illegal sewage dumping (Fig. 1), but as poor protection of the pipeline from sewage leakage risked to reproduce faecal pollution episodes, renovation of the municipal water system was also advised, as well as systematic control of illegal sewage dumping. The additional chlorination device was apparently the most effective intervention, since right after its activation faecal bacterial pollution indicators disappeared. However, NoV genome was detected in a water sample taken 6 weeks after the last detection of bacterial indicators. Unfortunately, detection of faecal contamination of drinking water is only based on the use of bacterial indicators despite no correlation between levels of enteric bacteria and enteric viruses has been found [34–36]. According to ISGEV enteric virus circulation surveillance data, in 2016 the circulation of GII.P2/GII.2 NoV strains in Palermo, home town of the travellers involved in the outbreak, was the highest ever observed from the starting of local NoV surveillance in 2004. GII.P2/GII.2 was the fourth most common genotype in 2016 (12.2% prevalence), following GII.P4/GII.4 Sydney 2012 (28.4%), GII.P16/GII.2 (18.9%) and GII.Pe/GII.4 Sydney 2012 (17.6%). GII.P2/GII.2 strains were mostly observed starting from July 2016, leaving space for the possibility that their increased circulation was linked to the introduction of the ‘Mascali clone’. However, a couple of GII.P2/GII.2 strains (including PA36/16) genetically close to the ‘Mascali clone’ had been already detected in January 2016 in Palermo and GII.P2/GII.2 NoV had sporadically been observed also in 2011 and 2014. Moreover, the circulation of GII.P2/GII.2 NoVs has been increasing worldwide over the last period [37]. Therefore, the unusually high circulation observed in Palermo might just reflect the global epidemiology of GII.P2/GII.2 strains. Noteworthy, the older GII.P2/GII.2 NoV strains detected in Palermo were closely related in the ORF2 (98.9% nt id) to the GII.P16/GII.2 Nashville strain (KY865307) which is supposed to have generated the recombinant GII.P16/GII.4 clone recently circulating epidemically worldwide [37, 38] and were more distantly associated (95.3% nt id) to the GII.2 capsid sequence (AB662868) of another putative GII.P16/GII.4 ancestor from Japan [39]. The increasing frequency of detection of waterborne NoV outbreaks calls for the introduction of routine monitoring for viruses of drinking waters and for the definition of affordable and reliable indicators for the presence of NoV in drinking waters. The use of ultrafiltration systems as a concentration method in order to recover viruses from environmental waters combined with RT-PCR to detect viral genomes has been demonstrated to grant a recovery rate of >50% of multiple viruses [40] and is at present one of the most sensitive approaches for the detection of NoVs, although the discrimination between viable and nonviable viruses remains a problem. The burden of NoV outbreaks, either food associated or waterborne, is still poorly defined due to the lack of specific diagnostic tools in many clinical laboratories. NoV epidemics should always be suspected when the main symptom is vomiting and the duration of disease is short (<48 h). Waterborne transmission is suggested by the rapid spread of disease and the high attack rate. Care should be taken to perform sensitive diagnostic tests on a sufficient number of stool samples in order to identify NoV as the causative agent of enteritis outbreaks, possibly preferring genome screening over antigen detection. Only the systematic detection of NoV outbreaks can guide in the choice of the safety procedures that could be most appropriate in containing such microbiological hazard.

What did they test for in the samples?

1558

Waterborne Norovirus outbreak at a seaside resort likely originating from municipal water distribution system failure

Abstract In May 2016 a Norovirus (NoV) gastroenteritis outbreak involved a high school class visiting a seaside resort near Taormina (Mascali, Sicily). Twenty-four students and a teacher were affected and 17 of them showed symptoms on the second day of the journey, while the others got ill within the following 2 days. Symptoms included vomiting, diarrhoea and fever, and 12 students required hospitalisation. Stool samples tested positive for NoV genome by Real-Time polymerase chain reaction assay in all 25 symptomatic subjects. The GII.P2/GII.2 NoV genotype was linked to the outbreak by ORF1/ORF2 sequence analysis. The epidemiological features of the outbreak were consistent with food/waterborne followed by person-to-person and/ or vomit transmission. Food consumed at a shared lunch on the first day of the trip was associated to illness and drinking un-bottled tap water was also considered as a risk factor. The analysis of water samples revealed the presence of bacterial indicators of faecal contamination in the water used in the resort as well as in other areas of the municipal water network, linking the NoV gastroenteritis outbreak to tap water pollution from sewage leakage. From a single water sample, an amplicon whose sequence corresponded to the capsid genotype recovered from patients could be obtained. Introduction Noroviruses (NoVs) are recognised as a leading cause of sporadic cases and outbreaks of acute gastroenteritis (AGE) across all age groups [1, 2]. They are transmitted primarily through the faecal-oral route either by direct person–person spread and with consumption of contaminated food or water [1, 3, 4]. NoVs are non-enveloped RNA viruses belonging to the Caliciviridae family and are classified into seven distinct genogroups (GI-GVII) on the basis of the major capsid protein VP1 sequence. GI, GII and GIV NoVs infect humans and they have been divided into more than 30 genotypes based on the sequences of polymerase (ORF1) and capsid proteins (ORF2) [5]. NoV GII.4 genotype has been associated with the vast majority of NoV related cases of AGE worldwide although many genotypes, belonging to both GI and GII genogroups are involved in foodborne and waterborne outbreaks [6, 7]. NoVs are highly contagious owing to their high stability in the environment and their ability to infect at low doses. In fact, just 10–100 virions of NoV may be sufficient to cause AGE [8, 9]. NoVs are considered the major cause of water-related disease in semi-closed institutions, including restaurants, schools, hospitals, nursing homes and cruise ships [10–12]. To confirm NoV as a cause of outbreaks, the presence of four epidemiological features of disease, vomiting in more than half of affected persons, a mean incubation period of 24/48 h, a mean duration of illness of 12–60 h and the absence of bacterial pathogens in stool culture, should be evaluated [13]. Waterborne NoV outbreaks can be caused by contaminated drinking water despite correct chlorine concentration and in the absence of coliform bacteria [10, 14–17]. The aim of this study was to investigate an outbreak of NoV gastroenteritis associated with drinking water involving a high school class travelling from Palermo (West Sicily) to Acireale (East Sicily) on an educational tour in May 2016. Stool samples were collected soon after the onset of AGE symptoms to ascertain the aetiology of illness. The source of infection was investigated collecting epidemiological data through an online questionnaire and, in order to explore a possible transmission route, laboratory tests on water from the municipal network were performed. Material and methods Epidemiological investigation From 20 to 22 May 2016 an AGE outbreak occurred in Mascali, a small town near Taormina, involving a school class of 28 students, 13 girls and 15 boys aged between 14 and 18 years and a 54-year-old female teacher, during a trip to participate in a theatre festival in Acireale. A total of 25/29 persons were affected by nausea, vomiting, abdominal cramps, diarrhoea and fever while they were staying at a seaside resort in Mascali, on the way to Acireale. To obtain data on symptoms and the correlated risk factors all participants filled an online questionnaire investigating personal data (age and sex), travel and accommodation details and food consumption over 48 h before symptoms presentation, time of onset and duration of symptoms, therapy and/or hospitalisation required and history of contact with people affected by diarrhoea and/or vomiting (Table 1). The severity of illness was analysed by the Vesikari severity scale allowing to distinguish non-severe (score <11) and severe (score ⩾11) AGE (Table 1). Vesikari score was calculated from maximum number of episodes/day and duration of both diarrhoea and vomiting (score 0–3 for each of the 4 parameters), temperature (score 0–3), level of dehydration (score 0–3) and treatment required (score 0–2) [18]. A case-control study was conducted to establish if food consumed before the onset of symptoms was associated with the disease. A case was defined as a student attending the tour and reporting diarrhoea and/or vomiting within 36 h from the first shared meal during the tour, while students who did not present any of the previous symptoms or those showing the first symptoms with a delay of ⩾48 h (secondary cases) were used as controls. Variables were studied using odds ratio (OR) and its 95% CI and statistical associations were determined using the χ2 test, with the level of statistical significance set at 0.05 (Table 2). NoV detection in stool samples A stool sample was obtained from each of the 29 travellers, soon after (3–5 days) the onset of the outbreak. No stool samples could be obtained from the food-handlers working at the resort. All samples were tested for NoV at the Enteric Viruses Laboratory of the University of Palermo, member of the Italian Study Group for Enteric Viruses (ISGEV), by a Real-Time polymerase chain reaction (RT-PCR) assay that allows to differentiate GI and GII NoVs [19]. Total RNA was extracted from stool samples using QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) and cDNA was generated by MMLV reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were considered positive for NoVs when cycle threshold (ct) of RT-PCR was ⩽30 and lowpositive when 30<ct⩽35. For all NoV-positive samples two genomic fragments were amplified by conventional RT-PCR using Superscript III One-Step RT-PCR system with Platinum Taq DNA polymerase (Invitrogen) and primers JV12a/JV13b, targeting 326 bps of the RdRp region A [20] and either COG1F/ GISKR or COG2F/GIISKR, amplifying a 315 bps 5′portion of GI or GII VP1 protein [19, 21]. Environmental investigation From May through August 2016, a total of 32 tap water samples were collected from the Mascali municipal network. In particular, seven samples were collected at the resort where the outbreak occurred and from its water entry (R), two in the city centre of Mascali (M), 15 in the village of Fondachello (F), on the seaside, three in Carrabba village (C), far from the coast, and two and three, respectively, from the water well ‘Carlino’ (W) and the source ‘Bufardo and Torregrossa’ (S) which feed the municipal system and provide drinking water to the Mascali area (Fig. 1, Table 3). One litre water samples were sent by the local health authorities to the SIALAB laboratory in Avola, Siracusa and 100 ml aliquots were screened for coliforms, E. coli and enterococci as faecal indicators, by standard bacteriological methods according to UNI EN ISO 9308-1:2014. Unfortunately, although the suspect of a water source for the viral outbreak was quickly communicated, only in August and November 2016, following the implementation of control measures, local authorities provided water samplings of appropriate volume to perform viral tests (10 l). Five water samples of 10 l each were collected from different sampling sites in F and R areas for viral screening (Table 3) and were concentrated according to the Italian National Institute of Health (Istituto Superiore di Sanità, ISS) protocols [22] through an ultra-filtration system (Sartoflow® Slice 200 Benchtop Crossflow System, Sartorius AG, Goettingen, Germany), using appropriate membranes (SG Hydrosart 10 kDa) pretreated with 300 ml of 3% Beef Extract (BE) at pH 7, by recirculating it for 10 min in the system, at a pressure not higher than 1.30 bar. Elution and recovery of viruses attached to the membranes were obtained by recirculating 200 ml of 3% BE at pH 9.5 for 3 h until a 10–12 ml volume of eluate was obtained. The pH of the eluate was then checked with litmus paper and if necessary brought to the value of 7 with HCl. Concentrated water samples were simultaneously tested for NoV in three laboratories: Enteric Viruses Laboratory, University of Palermo; Istituto Zooprofilattico Sperimentale della Sicilia (IZS), Palermo; and Istituto Superiore di Sanità (ISS), Rome. Total RNA was extracted from 500 and 560 µl of a concentrated water sample using BioMèrieux NucliSENS® Magnetic Extraction System (BioMèrieux) and QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) respectively, according to manufacturer’s instructions. Bonded RNA was eluted by 100 µl of Elution Buffer. Amplifications were performed at least in duplicate on whole and 1:10 diluted extracts. A Real-Time RT-PCR assay according to ISO/TS 15216- 2:2013, two in-house Real-Time RT-PCR protocols and the One-Step RT-PCR methods already used for amplifying ORF1 and ORF2 fragments in NoV-positive faecal samples were performed [19–21, 23]. Nucleotide sequences analysis and phylogenetic analysis PCR amplicons were purified by commercial columns (CleanSweep PCR Purification, Thermo Fisher Scientific, Vilnius, Lithuania) and sequenced by Sanger method and all nucleotide sequences were aligned using ClustalW. To identify NoV genotypes, RdRp and VP1 sequences were submitted to the RIVM NoV genotyping tool (http://www.rivm.nl/mpf/norovirus/typingtool). Phylogenetic trees were constructed using MEGA 7 and Neighbour-Joining method (bootstrap 1000 replicates). Sequences from stool and water samples are available at NCBI with the following accession numbers: MG517455 for the major capsid protein obtained from a water sample; MG517456 and MG517461, respectively, for the major capsid protein and RNA-dependent-RNA-polymerase, as representatives of the sequences obtained from the stool samples of those involved in the outbreak. Control measures After outbreak notification, on 21 May, the resort was closed for 2 days. On 25 May, following the first results of the bacteriological analyses on municipal water samples, the use of municipal water for human consumption was temporarily restricted (use of water was permitted only for cleaning purposes) in the Carrabba and Fondachello villages and in the area surrounding the resort. From 25 May the reservoir collecting the water from the Carlino well and its chlorination device were submitted to maintenance work and a new chlorination device was installed on the conduit bringing the water from the Bufardo source. Actually, the Bufardo source had been connected to the municipal network in 2003 but unfortunately a dedicated chlorination station had not been planned at that time. Therefore, unchlorinated water entered the system near Fondachello village on the seaside (Fig. 1), possibly diluting chlorine concentrations in water treated at the Carlino reservoir. Moreover, the exact path of the branch of the water network situated downstream the Bufardo source water connection was not marked in the maps of the municipality (Fig. 1, red line), indicating that this section had never been submitted to maintenance. The area surrounding the unmaintained section of the water pipeline was scattered with unauthorised cottages that were not connected to the sewerage systems. Likely, untreated waste from cesspools was unloaded directly into the soil, generating a risk of contamination of the municipal water system through leakage. No work had been made on the water pipelines before the outbreak but sections of the network had been temporarily closed during the winter season on the seaside (Fondachello). Local health authorities advised the connection of all households in the area to the sewerage system and the replacement of the water system section without maintenance. Health authorities also raised attention on AGE cases notification at all local Emergency Departments and Family Practitioners but there was no further evidence of AGE outbreaks in the months following the intervention. Results Data collected through the online questionnaire compiled 7–10 days after the outbreak revealed that none of the 28 students and the teacher taking part in the school trip showed gastroenteritis symptoms at the beginning of the journey and they had no contact with people who suffered from diarrhoea and/or vomiting in the previous days. Travellers had their first shared meal on 19 May (Day 1), 2016, at lunch on their arrival to the resort in Mascali. The list of foods included in the menu is shown in Table 2. Conversely, the dinner on Day 1 was not shared and the students went to dine in three different fast food restaurants in Acireale. On 20 May (Day 2), 17 students (58.6% of the travellers) presented with vomit and/or diarrhoea, falling ill 26–36 h after lunch on Day 1, strongly suggesting that exposure occurred on the first shared meal. Some students also vomited in the presence of the other travellers on the bus bringing them to Acireale. Five more students were symptomatic in the early hours of the following day (21 May, Day 3, from 0:30 am to 8:00 am), as a continuation of the wave of illnesses of the previous day. The teacher and one more student fell ill at 8:00 pm and 11:00 pm of Day 3, respectively, for a second wave of illnesses occurring >48 h from the first shared meal and <36 h from the first cases, with the last student falling sick on 22 May (Day 4) at 3:00 am (Fig. 2, Table 1). On the whole, the mean incubation period was 33.2 h. Based on the online survey, the severity of illness was evaluated by the Vesikari severity scale, grouping symptomatic travellers into severe (11/25 = 44%) and non-severe (14/25 = 56%) cases (Table 1). Vomiting and fever were the predominant symptoms, affecting 22 out of 25 (88%) symptomatic subjects, while diarrhoea was present in 14 out of 25 (56%). Hospitalisation for rehydration therapy was necessary for 13 students (91% [10/11] of severe and 21% [3/14] of non-severe cases). Emergency room (ER) or family practitioner visits were necessary for 21 students and 13 (62%) of them initiated an antibiotic therapy. Two students indicated that other family members had gastroenteric symptoms more than 24 h after their return home, on Day 3, 4 and 5. All those involved in the outbreak recovered within 4 days, with an average duration of illness of 38.4 h. A case-control study of the risk associated with eating foods at lunch on Day 1 or with attending different restaurants at dinner on Day 1 was attempted. A total of 19 subjects filled the section of the online survey concerning food consumption. Fourteen of the responders were among the 22 cases falling ill within 36 h from the first shared meal (first wave), while the remaining five responders were three travellers with supposed secondary infections (showing symptoms with a delay ⩾48 h; second wave) and two asymptomatic students that were used as controls. No statistically significant associations with any food items served at lunch on Day 1 was identified (P < 0.05; Table 2). However, all food consumed on that occasion was to some extent associated with the risk of disease, the highest OR being obtained with chicken (OR: 20.7143, P: 0.0678; Table 2). Actually, most of the cases ate all foods included in the menu, while one of the two asymptomatic controls did not participate in the meal and the other had a pasta dish only. No statistically significant association was found for attending a particular fast food restaurant at the dinner on Day 1 (OR: 2.5, P: 0.3635). The Real-Time RT-PCR assay allowed to detect GII genogroup NoV genome in all the 25 samples from symptomatic subjects, although two of them showed high Ct values, indicating lower shedding of viral genomes (Table 1). The samples from four asymptomatic students were negative for both NoV genogroups. By sequence analysis, a NoV genotype could be defined for 20/ 25 NoV-positive stool samples. All NoV strains from travellers involved in the outbreaks belonged to the same genotype, GII.P2/GII.2, showing >99% nucleotide identities (nt id) in both fragments (ORF1 and ORF2) analysed. Thirty-two water samples were collected and analysed for bacterial contamination (Table 3). Seven samples were taken from the resort (site R) and the remaining 25 from five other sites (C, F, M, S, W), representative of the entire municipal water network. The majority (10/12) of the water samples collected immediately after the onset of the outbreak, between 21 May and 16 June 2016, at the resort (site R) and at sites F and C, located close to the resort, on the seaside and in the interior, respectively, were positive for coliform bacteria. More than 1000 colony-forming units (CFU)/ 100 ml (up to 2700 CFU/100 ml) were detected in four samples, while their absence is required in 100 ml by the Standard of Water Quality in Italy and seven of 11 samples tested were also Escherichia coli-positive (up to 900 CFU/100 ml). Enterococci were searched in 11 samples and three of them were found positive for this faecal contamination indicator. From 7 July to 16 August 2016, after control measures on the water network were undertaken, 13 water samples were collected for follow-up laboratory examination of the water system at F, R and C sites. All of them were negative for bacterial indicators of faecal contamination. No faecal bacteria were identified in the seven samples collected from M, W and S sites, including the well and the source that fed the water network, at any sampling time, before and after implementing the control measures. The research of NoV genome was conducted on concentrated water samples collected on 1 August from R and F sites (Table 3). NoV genome was detected in a single sample collected from the water entry of the resort involved in the outbreak (site R). Actually, only two out of 26 One-Step RT-PCR repetitions were positive for NoV viral capsid gene. The two ORF2-positive results were obtained at the University of Palermo Laboratory using both 1:10 diluted and undiluted extracted RNA. The sequence analysis of the partial ORF2 portion (289 bp) indicated the presence of the same GII.2 capsid genotype detected in the stool samples of the students involved in the outbreak, showing 98.9% nt id. No ORF1 NoV sequence could be obtained from the NoV-positive water sample. Follow-up analyses for NoV genome, conducted on water samples collected on 20 November both at R and F sites found no positive samples. By sequence alignment, the NoV sequences obtained from Mascali outbreak samples were phylogenetically closely related in both ORF1 and ORF2 (99 and 98.9% of nt id, respectively) to a GII.P2/GII.2 strain responsible for a sporadic paediatric gastroenteritis case occurred in Palermo in January 2016 (PA 36/16, accession numbers MG517464 and MG517460, for ORF1 and ORF2, respectively) and previously detected at the Enteric Viruses Laboratory, University of Palermo. Whereas, three GII.P2/GII.2 Italian strains (PA 280/11, PA 366/11 and PA 3/ 14, accession numbers MG517462, MG517465 and MG517463, for ORF1 and MG517457 to −59, for ORF2, respectively), detected in Palermo in 2011 and 2014, were more distantly associated with the outbreak strain both in RdRp and VP1 (90% to 97.4% nt id and 96.1% to 98.1% nt id, respectively, in ORF1 and ORF2). Discussion Epidemiological features, such as the presence of vomiting in more than half (88%) of symptomatic subjects, the 33.2 h average incubation period and the 38.4 h average duration of illness, suggested a NoV aetiology for the investigated outbreak [13]. The detection of NoV genome in all 25 faecal samples from symptomatic subjects fully complies with the requirement of detection of at least one positive sample out of four tested that is considered to be sufficient to ascertain the aetiology of a NoV outbreak [24]. Moreover, molecular typing indicated that a single viral strain, belonging to genotype GII.P2/GII.2, infected all symptomatic subjects involved in the outbreak. NoVs are highly infectious and can easily spread from person to person but also through contaminated food or water. Therefore, settings such as health care institutions, schools and touristic resorts are mainly affected, since close contact and shared food services facilitate transmission. In Mascali, two waves of transmission were observed over three days, being consistent with food/waterborne followed by person-to-person and/or airborne (vomit) transmission (Fig. 2). The first wave of cases presented on 20 May 2016 after a single shared meal, while the last case was detected on 22 May. Exposure to vomit may have been responsible for some of the secondary cases occurred after the first wave of illness. Vomiting produces aerosols that can be inhaled and swallowed. The extremely low infectious dose of NoV makes the inhalation of vomit aerosols an effective transmission route in settings where there is close contact, like on a bus trip [25]. All food consumed in the only shared lunch on 19 May at the resort was associated with the risk of disease (Table 2), but the association was inconclusive since all cases shared the same dishes included in the lunch menu on Day 1. Although many foods were potentially involved, our attention focused on the tap water. In fact, trivial information from local health authorities reported that local population avoided tap water consumption and relied on bottled water for drinking and cooking. Many small outbreaks of gastroenteritis involving restaurants and family groups had been reported in the same area, especially at the beginning of the summer season, when vacation houses and recreational activities were re-opened after a period of low water demand. Illness caused by NoV is common and if the level of water contamination is low, the number of cases remains limited and water can persist as unrecognised source of infection. If polluted water was served for drinking and used for cooking and washing dishes in the Mascali resort at Day 1 a generalised contamination of food and surfaces could have occurred. The potential role of the foodhandlers as a source of contamination could not be analysed in this study but should also be taken into account. If water contamination has already been present in the area for a long time, infections may have occurred among the local food-handlers. The analysis of water samples taken immediately after the outbreak revealed the presence of bacterial indicators of faecal contamination in the water used in the resort, with the only exception of a sample of filter purified water used both for cooking and as drinking water. When extending the sampling to other areas of the municipal water network, many other sites, mostly on the seaside, showed faecal pollution, which demonstrates a poor protection of the water network from faecal leakage (Fig. 1). NoV contamination of waters generally originates from sewage since these enteric pathogens, as other enteric viruses affecting humans, are mostly species-specific and are shed at high concentrations in the stools of patients. When the analyses on the stools of the students made it clear that the outbreak had been caused by NoV, the viral genome was also searched in appropriate volumes of water samples, but this happened when correction measures on the water network had already been taken. Nevertheless, a single water sample, obtained 2 months after the outbreak (1 August 2016), still contained genome fragments that could be correlated to the same NoV genotype causing the outbreak. Although the finding of NoV genome does not demonstrate that entire infecting viral particles were present in the water, it could indicate that a NoV belonging to the same genotype involved in the outbreak was still circulating in the local population several weeks after the outbreak and it was someway able to penetrate the water network, probably indicating continuing faecal contamination through leakage from sewage systems. NoV is particularly resistant in the environment and also to chlorine, therefore transmission through chlorinated water is not exceptional [9, 12, 26]. Drinking water from municipal supplies fed by lake or groundwater has already been identified in the past as the source of waterborne NoV outbreaks in Italy [27–29] and NoV is now considered one of the pathogens causing the largest number of affected consumers in drinking waterborne outbreaks [12]. NoV can persist in water longer than 15 days [30] and it is resistant to chlorine disinfection when free chlorine levels or pre-treatment processes are inadequate [31, 32]. Moreover, microfiltration filters labelled as certified by NSF Standards 53 or 58 (with a pore size ranging from 0.05 to 5 µm) are not effective in removing viruses and can even decrease chlorine concentration in filtered water [33]. Actually, the owner of the resort involved in our outbreak was encouraged to serve tap water as drinking water at the restaurant relying on the effectiveness of such a filter purifier. Unfortunately, we did not have a tap water sample from the restaurant of the resort taken before correction measures and whose volume was appropriate for virus detection. Therefore, we could not demonstrate the failure to eliminate viral pollution of the microfiltration system. In Mascali, correction measures on the water network consisted in the maintenance of the old chlorination device and in the addition of a new one at a key point located upstream and close to suspected illegal sewage dumping (Fig. 1), but as poor protection of the pipeline from sewage leakage risked to reproduce faecal pollution episodes, renovation of the municipal water system was also advised, as well as systematic control of illegal sewage dumping. The additional chlorination device was apparently the most effective intervention, since right after its activation faecal bacterial pollution indicators disappeared. However, NoV genome was detected in a water sample taken 6 weeks after the last detection of bacterial indicators. Unfortunately, detection of faecal contamination of drinking water is only based on the use of bacterial indicators despite no correlation between levels of enteric bacteria and enteric viruses has been found [34–36]. According to ISGEV enteric virus circulation surveillance data, in 2016 the circulation of GII.P2/GII.2 NoV strains in Palermo, home town of the travellers involved in the outbreak, was the highest ever observed from the starting of local NoV surveillance in 2004. GII.P2/GII.2 was the fourth most common genotype in 2016 (12.2% prevalence), following GII.P4/GII.4 Sydney 2012 (28.4%), GII.P16/GII.2 (18.9%) and GII.Pe/GII.4 Sydney 2012 (17.6%). GII.P2/GII.2 strains were mostly observed starting from July 2016, leaving space for the possibility that their increased circulation was linked to the introduction of the ‘Mascali clone’. However, a couple of GII.P2/GII.2 strains (including PA36/16) genetically close to the ‘Mascali clone’ had been already detected in January 2016 in Palermo and GII.P2/GII.2 NoV had sporadically been observed also in 2011 and 2014. Moreover, the circulation of GII.P2/GII.2 NoVs has been increasing worldwide over the last period [37]. Therefore, the unusually high circulation observed in Palermo might just reflect the global epidemiology of GII.P2/GII.2 strains. Noteworthy, the older GII.P2/GII.2 NoV strains detected in Palermo were closely related in the ORF2 (98.9% nt id) to the GII.P16/GII.2 Nashville strain (KY865307) which is supposed to have generated the recombinant GII.P16/GII.4 clone recently circulating epidemically worldwide [37, 38] and were more distantly associated (95.3% nt id) to the GII.2 capsid sequence (AB662868) of another putative GII.P16/GII.4 ancestor from Japan [39]. The increasing frequency of detection of waterborne NoV outbreaks calls for the introduction of routine monitoring for viruses of drinking waters and for the definition of affordable and reliable indicators for the presence of NoV in drinking waters. The use of ultrafiltration systems as a concentration method in order to recover viruses from environmental waters combined with RT-PCR to detect viral genomes has been demonstrated to grant a recovery rate of >50% of multiple viruses [40] and is at present one of the most sensitive approaches for the detection of NoVs, although the discrimination between viable and nonviable viruses remains a problem. The burden of NoV outbreaks, either food associated or waterborne, is still poorly defined due to the lack of specific diagnostic tools in many clinical laboratories. NoV epidemics should always be suspected when the main symptom is vomiting and the duration of disease is short (<48 h). Waterborne transmission is suggested by the rapid spread of disease and the high attack rate. Care should be taken to perform sensitive diagnostic tests on a sufficient number of stool samples in order to identify NoV as the causative agent of enteritis outbreaks, possibly preferring genome screening over antigen detection. Only the systematic detection of NoV outbreaks can guide in the choice of the safety procedures that could be most appropriate in containing such microbiological hazard.

What is the concentration of the pathogens?

1559

Waterborne Norovirus outbreak at a seaside resort likely originating from municipal water distribution system failure

Abstract In May 2016 a Norovirus (NoV) gastroenteritis outbreak involved a high school class visiting a seaside resort near Taormina (Mascali, Sicily). Twenty-four students and a teacher were affected and 17 of them showed symptoms on the second day of the journey, while the others got ill within the following 2 days. Symptoms included vomiting, diarrhoea and fever, and 12 students required hospitalisation. Stool samples tested positive for NoV genome by Real-Time polymerase chain reaction assay in all 25 symptomatic subjects. The GII.P2/GII.2 NoV genotype was linked to the outbreak by ORF1/ORF2 sequence analysis. The epidemiological features of the outbreak were consistent with food/waterborne followed by person-to-person and/ or vomit transmission. Food consumed at a shared lunch on the first day of the trip was associated to illness and drinking un-bottled tap water was also considered as a risk factor. The analysis of water samples revealed the presence of bacterial indicators of faecal contamination in the water used in the resort as well as in other areas of the municipal water network, linking the NoV gastroenteritis outbreak to tap water pollution from sewage leakage. From a single water sample, an amplicon whose sequence corresponded to the capsid genotype recovered from patients could be obtained. Introduction Noroviruses (NoVs) are recognised as a leading cause of sporadic cases and outbreaks of acute gastroenteritis (AGE) across all age groups [1, 2]. They are transmitted primarily through the faecal-oral route either by direct person–person spread and with consumption of contaminated food or water [1, 3, 4]. NoVs are non-enveloped RNA viruses belonging to the Caliciviridae family and are classified into seven distinct genogroups (GI-GVII) on the basis of the major capsid protein VP1 sequence. GI, GII and GIV NoVs infect humans and they have been divided into more than 30 genotypes based on the sequences of polymerase (ORF1) and capsid proteins (ORF2) [5]. NoV GII.4 genotype has been associated with the vast majority of NoV related cases of AGE worldwide although many genotypes, belonging to both GI and GII genogroups are involved in foodborne and waterborne outbreaks [6, 7]. NoVs are highly contagious owing to their high stability in the environment and their ability to infect at low doses. In fact, just 10–100 virions of NoV may be sufficient to cause AGE [8, 9]. NoVs are considered the major cause of water-related disease in semi-closed institutions, including restaurants, schools, hospitals, nursing homes and cruise ships [10–12]. To confirm NoV as a cause of outbreaks, the presence of four epidemiological features of disease, vomiting in more than half of affected persons, a mean incubation period of 24/48 h, a mean duration of illness of 12–60 h and the absence of bacterial pathogens in stool culture, should be evaluated [13]. Waterborne NoV outbreaks can be caused by contaminated drinking water despite correct chlorine concentration and in the absence of coliform bacteria [10, 14–17]. The aim of this study was to investigate an outbreak of NoV gastroenteritis associated with drinking water involving a high school class travelling from Palermo (West Sicily) to Acireale (East Sicily) on an educational tour in May 2016. Stool samples were collected soon after the onset of AGE symptoms to ascertain the aetiology of illness. The source of infection was investigated collecting epidemiological data through an online questionnaire and, in order to explore a possible transmission route, laboratory tests on water from the municipal network were performed. Material and methods Epidemiological investigation From 20 to 22 May 2016 an AGE outbreak occurred in Mascali, a small town near Taormina, involving a school class of 28 students, 13 girls and 15 boys aged between 14 and 18 years and a 54-year-old female teacher, during a trip to participate in a theatre festival in Acireale. A total of 25/29 persons were affected by nausea, vomiting, abdominal cramps, diarrhoea and fever while they were staying at a seaside resort in Mascali, on the way to Acireale. To obtain data on symptoms and the correlated risk factors all participants filled an online questionnaire investigating personal data (age and sex), travel and accommodation details and food consumption over 48 h before symptoms presentation, time of onset and duration of symptoms, therapy and/or hospitalisation required and history of contact with people affected by diarrhoea and/or vomiting (Table 1). The severity of illness was analysed by the Vesikari severity scale allowing to distinguish non-severe (score <11) and severe (score ⩾11) AGE (Table 1). Vesikari score was calculated from maximum number of episodes/day and duration of both diarrhoea and vomiting (score 0–3 for each of the 4 parameters), temperature (score 0–3), level of dehydration (score 0–3) and treatment required (score 0–2) [18]. A case-control study was conducted to establish if food consumed before the onset of symptoms was associated with the disease. A case was defined as a student attending the tour and reporting diarrhoea and/or vomiting within 36 h from the first shared meal during the tour, while students who did not present any of the previous symptoms or those showing the first symptoms with a delay of ⩾48 h (secondary cases) were used as controls. Variables were studied using odds ratio (OR) and its 95% CI and statistical associations were determined using the χ2 test, with the level of statistical significance set at 0.05 (Table 2). NoV detection in stool samples A stool sample was obtained from each of the 29 travellers, soon after (3–5 days) the onset of the outbreak. No stool samples could be obtained from the food-handlers working at the resort. All samples were tested for NoV at the Enteric Viruses Laboratory of the University of Palermo, member of the Italian Study Group for Enteric Viruses (ISGEV), by a Real-Time polymerase chain reaction (RT-PCR) assay that allows to differentiate GI and GII NoVs [19]. Total RNA was extracted from stool samples using QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) and cDNA was generated by MMLV reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were considered positive for NoVs when cycle threshold (ct) of RT-PCR was ⩽30 and lowpositive when 30<ct⩽35. For all NoV-positive samples two genomic fragments were amplified by conventional RT-PCR using Superscript III One-Step RT-PCR system with Platinum Taq DNA polymerase (Invitrogen) and primers JV12a/JV13b, targeting 326 bps of the RdRp region A [20] and either COG1F/ GISKR or COG2F/GIISKR, amplifying a 315 bps 5′portion of GI or GII VP1 protein [19, 21]. Environmental investigation From May through August 2016, a total of 32 tap water samples were collected from the Mascali municipal network. In particular, seven samples were collected at the resort where the outbreak occurred and from its water entry (R), two in the city centre of Mascali (M), 15 in the village of Fondachello (F), on the seaside, three in Carrabba village (C), far from the coast, and two and three, respectively, from the water well ‘Carlino’ (W) and the source ‘Bufardo and Torregrossa’ (S) which feed the municipal system and provide drinking water to the Mascali area (Fig. 1, Table 3). One litre water samples were sent by the local health authorities to the SIALAB laboratory in Avola, Siracusa and 100 ml aliquots were screened for coliforms, E. coli and enterococci as faecal indicators, by standard bacteriological methods according to UNI EN ISO 9308-1:2014. Unfortunately, although the suspect of a water source for the viral outbreak was quickly communicated, only in August and November 2016, following the implementation of control measures, local authorities provided water samplings of appropriate volume to perform viral tests (10 l). Five water samples of 10 l each were collected from different sampling sites in F and R areas for viral screening (Table 3) and were concentrated according to the Italian National Institute of Health (Istituto Superiore di Sanità, ISS) protocols [22] through an ultra-filtration system (Sartoflow® Slice 200 Benchtop Crossflow System, Sartorius AG, Goettingen, Germany), using appropriate membranes (SG Hydrosart 10 kDa) pretreated with 300 ml of 3% Beef Extract (BE) at pH 7, by recirculating it for 10 min in the system, at a pressure not higher than 1.30 bar. Elution and recovery of viruses attached to the membranes were obtained by recirculating 200 ml of 3% BE at pH 9.5 for 3 h until a 10–12 ml volume of eluate was obtained. The pH of the eluate was then checked with litmus paper and if necessary brought to the value of 7 with HCl. Concentrated water samples were simultaneously tested for NoV in three laboratories: Enteric Viruses Laboratory, University of Palermo; Istituto Zooprofilattico Sperimentale della Sicilia (IZS), Palermo; and Istituto Superiore di Sanità (ISS), Rome. Total RNA was extracted from 500 and 560 µl of a concentrated water sample using BioMèrieux NucliSENS® Magnetic Extraction System (BioMèrieux) and QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) respectively, according to manufacturer’s instructions. Bonded RNA was eluted by 100 µl of Elution Buffer. Amplifications were performed at least in duplicate on whole and 1:10 diluted extracts. A Real-Time RT-PCR assay according to ISO/TS 15216- 2:2013, two in-house Real-Time RT-PCR protocols and the One-Step RT-PCR methods already used for amplifying ORF1 and ORF2 fragments in NoV-positive faecal samples were performed [19–21, 23]. Nucleotide sequences analysis and phylogenetic analysis PCR amplicons were purified by commercial columns (CleanSweep PCR Purification, Thermo Fisher Scientific, Vilnius, Lithuania) and sequenced by Sanger method and all nucleotide sequences were aligned using ClustalW. To identify NoV genotypes, RdRp and VP1 sequences were submitted to the RIVM NoV genotyping tool (http://www.rivm.nl/mpf/norovirus/typingtool). Phylogenetic trees were constructed using MEGA 7 and Neighbour-Joining method (bootstrap 1000 replicates). Sequences from stool and water samples are available at NCBI with the following accession numbers: MG517455 for the major capsid protein obtained from a water sample; MG517456 and MG517461, respectively, for the major capsid protein and RNA-dependent-RNA-polymerase, as representatives of the sequences obtained from the stool samples of those involved in the outbreak. Control measures After outbreak notification, on 21 May, the resort was closed for 2 days. On 25 May, following the first results of the bacteriological analyses on municipal water samples, the use of municipal water for human consumption was temporarily restricted (use of water was permitted only for cleaning purposes) in the Carrabba and Fondachello villages and in the area surrounding the resort. From 25 May the reservoir collecting the water from the Carlino well and its chlorination device were submitted to maintenance work and a new chlorination device was installed on the conduit bringing the water from the Bufardo source. Actually, the Bufardo source had been connected to the municipal network in 2003 but unfortunately a dedicated chlorination station had not been planned at that time. Therefore, unchlorinated water entered the system near Fondachello village on the seaside (Fig. 1), possibly diluting chlorine concentrations in water treated at the Carlino reservoir. Moreover, the exact path of the branch of the water network situated downstream the Bufardo source water connection was not marked in the maps of the municipality (Fig. 1, red line), indicating that this section had never been submitted to maintenance. The area surrounding the unmaintained section of the water pipeline was scattered with unauthorised cottages that were not connected to the sewerage systems. Likely, untreated waste from cesspools was unloaded directly into the soil, generating a risk of contamination of the municipal water system through leakage. No work had been made on the water pipelines before the outbreak but sections of the network had been temporarily closed during the winter season on the seaside (Fondachello). Local health authorities advised the connection of all households in the area to the sewerage system and the replacement of the water system section without maintenance. Health authorities also raised attention on AGE cases notification at all local Emergency Departments and Family Practitioners but there was no further evidence of AGE outbreaks in the months following the intervention. Results Data collected through the online questionnaire compiled 7–10 days after the outbreak revealed that none of the 28 students and the teacher taking part in the school trip showed gastroenteritis symptoms at the beginning of the journey and they had no contact with people who suffered from diarrhoea and/or vomiting in the previous days. Travellers had their first shared meal on 19 May (Day 1), 2016, at lunch on their arrival to the resort in Mascali. The list of foods included in the menu is shown in Table 2. Conversely, the dinner on Day 1 was not shared and the students went to dine in three different fast food restaurants in Acireale. On 20 May (Day 2), 17 students (58.6% of the travellers) presented with vomit and/or diarrhoea, falling ill 26–36 h after lunch on Day 1, strongly suggesting that exposure occurred on the first shared meal. Some students also vomited in the presence of the other travellers on the bus bringing them to Acireale. Five more students were symptomatic in the early hours of the following day (21 May, Day 3, from 0:30 am to 8:00 am), as a continuation of the wave of illnesses of the previous day. The teacher and one more student fell ill at 8:00 pm and 11:00 pm of Day 3, respectively, for a second wave of illnesses occurring >48 h from the first shared meal and <36 h from the first cases, with the last student falling sick on 22 May (Day 4) at 3:00 am (Fig. 2, Table 1). On the whole, the mean incubation period was 33.2 h. Based on the online survey, the severity of illness was evaluated by the Vesikari severity scale, grouping symptomatic travellers into severe (11/25 = 44%) and non-severe (14/25 = 56%) cases (Table 1). Vomiting and fever were the predominant symptoms, affecting 22 out of 25 (88%) symptomatic subjects, while diarrhoea was present in 14 out of 25 (56%). Hospitalisation for rehydration therapy was necessary for 13 students (91% [10/11] of severe and 21% [3/14] of non-severe cases). Emergency room (ER) or family practitioner visits were necessary for 21 students and 13 (62%) of them initiated an antibiotic therapy. Two students indicated that other family members had gastroenteric symptoms more than 24 h after their return home, on Day 3, 4 and 5. All those involved in the outbreak recovered within 4 days, with an average duration of illness of 38.4 h. A case-control study of the risk associated with eating foods at lunch on Day 1 or with attending different restaurants at dinner on Day 1 was attempted. A total of 19 subjects filled the section of the online survey concerning food consumption. Fourteen of the responders were among the 22 cases falling ill within 36 h from the first shared meal (first wave), while the remaining five responders were three travellers with supposed secondary infections (showing symptoms with a delay ⩾48 h; second wave) and two asymptomatic students that were used as controls. No statistically significant associations with any food items served at lunch on Day 1 was identified (P < 0.05; Table 2). However, all food consumed on that occasion was to some extent associated with the risk of disease, the highest OR being obtained with chicken (OR: 20.7143, P: 0.0678; Table 2). Actually, most of the cases ate all foods included in the menu, while one of the two asymptomatic controls did not participate in the meal and the other had a pasta dish only. No statistically significant association was found for attending a particular fast food restaurant at the dinner on Day 1 (OR: 2.5, P: 0.3635). The Real-Time RT-PCR assay allowed to detect GII genogroup NoV genome in all the 25 samples from symptomatic subjects, although two of them showed high Ct values, indicating lower shedding of viral genomes (Table 1). The samples from four asymptomatic students were negative for both NoV genogroups. By sequence analysis, a NoV genotype could be defined for 20/ 25 NoV-positive stool samples. All NoV strains from travellers involved in the outbreaks belonged to the same genotype, GII.P2/GII.2, showing >99% nucleotide identities (nt id) in both fragments (ORF1 and ORF2) analysed. Thirty-two water samples were collected and analysed for bacterial contamination (Table 3). Seven samples were taken from the resort (site R) and the remaining 25 from five other sites (C, F, M, S, W), representative of the entire municipal water network. The majority (10/12) of the water samples collected immediately after the onset of the outbreak, between 21 May and 16 June 2016, at the resort (site R) and at sites F and C, located close to the resort, on the seaside and in the interior, respectively, were positive for coliform bacteria. More than 1000 colony-forming units (CFU)/ 100 ml (up to 2700 CFU/100 ml) were detected in four samples, while their absence is required in 100 ml by the Standard of Water Quality in Italy and seven of 11 samples tested were also Escherichia coli-positive (up to 900 CFU/100 ml). Enterococci were searched in 11 samples and three of them were found positive for this faecal contamination indicator. From 7 July to 16 August 2016, after control measures on the water network were undertaken, 13 water samples were collected for follow-up laboratory examination of the water system at F, R and C sites. All of them were negative for bacterial indicators of faecal contamination. No faecal bacteria were identified in the seven samples collected from M, W and S sites, including the well and the source that fed the water network, at any sampling time, before and after implementing the control measures. The research of NoV genome was conducted on concentrated water samples collected on 1 August from R and F sites (Table 3). NoV genome was detected in a single sample collected from the water entry of the resort involved in the outbreak (site R). Actually, only two out of 26 One-Step RT-PCR repetitions were positive for NoV viral capsid gene. The two ORF2-positive results were obtained at the University of Palermo Laboratory using both 1:10 diluted and undiluted extracted RNA. The sequence analysis of the partial ORF2 portion (289 bp) indicated the presence of the same GII.2 capsid genotype detected in the stool samples of the students involved in the outbreak, showing 98.9% nt id. No ORF1 NoV sequence could be obtained from the NoV-positive water sample. Follow-up analyses for NoV genome, conducted on water samples collected on 20 November both at R and F sites found no positive samples. By sequence alignment, the NoV sequences obtained from Mascali outbreak samples were phylogenetically closely related in both ORF1 and ORF2 (99 and 98.9% of nt id, respectively) to a GII.P2/GII.2 strain responsible for a sporadic paediatric gastroenteritis case occurred in Palermo in January 2016 (PA 36/16, accession numbers MG517464 and MG517460, for ORF1 and ORF2, respectively) and previously detected at the Enteric Viruses Laboratory, University of Palermo. Whereas, three GII.P2/GII.2 Italian strains (PA 280/11, PA 366/11 and PA 3/ 14, accession numbers MG517462, MG517465 and MG517463, for ORF1 and MG517457 to −59, for ORF2, respectively), detected in Palermo in 2011 and 2014, were more distantly associated with the outbreak strain both in RdRp and VP1 (90% to 97.4% nt id and 96.1% to 98.1% nt id, respectively, in ORF1 and ORF2). Discussion Epidemiological features, such as the presence of vomiting in more than half (88%) of symptomatic subjects, the 33.2 h average incubation period and the 38.4 h average duration of illness, suggested a NoV aetiology for the investigated outbreak [13]. The detection of NoV genome in all 25 faecal samples from symptomatic subjects fully complies with the requirement of detection of at least one positive sample out of four tested that is considered to be sufficient to ascertain the aetiology of a NoV outbreak [24]. Moreover, molecular typing indicated that a single viral strain, belonging to genotype GII.P2/GII.2, infected all symptomatic subjects involved in the outbreak. NoVs are highly infectious and can easily spread from person to person but also through contaminated food or water. Therefore, settings such as health care institutions, schools and touristic resorts are mainly affected, since close contact and shared food services facilitate transmission. In Mascali, two waves of transmission were observed over three days, being consistent with food/waterborne followed by person-to-person and/or airborne (vomit) transmission (Fig. 2). The first wave of cases presented on 20 May 2016 after a single shared meal, while the last case was detected on 22 May. Exposure to vomit may have been responsible for some of the secondary cases occurred after the first wave of illness. Vomiting produces aerosols that can be inhaled and swallowed. The extremely low infectious dose of NoV makes the inhalation of vomit aerosols an effective transmission route in settings where there is close contact, like on a bus trip [25]. All food consumed in the only shared lunch on 19 May at the resort was associated with the risk of disease (Table 2), but the association was inconclusive since all cases shared the same dishes included in the lunch menu on Day 1. Although many foods were potentially involved, our attention focused on the tap water. In fact, trivial information from local health authorities reported that local population avoided tap water consumption and relied on bottled water for drinking and cooking. Many small outbreaks of gastroenteritis involving restaurants and family groups had been reported in the same area, especially at the beginning of the summer season, when vacation houses and recreational activities were re-opened after a period of low water demand. Illness caused by NoV is common and if the level of water contamination is low, the number of cases remains limited and water can persist as unrecognised source of infection. If polluted water was served for drinking and used for cooking and washing dishes in the Mascali resort at Day 1 a generalised contamination of food and surfaces could have occurred. The potential role of the foodhandlers as a source of contamination could not be analysed in this study but should also be taken into account. If water contamination has already been present in the area for a long time, infections may have occurred among the local food-handlers. The analysis of water samples taken immediately after the outbreak revealed the presence of bacterial indicators of faecal contamination in the water used in the resort, with the only exception of a sample of filter purified water used both for cooking and as drinking water. When extending the sampling to other areas of the municipal water network, many other sites, mostly on the seaside, showed faecal pollution, which demonstrates a poor protection of the water network from faecal leakage (Fig. 1). NoV contamination of waters generally originates from sewage since these enteric pathogens, as other enteric viruses affecting humans, are mostly species-specific and are shed at high concentrations in the stools of patients. When the analyses on the stools of the students made it clear that the outbreak had been caused by NoV, the viral genome was also searched in appropriate volumes of water samples, but this happened when correction measures on the water network had already been taken. Nevertheless, a single water sample, obtained 2 months after the outbreak (1 August 2016), still contained genome fragments that could be correlated to the same NoV genotype causing the outbreak. Although the finding of NoV genome does not demonstrate that entire infecting viral particles were present in the water, it could indicate that a NoV belonging to the same genotype involved in the outbreak was still circulating in the local population several weeks after the outbreak and it was someway able to penetrate the water network, probably indicating continuing faecal contamination through leakage from sewage systems. NoV is particularly resistant in the environment and also to chlorine, therefore transmission through chlorinated water is not exceptional [9, 12, 26]. Drinking water from municipal supplies fed by lake or groundwater has already been identified in the past as the source of waterborne NoV outbreaks in Italy [27–29] and NoV is now considered one of the pathogens causing the largest number of affected consumers in drinking waterborne outbreaks [12]. NoV can persist in water longer than 15 days [30] and it is resistant to chlorine disinfection when free chlorine levels or pre-treatment processes are inadequate [31, 32]. Moreover, microfiltration filters labelled as certified by NSF Standards 53 or 58 (with a pore size ranging from 0.05 to 5 µm) are not effective in removing viruses and can even decrease chlorine concentration in filtered water [33]. Actually, the owner of the resort involved in our outbreak was encouraged to serve tap water as drinking water at the restaurant relying on the effectiveness of such a filter purifier. Unfortunately, we did not have a tap water sample from the restaurant of the resort taken before correction measures and whose volume was appropriate for virus detection. Therefore, we could not demonstrate the failure to eliminate viral pollution of the microfiltration system. In Mascali, correction measures on the water network consisted in the maintenance of the old chlorination device and in the addition of a new one at a key point located upstream and close to suspected illegal sewage dumping (Fig. 1), but as poor protection of the pipeline from sewage leakage risked to reproduce faecal pollution episodes, renovation of the municipal water system was also advised, as well as systematic control of illegal sewage dumping. The additional chlorination device was apparently the most effective intervention, since right after its activation faecal bacterial pollution indicators disappeared. However, NoV genome was detected in a water sample taken 6 weeks after the last detection of bacterial indicators. Unfortunately, detection of faecal contamination of drinking water is only based on the use of bacterial indicators despite no correlation between levels of enteric bacteria and enteric viruses has been found [34–36]. According to ISGEV enteric virus circulation surveillance data, in 2016 the circulation of GII.P2/GII.2 NoV strains in Palermo, home town of the travellers involved in the outbreak, was the highest ever observed from the starting of local NoV surveillance in 2004. GII.P2/GII.2 was the fourth most common genotype in 2016 (12.2% prevalence), following GII.P4/GII.4 Sydney 2012 (28.4%), GII.P16/GII.2 (18.9%) and GII.Pe/GII.4 Sydney 2012 (17.6%). GII.P2/GII.2 strains were mostly observed starting from July 2016, leaving space for the possibility that their increased circulation was linked to the introduction of the ‘Mascali clone’. However, a couple of GII.P2/GII.2 strains (including PA36/16) genetically close to the ‘Mascali clone’ had been already detected in January 2016 in Palermo and GII.P2/GII.2 NoV had sporadically been observed also in 2011 and 2014. Moreover, the circulation of GII.P2/GII.2 NoVs has been increasing worldwide over the last period [37]. Therefore, the unusually high circulation observed in Palermo might just reflect the global epidemiology of GII.P2/GII.2 strains. Noteworthy, the older GII.P2/GII.2 NoV strains detected in Palermo were closely related in the ORF2 (98.9% nt id) to the GII.P16/GII.2 Nashville strain (KY865307) which is supposed to have generated the recombinant GII.P16/GII.4 clone recently circulating epidemically worldwide [37, 38] and were more distantly associated (95.3% nt id) to the GII.2 capsid sequence (AB662868) of another putative GII.P16/GII.4 ancestor from Japan [39]. The increasing frequency of detection of waterborne NoV outbreaks calls for the introduction of routine monitoring for viruses of drinking waters and for the definition of affordable and reliable indicators for the presence of NoV in drinking waters. The use of ultrafiltration systems as a concentration method in order to recover viruses from environmental waters combined with RT-PCR to detect viral genomes has been demonstrated to grant a recovery rate of >50% of multiple viruses [40] and is at present one of the most sensitive approaches for the detection of NoVs, although the discrimination between viable and nonviable viruses remains a problem. The burden of NoV outbreaks, either food associated or waterborne, is still poorly defined due to the lack of specific diagnostic tools in many clinical laboratories. NoV epidemics should always be suspected when the main symptom is vomiting and the duration of disease is short (<48 h). Waterborne transmission is suggested by the rapid spread of disease and the high attack rate. Care should be taken to perform sensitive diagnostic tests on a sufficient number of stool samples in order to identify NoV as the causative agent of enteritis outbreaks, possibly preferring genome screening over antigen detection. Only the systematic detection of NoV outbreaks can guide in the choice of the safety procedures that could be most appropriate in containing such microbiological hazard.

What steps were taken to restore the problem?

1560

Waterborne Norovirus outbreak at a seaside resort likely originating from municipal water distribution system failure

Abstract In May 2016 a Norovirus (NoV) gastroenteritis outbreak involved a high school class visiting a seaside resort near Taormina (Mascali, Sicily). Twenty-four students and a teacher were affected and 17 of them showed symptoms on the second day of the journey, while the others got ill within the following 2 days. Symptoms included vomiting, diarrhoea and fever, and 12 students required hospitalisation. Stool samples tested positive for NoV genome by Real-Time polymerase chain reaction assay in all 25 symptomatic subjects. The GII.P2/GII.2 NoV genotype was linked to the outbreak by ORF1/ORF2 sequence analysis. The epidemiological features of the outbreak were consistent with food/waterborne followed by person-to-person and/ or vomit transmission. Food consumed at a shared lunch on the first day of the trip was associated to illness and drinking un-bottled tap water was also considered as a risk factor. The analysis of water samples revealed the presence of bacterial indicators of faecal contamination in the water used in the resort as well as in other areas of the municipal water network, linking the NoV gastroenteritis outbreak to tap water pollution from sewage leakage. From a single water sample, an amplicon whose sequence corresponded to the capsid genotype recovered from patients could be obtained. Introduction Noroviruses (NoVs) are recognised as a leading cause of sporadic cases and outbreaks of acute gastroenteritis (AGE) across all age groups [1, 2]. They are transmitted primarily through the faecal-oral route either by direct person–person spread and with consumption of contaminated food or water [1, 3, 4]. NoVs are non-enveloped RNA viruses belonging to the Caliciviridae family and are classified into seven distinct genogroups (GI-GVII) on the basis of the major capsid protein VP1 sequence. GI, GII and GIV NoVs infect humans and they have been divided into more than 30 genotypes based on the sequences of polymerase (ORF1) and capsid proteins (ORF2) [5]. NoV GII.4 genotype has been associated with the vast majority of NoV related cases of AGE worldwide although many genotypes, belonging to both GI and GII genogroups are involved in foodborne and waterborne outbreaks [6, 7]. NoVs are highly contagious owing to their high stability in the environment and their ability to infect at low doses. In fact, just 10–100 virions of NoV may be sufficient to cause AGE [8, 9]. NoVs are considered the major cause of water-related disease in semi-closed institutions, including restaurants, schools, hospitals, nursing homes and cruise ships [10–12]. To confirm NoV as a cause of outbreaks, the presence of four epidemiological features of disease, vomiting in more than half of affected persons, a mean incubation period of 24/48 h, a mean duration of illness of 12–60 h and the absence of bacterial pathogens in stool culture, should be evaluated [13]. Waterborne NoV outbreaks can be caused by contaminated drinking water despite correct chlorine concentration and in the absence of coliform bacteria [10, 14–17]. The aim of this study was to investigate an outbreak of NoV gastroenteritis associated with drinking water involving a high school class travelling from Palermo (West Sicily) to Acireale (East Sicily) on an educational tour in May 2016. Stool samples were collected soon after the onset of AGE symptoms to ascertain the aetiology of illness. The source of infection was investigated collecting epidemiological data through an online questionnaire and, in order to explore a possible transmission route, laboratory tests on water from the municipal network were performed. Material and methods Epidemiological investigation From 20 to 22 May 2016 an AGE outbreak occurred in Mascali, a small town near Taormina, involving a school class of 28 students, 13 girls and 15 boys aged between 14 and 18 years and a 54-year-old female teacher, during a trip to participate in a theatre festival in Acireale. A total of 25/29 persons were affected by nausea, vomiting, abdominal cramps, diarrhoea and fever while they were staying at a seaside resort in Mascali, on the way to Acireale. To obtain data on symptoms and the correlated risk factors all participants filled an online questionnaire investigating personal data (age and sex), travel and accommodation details and food consumption over 48 h before symptoms presentation, time of onset and duration of symptoms, therapy and/or hospitalisation required and history of contact with people affected by diarrhoea and/or vomiting (Table 1). The severity of illness was analysed by the Vesikari severity scale allowing to distinguish non-severe (score <11) and severe (score ⩾11) AGE (Table 1). Vesikari score was calculated from maximum number of episodes/day and duration of both diarrhoea and vomiting (score 0–3 for each of the 4 parameters), temperature (score 0–3), level of dehydration (score 0–3) and treatment required (score 0–2) [18]. A case-control study was conducted to establish if food consumed before the onset of symptoms was associated with the disease. A case was defined as a student attending the tour and reporting diarrhoea and/or vomiting within 36 h from the first shared meal during the tour, while students who did not present any of the previous symptoms or those showing the first symptoms with a delay of ⩾48 h (secondary cases) were used as controls. Variables were studied using odds ratio (OR) and its 95% CI and statistical associations were determined using the χ2 test, with the level of statistical significance set at 0.05 (Table 2). NoV detection in stool samples A stool sample was obtained from each of the 29 travellers, soon after (3–5 days) the onset of the outbreak. No stool samples could be obtained from the food-handlers working at the resort. All samples were tested for NoV at the Enteric Viruses Laboratory of the University of Palermo, member of the Italian Study Group for Enteric Viruses (ISGEV), by a Real-Time polymerase chain reaction (RT-PCR) assay that allows to differentiate GI and GII NoVs [19]. Total RNA was extracted from stool samples using QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) and cDNA was generated by MMLV reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were considered positive for NoVs when cycle threshold (ct) of RT-PCR was ⩽30 and lowpositive when 30<ct⩽35. For all NoV-positive samples two genomic fragments were amplified by conventional RT-PCR using Superscript III One-Step RT-PCR system with Platinum Taq DNA polymerase (Invitrogen) and primers JV12a/JV13b, targeting 326 bps of the RdRp region A [20] and either COG1F/ GISKR or COG2F/GIISKR, amplifying a 315 bps 5′portion of GI or GII VP1 protein [19, 21]. Environmental investigation From May through August 2016, a total of 32 tap water samples were collected from the Mascali municipal network. In particular, seven samples were collected at the resort where the outbreak occurred and from its water entry (R), two in the city centre of Mascali (M), 15 in the village of Fondachello (F), on the seaside, three in Carrabba village (C), far from the coast, and two and three, respectively, from the water well ‘Carlino’ (W) and the source ‘Bufardo and Torregrossa’ (S) which feed the municipal system and provide drinking water to the Mascali area (Fig. 1, Table 3). One litre water samples were sent by the local health authorities to the SIALAB laboratory in Avola, Siracusa and 100 ml aliquots were screened for coliforms, E. coli and enterococci as faecal indicators, by standard bacteriological methods according to UNI EN ISO 9308-1:2014. Unfortunately, although the suspect of a water source for the viral outbreak was quickly communicated, only in August and November 2016, following the implementation of control measures, local authorities provided water samplings of appropriate volume to perform viral tests (10 l). Five water samples of 10 l each were collected from different sampling sites in F and R areas for viral screening (Table 3) and were concentrated according to the Italian National Institute of Health (Istituto Superiore di Sanità, ISS) protocols [22] through an ultra-filtration system (Sartoflow® Slice 200 Benchtop Crossflow System, Sartorius AG, Goettingen, Germany), using appropriate membranes (SG Hydrosart 10 kDa) pretreated with 300 ml of 3% Beef Extract (BE) at pH 7, by recirculating it for 10 min in the system, at a pressure not higher than 1.30 bar. Elution and recovery of viruses attached to the membranes were obtained by recirculating 200 ml of 3% BE at pH 9.5 for 3 h until a 10–12 ml volume of eluate was obtained. The pH of the eluate was then checked with litmus paper and if necessary brought to the value of 7 with HCl. Concentrated water samples were simultaneously tested for NoV in three laboratories: Enteric Viruses Laboratory, University of Palermo; Istituto Zooprofilattico Sperimentale della Sicilia (IZS), Palermo; and Istituto Superiore di Sanità (ISS), Rome. Total RNA was extracted from 500 and 560 µl of a concentrated water sample using BioMèrieux NucliSENS® Magnetic Extraction System (BioMèrieux) and QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) respectively, according to manufacturer’s instructions. Bonded RNA was eluted by 100 µl of Elution Buffer. Amplifications were performed at least in duplicate on whole and 1:10 diluted extracts. A Real-Time RT-PCR assay according to ISO/TS 15216- 2:2013, two in-house Real-Time RT-PCR protocols and the One-Step RT-PCR methods already used for amplifying ORF1 and ORF2 fragments in NoV-positive faecal samples were performed [19–21, 23]. Nucleotide sequences analysis and phylogenetic analysis PCR amplicons were purified by commercial columns (CleanSweep PCR Purification, Thermo Fisher Scientific, Vilnius, Lithuania) and sequenced by Sanger method and all nucleotide sequences were aligned using ClustalW. To identify NoV genotypes, RdRp and VP1 sequences were submitted to the RIVM NoV genotyping tool (http://www.rivm.nl/mpf/norovirus/typingtool). Phylogenetic trees were constructed using MEGA 7 and Neighbour-Joining method (bootstrap 1000 replicates). Sequences from stool and water samples are available at NCBI with the following accession numbers: MG517455 for the major capsid protein obtained from a water sample; MG517456 and MG517461, respectively, for the major capsid protein and RNA-dependent-RNA-polymerase, as representatives of the sequences obtained from the stool samples of those involved in the outbreak. Control measures After outbreak notification, on 21 May, the resort was closed for 2 days. On 25 May, following the first results of the bacteriological analyses on municipal water samples, the use of municipal water for human consumption was temporarily restricted (use of water was permitted only for cleaning purposes) in the Carrabba and Fondachello villages and in the area surrounding the resort. From 25 May the reservoir collecting the water from the Carlino well and its chlorination device were submitted to maintenance work and a new chlorination device was installed on the conduit bringing the water from the Bufardo source. Actually, the Bufardo source had been connected to the municipal network in 2003 but unfortunately a dedicated chlorination station had not been planned at that time. Therefore, unchlorinated water entered the system near Fondachello village on the seaside (Fig. 1), possibly diluting chlorine concentrations in water treated at the Carlino reservoir. Moreover, the exact path of the branch of the water network situated downstream the Bufardo source water connection was not marked in the maps of the municipality (Fig. 1, red line), indicating that this section had never been submitted to maintenance. The area surrounding the unmaintained section of the water pipeline was scattered with unauthorised cottages that were not connected to the sewerage systems. Likely, untreated waste from cesspools was unloaded directly into the soil, generating a risk of contamination of the municipal water system through leakage. No work had been made on the water pipelines before the outbreak but sections of the network had been temporarily closed during the winter season on the seaside (Fondachello). Local health authorities advised the connection of all households in the area to the sewerage system and the replacement of the water system section without maintenance. Health authorities also raised attention on AGE cases notification at all local Emergency Departments and Family Practitioners but there was no further evidence of AGE outbreaks in the months following the intervention. Results Data collected through the online questionnaire compiled 7–10 days after the outbreak revealed that none of the 28 students and the teacher taking part in the school trip showed gastroenteritis symptoms at the beginning of the journey and they had no contact with people who suffered from diarrhoea and/or vomiting in the previous days. Travellers had their first shared meal on 19 May (Day 1), 2016, at lunch on their arrival to the resort in Mascali. The list of foods included in the menu is shown in Table 2. Conversely, the dinner on Day 1 was not shared and the students went to dine in three different fast food restaurants in Acireale. On 20 May (Day 2), 17 students (58.6% of the travellers) presented with vomit and/or diarrhoea, falling ill 26–36 h after lunch on Day 1, strongly suggesting that exposure occurred on the first shared meal. Some students also vomited in the presence of the other travellers on the bus bringing them to Acireale. Five more students were symptomatic in the early hours of the following day (21 May, Day 3, from 0:30 am to 8:00 am), as a continuation of the wave of illnesses of the previous day. The teacher and one more student fell ill at 8:00 pm and 11:00 pm of Day 3, respectively, for a second wave of illnesses occurring >48 h from the first shared meal and <36 h from the first cases, with the last student falling sick on 22 May (Day 4) at 3:00 am (Fig. 2, Table 1). On the whole, the mean incubation period was 33.2 h. Based on the online survey, the severity of illness was evaluated by the Vesikari severity scale, grouping symptomatic travellers into severe (11/25 = 44%) and non-severe (14/25 = 56%) cases (Table 1). Vomiting and fever were the predominant symptoms, affecting 22 out of 25 (88%) symptomatic subjects, while diarrhoea was present in 14 out of 25 (56%). Hospitalisation for rehydration therapy was necessary for 13 students (91% [10/11] of severe and 21% [3/14] of non-severe cases). Emergency room (ER) or family practitioner visits were necessary for 21 students and 13 (62%) of them initiated an antibiotic therapy. Two students indicated that other family members had gastroenteric symptoms more than 24 h after their return home, on Day 3, 4 and 5. All those involved in the outbreak recovered within 4 days, with an average duration of illness of 38.4 h. A case-control study of the risk associated with eating foods at lunch on Day 1 or with attending different restaurants at dinner on Day 1 was attempted. A total of 19 subjects filled the section of the online survey concerning food consumption. Fourteen of the responders were among the 22 cases falling ill within 36 h from the first shared meal (first wave), while the remaining five responders were three travellers with supposed secondary infections (showing symptoms with a delay ⩾48 h; second wave) and two asymptomatic students that were used as controls. No statistically significant associations with any food items served at lunch on Day 1 was identified (P < 0.05; Table 2). However, all food consumed on that occasion was to some extent associated with the risk of disease, the highest OR being obtained with chicken (OR: 20.7143, P: 0.0678; Table 2). Actually, most of the cases ate all foods included in the menu, while one of the two asymptomatic controls did not participate in the meal and the other had a pasta dish only. No statistically significant association was found for attending a particular fast food restaurant at the dinner on Day 1 (OR: 2.5, P: 0.3635). The Real-Time RT-PCR assay allowed to detect GII genogroup NoV genome in all the 25 samples from symptomatic subjects, although two of them showed high Ct values, indicating lower shedding of viral genomes (Table 1). The samples from four asymptomatic students were negative for both NoV genogroups. By sequence analysis, a NoV genotype could be defined for 20/ 25 NoV-positive stool samples. All NoV strains from travellers involved in the outbreaks belonged to the same genotype, GII.P2/GII.2, showing >99% nucleotide identities (nt id) in both fragments (ORF1 and ORF2) analysed. Thirty-two water samples were collected and analysed for bacterial contamination (Table 3). Seven samples were taken from the resort (site R) and the remaining 25 from five other sites (C, F, M, S, W), representative of the entire municipal water network. The majority (10/12) of the water samples collected immediately after the onset of the outbreak, between 21 May and 16 June 2016, at the resort (site R) and at sites F and C, located close to the resort, on the seaside and in the interior, respectively, were positive for coliform bacteria. More than 1000 colony-forming units (CFU)/ 100 ml (up to 2700 CFU/100 ml) were detected in four samples, while their absence is required in 100 ml by the Standard of Water Quality in Italy and seven of 11 samples tested were also Escherichia coli-positive (up to 900 CFU/100 ml). Enterococci were searched in 11 samples and three of them were found positive for this faecal contamination indicator. From 7 July to 16 August 2016, after control measures on the water network were undertaken, 13 water samples were collected for follow-up laboratory examination of the water system at F, R and C sites. All of them were negative for bacterial indicators of faecal contamination. No faecal bacteria were identified in the seven samples collected from M, W and S sites, including the well and the source that fed the water network, at any sampling time, before and after implementing the control measures. The research of NoV genome was conducted on concentrated water samples collected on 1 August from R and F sites (Table 3). NoV genome was detected in a single sample collected from the water entry of the resort involved in the outbreak (site R). Actually, only two out of 26 One-Step RT-PCR repetitions were positive for NoV viral capsid gene. The two ORF2-positive results were obtained at the University of Palermo Laboratory using both 1:10 diluted and undiluted extracted RNA. The sequence analysis of the partial ORF2 portion (289 bp) indicated the presence of the same GII.2 capsid genotype detected in the stool samples of the students involved in the outbreak, showing 98.9% nt id. No ORF1 NoV sequence could be obtained from the NoV-positive water sample. Follow-up analyses for NoV genome, conducted on water samples collected on 20 November both at R and F sites found no positive samples. By sequence alignment, the NoV sequences obtained from Mascali outbreak samples were phylogenetically closely related in both ORF1 and ORF2 (99 and 98.9% of nt id, respectively) to a GII.P2/GII.2 strain responsible for a sporadic paediatric gastroenteritis case occurred in Palermo in January 2016 (PA 36/16, accession numbers MG517464 and MG517460, for ORF1 and ORF2, respectively) and previously detected at the Enteric Viruses Laboratory, University of Palermo. Whereas, three GII.P2/GII.2 Italian strains (PA 280/11, PA 366/11 and PA 3/ 14, accession numbers MG517462, MG517465 and MG517463, for ORF1 and MG517457 to −59, for ORF2, respectively), detected in Palermo in 2011 and 2014, were more distantly associated with the outbreak strain both in RdRp and VP1 (90% to 97.4% nt id and 96.1% to 98.1% nt id, respectively, in ORF1 and ORF2). Discussion Epidemiological features, such as the presence of vomiting in more than half (88%) of symptomatic subjects, the 33.2 h average incubation period and the 38.4 h average duration of illness, suggested a NoV aetiology for the investigated outbreak [13]. The detection of NoV genome in all 25 faecal samples from symptomatic subjects fully complies with the requirement of detection of at least one positive sample out of four tested that is considered to be sufficient to ascertain the aetiology of a NoV outbreak [24]. Moreover, molecular typing indicated that a single viral strain, belonging to genotype GII.P2/GII.2, infected all symptomatic subjects involved in the outbreak. NoVs are highly infectious and can easily spread from person to person but also through contaminated food or water. Therefore, settings such as health care institutions, schools and touristic resorts are mainly affected, since close contact and shared food services facilitate transmission. In Mascali, two waves of transmission were observed over three days, being consistent with food/waterborne followed by person-to-person and/or airborne (vomit) transmission (Fig. 2). The first wave of cases presented on 20 May 2016 after a single shared meal, while the last case was detected on 22 May. Exposure to vomit may have been responsible for some of the secondary cases occurred after the first wave of illness. Vomiting produces aerosols that can be inhaled and swallowed. The extremely low infectious dose of NoV makes the inhalation of vomit aerosols an effective transmission route in settings where there is close contact, like on a bus trip [25]. All food consumed in the only shared lunch on 19 May at the resort was associated with the risk of disease (Table 2), but the association was inconclusive since all cases shared the same dishes included in the lunch menu on Day 1. Although many foods were potentially involved, our attention focused on the tap water. In fact, trivial information from local health authorities reported that local population avoided tap water consumption and relied on bottled water for drinking and cooking. Many small outbreaks of gastroenteritis involving restaurants and family groups had been reported in the same area, especially at the beginning of the summer season, when vacation houses and recreational activities were re-opened after a period of low water demand. Illness caused by NoV is common and if the level of water contamination is low, the number of cases remains limited and water can persist as unrecognised source of infection. If polluted water was served for drinking and used for cooking and washing dishes in the Mascali resort at Day 1 a generalised contamination of food and surfaces could have occurred. The potential role of the foodhandlers as a source of contamination could not be analysed in this study but should also be taken into account. If water contamination has already been present in the area for a long time, infections may have occurred among the local food-handlers. The analysis of water samples taken immediately after the outbreak revealed the presence of bacterial indicators of faecal contamination in the water used in the resort, with the only exception of a sample of filter purified water used both for cooking and as drinking water. When extending the sampling to other areas of the municipal water network, many other sites, mostly on the seaside, showed faecal pollution, which demonstrates a poor protection of the water network from faecal leakage (Fig. 1). NoV contamination of waters generally originates from sewage since these enteric pathogens, as other enteric viruses affecting humans, are mostly species-specific and are shed at high concentrations in the stools of patients. When the analyses on the stools of the students made it clear that the outbreak had been caused by NoV, the viral genome was also searched in appropriate volumes of water samples, but this happened when correction measures on the water network had already been taken. Nevertheless, a single water sample, obtained 2 months after the outbreak (1 August 2016), still contained genome fragments that could be correlated to the same NoV genotype causing the outbreak. Although the finding of NoV genome does not demonstrate that entire infecting viral particles were present in the water, it could indicate that a NoV belonging to the same genotype involved in the outbreak was still circulating in the local population several weeks after the outbreak and it was someway able to penetrate the water network, probably indicating continuing faecal contamination through leakage from sewage systems. NoV is particularly resistant in the environment and also to chlorine, therefore transmission through chlorinated water is not exceptional [9, 12, 26]. Drinking water from municipal supplies fed by lake or groundwater has already been identified in the past as the source of waterborne NoV outbreaks in Italy [27–29] and NoV is now considered one of the pathogens causing the largest number of affected consumers in drinking waterborne outbreaks [12]. NoV can persist in water longer than 15 days [30] and it is resistant to chlorine disinfection when free chlorine levels or pre-treatment processes are inadequate [31, 32]. Moreover, microfiltration filters labelled as certified by NSF Standards 53 or 58 (with a pore size ranging from 0.05 to 5 µm) are not effective in removing viruses and can even decrease chlorine concentration in filtered water [33]. Actually, the owner of the resort involved in our outbreak was encouraged to serve tap water as drinking water at the restaurant relying on the effectiveness of such a filter purifier. Unfortunately, we did not have a tap water sample from the restaurant of the resort taken before correction measures and whose volume was appropriate for virus detection. Therefore, we could not demonstrate the failure to eliminate viral pollution of the microfiltration system. In Mascali, correction measures on the water network consisted in the maintenance of the old chlorination device and in the addition of a new one at a key point located upstream and close to suspected illegal sewage dumping (Fig. 1), but as poor protection of the pipeline from sewage leakage risked to reproduce faecal pollution episodes, renovation of the municipal water system was also advised, as well as systematic control of illegal sewage dumping. The additional chlorination device was apparently the most effective intervention, since right after its activation faecal bacterial pollution indicators disappeared. However, NoV genome was detected in a water sample taken 6 weeks after the last detection of bacterial indicators. Unfortunately, detection of faecal contamination of drinking water is only based on the use of bacterial indicators despite no correlation between levels of enteric bacteria and enteric viruses has been found [34–36]. According to ISGEV enteric virus circulation surveillance data, in 2016 the circulation of GII.P2/GII.2 NoV strains in Palermo, home town of the travellers involved in the outbreak, was the highest ever observed from the starting of local NoV surveillance in 2004. GII.P2/GII.2 was the fourth most common genotype in 2016 (12.2% prevalence), following GII.P4/GII.4 Sydney 2012 (28.4%), GII.P16/GII.2 (18.9%) and GII.Pe/GII.4 Sydney 2012 (17.6%). GII.P2/GII.2 strains were mostly observed starting from July 2016, leaving space for the possibility that their increased circulation was linked to the introduction of the ‘Mascali clone’. However, a couple of GII.P2/GII.2 strains (including PA36/16) genetically close to the ‘Mascali clone’ had been already detected in January 2016 in Palermo and GII.P2/GII.2 NoV had sporadically been observed also in 2011 and 2014. Moreover, the circulation of GII.P2/GII.2 NoVs has been increasing worldwide over the last period [37]. Therefore, the unusually high circulation observed in Palermo might just reflect the global epidemiology of GII.P2/GII.2 strains. Noteworthy, the older GII.P2/GII.2 NoV strains detected in Palermo were closely related in the ORF2 (98.9% nt id) to the GII.P16/GII.2 Nashville strain (KY865307) which is supposed to have generated the recombinant GII.P16/GII.4 clone recently circulating epidemically worldwide [37, 38] and were more distantly associated (95.3% nt id) to the GII.2 capsid sequence (AB662868) of another putative GII.P16/GII.4 ancestor from Japan [39]. The increasing frequency of detection of waterborne NoV outbreaks calls for the introduction of routine monitoring for viruses of drinking waters and for the definition of affordable and reliable indicators for the presence of NoV in drinking waters. The use of ultrafiltration systems as a concentration method in order to recover viruses from environmental waters combined with RT-PCR to detect viral genomes has been demonstrated to grant a recovery rate of >50% of multiple viruses [40] and is at present one of the most sensitive approaches for the detection of NoVs, although the discrimination between viable and nonviable viruses remains a problem. The burden of NoV outbreaks, either food associated or waterborne, is still poorly defined due to the lack of specific diagnostic tools in many clinical laboratories. NoV epidemics should always be suspected when the main symptom is vomiting and the duration of disease is short (<48 h). Waterborne transmission is suggested by the rapid spread of disease and the high attack rate. Care should be taken to perform sensitive diagnostic tests on a sufficient number of stool samples in order to identify NoV as the causative agent of enteritis outbreaks, possibly preferring genome screening over antigen detection. Only the systematic detection of NoV outbreaks can guide in the choice of the safety procedures that could be most appropriate in containing such microbiological hazard.

What was done to fix the problem?

1561

Waterborne Norovirus outbreak at a seaside resort likely originating from municipal water distribution system failure

Abstract In May 2016 a Norovirus (NoV) gastroenteritis outbreak involved a high school class visiting a seaside resort near Taormina (Mascali, Sicily). Twenty-four students and a teacher were affected and 17 of them showed symptoms on the second day of the journey, while the others got ill within the following 2 days. Symptoms included vomiting, diarrhoea and fever, and 12 students required hospitalisation. Stool samples tested positive for NoV genome by Real-Time polymerase chain reaction assay in all 25 symptomatic subjects. The GII.P2/GII.2 NoV genotype was linked to the outbreak by ORF1/ORF2 sequence analysis. The epidemiological features of the outbreak were consistent with food/waterborne followed by person-to-person and/ or vomit transmission. Food consumed at a shared lunch on the first day of the trip was associated to illness and drinking un-bottled tap water was also considered as a risk factor. The analysis of water samples revealed the presence of bacterial indicators of faecal contamination in the water used in the resort as well as in other areas of the municipal water network, linking the NoV gastroenteritis outbreak to tap water pollution from sewage leakage. From a single water sample, an amplicon whose sequence corresponded to the capsid genotype recovered from patients could be obtained. Introduction Noroviruses (NoVs) are recognised as a leading cause of sporadic cases and outbreaks of acute gastroenteritis (AGE) across all age groups [1, 2]. They are transmitted primarily through the faecal-oral route either by direct person–person spread and with consumption of contaminated food or water [1, 3, 4]. NoVs are non-enveloped RNA viruses belonging to the Caliciviridae family and are classified into seven distinct genogroups (GI-GVII) on the basis of the major capsid protein VP1 sequence. GI, GII and GIV NoVs infect humans and they have been divided into more than 30 genotypes based on the sequences of polymerase (ORF1) and capsid proteins (ORF2) [5]. NoV GII.4 genotype has been associated with the vast majority of NoV related cases of AGE worldwide although many genotypes, belonging to both GI and GII genogroups are involved in foodborne and waterborne outbreaks [6, 7]. NoVs are highly contagious owing to their high stability in the environment and their ability to infect at low doses. In fact, just 10–100 virions of NoV may be sufficient to cause AGE [8, 9]. NoVs are considered the major cause of water-related disease in semi-closed institutions, including restaurants, schools, hospitals, nursing homes and cruise ships [10–12]. To confirm NoV as a cause of outbreaks, the presence of four epidemiological features of disease, vomiting in more than half of affected persons, a mean incubation period of 24/48 h, a mean duration of illness of 12–60 h and the absence of bacterial pathogens in stool culture, should be evaluated [13]. Waterborne NoV outbreaks can be caused by contaminated drinking water despite correct chlorine concentration and in the absence of coliform bacteria [10, 14–17]. The aim of this study was to investigate an outbreak of NoV gastroenteritis associated with drinking water involving a high school class travelling from Palermo (West Sicily) to Acireale (East Sicily) on an educational tour in May 2016. Stool samples were collected soon after the onset of AGE symptoms to ascertain the aetiology of illness. The source of infection was investigated collecting epidemiological data through an online questionnaire and, in order to explore a possible transmission route, laboratory tests on water from the municipal network were performed. Material and methods Epidemiological investigation From 20 to 22 May 2016 an AGE outbreak occurred in Mascali, a small town near Taormina, involving a school class of 28 students, 13 girls and 15 boys aged between 14 and 18 years and a 54-year-old female teacher, during a trip to participate in a theatre festival in Acireale. A total of 25/29 persons were affected by nausea, vomiting, abdominal cramps, diarrhoea and fever while they were staying at a seaside resort in Mascali, on the way to Acireale. To obtain data on symptoms and the correlated risk factors all participants filled an online questionnaire investigating personal data (age and sex), travel and accommodation details and food consumption over 48 h before symptoms presentation, time of onset and duration of symptoms, therapy and/or hospitalisation required and history of contact with people affected by diarrhoea and/or vomiting (Table 1). The severity of illness was analysed by the Vesikari severity scale allowing to distinguish non-severe (score <11) and severe (score ⩾11) AGE (Table 1). Vesikari score was calculated from maximum number of episodes/day and duration of both diarrhoea and vomiting (score 0–3 for each of the 4 parameters), temperature (score 0–3), level of dehydration (score 0–3) and treatment required (score 0–2) [18]. A case-control study was conducted to establish if food consumed before the onset of symptoms was associated with the disease. A case was defined as a student attending the tour and reporting diarrhoea and/or vomiting within 36 h from the first shared meal during the tour, while students who did not present any of the previous symptoms or those showing the first symptoms with a delay of ⩾48 h (secondary cases) were used as controls. Variables were studied using odds ratio (OR) and its 95% CI and statistical associations were determined using the χ2 test, with the level of statistical significance set at 0.05 (Table 2). NoV detection in stool samples A stool sample was obtained from each of the 29 travellers, soon after (3–5 days) the onset of the outbreak. No stool samples could be obtained from the food-handlers working at the resort. All samples were tested for NoV at the Enteric Viruses Laboratory of the University of Palermo, member of the Italian Study Group for Enteric Viruses (ISGEV), by a Real-Time polymerase chain reaction (RT-PCR) assay that allows to differentiate GI and GII NoVs [19]. Total RNA was extracted from stool samples using QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) and cDNA was generated by MMLV reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were considered positive for NoVs when cycle threshold (ct) of RT-PCR was ⩽30 and lowpositive when 30<ct⩽35. For all NoV-positive samples two genomic fragments were amplified by conventional RT-PCR using Superscript III One-Step RT-PCR system with Platinum Taq DNA polymerase (Invitrogen) and primers JV12a/JV13b, targeting 326 bps of the RdRp region A [20] and either COG1F/ GISKR or COG2F/GIISKR, amplifying a 315 bps 5′portion of GI or GII VP1 protein [19, 21]. Environmental investigation From May through August 2016, a total of 32 tap water samples were collected from the Mascali municipal network. In particular, seven samples were collected at the resort where the outbreak occurred and from its water entry (R), two in the city centre of Mascali (M), 15 in the village of Fondachello (F), on the seaside, three in Carrabba village (C), far from the coast, and two and three, respectively, from the water well ‘Carlino’ (W) and the source ‘Bufardo and Torregrossa’ (S) which feed the municipal system and provide drinking water to the Mascali area (Fig. 1, Table 3). One litre water samples were sent by the local health authorities to the SIALAB laboratory in Avola, Siracusa and 100 ml aliquots were screened for coliforms, E. coli and enterococci as faecal indicators, by standard bacteriological methods according to UNI EN ISO 9308-1:2014. Unfortunately, although the suspect of a water source for the viral outbreak was quickly communicated, only in August and November 2016, following the implementation of control measures, local authorities provided water samplings of appropriate volume to perform viral tests (10 l). Five water samples of 10 l each were collected from different sampling sites in F and R areas for viral screening (Table 3) and were concentrated according to the Italian National Institute of Health (Istituto Superiore di Sanità, ISS) protocols [22] through an ultra-filtration system (Sartoflow® Slice 200 Benchtop Crossflow System, Sartorius AG, Goettingen, Germany), using appropriate membranes (SG Hydrosart 10 kDa) pretreated with 300 ml of 3% Beef Extract (BE) at pH 7, by recirculating it for 10 min in the system, at a pressure not higher than 1.30 bar. Elution and recovery of viruses attached to the membranes were obtained by recirculating 200 ml of 3% BE at pH 9.5 for 3 h until a 10–12 ml volume of eluate was obtained. The pH of the eluate was then checked with litmus paper and if necessary brought to the value of 7 with HCl. Concentrated water samples were simultaneously tested for NoV in three laboratories: Enteric Viruses Laboratory, University of Palermo; Istituto Zooprofilattico Sperimentale della Sicilia (IZS), Palermo; and Istituto Superiore di Sanità (ISS), Rome. Total RNA was extracted from 500 and 560 µl of a concentrated water sample using BioMèrieux NucliSENS® Magnetic Extraction System (BioMèrieux) and QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) respectively, according to manufacturer’s instructions. Bonded RNA was eluted by 100 µl of Elution Buffer. Amplifications were performed at least in duplicate on whole and 1:10 diluted extracts. A Real-Time RT-PCR assay according to ISO/TS 15216- 2:2013, two in-house Real-Time RT-PCR protocols and the One-Step RT-PCR methods already used for amplifying ORF1 and ORF2 fragments in NoV-positive faecal samples were performed [19–21, 23]. Nucleotide sequences analysis and phylogenetic analysis PCR amplicons were purified by commercial columns (CleanSweep PCR Purification, Thermo Fisher Scientific, Vilnius, Lithuania) and sequenced by Sanger method and all nucleotide sequences were aligned using ClustalW. To identify NoV genotypes, RdRp and VP1 sequences were submitted to the RIVM NoV genotyping tool (http://www.rivm.nl/mpf/norovirus/typingtool). Phylogenetic trees were constructed using MEGA 7 and Neighbour-Joining method (bootstrap 1000 replicates). Sequences from stool and water samples are available at NCBI with the following accession numbers: MG517455 for the major capsid protein obtained from a water sample; MG517456 and MG517461, respectively, for the major capsid protein and RNA-dependent-RNA-polymerase, as representatives of the sequences obtained from the stool samples of those involved in the outbreak. Control measures After outbreak notification, on 21 May, the resort was closed for 2 days. On 25 May, following the first results of the bacteriological analyses on municipal water samples, the use of municipal water for human consumption was temporarily restricted (use of water was permitted only for cleaning purposes) in the Carrabba and Fondachello villages and in the area surrounding the resort. From 25 May the reservoir collecting the water from the Carlino well and its chlorination device were submitted to maintenance work and a new chlorination device was installed on the conduit bringing the water from the Bufardo source. Actually, the Bufardo source had been connected to the municipal network in 2003 but unfortunately a dedicated chlorination station had not been planned at that time. Therefore, unchlorinated water entered the system near Fondachello village on the seaside (Fig. 1), possibly diluting chlorine concentrations in water treated at the Carlino reservoir. Moreover, the exact path of the branch of the water network situated downstream the Bufardo source water connection was not marked in the maps of the municipality (Fig. 1, red line), indicating that this section had never been submitted to maintenance. The area surrounding the unmaintained section of the water pipeline was scattered with unauthorised cottages that were not connected to the sewerage systems. Likely, untreated waste from cesspools was unloaded directly into the soil, generating a risk of contamination of the municipal water system through leakage. No work had been made on the water pipelines before the outbreak but sections of the network had been temporarily closed during the winter season on the seaside (Fondachello). Local health authorities advised the connection of all households in the area to the sewerage system and the replacement of the water system section without maintenance. Health authorities also raised attention on AGE cases notification at all local Emergency Departments and Family Practitioners but there was no further evidence of AGE outbreaks in the months following the intervention. Results Data collected through the online questionnaire compiled 7–10 days after the outbreak revealed that none of the 28 students and the teacher taking part in the school trip showed gastroenteritis symptoms at the beginning of the journey and they had no contact with people who suffered from diarrhoea and/or vomiting in the previous days. Travellers had their first shared meal on 19 May (Day 1), 2016, at lunch on their arrival to the resort in Mascali. The list of foods included in the menu is shown in Table 2. Conversely, the dinner on Day 1 was not shared and the students went to dine in three different fast food restaurants in Acireale. On 20 May (Day 2), 17 students (58.6% of the travellers) presented with vomit and/or diarrhoea, falling ill 26–36 h after lunch on Day 1, strongly suggesting that exposure occurred on the first shared meal. Some students also vomited in the presence of the other travellers on the bus bringing them to Acireale. Five more students were symptomatic in the early hours of the following day (21 May, Day 3, from 0:30 am to 8:00 am), as a continuation of the wave of illnesses of the previous day. The teacher and one more student fell ill at 8:00 pm and 11:00 pm of Day 3, respectively, for a second wave of illnesses occurring >48 h from the first shared meal and <36 h from the first cases, with the last student falling sick on 22 May (Day 4) at 3:00 am (Fig. 2, Table 1). On the whole, the mean incubation period was 33.2 h. Based on the online survey, the severity of illness was evaluated by the Vesikari severity scale, grouping symptomatic travellers into severe (11/25 = 44%) and non-severe (14/25 = 56%) cases (Table 1). Vomiting and fever were the predominant symptoms, affecting 22 out of 25 (88%) symptomatic subjects, while diarrhoea was present in 14 out of 25 (56%). Hospitalisation for rehydration therapy was necessary for 13 students (91% [10/11] of severe and 21% [3/14] of non-severe cases). Emergency room (ER) or family practitioner visits were necessary for 21 students and 13 (62%) of them initiated an antibiotic therapy. Two students indicated that other family members had gastroenteric symptoms more than 24 h after their return home, on Day 3, 4 and 5. All those involved in the outbreak recovered within 4 days, with an average duration of illness of 38.4 h. A case-control study of the risk associated with eating foods at lunch on Day 1 or with attending different restaurants at dinner on Day 1 was attempted. A total of 19 subjects filled the section of the online survey concerning food consumption. Fourteen of the responders were among the 22 cases falling ill within 36 h from the first shared meal (first wave), while the remaining five responders were three travellers with supposed secondary infections (showing symptoms with a delay ⩾48 h; second wave) and two asymptomatic students that were used as controls. No statistically significant associations with any food items served at lunch on Day 1 was identified (P < 0.05; Table 2). However, all food consumed on that occasion was to some extent associated with the risk of disease, the highest OR being obtained with chicken (OR: 20.7143, P: 0.0678; Table 2). Actually, most of the cases ate all foods included in the menu, while one of the two asymptomatic controls did not participate in the meal and the other had a pasta dish only. No statistically significant association was found for attending a particular fast food restaurant at the dinner on Day 1 (OR: 2.5, P: 0.3635). The Real-Time RT-PCR assay allowed to detect GII genogroup NoV genome in all the 25 samples from symptomatic subjects, although two of them showed high Ct values, indicating lower shedding of viral genomes (Table 1). The samples from four asymptomatic students were negative for both NoV genogroups. By sequence analysis, a NoV genotype could be defined for 20/ 25 NoV-positive stool samples. All NoV strains from travellers involved in the outbreaks belonged to the same genotype, GII.P2/GII.2, showing >99% nucleotide identities (nt id) in both fragments (ORF1 and ORF2) analysed. Thirty-two water samples were collected and analysed for bacterial contamination (Table 3). Seven samples were taken from the resort (site R) and the remaining 25 from five other sites (C, F, M, S, W), representative of the entire municipal water network. The majority (10/12) of the water samples collected immediately after the onset of the outbreak, between 21 May and 16 June 2016, at the resort (site R) and at sites F and C, located close to the resort, on the seaside and in the interior, respectively, were positive for coliform bacteria. More than 1000 colony-forming units (CFU)/ 100 ml (up to 2700 CFU/100 ml) were detected in four samples, while their absence is required in 100 ml by the Standard of Water Quality in Italy and seven of 11 samples tested were also Escherichia coli-positive (up to 900 CFU/100 ml). Enterococci were searched in 11 samples and three of them were found positive for this faecal contamination indicator. From 7 July to 16 August 2016, after control measures on the water network were undertaken, 13 water samples were collected for follow-up laboratory examination of the water system at F, R and C sites. All of them were negative for bacterial indicators of faecal contamination. No faecal bacteria were identified in the seven samples collected from M, W and S sites, including the well and the source that fed the water network, at any sampling time, before and after implementing the control measures. The research of NoV genome was conducted on concentrated water samples collected on 1 August from R and F sites (Table 3). NoV genome was detected in a single sample collected from the water entry of the resort involved in the outbreak (site R). Actually, only two out of 26 One-Step RT-PCR repetitions were positive for NoV viral capsid gene. The two ORF2-positive results were obtained at the University of Palermo Laboratory using both 1:10 diluted and undiluted extracted RNA. The sequence analysis of the partial ORF2 portion (289 bp) indicated the presence of the same GII.2 capsid genotype detected in the stool samples of the students involved in the outbreak, showing 98.9% nt id. No ORF1 NoV sequence could be obtained from the NoV-positive water sample. Follow-up analyses for NoV genome, conducted on water samples collected on 20 November both at R and F sites found no positive samples. By sequence alignment, the NoV sequences obtained from Mascali outbreak samples were phylogenetically closely related in both ORF1 and ORF2 (99 and 98.9% of nt id, respectively) to a GII.P2/GII.2 strain responsible for a sporadic paediatric gastroenteritis case occurred in Palermo in January 2016 (PA 36/16, accession numbers MG517464 and MG517460, for ORF1 and ORF2, respectively) and previously detected at the Enteric Viruses Laboratory, University of Palermo. Whereas, three GII.P2/GII.2 Italian strains (PA 280/11, PA 366/11 and PA 3/ 14, accession numbers MG517462, MG517465 and MG517463, for ORF1 and MG517457 to −59, for ORF2, respectively), detected in Palermo in 2011 and 2014, were more distantly associated with the outbreak strain both in RdRp and VP1 (90% to 97.4% nt id and 96.1% to 98.1% nt id, respectively, in ORF1 and ORF2). Discussion Epidemiological features, such as the presence of vomiting in more than half (88%) of symptomatic subjects, the 33.2 h average incubation period and the 38.4 h average duration of illness, suggested a NoV aetiology for the investigated outbreak [13]. The detection of NoV genome in all 25 faecal samples from symptomatic subjects fully complies with the requirement of detection of at least one positive sample out of four tested that is considered to be sufficient to ascertain the aetiology of a NoV outbreak [24]. Moreover, molecular typing indicated that a single viral strain, belonging to genotype GII.P2/GII.2, infected all symptomatic subjects involved in the outbreak. NoVs are highly infectious and can easily spread from person to person but also through contaminated food or water. Therefore, settings such as health care institutions, schools and touristic resorts are mainly affected, since close contact and shared food services facilitate transmission. In Mascali, two waves of transmission were observed over three days, being consistent with food/waterborne followed by person-to-person and/or airborne (vomit) transmission (Fig. 2). The first wave of cases presented on 20 May 2016 after a single shared meal, while the last case was detected on 22 May. Exposure to vomit may have been responsible for some of the secondary cases occurred after the first wave of illness. Vomiting produces aerosols that can be inhaled and swallowed. The extremely low infectious dose of NoV makes the inhalation of vomit aerosols an effective transmission route in settings where there is close contact, like on a bus trip [25]. All food consumed in the only shared lunch on 19 May at the resort was associated with the risk of disease (Table 2), but the association was inconclusive since all cases shared the same dishes included in the lunch menu on Day 1. Although many foods were potentially involved, our attention focused on the tap water. In fact, trivial information from local health authorities reported that local population avoided tap water consumption and relied on bottled water for drinking and cooking. Many small outbreaks of gastroenteritis involving restaurants and family groups had been reported in the same area, especially at the beginning of the summer season, when vacation houses and recreational activities were re-opened after a period of low water demand. Illness caused by NoV is common and if the level of water contamination is low, the number of cases remains limited and water can persist as unrecognised source of infection. If polluted water was served for drinking and used for cooking and washing dishes in the Mascali resort at Day 1 a generalised contamination of food and surfaces could have occurred. The potential role of the foodhandlers as a source of contamination could not be analysed in this study but should also be taken into account. If water contamination has already been present in the area for a long time, infections may have occurred among the local food-handlers. The analysis of water samples taken immediately after the outbreak revealed the presence of bacterial indicators of faecal contamination in the water used in the resort, with the only exception of a sample of filter purified water used both for cooking and as drinking water. When extending the sampling to other areas of the municipal water network, many other sites, mostly on the seaside, showed faecal pollution, which demonstrates a poor protection of the water network from faecal leakage (Fig. 1). NoV contamination of waters generally originates from sewage since these enteric pathogens, as other enteric viruses affecting humans, are mostly species-specific and are shed at high concentrations in the stools of patients. When the analyses on the stools of the students made it clear that the outbreak had been caused by NoV, the viral genome was also searched in appropriate volumes of water samples, but this happened when correction measures on the water network had already been taken. Nevertheless, a single water sample, obtained 2 months after the outbreak (1 August 2016), still contained genome fragments that could be correlated to the same NoV genotype causing the outbreak. Although the finding of NoV genome does not demonstrate that entire infecting viral particles were present in the water, it could indicate that a NoV belonging to the same genotype involved in the outbreak was still circulating in the local population several weeks after the outbreak and it was someway able to penetrate the water network, probably indicating continuing faecal contamination through leakage from sewage systems. NoV is particularly resistant in the environment and also to chlorine, therefore transmission through chlorinated water is not exceptional [9, 12, 26]. Drinking water from municipal supplies fed by lake or groundwater has already been identified in the past as the source of waterborne NoV outbreaks in Italy [27–29] and NoV is now considered one of the pathogens causing the largest number of affected consumers in drinking waterborne outbreaks [12]. NoV can persist in water longer than 15 days [30] and it is resistant to chlorine disinfection when free chlorine levels or pre-treatment processes are inadequate [31, 32]. Moreover, microfiltration filters labelled as certified by NSF Standards 53 or 58 (with a pore size ranging from 0.05 to 5 µm) are not effective in removing viruses and can even decrease chlorine concentration in filtered water [33]. Actually, the owner of the resort involved in our outbreak was encouraged to serve tap water as drinking water at the restaurant relying on the effectiveness of such a filter purifier. Unfortunately, we did not have a tap water sample from the restaurant of the resort taken before correction measures and whose volume was appropriate for virus detection. Therefore, we could not demonstrate the failure to eliminate viral pollution of the microfiltration system. In Mascali, correction measures on the water network consisted in the maintenance of the old chlorination device and in the addition of a new one at a key point located upstream and close to suspected illegal sewage dumping (Fig. 1), but as poor protection of the pipeline from sewage leakage risked to reproduce faecal pollution episodes, renovation of the municipal water system was also advised, as well as systematic control of illegal sewage dumping. The additional chlorination device was apparently the most effective intervention, since right after its activation faecal bacterial pollution indicators disappeared. However, NoV genome was detected in a water sample taken 6 weeks after the last detection of bacterial indicators. Unfortunately, detection of faecal contamination of drinking water is only based on the use of bacterial indicators despite no correlation between levels of enteric bacteria and enteric viruses has been found [34–36]. According to ISGEV enteric virus circulation surveillance data, in 2016 the circulation of GII.P2/GII.2 NoV strains in Palermo, home town of the travellers involved in the outbreak, was the highest ever observed from the starting of local NoV surveillance in 2004. GII.P2/GII.2 was the fourth most common genotype in 2016 (12.2% prevalence), following GII.P4/GII.4 Sydney 2012 (28.4%), GII.P16/GII.2 (18.9%) and GII.Pe/GII.4 Sydney 2012 (17.6%). GII.P2/GII.2 strains were mostly observed starting from July 2016, leaving space for the possibility that their increased circulation was linked to the introduction of the ‘Mascali clone’. However, a couple of GII.P2/GII.2 strains (including PA36/16) genetically close to the ‘Mascali clone’ had been already detected in January 2016 in Palermo and GII.P2/GII.2 NoV had sporadically been observed also in 2011 and 2014. Moreover, the circulation of GII.P2/GII.2 NoVs has been increasing worldwide over the last period [37]. Therefore, the unusually high circulation observed in Palermo might just reflect the global epidemiology of GII.P2/GII.2 strains. Noteworthy, the older GII.P2/GII.2 NoV strains detected in Palermo were closely related in the ORF2 (98.9% nt id) to the GII.P16/GII.2 Nashville strain (KY865307) which is supposed to have generated the recombinant GII.P16/GII.4 clone recently circulating epidemically worldwide [37, 38] and were more distantly associated (95.3% nt id) to the GII.2 capsid sequence (AB662868) of another putative GII.P16/GII.4 ancestor from Japan [39]. The increasing frequency of detection of waterborne NoV outbreaks calls for the introduction of routine monitoring for viruses of drinking waters and for the definition of affordable and reliable indicators for the presence of NoV in drinking waters. The use of ultrafiltration systems as a concentration method in order to recover viruses from environmental waters combined with RT-PCR to detect viral genomes has been demonstrated to grant a recovery rate of >50% of multiple viruses [40] and is at present one of the most sensitive approaches for the detection of NoVs, although the discrimination between viable and nonviable viruses remains a problem. The burden of NoV outbreaks, either food associated or waterborne, is still poorly defined due to the lack of specific diagnostic tools in many clinical laboratories. NoV epidemics should always be suspected when the main symptom is vomiting and the duration of disease is short (<48 h). Waterborne transmission is suggested by the rapid spread of disease and the high attack rate. Care should be taken to perform sensitive diagnostic tests on a sufficient number of stool samples in order to identify NoV as the causative agent of enteritis outbreaks, possibly preferring genome screening over antigen detection. Only the systematic detection of NoV outbreaks can guide in the choice of the safety procedures that could be most appropriate in containing such microbiological hazard.

What could have been done to prevent the event?

1562

Waterborne Norovirus outbreak at a seaside resort likely originating from municipal water distribution system failure

Abstract In May 2016 a Norovirus (NoV) gastroenteritis outbreak involved a high school class visiting a seaside resort near Taormina (Mascali, Sicily). Twenty-four students and a teacher were affected and 17 of them showed symptoms on the second day of the journey, while the others got ill within the following 2 days. Symptoms included vomiting, diarrhoea and fever, and 12 students required hospitalisation. Stool samples tested positive for NoV genome by Real-Time polymerase chain reaction assay in all 25 symptomatic subjects. The GII.P2/GII.2 NoV genotype was linked to the outbreak by ORF1/ORF2 sequence analysis. The epidemiological features of the outbreak were consistent with food/waterborne followed by person-to-person and/ or vomit transmission. Food consumed at a shared lunch on the first day of the trip was associated to illness and drinking un-bottled tap water was also considered as a risk factor. The analysis of water samples revealed the presence of bacterial indicators of faecal contamination in the water used in the resort as well as in other areas of the municipal water network, linking the NoV gastroenteritis outbreak to tap water pollution from sewage leakage. From a single water sample, an amplicon whose sequence corresponded to the capsid genotype recovered from patients could be obtained. Introduction Noroviruses (NoVs) are recognised as a leading cause of sporadic cases and outbreaks of acute gastroenteritis (AGE) across all age groups [1, 2]. They are transmitted primarily through the faecal-oral route either by direct person–person spread and with consumption of contaminated food or water [1, 3, 4]. NoVs are non-enveloped RNA viruses belonging to the Caliciviridae family and are classified into seven distinct genogroups (GI-GVII) on the basis of the major capsid protein VP1 sequence. GI, GII and GIV NoVs infect humans and they have been divided into more than 30 genotypes based on the sequences of polymerase (ORF1) and capsid proteins (ORF2) [5]. NoV GII.4 genotype has been associated with the vast majority of NoV related cases of AGE worldwide although many genotypes, belonging to both GI and GII genogroups are involved in foodborne and waterborne outbreaks [6, 7]. NoVs are highly contagious owing to their high stability in the environment and their ability to infect at low doses. In fact, just 10–100 virions of NoV may be sufficient to cause AGE [8, 9]. NoVs are considered the major cause of water-related disease in semi-closed institutions, including restaurants, schools, hospitals, nursing homes and cruise ships [10–12]. To confirm NoV as a cause of outbreaks, the presence of four epidemiological features of disease, vomiting in more than half of affected persons, a mean incubation period of 24/48 h, a mean duration of illness of 12–60 h and the absence of bacterial pathogens in stool culture, should be evaluated [13]. Waterborne NoV outbreaks can be caused by contaminated drinking water despite correct chlorine concentration and in the absence of coliform bacteria [10, 14–17]. The aim of this study was to investigate an outbreak of NoV gastroenteritis associated with drinking water involving a high school class travelling from Palermo (West Sicily) to Acireale (East Sicily) on an educational tour in May 2016. Stool samples were collected soon after the onset of AGE symptoms to ascertain the aetiology of illness. The source of infection was investigated collecting epidemiological data through an online questionnaire and, in order to explore a possible transmission route, laboratory tests on water from the municipal network were performed. Material and methods Epidemiological investigation From 20 to 22 May 2016 an AGE outbreak occurred in Mascali, a small town near Taormina, involving a school class of 28 students, 13 girls and 15 boys aged between 14 and 18 years and a 54-year-old female teacher, during a trip to participate in a theatre festival in Acireale. A total of 25/29 persons were affected by nausea, vomiting, abdominal cramps, diarrhoea and fever while they were staying at a seaside resort in Mascali, on the way to Acireale. To obtain data on symptoms and the correlated risk factors all participants filled an online questionnaire investigating personal data (age and sex), travel and accommodation details and food consumption over 48 h before symptoms presentation, time of onset and duration of symptoms, therapy and/or hospitalisation required and history of contact with people affected by diarrhoea and/or vomiting (Table 1). The severity of illness was analysed by the Vesikari severity scale allowing to distinguish non-severe (score <11) and severe (score ⩾11) AGE (Table 1). Vesikari score was calculated from maximum number of episodes/day and duration of both diarrhoea and vomiting (score 0–3 for each of the 4 parameters), temperature (score 0–3), level of dehydration (score 0–3) and treatment required (score 0–2) [18]. A case-control study was conducted to establish if food consumed before the onset of symptoms was associated with the disease. A case was defined as a student attending the tour and reporting diarrhoea and/or vomiting within 36 h from the first shared meal during the tour, while students who did not present any of the previous symptoms or those showing the first symptoms with a delay of ⩾48 h (secondary cases) were used as controls. Variables were studied using odds ratio (OR) and its 95% CI and statistical associations were determined using the χ2 test, with the level of statistical significance set at 0.05 (Table 2). NoV detection in stool samples A stool sample was obtained from each of the 29 travellers, soon after (3–5 days) the onset of the outbreak. No stool samples could be obtained from the food-handlers working at the resort. All samples were tested for NoV at the Enteric Viruses Laboratory of the University of Palermo, member of the Italian Study Group for Enteric Viruses (ISGEV), by a Real-Time polymerase chain reaction (RT-PCR) assay that allows to differentiate GI and GII NoVs [19]. Total RNA was extracted from stool samples using QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) and cDNA was generated by MMLV reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were considered positive for NoVs when cycle threshold (ct) of RT-PCR was ⩽30 and lowpositive when 30<ct⩽35. For all NoV-positive samples two genomic fragments were amplified by conventional RT-PCR using Superscript III One-Step RT-PCR system with Platinum Taq DNA polymerase (Invitrogen) and primers JV12a/JV13b, targeting 326 bps of the RdRp region A [20] and either COG1F/ GISKR or COG2F/GIISKR, amplifying a 315 bps 5′portion of GI or GII VP1 protein [19, 21]. Environmental investigation From May through August 2016, a total of 32 tap water samples were collected from the Mascali municipal network. In particular, seven samples were collected at the resort where the outbreak occurred and from its water entry (R), two in the city centre of Mascali (M), 15 in the village of Fondachello (F), on the seaside, three in Carrabba village (C), far from the coast, and two and three, respectively, from the water well ‘Carlino’ (W) and the source ‘Bufardo and Torregrossa’ (S) which feed the municipal system and provide drinking water to the Mascali area (Fig. 1, Table 3). One litre water samples were sent by the local health authorities to the SIALAB laboratory in Avola, Siracusa and 100 ml aliquots were screened for coliforms, E. coli and enterococci as faecal indicators, by standard bacteriological methods according to UNI EN ISO 9308-1:2014. Unfortunately, although the suspect of a water source for the viral outbreak was quickly communicated, only in August and November 2016, following the implementation of control measures, local authorities provided water samplings of appropriate volume to perform viral tests (10 l). Five water samples of 10 l each were collected from different sampling sites in F and R areas for viral screening (Table 3) and were concentrated according to the Italian National Institute of Health (Istituto Superiore di Sanità, ISS) protocols [22] through an ultra-filtration system (Sartoflow® Slice 200 Benchtop Crossflow System, Sartorius AG, Goettingen, Germany), using appropriate membranes (SG Hydrosart 10 kDa) pretreated with 300 ml of 3% Beef Extract (BE) at pH 7, by recirculating it for 10 min in the system, at a pressure not higher than 1.30 bar. Elution and recovery of viruses attached to the membranes were obtained by recirculating 200 ml of 3% BE at pH 9.5 for 3 h until a 10–12 ml volume of eluate was obtained. The pH of the eluate was then checked with litmus paper and if necessary brought to the value of 7 with HCl. Concentrated water samples were simultaneously tested for NoV in three laboratories: Enteric Viruses Laboratory, University of Palermo; Istituto Zooprofilattico Sperimentale della Sicilia (IZS), Palermo; and Istituto Superiore di Sanità (ISS), Rome. Total RNA was extracted from 500 and 560 µl of a concentrated water sample using BioMèrieux NucliSENS® Magnetic Extraction System (BioMèrieux) and QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) respectively, according to manufacturer’s instructions. Bonded RNA was eluted by 100 µl of Elution Buffer. Amplifications were performed at least in duplicate on whole and 1:10 diluted extracts. A Real-Time RT-PCR assay according to ISO/TS 15216- 2:2013, two in-house Real-Time RT-PCR protocols and the One-Step RT-PCR methods already used for amplifying ORF1 and ORF2 fragments in NoV-positive faecal samples were performed [19–21, 23]. Nucleotide sequences analysis and phylogenetic analysis PCR amplicons were purified by commercial columns (CleanSweep PCR Purification, Thermo Fisher Scientific, Vilnius, Lithuania) and sequenced by Sanger method and all nucleotide sequences were aligned using ClustalW. To identify NoV genotypes, RdRp and VP1 sequences were submitted to the RIVM NoV genotyping tool (http://www.rivm.nl/mpf/norovirus/typingtool). Phylogenetic trees were constructed using MEGA 7 and Neighbour-Joining method (bootstrap 1000 replicates). Sequences from stool and water samples are available at NCBI with the following accession numbers: MG517455 for the major capsid protein obtained from a water sample; MG517456 and MG517461, respectively, for the major capsid protein and RNA-dependent-RNA-polymerase, as representatives of the sequences obtained from the stool samples of those involved in the outbreak. Control measures After outbreak notification, on 21 May, the resort was closed for 2 days. On 25 May, following the first results of the bacteriological analyses on municipal water samples, the use of municipal water for human consumption was temporarily restricted (use of water was permitted only for cleaning purposes) in the Carrabba and Fondachello villages and in the area surrounding the resort. From 25 May the reservoir collecting the water from the Carlino well and its chlorination device were submitted to maintenance work and a new chlorination device was installed on the conduit bringing the water from the Bufardo source. Actually, the Bufardo source had been connected to the municipal network in 2003 but unfortunately a dedicated chlorination station had not been planned at that time. Therefore, unchlorinated water entered the system near Fondachello village on the seaside (Fig. 1), possibly diluting chlorine concentrations in water treated at the Carlino reservoir. Moreover, the exact path of the branch of the water network situated downstream the Bufardo source water connection was not marked in the maps of the municipality (Fig. 1, red line), indicating that this section had never been submitted to maintenance. The area surrounding the unmaintained section of the water pipeline was scattered with unauthorised cottages that were not connected to the sewerage systems. Likely, untreated waste from cesspools was unloaded directly into the soil, generating a risk of contamination of the municipal water system through leakage. No work had been made on the water pipelines before the outbreak but sections of the network had been temporarily closed during the winter season on the seaside (Fondachello). Local health authorities advised the connection of all households in the area to the sewerage system and the replacement of the water system section without maintenance. Health authorities also raised attention on AGE cases notification at all local Emergency Departments and Family Practitioners but there was no further evidence of AGE outbreaks in the months following the intervention. Results Data collected through the online questionnaire compiled 7–10 days after the outbreak revealed that none of the 28 students and the teacher taking part in the school trip showed gastroenteritis symptoms at the beginning of the journey and they had no contact with people who suffered from diarrhoea and/or vomiting in the previous days. Travellers had their first shared meal on 19 May (Day 1), 2016, at lunch on their arrival to the resort in Mascali. The list of foods included in the menu is shown in Table 2. Conversely, the dinner on Day 1 was not shared and the students went to dine in three different fast food restaurants in Acireale. On 20 May (Day 2), 17 students (58.6% of the travellers) presented with vomit and/or diarrhoea, falling ill 26–36 h after lunch on Day 1, strongly suggesting that exposure occurred on the first shared meal. Some students also vomited in the presence of the other travellers on the bus bringing them to Acireale. Five more students were symptomatic in the early hours of the following day (21 May, Day 3, from 0:30 am to 8:00 am), as a continuation of the wave of illnesses of the previous day. The teacher and one more student fell ill at 8:00 pm and 11:00 pm of Day 3, respectively, for a second wave of illnesses occurring >48 h from the first shared meal and <36 h from the first cases, with the last student falling sick on 22 May (Day 4) at 3:00 am (Fig. 2, Table 1). On the whole, the mean incubation period was 33.2 h. Based on the online survey, the severity of illness was evaluated by the Vesikari severity scale, grouping symptomatic travellers into severe (11/25 = 44%) and non-severe (14/25 = 56%) cases (Table 1). Vomiting and fever were the predominant symptoms, affecting 22 out of 25 (88%) symptomatic subjects, while diarrhoea was present in 14 out of 25 (56%). Hospitalisation for rehydration therapy was necessary for 13 students (91% [10/11] of severe and 21% [3/14] of non-severe cases). Emergency room (ER) or family practitioner visits were necessary for 21 students and 13 (62%) of them initiated an antibiotic therapy. Two students indicated that other family members had gastroenteric symptoms more than 24 h after their return home, on Day 3, 4 and 5. All those involved in the outbreak recovered within 4 days, with an average duration of illness of 38.4 h. A case-control study of the risk associated with eating foods at lunch on Day 1 or with attending different restaurants at dinner on Day 1 was attempted. A total of 19 subjects filled the section of the online survey concerning food consumption. Fourteen of the responders were among the 22 cases falling ill within 36 h from the first shared meal (first wave), while the remaining five responders were three travellers with supposed secondary infections (showing symptoms with a delay ⩾48 h; second wave) and two asymptomatic students that were used as controls. No statistically significant associations with any food items served at lunch on Day 1 was identified (P < 0.05; Table 2). However, all food consumed on that occasion was to some extent associated with the risk of disease, the highest OR being obtained with chicken (OR: 20.7143, P: 0.0678; Table 2). Actually, most of the cases ate all foods included in the menu, while one of the two asymptomatic controls did not participate in the meal and the other had a pasta dish only. No statistically significant association was found for attending a particular fast food restaurant at the dinner on Day 1 (OR: 2.5, P: 0.3635). The Real-Time RT-PCR assay allowed to detect GII genogroup NoV genome in all the 25 samples from symptomatic subjects, although two of them showed high Ct values, indicating lower shedding of viral genomes (Table 1). The samples from four asymptomatic students were negative for both NoV genogroups. By sequence analysis, a NoV genotype could be defined for 20/ 25 NoV-positive stool samples. All NoV strains from travellers involved in the outbreaks belonged to the same genotype, GII.P2/GII.2, showing >99% nucleotide identities (nt id) in both fragments (ORF1 and ORF2) analysed. Thirty-two water samples were collected and analysed for bacterial contamination (Table 3). Seven samples were taken from the resort (site R) and the remaining 25 from five other sites (C, F, M, S, W), representative of the entire municipal water network. The majority (10/12) of the water samples collected immediately after the onset of the outbreak, between 21 May and 16 June 2016, at the resort (site R) and at sites F and C, located close to the resort, on the seaside and in the interior, respectively, were positive for coliform bacteria. More than 1000 colony-forming units (CFU)/ 100 ml (up to 2700 CFU/100 ml) were detected in four samples, while their absence is required in 100 ml by the Standard of Water Quality in Italy and seven of 11 samples tested were also Escherichia coli-positive (up to 900 CFU/100 ml). Enterococci were searched in 11 samples and three of them were found positive for this faecal contamination indicator. From 7 July to 16 August 2016, after control measures on the water network were undertaken, 13 water samples were collected for follow-up laboratory examination of the water system at F, R and C sites. All of them were negative for bacterial indicators of faecal contamination. No faecal bacteria were identified in the seven samples collected from M, W and S sites, including the well and the source that fed the water network, at any sampling time, before and after implementing the control measures. The research of NoV genome was conducted on concentrated water samples collected on 1 August from R and F sites (Table 3). NoV genome was detected in a single sample collected from the water entry of the resort involved in the outbreak (site R). Actually, only two out of 26 One-Step RT-PCR repetitions were positive for NoV viral capsid gene. The two ORF2-positive results were obtained at the University of Palermo Laboratory using both 1:10 diluted and undiluted extracted RNA. The sequence analysis of the partial ORF2 portion (289 bp) indicated the presence of the same GII.2 capsid genotype detected in the stool samples of the students involved in the outbreak, showing 98.9% nt id. No ORF1 NoV sequence could be obtained from the NoV-positive water sample. Follow-up analyses for NoV genome, conducted on water samples collected on 20 November both at R and F sites found no positive samples. By sequence alignment, the NoV sequences obtained from Mascali outbreak samples were phylogenetically closely related in both ORF1 and ORF2 (99 and 98.9% of nt id, respectively) to a GII.P2/GII.2 strain responsible for a sporadic paediatric gastroenteritis case occurred in Palermo in January 2016 (PA 36/16, accession numbers MG517464 and MG517460, for ORF1 and ORF2, respectively) and previously detected at the Enteric Viruses Laboratory, University of Palermo. Whereas, three GII.P2/GII.2 Italian strains (PA 280/11, PA 366/11 and PA 3/ 14, accession numbers MG517462, MG517465 and MG517463, for ORF1 and MG517457 to −59, for ORF2, respectively), detected in Palermo in 2011 and 2014, were more distantly associated with the outbreak strain both in RdRp and VP1 (90% to 97.4% nt id and 96.1% to 98.1% nt id, respectively, in ORF1 and ORF2). Discussion Epidemiological features, such as the presence of vomiting in more than half (88%) of symptomatic subjects, the 33.2 h average incubation period and the 38.4 h average duration of illness, suggested a NoV aetiology for the investigated outbreak [13]. The detection of NoV genome in all 25 faecal samples from symptomatic subjects fully complies with the requirement of detection of at least one positive sample out of four tested that is considered to be sufficient to ascertain the aetiology of a NoV outbreak [24]. Moreover, molecular typing indicated that a single viral strain, belonging to genotype GII.P2/GII.2, infected all symptomatic subjects involved in the outbreak. NoVs are highly infectious and can easily spread from person to person but also through contaminated food or water. Therefore, settings such as health care institutions, schools and touristic resorts are mainly affected, since close contact and shared food services facilitate transmission. In Mascali, two waves of transmission were observed over three days, being consistent with food/waterborne followed by person-to-person and/or airborne (vomit) transmission (Fig. 2). The first wave of cases presented on 20 May 2016 after a single shared meal, while the last case was detected on 22 May. Exposure to vomit may have been responsible for some of the secondary cases occurred after the first wave of illness. Vomiting produces aerosols that can be inhaled and swallowed. The extremely low infectious dose of NoV makes the inhalation of vomit aerosols an effective transmission route in settings where there is close contact, like on a bus trip [25]. All food consumed in the only shared lunch on 19 May at the resort was associated with the risk of disease (Table 2), but the association was inconclusive since all cases shared the same dishes included in the lunch menu on Day 1. Although many foods were potentially involved, our attention focused on the tap water. In fact, trivial information from local health authorities reported that local population avoided tap water consumption and relied on bottled water for drinking and cooking. Many small outbreaks of gastroenteritis involving restaurants and family groups had been reported in the same area, especially at the beginning of the summer season, when vacation houses and recreational activities were re-opened after a period of low water demand. Illness caused by NoV is common and if the level of water contamination is low, the number of cases remains limited and water can persist as unrecognised source of infection. If polluted water was served for drinking and used for cooking and washing dishes in the Mascali resort at Day 1 a generalised contamination of food and surfaces could have occurred. The potential role of the foodhandlers as a source of contamination could not be analysed in this study but should also be taken into account. If water contamination has already been present in the area for a long time, infections may have occurred among the local food-handlers. The analysis of water samples taken immediately after the outbreak revealed the presence of bacterial indicators of faecal contamination in the water used in the resort, with the only exception of a sample of filter purified water used both for cooking and as drinking water. When extending the sampling to other areas of the municipal water network, many other sites, mostly on the seaside, showed faecal pollution, which demonstrates a poor protection of the water network from faecal leakage (Fig. 1). NoV contamination of waters generally originates from sewage since these enteric pathogens, as other enteric viruses affecting humans, are mostly species-specific and are shed at high concentrations in the stools of patients. When the analyses on the stools of the students made it clear that the outbreak had been caused by NoV, the viral genome was also searched in appropriate volumes of water samples, but this happened when correction measures on the water network had already been taken. Nevertheless, a single water sample, obtained 2 months after the outbreak (1 August 2016), still contained genome fragments that could be correlated to the same NoV genotype causing the outbreak. Although the finding of NoV genome does not demonstrate that entire infecting viral particles were present in the water, it could indicate that a NoV belonging to the same genotype involved in the outbreak was still circulating in the local population several weeks after the outbreak and it was someway able to penetrate the water network, probably indicating continuing faecal contamination through leakage from sewage systems. NoV is particularly resistant in the environment and also to chlorine, therefore transmission through chlorinated water is not exceptional [9, 12, 26]. Drinking water from municipal supplies fed by lake or groundwater has already been identified in the past as the source of waterborne NoV outbreaks in Italy [27–29] and NoV is now considered one of the pathogens causing the largest number of affected consumers in drinking waterborne outbreaks [12]. NoV can persist in water longer than 15 days [30] and it is resistant to chlorine disinfection when free chlorine levels or pre-treatment processes are inadequate [31, 32]. Moreover, microfiltration filters labelled as certified by NSF Standards 53 or 58 (with a pore size ranging from 0.05 to 5 µm) are not effective in removing viruses and can even decrease chlorine concentration in filtered water [33]. Actually, the owner of the resort involved in our outbreak was encouraged to serve tap water as drinking water at the restaurant relying on the effectiveness of such a filter purifier. Unfortunately, we did not have a tap water sample from the restaurant of the resort taken before correction measures and whose volume was appropriate for virus detection. Therefore, we could not demonstrate the failure to eliminate viral pollution of the microfiltration system. In Mascali, correction measures on the water network consisted in the maintenance of the old chlorination device and in the addition of a new one at a key point located upstream and close to suspected illegal sewage dumping (Fig. 1), but as poor protection of the pipeline from sewage leakage risked to reproduce faecal pollution episodes, renovation of the municipal water system was also advised, as well as systematic control of illegal sewage dumping. The additional chlorination device was apparently the most effective intervention, since right after its activation faecal bacterial pollution indicators disappeared. However, NoV genome was detected in a water sample taken 6 weeks after the last detection of bacterial indicators. Unfortunately, detection of faecal contamination of drinking water is only based on the use of bacterial indicators despite no correlation between levels of enteric bacteria and enteric viruses has been found [34–36]. According to ISGEV enteric virus circulation surveillance data, in 2016 the circulation of GII.P2/GII.2 NoV strains in Palermo, home town of the travellers involved in the outbreak, was the highest ever observed from the starting of local NoV surveillance in 2004. GII.P2/GII.2 was the fourth most common genotype in 2016 (12.2% prevalence), following GII.P4/GII.4 Sydney 2012 (28.4%), GII.P16/GII.2 (18.9%) and GII.Pe/GII.4 Sydney 2012 (17.6%). GII.P2/GII.2 strains were mostly observed starting from July 2016, leaving space for the possibility that their increased circulation was linked to the introduction of the ‘Mascali clone’. However, a couple of GII.P2/GII.2 strains (including PA36/16) genetically close to the ‘Mascali clone’ had been already detected in January 2016 in Palermo and GII.P2/GII.2 NoV had sporadically been observed also in 2011 and 2014. Moreover, the circulation of GII.P2/GII.2 NoVs has been increasing worldwide over the last period [37]. Therefore, the unusually high circulation observed in Palermo might just reflect the global epidemiology of GII.P2/GII.2 strains. Noteworthy, the older GII.P2/GII.2 NoV strains detected in Palermo were closely related in the ORF2 (98.9% nt id) to the GII.P16/GII.2 Nashville strain (KY865307) which is supposed to have generated the recombinant GII.P16/GII.4 clone recently circulating epidemically worldwide [37, 38] and were more distantly associated (95.3% nt id) to the GII.2 capsid sequence (AB662868) of another putative GII.P16/GII.4 ancestor from Japan [39]. The increasing frequency of detection of waterborne NoV outbreaks calls for the introduction of routine monitoring for viruses of drinking waters and for the definition of affordable and reliable indicators for the presence of NoV in drinking waters. The use of ultrafiltration systems as a concentration method in order to recover viruses from environmental waters combined with RT-PCR to detect viral genomes has been demonstrated to grant a recovery rate of >50% of multiple viruses [40] and is at present one of the most sensitive approaches for the detection of NoVs, although the discrimination between viable and nonviable viruses remains a problem. The burden of NoV outbreaks, either food associated or waterborne, is still poorly defined due to the lack of specific diagnostic tools in many clinical laboratories. NoV epidemics should always be suspected when the main symptom is vomiting and the duration of disease is short (<48 h). Waterborne transmission is suggested by the rapid spread of disease and the high attack rate. Care should be taken to perform sensitive diagnostic tests on a sufficient number of stool samples in order to identify NoV as the causative agent of enteritis outbreaks, possibly preferring genome screening over antigen detection. Only the systematic detection of NoV outbreaks can guide in the choice of the safety procedures that could be most appropriate in containing such microbiological hazard.

How to prevent this?

1563

Waterborne Norovirus outbreak at a seaside resort likely originating from municipal water distribution system failure

Abstract In May 2016 a Norovirus (NoV) gastroenteritis outbreak involved a high school class visiting a seaside resort near Taormina (Mascali, Sicily). Twenty-four students and a teacher were affected and 17 of them showed symptoms on the second day of the journey, while the others got ill within the following 2 days. Symptoms included vomiting, diarrhoea and fever, and 12 students required hospitalisation. Stool samples tested positive for NoV genome by Real-Time polymerase chain reaction assay in all 25 symptomatic subjects. The GII.P2/GII.2 NoV genotype was linked to the outbreak by ORF1/ORF2 sequence analysis. The epidemiological features of the outbreak were consistent with food/waterborne followed by person-to-person and/ or vomit transmission. Food consumed at a shared lunch on the first day of the trip was associated to illness and drinking un-bottled tap water was also considered as a risk factor. The analysis of water samples revealed the presence of bacterial indicators of faecal contamination in the water used in the resort as well as in other areas of the municipal water network, linking the NoV gastroenteritis outbreak to tap water pollution from sewage leakage. From a single water sample, an amplicon whose sequence corresponded to the capsid genotype recovered from patients could be obtained. Introduction Noroviruses (NoVs) are recognised as a leading cause of sporadic cases and outbreaks of acute gastroenteritis (AGE) across all age groups [1, 2]. They are transmitted primarily through the faecal-oral route either by direct person–person spread and with consumption of contaminated food or water [1, 3, 4]. NoVs are non-enveloped RNA viruses belonging to the Caliciviridae family and are classified into seven distinct genogroups (GI-GVII) on the basis of the major capsid protein VP1 sequence. GI, GII and GIV NoVs infect humans and they have been divided into more than 30 genotypes based on the sequences of polymerase (ORF1) and capsid proteins (ORF2) [5]. NoV GII.4 genotype has been associated with the vast majority of NoV related cases of AGE worldwide although many genotypes, belonging to both GI and GII genogroups are involved in foodborne and waterborne outbreaks [6, 7]. NoVs are highly contagious owing to their high stability in the environment and their ability to infect at low doses. In fact, just 10–100 virions of NoV may be sufficient to cause AGE [8, 9]. NoVs are considered the major cause of water-related disease in semi-closed institutions, including restaurants, schools, hospitals, nursing homes and cruise ships [10–12]. To confirm NoV as a cause of outbreaks, the presence of four epidemiological features of disease, vomiting in more than half of affected persons, a mean incubation period of 24/48 h, a mean duration of illness of 12–60 h and the absence of bacterial pathogens in stool culture, should be evaluated [13]. Waterborne NoV outbreaks can be caused by contaminated drinking water despite correct chlorine concentration and in the absence of coliform bacteria [10, 14–17]. The aim of this study was to investigate an outbreak of NoV gastroenteritis associated with drinking water involving a high school class travelling from Palermo (West Sicily) to Acireale (East Sicily) on an educational tour in May 2016. Stool samples were collected soon after the onset of AGE symptoms to ascertain the aetiology of illness. The source of infection was investigated collecting epidemiological data through an online questionnaire and, in order to explore a possible transmission route, laboratory tests on water from the municipal network were performed. Material and methods Epidemiological investigation From 20 to 22 May 2016 an AGE outbreak occurred in Mascali, a small town near Taormina, involving a school class of 28 students, 13 girls and 15 boys aged between 14 and 18 years and a 54-year-old female teacher, during a trip to participate in a theatre festival in Acireale. A total of 25/29 persons were affected by nausea, vomiting, abdominal cramps, diarrhoea and fever while they were staying at a seaside resort in Mascali, on the way to Acireale. To obtain data on symptoms and the correlated risk factors all participants filled an online questionnaire investigating personal data (age and sex), travel and accommodation details and food consumption over 48 h before symptoms presentation, time of onset and duration of symptoms, therapy and/or hospitalisation required and history of contact with people affected by diarrhoea and/or vomiting (Table 1). The severity of illness was analysed by the Vesikari severity scale allowing to distinguish non-severe (score <11) and severe (score ⩾11) AGE (Table 1). Vesikari score was calculated from maximum number of episodes/day and duration of both diarrhoea and vomiting (score 0–3 for each of the 4 parameters), temperature (score 0–3), level of dehydration (score 0–3) and treatment required (score 0–2) [18]. A case-control study was conducted to establish if food consumed before the onset of symptoms was associated with the disease. A case was defined as a student attending the tour and reporting diarrhoea and/or vomiting within 36 h from the first shared meal during the tour, while students who did not present any of the previous symptoms or those showing the first symptoms with a delay of ⩾48 h (secondary cases) were used as controls. Variables were studied using odds ratio (OR) and its 95% CI and statistical associations were determined using the χ2 test, with the level of statistical significance set at 0.05 (Table 2). NoV detection in stool samples A stool sample was obtained from each of the 29 travellers, soon after (3–5 days) the onset of the outbreak. No stool samples could be obtained from the food-handlers working at the resort. All samples were tested for NoV at the Enteric Viruses Laboratory of the University of Palermo, member of the Italian Study Group for Enteric Viruses (ISGEV), by a Real-Time polymerase chain reaction (RT-PCR) assay that allows to differentiate GI and GII NoVs [19]. Total RNA was extracted from stool samples using QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) and cDNA was generated by MMLV reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were considered positive for NoVs when cycle threshold (ct) of RT-PCR was ⩽30 and lowpositive when 30<ct⩽35. For all NoV-positive samples two genomic fragments were amplified by conventional RT-PCR using Superscript III One-Step RT-PCR system with Platinum Taq DNA polymerase (Invitrogen) and primers JV12a/JV13b, targeting 326 bps of the RdRp region A [20] and either COG1F/ GISKR or COG2F/GIISKR, amplifying a 315 bps 5′portion of GI or GII VP1 protein [19, 21]. Environmental investigation From May through August 2016, a total of 32 tap water samples were collected from the Mascali municipal network. In particular, seven samples were collected at the resort where the outbreak occurred and from its water entry (R), two in the city centre of Mascali (M), 15 in the village of Fondachello (F), on the seaside, three in Carrabba village (C), far from the coast, and two and three, respectively, from the water well ‘Carlino’ (W) and the source ‘Bufardo and Torregrossa’ (S) which feed the municipal system and provide drinking water to the Mascali area (Fig. 1, Table 3). One litre water samples were sent by the local health authorities to the SIALAB laboratory in Avola, Siracusa and 100 ml aliquots were screened for coliforms, E. coli and enterococci as faecal indicators, by standard bacteriological methods according to UNI EN ISO 9308-1:2014. Unfortunately, although the suspect of a water source for the viral outbreak was quickly communicated, only in August and November 2016, following the implementation of control measures, local authorities provided water samplings of appropriate volume to perform viral tests (10 l). Five water samples of 10 l each were collected from different sampling sites in F and R areas for viral screening (Table 3) and were concentrated according to the Italian National Institute of Health (Istituto Superiore di Sanità, ISS) protocols [22] through an ultra-filtration system (Sartoflow® Slice 200 Benchtop Crossflow System, Sartorius AG, Goettingen, Germany), using appropriate membranes (SG Hydrosart 10 kDa) pretreated with 300 ml of 3% Beef Extract (BE) at pH 7, by recirculating it for 10 min in the system, at a pressure not higher than 1.30 bar. Elution and recovery of viruses attached to the membranes were obtained by recirculating 200 ml of 3% BE at pH 9.5 for 3 h until a 10–12 ml volume of eluate was obtained. The pH of the eluate was then checked with litmus paper and if necessary brought to the value of 7 with HCl. Concentrated water samples were simultaneously tested for NoV in three laboratories: Enteric Viruses Laboratory, University of Palermo; Istituto Zooprofilattico Sperimentale della Sicilia (IZS), Palermo; and Istituto Superiore di Sanità (ISS), Rome. Total RNA was extracted from 500 and 560 µl of a concentrated water sample using BioMèrieux NucliSENS® Magnetic Extraction System (BioMèrieux) and QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) respectively, according to manufacturer’s instructions. Bonded RNA was eluted by 100 µl of Elution Buffer. Amplifications were performed at least in duplicate on whole and 1:10 diluted extracts. A Real-Time RT-PCR assay according to ISO/TS 15216- 2:2013, two in-house Real-Time RT-PCR protocols and the One-Step RT-PCR methods already used for amplifying ORF1 and ORF2 fragments in NoV-positive faecal samples were performed [19–21, 23]. Nucleotide sequences analysis and phylogenetic analysis PCR amplicons were purified by commercial columns (CleanSweep PCR Purification, Thermo Fisher Scientific, Vilnius, Lithuania) and sequenced by Sanger method and all nucleotide sequences were aligned using ClustalW. To identify NoV genotypes, RdRp and VP1 sequences were submitted to the RIVM NoV genotyping tool (http://www.rivm.nl/mpf/norovirus/typingtool). Phylogenetic trees were constructed using MEGA 7 and Neighbour-Joining method (bootstrap 1000 replicates). Sequences from stool and water samples are available at NCBI with the following accession numbers: MG517455 for the major capsid protein obtained from a water sample; MG517456 and MG517461, respectively, for the major capsid protein and RNA-dependent-RNA-polymerase, as representatives of the sequences obtained from the stool samples of those involved in the outbreak. Control measures After outbreak notification, on 21 May, the resort was closed for 2 days. On 25 May, following the first results of the bacteriological analyses on municipal water samples, the use of municipal water for human consumption was temporarily restricted (use of water was permitted only for cleaning purposes) in the Carrabba and Fondachello villages and in the area surrounding the resort. From 25 May the reservoir collecting the water from the Carlino well and its chlorination device were submitted to maintenance work and a new chlorination device was installed on the conduit bringing the water from the Bufardo source. Actually, the Bufardo source had been connected to the municipal network in 2003 but unfortunately a dedicated chlorination station had not been planned at that time. Therefore, unchlorinated water entered the system near Fondachello village on the seaside (Fig. 1), possibly diluting chlorine concentrations in water treated at the Carlino reservoir. Moreover, the exact path of the branch of the water network situated downstream the Bufardo source water connection was not marked in the maps of the municipality (Fig. 1, red line), indicating that this section had never been submitted to maintenance. The area surrounding the unmaintained section of the water pipeline was scattered with unauthorised cottages that were not connected to the sewerage systems. Likely, untreated waste from cesspools was unloaded directly into the soil, generating a risk of contamination of the municipal water system through leakage. No work had been made on the water pipelines before the outbreak but sections of the network had been temporarily closed during the winter season on the seaside (Fondachello). Local health authorities advised the connection of all households in the area to the sewerage system and the replacement of the water system section without maintenance. Health authorities also raised attention on AGE cases notification at all local Emergency Departments and Family Practitioners but there was no further evidence of AGE outbreaks in the months following the intervention. Results Data collected through the online questionnaire compiled 7–10 days after the outbreak revealed that none of the 28 students and the teacher taking part in the school trip showed gastroenteritis symptoms at the beginning of the journey and they had no contact with people who suffered from diarrhoea and/or vomiting in the previous days. Travellers had their first shared meal on 19 May (Day 1), 2016, at lunch on their arrival to the resort in Mascali. The list of foods included in the menu is shown in Table 2. Conversely, the dinner on Day 1 was not shared and the students went to dine in three different fast food restaurants in Acireale. On 20 May (Day 2), 17 students (58.6% of the travellers) presented with vomit and/or diarrhoea, falling ill 26–36 h after lunch on Day 1, strongly suggesting that exposure occurred on the first shared meal. Some students also vomited in the presence of the other travellers on the bus bringing them to Acireale. Five more students were symptomatic in the early hours of the following day (21 May, Day 3, from 0:30 am to 8:00 am), as a continuation of the wave of illnesses of the previous day. The teacher and one more student fell ill at 8:00 pm and 11:00 pm of Day 3, respectively, for a second wave of illnesses occurring >48 h from the first shared meal and <36 h from the first cases, with the last student falling sick on 22 May (Day 4) at 3:00 am (Fig. 2, Table 1). On the whole, the mean incubation period was 33.2 h. Based on the online survey, the severity of illness was evaluated by the Vesikari severity scale, grouping symptomatic travellers into severe (11/25 = 44%) and non-severe (14/25 = 56%) cases (Table 1). Vomiting and fever were the predominant symptoms, affecting 22 out of 25 (88%) symptomatic subjects, while diarrhoea was present in 14 out of 25 (56%). Hospitalisation for rehydration therapy was necessary for 13 students (91% [10/11] of severe and 21% [3/14] of non-severe cases). Emergency room (ER) or family practitioner visits were necessary for 21 students and 13 (62%) of them initiated an antibiotic therapy. Two students indicated that other family members had gastroenteric symptoms more than 24 h after their return home, on Day 3, 4 and 5. All those involved in the outbreak recovered within 4 days, with an average duration of illness of 38.4 h. A case-control study of the risk associated with eating foods at lunch on Day 1 or with attending different restaurants at dinner on Day 1 was attempted. A total of 19 subjects filled the section of the online survey concerning food consumption. Fourteen of the responders were among the 22 cases falling ill within 36 h from the first shared meal (first wave), while the remaining five responders were three travellers with supposed secondary infections (showing symptoms with a delay ⩾48 h; second wave) and two asymptomatic students that were used as controls. No statistically significant associations with any food items served at lunch on Day 1 was identified (P < 0.05; Table 2). However, all food consumed on that occasion was to some extent associated with the risk of disease, the highest OR being obtained with chicken (OR: 20.7143, P: 0.0678; Table 2). Actually, most of the cases ate all foods included in the menu, while one of the two asymptomatic controls did not participate in the meal and the other had a pasta dish only. No statistically significant association was found for attending a particular fast food restaurant at the dinner on Day 1 (OR: 2.5, P: 0.3635). The Real-Time RT-PCR assay allowed to detect GII genogroup NoV genome in all the 25 samples from symptomatic subjects, although two of them showed high Ct values, indicating lower shedding of viral genomes (Table 1). The samples from four asymptomatic students were negative for both NoV genogroups. By sequence analysis, a NoV genotype could be defined for 20/ 25 NoV-positive stool samples. All NoV strains from travellers involved in the outbreaks belonged to the same genotype, GII.P2/GII.2, showing >99% nucleotide identities (nt id) in both fragments (ORF1 and ORF2) analysed. Thirty-two water samples were collected and analysed for bacterial contamination (Table 3). Seven samples were taken from the resort (site R) and the remaining 25 from five other sites (C, F, M, S, W), representative of the entire municipal water network. The majority (10/12) of the water samples collected immediately after the onset of the outbreak, between 21 May and 16 June 2016, at the resort (site R) and at sites F and C, located close to the resort, on the seaside and in the interior, respectively, were positive for coliform bacteria. More than 1000 colony-forming units (CFU)/ 100 ml (up to 2700 CFU/100 ml) were detected in four samples, while their absence is required in 100 ml by the Standard of Water Quality in Italy and seven of 11 samples tested were also Escherichia coli-positive (up to 900 CFU/100 ml). Enterococci were searched in 11 samples and three of them were found positive for this faecal contamination indicator. From 7 July to 16 August 2016, after control measures on the water network were undertaken, 13 water samples were collected for follow-up laboratory examination of the water system at F, R and C sites. All of them were negative for bacterial indicators of faecal contamination. No faecal bacteria were identified in the seven samples collected from M, W and S sites, including the well and the source that fed the water network, at any sampling time, before and after implementing the control measures. The research of NoV genome was conducted on concentrated water samples collected on 1 August from R and F sites (Table 3). NoV genome was detected in a single sample collected from the water entry of the resort involved in the outbreak (site R). Actually, only two out of 26 One-Step RT-PCR repetitions were positive for NoV viral capsid gene. The two ORF2-positive results were obtained at the University of Palermo Laboratory using both 1:10 diluted and undiluted extracted RNA. The sequence analysis of the partial ORF2 portion (289 bp) indicated the presence of the same GII.2 capsid genotype detected in the stool samples of the students involved in the outbreak, showing 98.9% nt id. No ORF1 NoV sequence could be obtained from the NoV-positive water sample. Follow-up analyses for NoV genome, conducted on water samples collected on 20 November both at R and F sites found no positive samples. By sequence alignment, the NoV sequences obtained from Mascali outbreak samples were phylogenetically closely related in both ORF1 and ORF2 (99 and 98.9% of nt id, respectively) to a GII.P2/GII.2 strain responsible for a sporadic paediatric gastroenteritis case occurred in Palermo in January 2016 (PA 36/16, accession numbers MG517464 and MG517460, for ORF1 and ORF2, respectively) and previously detected at the Enteric Viruses Laboratory, University of Palermo. Whereas, three GII.P2/GII.2 Italian strains (PA 280/11, PA 366/11 and PA 3/ 14, accession numbers MG517462, MG517465 and MG517463, for ORF1 and MG517457 to −59, for ORF2, respectively), detected in Palermo in 2011 and 2014, were more distantly associated with the outbreak strain both in RdRp and VP1 (90% to 97.4% nt id and 96.1% to 98.1% nt id, respectively, in ORF1 and ORF2). Discussion Epidemiological features, such as the presence of vomiting in more than half (88%) of symptomatic subjects, the 33.2 h average incubation period and the 38.4 h average duration of illness, suggested a NoV aetiology for the investigated outbreak [13]. The detection of NoV genome in all 25 faecal samples from symptomatic subjects fully complies with the requirement of detection of at least one positive sample out of four tested that is considered to be sufficient to ascertain the aetiology of a NoV outbreak [24]. Moreover, molecular typing indicated that a single viral strain, belonging to genotype GII.P2/GII.2, infected all symptomatic subjects involved in the outbreak. NoVs are highly infectious and can easily spread from person to person but also through contaminated food or water. Therefore, settings such as health care institutions, schools and touristic resorts are mainly affected, since close contact and shared food services facilitate transmission. In Mascali, two waves of transmission were observed over three days, being consistent with food/waterborne followed by person-to-person and/or airborne (vomit) transmission (Fig. 2). The first wave of cases presented on 20 May 2016 after a single shared meal, while the last case was detected on 22 May. Exposure to vomit may have been responsible for some of the secondary cases occurred after the first wave of illness. Vomiting produces aerosols that can be inhaled and swallowed. The extremely low infectious dose of NoV makes the inhalation of vomit aerosols an effective transmission route in settings where there is close contact, like on a bus trip [25]. All food consumed in the only shared lunch on 19 May at the resort was associated with the risk of disease (Table 2), but the association was inconclusive since all cases shared the same dishes included in the lunch menu on Day 1. Although many foods were potentially involved, our attention focused on the tap water. In fact, trivial information from local health authorities reported that local population avoided tap water consumption and relied on bottled water for drinking and cooking. Many small outbreaks of gastroenteritis involving restaurants and family groups had been reported in the same area, especially at the beginning of the summer season, when vacation houses and recreational activities were re-opened after a period of low water demand. Illness caused by NoV is common and if the level of water contamination is low, the number of cases remains limited and water can persist as unrecognised source of infection. If polluted water was served for drinking and used for cooking and washing dishes in the Mascali resort at Day 1 a generalised contamination of food and surfaces could have occurred. The potential role of the foodhandlers as a source of contamination could not be analysed in this study but should also be taken into account. If water contamination has already been present in the area for a long time, infections may have occurred among the local food-handlers. The analysis of water samples taken immediately after the outbreak revealed the presence of bacterial indicators of faecal contamination in the water used in the resort, with the only exception of a sample of filter purified water used both for cooking and as drinking water. When extending the sampling to other areas of the municipal water network, many other sites, mostly on the seaside, showed faecal pollution, which demonstrates a poor protection of the water network from faecal leakage (Fig. 1). NoV contamination of waters generally originates from sewage since these enteric pathogens, as other enteric viruses affecting humans, are mostly species-specific and are shed at high concentrations in the stools of patients. When the analyses on the stools of the students made it clear that the outbreak had been caused by NoV, the viral genome was also searched in appropriate volumes of water samples, but this happened when correction measures on the water network had already been taken. Nevertheless, a single water sample, obtained 2 months after the outbreak (1 August 2016), still contained genome fragments that could be correlated to the same NoV genotype causing the outbreak. Although the finding of NoV genome does not demonstrate that entire infecting viral particles were present in the water, it could indicate that a NoV belonging to the same genotype involved in the outbreak was still circulating in the local population several weeks after the outbreak and it was someway able to penetrate the water network, probably indicating continuing faecal contamination through leakage from sewage systems. NoV is particularly resistant in the environment and also to chlorine, therefore transmission through chlorinated water is not exceptional [9, 12, 26]. Drinking water from municipal supplies fed by lake or groundwater has already been identified in the past as the source of waterborne NoV outbreaks in Italy [27–29] and NoV is now considered one of the pathogens causing the largest number of affected consumers in drinking waterborne outbreaks [12]. NoV can persist in water longer than 15 days [30] and it is resistant to chlorine disinfection when free chlorine levels or pre-treatment processes are inadequate [31, 32]. Moreover, microfiltration filters labelled as certified by NSF Standards 53 or 58 (with a pore size ranging from 0.05 to 5 µm) are not effective in removing viruses and can even decrease chlorine concentration in filtered water [33]. Actually, the owner of the resort involved in our outbreak was encouraged to serve tap water as drinking water at the restaurant relying on the effectiveness of such a filter purifier. Unfortunately, we did not have a tap water sample from the restaurant of the resort taken before correction measures and whose volume was appropriate for virus detection. Therefore, we could not demonstrate the failure to eliminate viral pollution of the microfiltration system. In Mascali, correction measures on the water network consisted in the maintenance of the old chlorination device and in the addition of a new one at a key point located upstream and close to suspected illegal sewage dumping (Fig. 1), but as poor protection of the pipeline from sewage leakage risked to reproduce faecal pollution episodes, renovation of the municipal water system was also advised, as well as systematic control of illegal sewage dumping. The additional chlorination device was apparently the most effective intervention, since right after its activation faecal bacterial pollution indicators disappeared. However, NoV genome was detected in a water sample taken 6 weeks after the last detection of bacterial indicators. Unfortunately, detection of faecal contamination of drinking water is only based on the use of bacterial indicators despite no correlation between levels of enteric bacteria and enteric viruses has been found [34–36]. According to ISGEV enteric virus circulation surveillance data, in 2016 the circulation of GII.P2/GII.2 NoV strains in Palermo, home town of the travellers involved in the outbreak, was the highest ever observed from the starting of local NoV surveillance in 2004. GII.P2/GII.2 was the fourth most common genotype in 2016 (12.2% prevalence), following GII.P4/GII.4 Sydney 2012 (28.4%), GII.P16/GII.2 (18.9%) and GII.Pe/GII.4 Sydney 2012 (17.6%). GII.P2/GII.2 strains were mostly observed starting from July 2016, leaving space for the possibility that their increased circulation was linked to the introduction of the ‘Mascali clone’. However, a couple of GII.P2/GII.2 strains (including PA36/16) genetically close to the ‘Mascali clone’ had been already detected in January 2016 in Palermo and GII.P2/GII.2 NoV had sporadically been observed also in 2011 and 2014. Moreover, the circulation of GII.P2/GII.2 NoVs has been increasing worldwide over the last period [37]. Therefore, the unusually high circulation observed in Palermo might just reflect the global epidemiology of GII.P2/GII.2 strains. Noteworthy, the older GII.P2/GII.2 NoV strains detected in Palermo were closely related in the ORF2 (98.9% nt id) to the GII.P16/GII.2 Nashville strain (KY865307) which is supposed to have generated the recombinant GII.P16/GII.4 clone recently circulating epidemically worldwide [37, 38] and were more distantly associated (95.3% nt id) to the GII.2 capsid sequence (AB662868) of another putative GII.P16/GII.4 ancestor from Japan [39]. The increasing frequency of detection of waterborne NoV outbreaks calls for the introduction of routine monitoring for viruses of drinking waters and for the definition of affordable and reliable indicators for the presence of NoV in drinking waters. The use of ultrafiltration systems as a concentration method in order to recover viruses from environmental waters combined with RT-PCR to detect viral genomes has been demonstrated to grant a recovery rate of >50% of multiple viruses [40] and is at present one of the most sensitive approaches for the detection of NoVs, although the discrimination between viable and nonviable viruses remains a problem. The burden of NoV outbreaks, either food associated or waterborne, is still poorly defined due to the lack of specific diagnostic tools in many clinical laboratories. NoV epidemics should always be suspected when the main symptom is vomiting and the duration of disease is short (<48 h). Waterborne transmission is suggested by the rapid spread of disease and the high attack rate. Care should be taken to perform sensitive diagnostic tests on a sufficient number of stool samples in order to identify NoV as the causative agent of enteritis outbreaks, possibly preferring genome screening over antigen detection. Only the systematic detection of NoV outbreaks can guide in the choice of the safety procedures that could be most appropriate in containing such microbiological hazard.

What were the investigation steps?

1564

Waterborne Norovirus outbreak at a seaside resort likely originating from municipal water distribution system failure

Abstract In May 2016 a Norovirus (NoV) gastroenteritis outbreak involved a high school class visiting a seaside resort near Taormina (Mascali, Sicily). Twenty-four students and a teacher were affected and 17 of them showed symptoms on the second day of the journey, while the others got ill within the following 2 days. Symptoms included vomiting, diarrhoea and fever, and 12 students required hospitalisation. Stool samples tested positive for NoV genome by Real-Time polymerase chain reaction assay in all 25 symptomatic subjects. The GII.P2/GII.2 NoV genotype was linked to the outbreak by ORF1/ORF2 sequence analysis. The epidemiological features of the outbreak were consistent with food/waterborne followed by person-to-person and/ or vomit transmission. Food consumed at a shared lunch on the first day of the trip was associated to illness and drinking un-bottled tap water was also considered as a risk factor. The analysis of water samples revealed the presence of bacterial indicators of faecal contamination in the water used in the resort as well as in other areas of the municipal water network, linking the NoV gastroenteritis outbreak to tap water pollution from sewage leakage. From a single water sample, an amplicon whose sequence corresponded to the capsid genotype recovered from patients could be obtained. Introduction Noroviruses (NoVs) are recognised as a leading cause of sporadic cases and outbreaks of acute gastroenteritis (AGE) across all age groups [1, 2]. They are transmitted primarily through the faecal-oral route either by direct person–person spread and with consumption of contaminated food or water [1, 3, 4]. NoVs are non-enveloped RNA viruses belonging to the Caliciviridae family and are classified into seven distinct genogroups (GI-GVII) on the basis of the major capsid protein VP1 sequence. GI, GII and GIV NoVs infect humans and they have been divided into more than 30 genotypes based on the sequences of polymerase (ORF1) and capsid proteins (ORF2) [5]. NoV GII.4 genotype has been associated with the vast majority of NoV related cases of AGE worldwide although many genotypes, belonging to both GI and GII genogroups are involved in foodborne and waterborne outbreaks [6, 7]. NoVs are highly contagious owing to their high stability in the environment and their ability to infect at low doses. In fact, just 10–100 virions of NoV may be sufficient to cause AGE [8, 9]. NoVs are considered the major cause of water-related disease in semi-closed institutions, including restaurants, schools, hospitals, nursing homes and cruise ships [10–12]. To confirm NoV as a cause of outbreaks, the presence of four epidemiological features of disease, vomiting in more than half of affected persons, a mean incubation period of 24/48 h, a mean duration of illness of 12–60 h and the absence of bacterial pathogens in stool culture, should be evaluated [13]. Waterborne NoV outbreaks can be caused by contaminated drinking water despite correct chlorine concentration and in the absence of coliform bacteria [10, 14–17]. The aim of this study was to investigate an outbreak of NoV gastroenteritis associated with drinking water involving a high school class travelling from Palermo (West Sicily) to Acireale (East Sicily) on an educational tour in May 2016. Stool samples were collected soon after the onset of AGE symptoms to ascertain the aetiology of illness. The source of infection was investigated collecting epidemiological data through an online questionnaire and, in order to explore a possible transmission route, laboratory tests on water from the municipal network were performed. Material and methods Epidemiological investigation From 20 to 22 May 2016 an AGE outbreak occurred in Mascali, a small town near Taormina, involving a school class of 28 students, 13 girls and 15 boys aged between 14 and 18 years and a 54-year-old female teacher, during a trip to participate in a theatre festival in Acireale. A total of 25/29 persons were affected by nausea, vomiting, abdominal cramps, diarrhoea and fever while they were staying at a seaside resort in Mascali, on the way to Acireale. To obtain data on symptoms and the correlated risk factors all participants filled an online questionnaire investigating personal data (age and sex), travel and accommodation details and food consumption over 48 h before symptoms presentation, time of onset and duration of symptoms, therapy and/or hospitalisation required and history of contact with people affected by diarrhoea and/or vomiting (Table 1). The severity of illness was analysed by the Vesikari severity scale allowing to distinguish non-severe (score <11) and severe (score ⩾11) AGE (Table 1). Vesikari score was calculated from maximum number of episodes/day and duration of both diarrhoea and vomiting (score 0–3 for each of the 4 parameters), temperature (score 0–3), level of dehydration (score 0–3) and treatment required (score 0–2) [18]. A case-control study was conducted to establish if food consumed before the onset of symptoms was associated with the disease. A case was defined as a student attending the tour and reporting diarrhoea and/or vomiting within 36 h from the first shared meal during the tour, while students who did not present any of the previous symptoms or those showing the first symptoms with a delay of ⩾48 h (secondary cases) were used as controls. Variables were studied using odds ratio (OR) and its 95% CI and statistical associations were determined using the χ2 test, with the level of statistical significance set at 0.05 (Table 2). NoV detection in stool samples A stool sample was obtained from each of the 29 travellers, soon after (3–5 days) the onset of the outbreak. No stool samples could be obtained from the food-handlers working at the resort. All samples were tested for NoV at the Enteric Viruses Laboratory of the University of Palermo, member of the Italian Study Group for Enteric Viruses (ISGEV), by a Real-Time polymerase chain reaction (RT-PCR) assay that allows to differentiate GI and GII NoVs [19]. Total RNA was extracted from stool samples using QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) and cDNA was generated by MMLV reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were considered positive for NoVs when cycle threshold (ct) of RT-PCR was ⩽30 and lowpositive when 30<ct⩽35. For all NoV-positive samples two genomic fragments were amplified by conventional RT-PCR using Superscript III One-Step RT-PCR system with Platinum Taq DNA polymerase (Invitrogen) and primers JV12a/JV13b, targeting 326 bps of the RdRp region A [20] and either COG1F/ GISKR or COG2F/GIISKR, amplifying a 315 bps 5′portion of GI or GII VP1 protein [19, 21]. Environmental investigation From May through August 2016, a total of 32 tap water samples were collected from the Mascali municipal network. In particular, seven samples were collected at the resort where the outbreak occurred and from its water entry (R), two in the city centre of Mascali (M), 15 in the village of Fondachello (F), on the seaside, three in Carrabba village (C), far from the coast, and two and three, respectively, from the water well ‘Carlino’ (W) and the source ‘Bufardo and Torregrossa’ (S) which feed the municipal system and provide drinking water to the Mascali area (Fig. 1, Table 3). One litre water samples were sent by the local health authorities to the SIALAB laboratory in Avola, Siracusa and 100 ml aliquots were screened for coliforms, E. coli and enterococci as faecal indicators, by standard bacteriological methods according to UNI EN ISO 9308-1:2014. Unfortunately, although the suspect of a water source for the viral outbreak was quickly communicated, only in August and November 2016, following the implementation of control measures, local authorities provided water samplings of appropriate volume to perform viral tests (10 l). Five water samples of 10 l each were collected from different sampling sites in F and R areas for viral screening (Table 3) and were concentrated according to the Italian National Institute of Health (Istituto Superiore di Sanità, ISS) protocols [22] through an ultra-filtration system (Sartoflow® Slice 200 Benchtop Crossflow System, Sartorius AG, Goettingen, Germany), using appropriate membranes (SG Hydrosart 10 kDa) pretreated with 300 ml of 3% Beef Extract (BE) at pH 7, by recirculating it for 10 min in the system, at a pressure not higher than 1.30 bar. Elution and recovery of viruses attached to the membranes were obtained by recirculating 200 ml of 3% BE at pH 9.5 for 3 h until a 10–12 ml volume of eluate was obtained. The pH of the eluate was then checked with litmus paper and if necessary brought to the value of 7 with HCl. Concentrated water samples were simultaneously tested for NoV in three laboratories: Enteric Viruses Laboratory, University of Palermo; Istituto Zooprofilattico Sperimentale della Sicilia (IZS), Palermo; and Istituto Superiore di Sanità (ISS), Rome. Total RNA was extracted from 500 and 560 µl of a concentrated water sample using BioMèrieux NucliSENS® Magnetic Extraction System (BioMèrieux) and QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) respectively, according to manufacturer’s instructions. Bonded RNA was eluted by 100 µl of Elution Buffer. Amplifications were performed at least in duplicate on whole and 1:10 diluted extracts. A Real-Time RT-PCR assay according to ISO/TS 15216- 2:2013, two in-house Real-Time RT-PCR protocols and the One-Step RT-PCR methods already used for amplifying ORF1 and ORF2 fragments in NoV-positive faecal samples were performed [19–21, 23]. Nucleotide sequences analysis and phylogenetic analysis PCR amplicons were purified by commercial columns (CleanSweep PCR Purification, Thermo Fisher Scientific, Vilnius, Lithuania) and sequenced by Sanger method and all nucleotide sequences were aligned using ClustalW. To identify NoV genotypes, RdRp and VP1 sequences were submitted to the RIVM NoV genotyping tool (http://www.rivm.nl/mpf/norovirus/typingtool). Phylogenetic trees were constructed using MEGA 7 and Neighbour-Joining method (bootstrap 1000 replicates). Sequences from stool and water samples are available at NCBI with the following accession numbers: MG517455 for the major capsid protein obtained from a water sample; MG517456 and MG517461, respectively, for the major capsid protein and RNA-dependent-RNA-polymerase, as representatives of the sequences obtained from the stool samples of those involved in the outbreak. Control measures After outbreak notification, on 21 May, the resort was closed for 2 days. On 25 May, following the first results of the bacteriological analyses on municipal water samples, the use of municipal water for human consumption was temporarily restricted (use of water was permitted only for cleaning purposes) in the Carrabba and Fondachello villages and in the area surrounding the resort. From 25 May the reservoir collecting the water from the Carlino well and its chlorination device were submitted to maintenance work and a new chlorination device was installed on the conduit bringing the water from the Bufardo source. Actually, the Bufardo source had been connected to the municipal network in 2003 but unfortunately a dedicated chlorination station had not been planned at that time. Therefore, unchlorinated water entered the system near Fondachello village on the seaside (Fig. 1), possibly diluting chlorine concentrations in water treated at the Carlino reservoir. Moreover, the exact path of the branch of the water network situated downstream the Bufardo source water connection was not marked in the maps of the municipality (Fig. 1, red line), indicating that this section had never been submitted to maintenance. The area surrounding the unmaintained section of the water pipeline was scattered with unauthorised cottages that were not connected to the sewerage systems. Likely, untreated waste from cesspools was unloaded directly into the soil, generating a risk of contamination of the municipal water system through leakage. No work had been made on the water pipelines before the outbreak but sections of the network had been temporarily closed during the winter season on the seaside (Fondachello). Local health authorities advised the connection of all households in the area to the sewerage system and the replacement of the water system section without maintenance. Health authorities also raised attention on AGE cases notification at all local Emergency Departments and Family Practitioners but there was no further evidence of AGE outbreaks in the months following the intervention. Results Data collected through the online questionnaire compiled 7–10 days after the outbreak revealed that none of the 28 students and the teacher taking part in the school trip showed gastroenteritis symptoms at the beginning of the journey and they had no contact with people who suffered from diarrhoea and/or vomiting in the previous days. Travellers had their first shared meal on 19 May (Day 1), 2016, at lunch on their arrival to the resort in Mascali. The list of foods included in the menu is shown in Table 2. Conversely, the dinner on Day 1 was not shared and the students went to dine in three different fast food restaurants in Acireale. On 20 May (Day 2), 17 students (58.6% of the travellers) presented with vomit and/or diarrhoea, falling ill 26–36 h after lunch on Day 1, strongly suggesting that exposure occurred on the first shared meal. Some students also vomited in the presence of the other travellers on the bus bringing them to Acireale. Five more students were symptomatic in the early hours of the following day (21 May, Day 3, from 0:30 am to 8:00 am), as a continuation of the wave of illnesses of the previous day. The teacher and one more student fell ill at 8:00 pm and 11:00 pm of Day 3, respectively, for a second wave of illnesses occurring >48 h from the first shared meal and <36 h from the first cases, with the last student falling sick on 22 May (Day 4) at 3:00 am (Fig. 2, Table 1). On the whole, the mean incubation period was 33.2 h. Based on the online survey, the severity of illness was evaluated by the Vesikari severity scale, grouping symptomatic travellers into severe (11/25 = 44%) and non-severe (14/25 = 56%) cases (Table 1). Vomiting and fever were the predominant symptoms, affecting 22 out of 25 (88%) symptomatic subjects, while diarrhoea was present in 14 out of 25 (56%). Hospitalisation for rehydration therapy was necessary for 13 students (91% [10/11] of severe and 21% [3/14] of non-severe cases). Emergency room (ER) or family practitioner visits were necessary for 21 students and 13 (62%) of them initiated an antibiotic therapy. Two students indicated that other family members had gastroenteric symptoms more than 24 h after their return home, on Day 3, 4 and 5. All those involved in the outbreak recovered within 4 days, with an average duration of illness of 38.4 h. A case-control study of the risk associated with eating foods at lunch on Day 1 or with attending different restaurants at dinner on Day 1 was attempted. A total of 19 subjects filled the section of the online survey concerning food consumption. Fourteen of the responders were among the 22 cases falling ill within 36 h from the first shared meal (first wave), while the remaining five responders were three travellers with supposed secondary infections (showing symptoms with a delay ⩾48 h; second wave) and two asymptomatic students that were used as controls. No statistically significant associations with any food items served at lunch on Day 1 was identified (P < 0.05; Table 2). However, all food consumed on that occasion was to some extent associated with the risk of disease, the highest OR being obtained with chicken (OR: 20.7143, P: 0.0678; Table 2). Actually, most of the cases ate all foods included in the menu, while one of the two asymptomatic controls did not participate in the meal and the other had a pasta dish only. No statistically significant association was found for attending a particular fast food restaurant at the dinner on Day 1 (OR: 2.5, P: 0.3635). The Real-Time RT-PCR assay allowed to detect GII genogroup NoV genome in all the 25 samples from symptomatic subjects, although two of them showed high Ct values, indicating lower shedding of viral genomes (Table 1). The samples from four asymptomatic students were negative for both NoV genogroups. By sequence analysis, a NoV genotype could be defined for 20/ 25 NoV-positive stool samples. All NoV strains from travellers involved in the outbreaks belonged to the same genotype, GII.P2/GII.2, showing >99% nucleotide identities (nt id) in both fragments (ORF1 and ORF2) analysed. Thirty-two water samples were collected and analysed for bacterial contamination (Table 3). Seven samples were taken from the resort (site R) and the remaining 25 from five other sites (C, F, M, S, W), representative of the entire municipal water network. The majority (10/12) of the water samples collected immediately after the onset of the outbreak, between 21 May and 16 June 2016, at the resort (site R) and at sites F and C, located close to the resort, on the seaside and in the interior, respectively, were positive for coliform bacteria. More than 1000 colony-forming units (CFU)/ 100 ml (up to 2700 CFU/100 ml) were detected in four samples, while their absence is required in 100 ml by the Standard of Water Quality in Italy and seven of 11 samples tested were also Escherichia coli-positive (up to 900 CFU/100 ml). Enterococci were searched in 11 samples and three of them were found positive for this faecal contamination indicator. From 7 July to 16 August 2016, after control measures on the water network were undertaken, 13 water samples were collected for follow-up laboratory examination of the water system at F, R and C sites. All of them were negative for bacterial indicators of faecal contamination. No faecal bacteria were identified in the seven samples collected from M, W and S sites, including the well and the source that fed the water network, at any sampling time, before and after implementing the control measures. The research of NoV genome was conducted on concentrated water samples collected on 1 August from R and F sites (Table 3). NoV genome was detected in a single sample collected from the water entry of the resort involved in the outbreak (site R). Actually, only two out of 26 One-Step RT-PCR repetitions were positive for NoV viral capsid gene. The two ORF2-positive results were obtained at the University of Palermo Laboratory using both 1:10 diluted and undiluted extracted RNA. The sequence analysis of the partial ORF2 portion (289 bp) indicated the presence of the same GII.2 capsid genotype detected in the stool samples of the students involved in the outbreak, showing 98.9% nt id. No ORF1 NoV sequence could be obtained from the NoV-positive water sample. Follow-up analyses for NoV genome, conducted on water samples collected on 20 November both at R and F sites found no positive samples. By sequence alignment, the NoV sequences obtained from Mascali outbreak samples were phylogenetically closely related in both ORF1 and ORF2 (99 and 98.9% of nt id, respectively) to a GII.P2/GII.2 strain responsible for a sporadic paediatric gastroenteritis case occurred in Palermo in January 2016 (PA 36/16, accession numbers MG517464 and MG517460, for ORF1 and ORF2, respectively) and previously detected at the Enteric Viruses Laboratory, University of Palermo. Whereas, three GII.P2/GII.2 Italian strains (PA 280/11, PA 366/11 and PA 3/ 14, accession numbers MG517462, MG517465 and MG517463, for ORF1 and MG517457 to −59, for ORF2, respectively), detected in Palermo in 2011 and 2014, were more distantly associated with the outbreak strain both in RdRp and VP1 (90% to 97.4% nt id and 96.1% to 98.1% nt id, respectively, in ORF1 and ORF2). Discussion Epidemiological features, such as the presence of vomiting in more than half (88%) of symptomatic subjects, the 33.2 h average incubation period and the 38.4 h average duration of illness, suggested a NoV aetiology for the investigated outbreak [13]. The detection of NoV genome in all 25 faecal samples from symptomatic subjects fully complies with the requirement of detection of at least one positive sample out of four tested that is considered to be sufficient to ascertain the aetiology of a NoV outbreak [24]. Moreover, molecular typing indicated that a single viral strain, belonging to genotype GII.P2/GII.2, infected all symptomatic subjects involved in the outbreak. NoVs are highly infectious and can easily spread from person to person but also through contaminated food or water. Therefore, settings such as health care institutions, schools and touristic resorts are mainly affected, since close contact and shared food services facilitate transmission. In Mascali, two waves of transmission were observed over three days, being consistent with food/waterborne followed by person-to-person and/or airborne (vomit) transmission (Fig. 2). The first wave of cases presented on 20 May 2016 after a single shared meal, while the last case was detected on 22 May. Exposure to vomit may have been responsible for some of the secondary cases occurred after the first wave of illness. Vomiting produces aerosols that can be inhaled and swallowed. The extremely low infectious dose of NoV makes the inhalation of vomit aerosols an effective transmission route in settings where there is close contact, like on a bus trip [25]. All food consumed in the only shared lunch on 19 May at the resort was associated with the risk of disease (Table 2), but the association was inconclusive since all cases shared the same dishes included in the lunch menu on Day 1. Although many foods were potentially involved, our attention focused on the tap water. In fact, trivial information from local health authorities reported that local population avoided tap water consumption and relied on bottled water for drinking and cooking. Many small outbreaks of gastroenteritis involving restaurants and family groups had been reported in the same area, especially at the beginning of the summer season, when vacation houses and recreational activities were re-opened after a period of low water demand. Illness caused by NoV is common and if the level of water contamination is low, the number of cases remains limited and water can persist as unrecognised source of infection. If polluted water was served for drinking and used for cooking and washing dishes in the Mascali resort at Day 1 a generalised contamination of food and surfaces could have occurred. The potential role of the foodhandlers as a source of contamination could not be analysed in this study but should also be taken into account. If water contamination has already been present in the area for a long time, infections may have occurred among the local food-handlers. The analysis of water samples taken immediately after the outbreak revealed the presence of bacterial indicators of faecal contamination in the water used in the resort, with the only exception of a sample of filter purified water used both for cooking and as drinking water. When extending the sampling to other areas of the municipal water network, many other sites, mostly on the seaside, showed faecal pollution, which demonstrates a poor protection of the water network from faecal leakage (Fig. 1). NoV contamination of waters generally originates from sewage since these enteric pathogens, as other enteric viruses affecting humans, are mostly species-specific and are shed at high concentrations in the stools of patients. When the analyses on the stools of the students made it clear that the outbreak had been caused by NoV, the viral genome was also searched in appropriate volumes of water samples, but this happened when correction measures on the water network had already been taken. Nevertheless, a single water sample, obtained 2 months after the outbreak (1 August 2016), still contained genome fragments that could be correlated to the same NoV genotype causing the outbreak. Although the finding of NoV genome does not demonstrate that entire infecting viral particles were present in the water, it could indicate that a NoV belonging to the same genotype involved in the outbreak was still circulating in the local population several weeks after the outbreak and it was someway able to penetrate the water network, probably indicating continuing faecal contamination through leakage from sewage systems. NoV is particularly resistant in the environment and also to chlorine, therefore transmission through chlorinated water is not exceptional [9, 12, 26]. Drinking water from municipal supplies fed by lake or groundwater has already been identified in the past as the source of waterborne NoV outbreaks in Italy [27–29] and NoV is now considered one of the pathogens causing the largest number of affected consumers in drinking waterborne outbreaks [12]. NoV can persist in water longer than 15 days [30] and it is resistant to chlorine disinfection when free chlorine levels or pre-treatment processes are inadequate [31, 32]. Moreover, microfiltration filters labelled as certified by NSF Standards 53 or 58 (with a pore size ranging from 0.05 to 5 µm) are not effective in removing viruses and can even decrease chlorine concentration in filtered water [33]. Actually, the owner of the resort involved in our outbreak was encouraged to serve tap water as drinking water at the restaurant relying on the effectiveness of such a filter purifier. Unfortunately, we did not have a tap water sample from the restaurant of the resort taken before correction measures and whose volume was appropriate for virus detection. Therefore, we could not demonstrate the failure to eliminate viral pollution of the microfiltration system. In Mascali, correction measures on the water network consisted in the maintenance of the old chlorination device and in the addition of a new one at a key point located upstream and close to suspected illegal sewage dumping (Fig. 1), but as poor protection of the pipeline from sewage leakage risked to reproduce faecal pollution episodes, renovation of the municipal water system was also advised, as well as systematic control of illegal sewage dumping. The additional chlorination device was apparently the most effective intervention, since right after its activation faecal bacterial pollution indicators disappeared. However, NoV genome was detected in a water sample taken 6 weeks after the last detection of bacterial indicators. Unfortunately, detection of faecal contamination of drinking water is only based on the use of bacterial indicators despite no correlation between levels of enteric bacteria and enteric viruses has been found [34–36]. According to ISGEV enteric virus circulation surveillance data, in 2016 the circulation of GII.P2/GII.2 NoV strains in Palermo, home town of the travellers involved in the outbreak, was the highest ever observed from the starting of local NoV surveillance in 2004. GII.P2/GII.2 was the fourth most common genotype in 2016 (12.2% prevalence), following GII.P4/GII.4 Sydney 2012 (28.4%), GII.P16/GII.2 (18.9%) and GII.Pe/GII.4 Sydney 2012 (17.6%). GII.P2/GII.2 strains were mostly observed starting from July 2016, leaving space for the possibility that their increased circulation was linked to the introduction of the ‘Mascali clone’. However, a couple of GII.P2/GII.2 strains (including PA36/16) genetically close to the ‘Mascali clone’ had been already detected in January 2016 in Palermo and GII.P2/GII.2 NoV had sporadically been observed also in 2011 and 2014. Moreover, the circulation of GII.P2/GII.2 NoVs has been increasing worldwide over the last period [37]. Therefore, the unusually high circulation observed in Palermo might just reflect the global epidemiology of GII.P2/GII.2 strains. Noteworthy, the older GII.P2/GII.2 NoV strains detected in Palermo were closely related in the ORF2 (98.9% nt id) to the GII.P16/GII.2 Nashville strain (KY865307) which is supposed to have generated the recombinant GII.P16/GII.4 clone recently circulating epidemically worldwide [37, 38] and were more distantly associated (95.3% nt id) to the GII.2 capsid sequence (AB662868) of another putative GII.P16/GII.4 ancestor from Japan [39]. The increasing frequency of detection of waterborne NoV outbreaks calls for the introduction of routine monitoring for viruses of drinking waters and for the definition of affordable and reliable indicators for the presence of NoV in drinking waters. The use of ultrafiltration systems as a concentration method in order to recover viruses from environmental waters combined with RT-PCR to detect viral genomes has been demonstrated to grant a recovery rate of >50% of multiple viruses [40] and is at present one of the most sensitive approaches for the detection of NoVs, although the discrimination between viable and nonviable viruses remains a problem. The burden of NoV outbreaks, either food associated or waterborne, is still poorly defined due to the lack of specific diagnostic tools in many clinical laboratories. NoV epidemics should always be suspected when the main symptom is vomiting and the duration of disease is short (<48 h). Waterborne transmission is suggested by the rapid spread of disease and the high attack rate. Care should be taken to perform sensitive diagnostic tests on a sufficient number of stool samples in order to identify NoV as the causative agent of enteritis outbreaks, possibly preferring genome screening over antigen detection. Only the systematic detection of NoV outbreaks can guide in the choice of the safety procedures that could be most appropriate in containing such microbiological hazard.

What did the investigation find?

1565

Waterborne Norovirus outbreak at a seaside resort likely originating from municipal water distribution system failure

Abstract In May 2016 a Norovirus (NoV) gastroenteritis outbreak involved a high school class visiting a seaside resort near Taormina (Mascali, Sicily). Twenty-four students and a teacher were affected and 17 of them showed symptoms on the second day of the journey, while the others got ill within the following 2 days. Symptoms included vomiting, diarrhoea and fever, and 12 students required hospitalisation. Stool samples tested positive for NoV genome by Real-Time polymerase chain reaction assay in all 25 symptomatic subjects. The GII.P2/GII.2 NoV genotype was linked to the outbreak by ORF1/ORF2 sequence analysis. The epidemiological features of the outbreak were consistent with food/waterborne followed by person-to-person and/ or vomit transmission. Food consumed at a shared lunch on the first day of the trip was associated to illness and drinking un-bottled tap water was also considered as a risk factor. The analysis of water samples revealed the presence of bacterial indicators of faecal contamination in the water used in the resort as well as in other areas of the municipal water network, linking the NoV gastroenteritis outbreak to tap water pollution from sewage leakage. From a single water sample, an amplicon whose sequence corresponded to the capsid genotype recovered from patients could be obtained. Introduction Noroviruses (NoVs) are recognised as a leading cause of sporadic cases and outbreaks of acute gastroenteritis (AGE) across all age groups [1, 2]. They are transmitted primarily through the faecal-oral route either by direct person–person spread and with consumption of contaminated food or water [1, 3, 4]. NoVs are non-enveloped RNA viruses belonging to the Caliciviridae family and are classified into seven distinct genogroups (GI-GVII) on the basis of the major capsid protein VP1 sequence. GI, GII and GIV NoVs infect humans and they have been divided into more than 30 genotypes based on the sequences of polymerase (ORF1) and capsid proteins (ORF2) [5]. NoV GII.4 genotype has been associated with the vast majority of NoV related cases of AGE worldwide although many genotypes, belonging to both GI and GII genogroups are involved in foodborne and waterborne outbreaks [6, 7]. NoVs are highly contagious owing to their high stability in the environment and their ability to infect at low doses. In fact, just 10–100 virions of NoV may be sufficient to cause AGE [8, 9]. NoVs are considered the major cause of water-related disease in semi-closed institutions, including restaurants, schools, hospitals, nursing homes and cruise ships [10–12]. To confirm NoV as a cause of outbreaks, the presence of four epidemiological features of disease, vomiting in more than half of affected persons, a mean incubation period of 24/48 h, a mean duration of illness of 12–60 h and the absence of bacterial pathogens in stool culture, should be evaluated [13]. Waterborne NoV outbreaks can be caused by contaminated drinking water despite correct chlorine concentration and in the absence of coliform bacteria [10, 14–17]. The aim of this study was to investigate an outbreak of NoV gastroenteritis associated with drinking water involving a high school class travelling from Palermo (West Sicily) to Acireale (East Sicily) on an educational tour in May 2016. Stool samples were collected soon after the onset of AGE symptoms to ascertain the aetiology of illness. The source of infection was investigated collecting epidemiological data through an online questionnaire and, in order to explore a possible transmission route, laboratory tests on water from the municipal network were performed. Material and methods Epidemiological investigation From 20 to 22 May 2016 an AGE outbreak occurred in Mascali, a small town near Taormina, involving a school class of 28 students, 13 girls and 15 boys aged between 14 and 18 years and a 54-year-old female teacher, during a trip to participate in a theatre festival in Acireale. A total of 25/29 persons were affected by nausea, vomiting, abdominal cramps, diarrhoea and fever while they were staying at a seaside resort in Mascali, on the way to Acireale. To obtain data on symptoms and the correlated risk factors all participants filled an online questionnaire investigating personal data (age and sex), travel and accommodation details and food consumption over 48 h before symptoms presentation, time of onset and duration of symptoms, therapy and/or hospitalisation required and history of contact with people affected by diarrhoea and/or vomiting (Table 1). The severity of illness was analysed by the Vesikari severity scale allowing to distinguish non-severe (score <11) and severe (score ⩾11) AGE (Table 1). Vesikari score was calculated from maximum number of episodes/day and duration of both diarrhoea and vomiting (score 0–3 for each of the 4 parameters), temperature (score 0–3), level of dehydration (score 0–3) and treatment required (score 0–2) [18]. A case-control study was conducted to establish if food consumed before the onset of symptoms was associated with the disease. A case was defined as a student attending the tour and reporting diarrhoea and/or vomiting within 36 h from the first shared meal during the tour, while students who did not present any of the previous symptoms or those showing the first symptoms with a delay of ⩾48 h (secondary cases) were used as controls. Variables were studied using odds ratio (OR) and its 95% CI and statistical associations were determined using the χ2 test, with the level of statistical significance set at 0.05 (Table 2). NoV detection in stool samples A stool sample was obtained from each of the 29 travellers, soon after (3–5 days) the onset of the outbreak. No stool samples could be obtained from the food-handlers working at the resort. All samples were tested for NoV at the Enteric Viruses Laboratory of the University of Palermo, member of the Italian Study Group for Enteric Viruses (ISGEV), by a Real-Time polymerase chain reaction (RT-PCR) assay that allows to differentiate GI and GII NoVs [19]. Total RNA was extracted from stool samples using QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) and cDNA was generated by MMLV reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were considered positive for NoVs when cycle threshold (ct) of RT-PCR was ⩽30 and lowpositive when 30<ct⩽35. For all NoV-positive samples two genomic fragments were amplified by conventional RT-PCR using Superscript III One-Step RT-PCR system with Platinum Taq DNA polymerase (Invitrogen) and primers JV12a/JV13b, targeting 326 bps of the RdRp region A [20] and either COG1F/ GISKR or COG2F/GIISKR, amplifying a 315 bps 5′portion of GI or GII VP1 protein [19, 21]. Environmental investigation From May through August 2016, a total of 32 tap water samples were collected from the Mascali municipal network. In particular, seven samples were collected at the resort where the outbreak occurred and from its water entry (R), two in the city centre of Mascali (M), 15 in the village of Fondachello (F), on the seaside, three in Carrabba village (C), far from the coast, and two and three, respectively, from the water well ‘Carlino’ (W) and the source ‘Bufardo and Torregrossa’ (S) which feed the municipal system and provide drinking water to the Mascali area (Fig. 1, Table 3). One litre water samples were sent by the local health authorities to the SIALAB laboratory in Avola, Siracusa and 100 ml aliquots were screened for coliforms, E. coli and enterococci as faecal indicators, by standard bacteriological methods according to UNI EN ISO 9308-1:2014. Unfortunately, although the suspect of a water source for the viral outbreak was quickly communicated, only in August and November 2016, following the implementation of control measures, local authorities provided water samplings of appropriate volume to perform viral tests (10 l). Five water samples of 10 l each were collected from different sampling sites in F and R areas for viral screening (Table 3) and were concentrated according to the Italian National Institute of Health (Istituto Superiore di Sanità, ISS) protocols [22] through an ultra-filtration system (Sartoflow® Slice 200 Benchtop Crossflow System, Sartorius AG, Goettingen, Germany), using appropriate membranes (SG Hydrosart 10 kDa) pretreated with 300 ml of 3% Beef Extract (BE) at pH 7, by recirculating it for 10 min in the system, at a pressure not higher than 1.30 bar. Elution and recovery of viruses attached to the membranes were obtained by recirculating 200 ml of 3% BE at pH 9.5 for 3 h until a 10–12 ml volume of eluate was obtained. The pH of the eluate was then checked with litmus paper and if necessary brought to the value of 7 with HCl. Concentrated water samples were simultaneously tested for NoV in three laboratories: Enteric Viruses Laboratory, University of Palermo; Istituto Zooprofilattico Sperimentale della Sicilia (IZS), Palermo; and Istituto Superiore di Sanità (ISS), Rome. Total RNA was extracted from 500 and 560 µl of a concentrated water sample using BioMèrieux NucliSENS® Magnetic Extraction System (BioMèrieux) and QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) respectively, according to manufacturer’s instructions. Bonded RNA was eluted by 100 µl of Elution Buffer. Amplifications were performed at least in duplicate on whole and 1:10 diluted extracts. A Real-Time RT-PCR assay according to ISO/TS 15216- 2:2013, two in-house Real-Time RT-PCR protocols and the One-Step RT-PCR methods already used for amplifying ORF1 and ORF2 fragments in NoV-positive faecal samples were performed [19–21, 23]. Nucleotide sequences analysis and phylogenetic analysis PCR amplicons were purified by commercial columns (CleanSweep PCR Purification, Thermo Fisher Scientific, Vilnius, Lithuania) and sequenced by Sanger method and all nucleotide sequences were aligned using ClustalW. To identify NoV genotypes, RdRp and VP1 sequences were submitted to the RIVM NoV genotyping tool (http://www.rivm.nl/mpf/norovirus/typingtool). Phylogenetic trees were constructed using MEGA 7 and Neighbour-Joining method (bootstrap 1000 replicates). Sequences from stool and water samples are available at NCBI with the following accession numbers: MG517455 for the major capsid protein obtained from a water sample; MG517456 and MG517461, respectively, for the major capsid protein and RNA-dependent-RNA-polymerase, as representatives of the sequences obtained from the stool samples of those involved in the outbreak. Control measures After outbreak notification, on 21 May, the resort was closed for 2 days. On 25 May, following the first results of the bacteriological analyses on municipal water samples, the use of municipal water for human consumption was temporarily restricted (use of water was permitted only for cleaning purposes) in the Carrabba and Fondachello villages and in the area surrounding the resort. From 25 May the reservoir collecting the water from the Carlino well and its chlorination device were submitted to maintenance work and a new chlorination device was installed on the conduit bringing the water from the Bufardo source. Actually, the Bufardo source had been connected to the municipal network in 2003 but unfortunately a dedicated chlorination station had not been planned at that time. Therefore, unchlorinated water entered the system near Fondachello village on the seaside (Fig. 1), possibly diluting chlorine concentrations in water treated at the Carlino reservoir. Moreover, the exact path of the branch of the water network situated downstream the Bufardo source water connection was not marked in the maps of the municipality (Fig. 1, red line), indicating that this section had never been submitted to maintenance. The area surrounding the unmaintained section of the water pipeline was scattered with unauthorised cottages that were not connected to the sewerage systems. Likely, untreated waste from cesspools was unloaded directly into the soil, generating a risk of contamination of the municipal water system through leakage. No work had been made on the water pipelines before the outbreak but sections of the network had been temporarily closed during the winter season on the seaside (Fondachello). Local health authorities advised the connection of all households in the area to the sewerage system and the replacement of the water system section without maintenance. Health authorities also raised attention on AGE cases notification at all local Emergency Departments and Family Practitioners but there was no further evidence of AGE outbreaks in the months following the intervention. Results Data collected through the online questionnaire compiled 7–10 days after the outbreak revealed that none of the 28 students and the teacher taking part in the school trip showed gastroenteritis symptoms at the beginning of the journey and they had no contact with people who suffered from diarrhoea and/or vomiting in the previous days. Travellers had their first shared meal on 19 May (Day 1), 2016, at lunch on their arrival to the resort in Mascali. The list of foods included in the menu is shown in Table 2. Conversely, the dinner on Day 1 was not shared and the students went to dine in three different fast food restaurants in Acireale. On 20 May (Day 2), 17 students (58.6% of the travellers) presented with vomit and/or diarrhoea, falling ill 26–36 h after lunch on Day 1, strongly suggesting that exposure occurred on the first shared meal. Some students also vomited in the presence of the other travellers on the bus bringing them to Acireale. Five more students were symptomatic in the early hours of the following day (21 May, Day 3, from 0:30 am to 8:00 am), as a continuation of the wave of illnesses of the previous day. The teacher and one more student fell ill at 8:00 pm and 11:00 pm of Day 3, respectively, for a second wave of illnesses occurring >48 h from the first shared meal and <36 h from the first cases, with the last student falling sick on 22 May (Day 4) at 3:00 am (Fig. 2, Table 1). On the whole, the mean incubation period was 33.2 h. Based on the online survey, the severity of illness was evaluated by the Vesikari severity scale, grouping symptomatic travellers into severe (11/25 = 44%) and non-severe (14/25 = 56%) cases (Table 1). Vomiting and fever were the predominant symptoms, affecting 22 out of 25 (88%) symptomatic subjects, while diarrhoea was present in 14 out of 25 (56%). Hospitalisation for rehydration therapy was necessary for 13 students (91% [10/11] of severe and 21% [3/14] of non-severe cases). Emergency room (ER) or family practitioner visits were necessary for 21 students and 13 (62%) of them initiated an antibiotic therapy. Two students indicated that other family members had gastroenteric symptoms more than 24 h after their return home, on Day 3, 4 and 5. All those involved in the outbreak recovered within 4 days, with an average duration of illness of 38.4 h. A case-control study of the risk associated with eating foods at lunch on Day 1 or with attending different restaurants at dinner on Day 1 was attempted. A total of 19 subjects filled the section of the online survey concerning food consumption. Fourteen of the responders were among the 22 cases falling ill within 36 h from the first shared meal (first wave), while the remaining five responders were three travellers with supposed secondary infections (showing symptoms with a delay ⩾48 h; second wave) and two asymptomatic students that were used as controls. No statistically significant associations with any food items served at lunch on Day 1 was identified (P < 0.05; Table 2). However, all food consumed on that occasion was to some extent associated with the risk of disease, the highest OR being obtained with chicken (OR: 20.7143, P: 0.0678; Table 2). Actually, most of the cases ate all foods included in the menu, while one of the two asymptomatic controls did not participate in the meal and the other had a pasta dish only. No statistically significant association was found for attending a particular fast food restaurant at the dinner on Day 1 (OR: 2.5, P: 0.3635). The Real-Time RT-PCR assay allowed to detect GII genogroup NoV genome in all the 25 samples from symptomatic subjects, although two of them showed high Ct values, indicating lower shedding of viral genomes (Table 1). The samples from four asymptomatic students were negative for both NoV genogroups. By sequence analysis, a NoV genotype could be defined for 20/ 25 NoV-positive stool samples. All NoV strains from travellers involved in the outbreaks belonged to the same genotype, GII.P2/GII.2, showing >99% nucleotide identities (nt id) in both fragments (ORF1 and ORF2) analysed. Thirty-two water samples were collected and analysed for bacterial contamination (Table 3). Seven samples were taken from the resort (site R) and the remaining 25 from five other sites (C, F, M, S, W), representative of the entire municipal water network. The majority (10/12) of the water samples collected immediately after the onset of the outbreak, between 21 May and 16 June 2016, at the resort (site R) and at sites F and C, located close to the resort, on the seaside and in the interior, respectively, were positive for coliform bacteria. More than 1000 colony-forming units (CFU)/ 100 ml (up to 2700 CFU/100 ml) were detected in four samples, while their absence is required in 100 ml by the Standard of Water Quality in Italy and seven of 11 samples tested were also Escherichia coli-positive (up to 900 CFU/100 ml). Enterococci were searched in 11 samples and three of them were found positive for this faecal contamination indicator. From 7 July to 16 August 2016, after control measures on the water network were undertaken, 13 water samples were collected for follow-up laboratory examination of the water system at F, R and C sites. All of them were negative for bacterial indicators of faecal contamination. No faecal bacteria were identified in the seven samples collected from M, W and S sites, including the well and the source that fed the water network, at any sampling time, before and after implementing the control measures. The research of NoV genome was conducted on concentrated water samples collected on 1 August from R and F sites (Table 3). NoV genome was detected in a single sample collected from the water entry of the resort involved in the outbreak (site R). Actually, only two out of 26 One-Step RT-PCR repetitions were positive for NoV viral capsid gene. The two ORF2-positive results were obtained at the University of Palermo Laboratory using both 1:10 diluted and undiluted extracted RNA. The sequence analysis of the partial ORF2 portion (289 bp) indicated the presence of the same GII.2 capsid genotype detected in the stool samples of the students involved in the outbreak, showing 98.9% nt id. No ORF1 NoV sequence could be obtained from the NoV-positive water sample. Follow-up analyses for NoV genome, conducted on water samples collected on 20 November both at R and F sites found no positive samples. By sequence alignment, the NoV sequences obtained from Mascali outbreak samples were phylogenetically closely related in both ORF1 and ORF2 (99 and 98.9% of nt id, respectively) to a GII.P2/GII.2 strain responsible for a sporadic paediatric gastroenteritis case occurred in Palermo in January 2016 (PA 36/16, accession numbers MG517464 and MG517460, for ORF1 and ORF2, respectively) and previously detected at the Enteric Viruses Laboratory, University of Palermo. Whereas, three GII.P2/GII.2 Italian strains (PA 280/11, PA 366/11 and PA 3/ 14, accession numbers MG517462, MG517465 and MG517463, for ORF1 and MG517457 to −59, for ORF2, respectively), detected in Palermo in 2011 and 2014, were more distantly associated with the outbreak strain both in RdRp and VP1 (90% to 97.4% nt id and 96.1% to 98.1% nt id, respectively, in ORF1 and ORF2). Discussion Epidemiological features, such as the presence of vomiting in more than half (88%) of symptomatic subjects, the 33.2 h average incubation period and the 38.4 h average duration of illness, suggested a NoV aetiology for the investigated outbreak [13]. The detection of NoV genome in all 25 faecal samples from symptomatic subjects fully complies with the requirement of detection of at least one positive sample out of four tested that is considered to be sufficient to ascertain the aetiology of a NoV outbreak [24]. Moreover, molecular typing indicated that a single viral strain, belonging to genotype GII.P2/GII.2, infected all symptomatic subjects involved in the outbreak. NoVs are highly infectious and can easily spread from person to person but also through contaminated food or water. Therefore, settings such as health care institutions, schools and touristic resorts are mainly affected, since close contact and shared food services facilitate transmission. In Mascali, two waves of transmission were observed over three days, being consistent with food/waterborne followed by person-to-person and/or airborne (vomit) transmission (Fig. 2). The first wave of cases presented on 20 May 2016 after a single shared meal, while the last case was detected on 22 May. Exposure to vomit may have been responsible for some of the secondary cases occurred after the first wave of illness. Vomiting produces aerosols that can be inhaled and swallowed. The extremely low infectious dose of NoV makes the inhalation of vomit aerosols an effective transmission route in settings where there is close contact, like on a bus trip [25]. All food consumed in the only shared lunch on 19 May at the resort was associated with the risk of disease (Table 2), but the association was inconclusive since all cases shared the same dishes included in the lunch menu on Day 1. Although many foods were potentially involved, our attention focused on the tap water. In fact, trivial information from local health authorities reported that local population avoided tap water consumption and relied on bottled water for drinking and cooking. Many small outbreaks of gastroenteritis involving restaurants and family groups had been reported in the same area, especially at the beginning of the summer season, when vacation houses and recreational activities were re-opened after a period of low water demand. Illness caused by NoV is common and if the level of water contamination is low, the number of cases remains limited and water can persist as unrecognised source of infection. If polluted water was served for drinking and used for cooking and washing dishes in the Mascali resort at Day 1 a generalised contamination of food and surfaces could have occurred. The potential role of the foodhandlers as a source of contamination could not be analysed in this study but should also be taken into account. If water contamination has already been present in the area for a long time, infections may have occurred among the local food-handlers. The analysis of water samples taken immediately after the outbreak revealed the presence of bacterial indicators of faecal contamination in the water used in the resort, with the only exception of a sample of filter purified water used both for cooking and as drinking water. When extending the sampling to other areas of the municipal water network, many other sites, mostly on the seaside, showed faecal pollution, which demonstrates a poor protection of the water network from faecal leakage (Fig. 1). NoV contamination of waters generally originates from sewage since these enteric pathogens, as other enteric viruses affecting humans, are mostly species-specific and are shed at high concentrations in the stools of patients. When the analyses on the stools of the students made it clear that the outbreak had been caused by NoV, the viral genome was also searched in appropriate volumes of water samples, but this happened when correction measures on the water network had already been taken. Nevertheless, a single water sample, obtained 2 months after the outbreak (1 August 2016), still contained genome fragments that could be correlated to the same NoV genotype causing the outbreak. Although the finding of NoV genome does not demonstrate that entire infecting viral particles were present in the water, it could indicate that a NoV belonging to the same genotype involved in the outbreak was still circulating in the local population several weeks after the outbreak and it was someway able to penetrate the water network, probably indicating continuing faecal contamination through leakage from sewage systems. NoV is particularly resistant in the environment and also to chlorine, therefore transmission through chlorinated water is not exceptional [9, 12, 26]. Drinking water from municipal supplies fed by lake or groundwater has already been identified in the past as the source of waterborne NoV outbreaks in Italy [27–29] and NoV is now considered one of the pathogens causing the largest number of affected consumers in drinking waterborne outbreaks [12]. NoV can persist in water longer than 15 days [30] and it is resistant to chlorine disinfection when free chlorine levels or pre-treatment processes are inadequate [31, 32]. Moreover, microfiltration filters labelled as certified by NSF Standards 53 or 58 (with a pore size ranging from 0.05 to 5 µm) are not effective in removing viruses and can even decrease chlorine concentration in filtered water [33]. Actually, the owner of the resort involved in our outbreak was encouraged to serve tap water as drinking water at the restaurant relying on the effectiveness of such a filter purifier. Unfortunately, we did not have a tap water sample from the restaurant of the resort taken before correction measures and whose volume was appropriate for virus detection. Therefore, we could not demonstrate the failure to eliminate viral pollution of the microfiltration system. In Mascali, correction measures on the water network consisted in the maintenance of the old chlorination device and in the addition of a new one at a key point located upstream and close to suspected illegal sewage dumping (Fig. 1), but as poor protection of the pipeline from sewage leakage risked to reproduce faecal pollution episodes, renovation of the municipal water system was also advised, as well as systematic control of illegal sewage dumping. The additional chlorination device was apparently the most effective intervention, since right after its activation faecal bacterial pollution indicators disappeared. However, NoV genome was detected in a water sample taken 6 weeks after the last detection of bacterial indicators. Unfortunately, detection of faecal contamination of drinking water is only based on the use of bacterial indicators despite no correlation between levels of enteric bacteria and enteric viruses has been found [34–36]. According to ISGEV enteric virus circulation surveillance data, in 2016 the circulation of GII.P2/GII.2 NoV strains in Palermo, home town of the travellers involved in the outbreak, was the highest ever observed from the starting of local NoV surveillance in 2004. GII.P2/GII.2 was the fourth most common genotype in 2016 (12.2% prevalence), following GII.P4/GII.4 Sydney 2012 (28.4%), GII.P16/GII.2 (18.9%) and GII.Pe/GII.4 Sydney 2012 (17.6%). GII.P2/GII.2 strains were mostly observed starting from July 2016, leaving space for the possibility that their increased circulation was linked to the introduction of the ‘Mascali clone’. However, a couple of GII.P2/GII.2 strains (including PA36/16) genetically close to the ‘Mascali clone’ had been already detected in January 2016 in Palermo and GII.P2/GII.2 NoV had sporadically been observed also in 2011 and 2014. Moreover, the circulation of GII.P2/GII.2 NoVs has been increasing worldwide over the last period [37]. Therefore, the unusually high circulation observed in Palermo might just reflect the global epidemiology of GII.P2/GII.2 strains. Noteworthy, the older GII.P2/GII.2 NoV strains detected in Palermo were closely related in the ORF2 (98.9% nt id) to the GII.P16/GII.2 Nashville strain (KY865307) which is supposed to have generated the recombinant GII.P16/GII.4 clone recently circulating epidemically worldwide [37, 38] and were more distantly associated (95.3% nt id) to the GII.2 capsid sequence (AB662868) of another putative GII.P16/GII.4 ancestor from Japan [39]. The increasing frequency of detection of waterborne NoV outbreaks calls for the introduction of routine monitoring for viruses of drinking waters and for the definition of affordable and reliable indicators for the presence of NoV in drinking waters. The use of ultrafiltration systems as a concentration method in order to recover viruses from environmental waters combined with RT-PCR to detect viral genomes has been demonstrated to grant a recovery rate of >50% of multiple viruses [40] and is at present one of the most sensitive approaches for the detection of NoVs, although the discrimination between viable and nonviable viruses remains a problem. The burden of NoV outbreaks, either food associated or waterborne, is still poorly defined due to the lack of specific diagnostic tools in many clinical laboratories. NoV epidemics should always be suspected when the main symptom is vomiting and the duration of disease is short (<48 h). Waterborne transmission is suggested by the rapid spread of disease and the high attack rate. Care should be taken to perform sensitive diagnostic tests on a sufficient number of stool samples in order to identify NoV as the causative agent of enteritis outbreaks, possibly preferring genome screening over antigen detection. Only the systematic detection of NoV outbreaks can guide in the choice of the safety procedures that could be most appropriate in containing such microbiological hazard.

How was the infrastructure affected?

1566

Waterborne Norovirus outbreak at a seaside resort likely originating from municipal water distribution system failure

Abstract In May 2016 a Norovirus (NoV) gastroenteritis outbreak involved a high school class visiting a seaside resort near Taormina (Mascali, Sicily). Twenty-four students and a teacher were affected and 17 of them showed symptoms on the second day of the journey, while the others got ill within the following 2 days. Symptoms included vomiting, diarrhoea and fever, and 12 students required hospitalisation. Stool samples tested positive for NoV genome by Real-Time polymerase chain reaction assay in all 25 symptomatic subjects. The GII.P2/GII.2 NoV genotype was linked to the outbreak by ORF1/ORF2 sequence analysis. The epidemiological features of the outbreak were consistent with food/waterborne followed by person-to-person and/ or vomit transmission. Food consumed at a shared lunch on the first day of the trip was associated to illness and drinking un-bottled tap water was also considered as a risk factor. The analysis of water samples revealed the presence of bacterial indicators of faecal contamination in the water used in the resort as well as in other areas of the municipal water network, linking the NoV gastroenteritis outbreak to tap water pollution from sewage leakage. From a single water sample, an amplicon whose sequence corresponded to the capsid genotype recovered from patients could be obtained. Introduction Noroviruses (NoVs) are recognised as a leading cause of sporadic cases and outbreaks of acute gastroenteritis (AGE) across all age groups [1, 2]. They are transmitted primarily through the faecal-oral route either by direct person–person spread and with consumption of contaminated food or water [1, 3, 4]. NoVs are non-enveloped RNA viruses belonging to the Caliciviridae family and are classified into seven distinct genogroups (GI-GVII) on the basis of the major capsid protein VP1 sequence. GI, GII and GIV NoVs infect humans and they have been divided into more than 30 genotypes based on the sequences of polymerase (ORF1) and capsid proteins (ORF2) [5]. NoV GII.4 genotype has been associated with the vast majority of NoV related cases of AGE worldwide although many genotypes, belonging to both GI and GII genogroups are involved in foodborne and waterborne outbreaks [6, 7]. NoVs are highly contagious owing to their high stability in the environment and their ability to infect at low doses. In fact, just 10–100 virions of NoV may be sufficient to cause AGE [8, 9]. NoVs are considered the major cause of water-related disease in semi-closed institutions, including restaurants, schools, hospitals, nursing homes and cruise ships [10–12]. To confirm NoV as a cause of outbreaks, the presence of four epidemiological features of disease, vomiting in more than half of affected persons, a mean incubation period of 24/48 h, a mean duration of illness of 12–60 h and the absence of bacterial pathogens in stool culture, should be evaluated [13]. Waterborne NoV outbreaks can be caused by contaminated drinking water despite correct chlorine concentration and in the absence of coliform bacteria [10, 14–17]. The aim of this study was to investigate an outbreak of NoV gastroenteritis associated with drinking water involving a high school class travelling from Palermo (West Sicily) to Acireale (East Sicily) on an educational tour in May 2016. Stool samples were collected soon after the onset of AGE symptoms to ascertain the aetiology of illness. The source of infection was investigated collecting epidemiological data through an online questionnaire and, in order to explore a possible transmission route, laboratory tests on water from the municipal network were performed. Material and methods Epidemiological investigation From 20 to 22 May 2016 an AGE outbreak occurred in Mascali, a small town near Taormina, involving a school class of 28 students, 13 girls and 15 boys aged between 14 and 18 years and a 54-year-old female teacher, during a trip to participate in a theatre festival in Acireale. A total of 25/29 persons were affected by nausea, vomiting, abdominal cramps, diarrhoea and fever while they were staying at a seaside resort in Mascali, on the way to Acireale. To obtain data on symptoms and the correlated risk factors all participants filled an online questionnaire investigating personal data (age and sex), travel and accommodation details and food consumption over 48 h before symptoms presentation, time of onset and duration of symptoms, therapy and/or hospitalisation required and history of contact with people affected by diarrhoea and/or vomiting (Table 1). The severity of illness was analysed by the Vesikari severity scale allowing to distinguish non-severe (score <11) and severe (score ⩾11) AGE (Table 1). Vesikari score was calculated from maximum number of episodes/day and duration of both diarrhoea and vomiting (score 0–3 for each of the 4 parameters), temperature (score 0–3), level of dehydration (score 0–3) and treatment required (score 0–2) [18]. A case-control study was conducted to establish if food consumed before the onset of symptoms was associated with the disease. A case was defined as a student attending the tour and reporting diarrhoea and/or vomiting within 36 h from the first shared meal during the tour, while students who did not present any of the previous symptoms or those showing the first symptoms with a delay of ⩾48 h (secondary cases) were used as controls. Variables were studied using odds ratio (OR) and its 95% CI and statistical associations were determined using the χ2 test, with the level of statistical significance set at 0.05 (Table 2). NoV detection in stool samples A stool sample was obtained from each of the 29 travellers, soon after (3–5 days) the onset of the outbreak. No stool samples could be obtained from the food-handlers working at the resort. All samples were tested for NoV at the Enteric Viruses Laboratory of the University of Palermo, member of the Italian Study Group for Enteric Viruses (ISGEV), by a Real-Time polymerase chain reaction (RT-PCR) assay that allows to differentiate GI and GII NoVs [19]. Total RNA was extracted from stool samples using QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) and cDNA was generated by MMLV reverse transcriptase (Invitrogen, Carlsbad, CA). Samples were considered positive for NoVs when cycle threshold (ct) of RT-PCR was ⩽30 and lowpositive when 30<ct⩽35. For all NoV-positive samples two genomic fragments were amplified by conventional RT-PCR using Superscript III One-Step RT-PCR system with Platinum Taq DNA polymerase (Invitrogen) and primers JV12a/JV13b, targeting 326 bps of the RdRp region A [20] and either COG1F/ GISKR or COG2F/GIISKR, amplifying a 315 bps 5′portion of GI or GII VP1 protein [19, 21]. Environmental investigation From May through August 2016, a total of 32 tap water samples were collected from the Mascali municipal network. In particular, seven samples were collected at the resort where the outbreak occurred and from its water entry (R), two in the city centre of Mascali (M), 15 in the village of Fondachello (F), on the seaside, three in Carrabba village (C), far from the coast, and two and three, respectively, from the water well ‘Carlino’ (W) and the source ‘Bufardo and Torregrossa’ (S) which feed the municipal system and provide drinking water to the Mascali area (Fig. 1, Table 3). One litre water samples were sent by the local health authorities to the SIALAB laboratory in Avola, Siracusa and 100 ml aliquots were screened for coliforms, E. coli and enterococci as faecal indicators, by standard bacteriological methods according to UNI EN ISO 9308-1:2014. Unfortunately, although the suspect of a water source for the viral outbreak was quickly communicated, only in August and November 2016, following the implementation of control measures, local authorities provided water samplings of appropriate volume to perform viral tests (10 l). Five water samples of 10 l each were collected from different sampling sites in F and R areas for viral screening (Table 3) and were concentrated according to the Italian National Institute of Health (Istituto Superiore di Sanità, ISS) protocols [22] through an ultra-filtration system (Sartoflow® Slice 200 Benchtop Crossflow System, Sartorius AG, Goettingen, Germany), using appropriate membranes (SG Hydrosart 10 kDa) pretreated with 300 ml of 3% Beef Extract (BE) at pH 7, by recirculating it for 10 min in the system, at a pressure not higher than 1.30 bar. Elution and recovery of viruses attached to the membranes were obtained by recirculating 200 ml of 3% BE at pH 9.5 for 3 h until a 10–12 ml volume of eluate was obtained. The pH of the eluate was then checked with litmus paper and if necessary brought to the value of 7 with HCl. Concentrated water samples were simultaneously tested for NoV in three laboratories: Enteric Viruses Laboratory, University of Palermo; Istituto Zooprofilattico Sperimentale della Sicilia (IZS), Palermo; and Istituto Superiore di Sanità (ISS), Rome. Total RNA was extracted from 500 and 560 µl of a concentrated water sample using BioMèrieux NucliSENS® Magnetic Extraction System (BioMèrieux) and QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) respectively, according to manufacturer’s instructions. Bonded RNA was eluted by 100 µl of Elution Buffer. Amplifications were performed at least in duplicate on whole and 1:10 diluted extracts. A Real-Time RT-PCR assay according to ISO/TS 15216- 2:2013, two in-house Real-Time RT-PCR protocols and the One-Step RT-PCR methods already used for amplifying ORF1 and ORF2 fragments in NoV-positive faecal samples were performed [19–21, 23]. Nucleotide sequences analysis and phylogenetic analysis PCR amplicons were purified by commercial columns (CleanSweep PCR Purification, Thermo Fisher Scientific, Vilnius, Lithuania) and sequenced by Sanger method and all nucleotide sequences were aligned using ClustalW. To identify NoV genotypes, RdRp and VP1 sequences were submitted to the RIVM NoV genotyping tool (http://www.rivm.nl/mpf/norovirus/typingtool). Phylogenetic trees were constructed using MEGA 7 and Neighbour-Joining method (bootstrap 1000 replicates). Sequences from stool and water samples are available at NCBI with the following accession numbers: MG517455 for the major capsid protein obtained from a water sample; MG517456 and MG517461, respectively, for the major capsid protein and RNA-dependent-RNA-polymerase, as representatives of the sequences obtained from the stool samples of those involved in the outbreak. Control measures After outbreak notification, on 21 May, the resort was closed for 2 days. On 25 May, following the first results of the bacteriological analyses on municipal water samples, the use of municipal water for human consumption was temporarily restricted (use of water was permitted only for cleaning purposes) in the Carrabba and Fondachello villages and in the area surrounding the resort. From 25 May the reservoir collecting the water from the Carlino well and its chlorination device were submitted to maintenance work and a new chlorination device was installed on the conduit bringing the water from the Bufardo source. Actually, the Bufardo source had been connected to the municipal network in 2003 but unfortunately a dedicated chlorination station had not been planned at that time. Therefore, unchlorinated water entered the system near Fondachello village on the seaside (Fig. 1), possibly diluting chlorine concentrations in water treated at the Carlino reservoir. Moreover, the exact path of the branch of the water network situated downstream the Bufardo source water connection was not marked in the maps of the municipality (Fig. 1, red line), indicating that this section had never been submitted to maintenance. The area surrounding the unmaintained section of the water pipeline was scattered with unauthorised cottages that were not connected to the sewerage systems. Likely, untreated waste from cesspools was unloaded directly into the soil, generating a risk of contamination of the municipal water system through leakage. No work had been made on the water pipelines before the outbreak but sections of the network had been temporarily closed during the winter season on the seaside (Fondachello). Local health authorities advised the connection of all households in the area to the sewerage system and the replacement of the water system section without maintenance. Health authorities also raised attention on AGE cases notification at all local Emergency Departments and Family Practitioners but there was no further evidence of AGE outbreaks in the months following the intervention. Results Data collected through the online questionnaire compiled 7–10 days after the outbreak revealed that none of the 28 students and the teacher taking part in the school trip showed gastroenteritis symptoms at the beginning of the journey and they had no contact with people who suffered from diarrhoea and/or vomiting in the previous days. Travellers had their first shared meal on 19 May (Day 1), 2016, at lunch on their arrival to the resort in Mascali. The list of foods included in the menu is shown in Table 2. Conversely, the dinner on Day 1 was not shared and the students went to dine in three different fast food restaurants in Acireale. On 20 May (Day 2), 17 students (58.6% of the travellers) presented with vomit and/or diarrhoea, falling ill 26–36 h after lunch on Day 1, strongly suggesting that exposure occurred on the first shared meal. Some students also vomited in the presence of the other travellers on the bus bringing them to Acireale. Five more students were symptomatic in the early hours of the following day (21 May, Day 3, from 0:30 am to 8:00 am), as a continuation of the wave of illnesses of the previous day. The teacher and one more student fell ill at 8:00 pm and 11:00 pm of Day 3, respectively, for a second wave of illnesses occurring >48 h from the first shared meal and <36 h from the first cases, with the last student falling sick on 22 May (Day 4) at 3:00 am (Fig. 2, Table 1). On the whole, the mean incubation period was 33.2 h. Based on the online survey, the severity of illness was evaluated by the Vesikari severity scale, grouping symptomatic travellers into severe (11/25 = 44%) and non-severe (14/25 = 56%) cases (Table 1). Vomiting and fever were the predominant symptoms, affecting 22 out of 25 (88%) symptomatic subjects, while diarrhoea was present in 14 out of 25 (56%). Hospitalisation for rehydration therapy was necessary for 13 students (91% [10/11] of severe and 21% [3/14] of non-severe cases). Emergency room (ER) or family practitioner visits were necessary for 21 students and 13 (62%) of them initiated an antibiotic therapy. Two students indicated that other family members had gastroenteric symptoms more than 24 h after their return home, on Day 3, 4 and 5. All those involved in the outbreak recovered within 4 days, with an average duration of illness of 38.4 h. A case-control study of the risk associated with eating foods at lunch on Day 1 or with attending different restaurants at dinner on Day 1 was attempted. A total of 19 subjects filled the section of the online survey concerning food consumption. Fourteen of the responders were among the 22 cases falling ill within 36 h from the first shared meal (first wave), while the remaining five responders were three travellers with supposed secondary infections (showing symptoms with a delay ⩾48 h; second wave) and two asymptomatic students that were used as controls. No statistically significant associations with any food items served at lunch on Day 1 was identified (P < 0.05; Table 2). However, all food consumed on that occasion was to some extent associated with the risk of disease, the highest OR being obtained with chicken (OR: 20.7143, P: 0.0678; Table 2). Actually, most of the cases ate all foods included in the menu, while one of the two asymptomatic controls did not participate in the meal and the other had a pasta dish only. No statistically significant association was found for attending a particular fast food restaurant at the dinner on Day 1 (OR: 2.5, P: 0.3635). The Real-Time RT-PCR assay allowed to detect GII genogroup NoV genome in all the 25 samples from symptomatic subjects, although two of them showed high Ct values, indicating lower shedding of viral genomes (Table 1). The samples from four asymptomatic students were negative for both NoV genogroups. By sequence analysis, a NoV genotype could be defined for 20/ 25 NoV-positive stool samples. All NoV strains from travellers involved in the outbreaks belonged to the same genotype, GII.P2/GII.2, showing >99% nucleotide identities (nt id) in both fragments (ORF1 and ORF2) analysed. Thirty-two water samples were collected and analysed for bacterial contamination (Table 3). Seven samples were taken from the resort (site R) and the remaining 25 from five other sites (C, F, M, S, W), representative of the entire municipal water network. The majority (10/12) of the water samples collected immediately after the onset of the outbreak, between 21 May and 16 June 2016, at the resort (site R) and at sites F and C, located close to the resort, on the seaside and in the interior, respectively, were positive for coliform bacteria. More than 1000 colony-forming units (CFU)/ 100 ml (up to 2700 CFU/100 ml) were detected in four samples, while their absence is required in 100 ml by the Standard of Water Quality in Italy and seven of 11 samples tested were also Escherichia coli-positive (up to 900 CFU/100 ml). Enterococci were searched in 11 samples and three of them were found positive for this faecal contamination indicator. From 7 July to 16 August 2016, after control measures on the water network were undertaken, 13 water samples were collected for follow-up laboratory examination of the water system at F, R and C sites. All of them were negative for bacterial indicators of faecal contamination. No faecal bacteria were identified in the seven samples collected from M, W and S sites, including the well and the source that fed the water network, at any sampling time, before and after implementing the control measures. The research of NoV genome was conducted on concentrated water samples collected on 1 August from R and F sites (Table 3). NoV genome was detected in a single sample collected from the water entry of the resort involved in the outbreak (site R). Actually, only two out of 26 One-Step RT-PCR repetitions were positive for NoV viral capsid gene. The two ORF2-positive results were obtained at the University of Palermo Laboratory using both 1:10 diluted and undiluted extracted RNA. The sequence analysis of the partial ORF2 portion (289 bp) indicated the presence of the same GII.2 capsid genotype detected in the stool samples of the students involved in the outbreak, showing 98.9% nt id. No ORF1 NoV sequence could be obtained from the NoV-positive water sample. Follow-up analyses for NoV genome, conducted on water samples collected on 20 November both at R and F sites found no positive samples. By sequence alignment, the NoV sequences obtained from Mascali outbreak samples were phylogenetically closely related in both ORF1 and ORF2 (99 and 98.9% of nt id, respectively) to a GII.P2/GII.2 strain responsible for a sporadic paediatric gastroenteritis case occurred in Palermo in January 2016 (PA 36/16, accession numbers MG517464 and MG517460, for ORF1 and ORF2, respectively) and previously detected at the Enteric Viruses Laboratory, University of Palermo. Whereas, three GII.P2/GII.2 Italian strains (PA 280/11, PA 366/11 and PA 3/ 14, accession numbers MG517462, MG517465 and MG517463, for ORF1 and MG517457 to −59, for ORF2, respectively), detected in Palermo in 2011 and 2014, were more distantly associated with the outbreak strain both in RdRp and VP1 (90% to 97.4% nt id and 96.1% to 98.1% nt id, respectively, in ORF1 and ORF2). Discussion Epidemiological features, such as the presence of vomiting in more than half (88%) of symptomatic subjects, the 33.2 h average incubation period and the 38.4 h average duration of illness, suggested a NoV aetiology for the investigated outbreak [13]. The detection of NoV genome in all 25 faecal samples from symptomatic subjects fully complies with the requirement of detection of at least one positive sample out of four tested that is considered to be sufficient to ascertain the aetiology of a NoV outbreak [24]. Moreover, molecular typing indicated that a single viral strain, belonging to genotype GII.P2/GII.2, infected all symptomatic subjects involved in the outbreak. NoVs are highly infectious and can easily spread from person to person but also through contaminated food or water. Therefore, settings such as health care institutions, schools and touristic resorts are mainly affected, since close contact and shared food services facilitate transmission. In Mascali, two waves of transmission were observed over three days, being consistent with food/waterborne followed by person-to-person and/or airborne (vomit) transmission (Fig. 2). The first wave of cases presented on 20 May 2016 after a single shared meal, while the last case was detected on 22 May. Exposure to vomit may have been responsible for some of the secondary cases occurred after the first wave of illness. Vomiting produces aerosols that can be inhaled and swallowed. The extremely low infectious dose of NoV makes the inhalation of vomit aerosols an effective transmission route in settings where there is close contact, like on a bus trip [25]. All food consumed in the only shared lunch on 19 May at the resort was associated with the risk of disease (Table 2), but the association was inconclusive since all cases shared the same dishes included in the lunch menu on Day 1. Although many foods were potentially involved, our attention focused on the tap water. In fact, trivial information from local health authorities reported that local population avoided tap water consumption and relied on bottled water for drinking and cooking. Many small outbreaks of gastroenteritis involving restaurants and family groups had been reported in the same area, especially at the beginning of the summer season, when vacation houses and recreational activities were re-opened after a period of low water demand. Illness caused by NoV is common and if the level of water contamination is low, the number of cases remains limited and water can persist as unrecognised source of infection. If polluted water was served for drinking and used for cooking and washing dishes in the Mascali resort at Day 1 a generalised contamination of food and surfaces could have occurred. The potential role of the foodhandlers as a source of contamination could not be analysed in this study but should also be taken into account. If water contamination has already been present in the area for a long time, infections may have occurred among the local food-handlers. The analysis of water samples taken immediately after the outbreak revealed the presence of bacterial indicators of faecal contamination in the water used in the resort, with the only exception of a sample of filter purified water used both for cooking and as drinking water. When extending the sampling to other areas of the municipal water network, many other sites, mostly on the seaside, showed faecal pollution, which demonstrates a poor protection of the water network from faecal leakage (Fig. 1). NoV contamination of waters generally originates from sewage since these enteric pathogens, as other enteric viruses affecting humans, are mostly species-specific and are shed at high concentrations in the stools of patients. When the analyses on the stools of the students made it clear that the outbreak had been caused by NoV, the viral genome was also searched in appropriate volumes of water samples, but this happened when correction measures on the water network had already been taken. Nevertheless, a single water sample, obtained 2 months after the outbreak (1 August 2016), still contained genome fragments that could be correlated to the same NoV genotype causing the outbreak. Although the finding of NoV genome does not demonstrate that entire infecting viral particles were present in the water, it could indicate that a NoV belonging to the same genotype involved in the outbreak was still circulating in the local population several weeks after the outbreak and it was someway able to penetrate the water network, probably indicating continuing faecal contamination through leakage from sewage systems. NoV is particularly resistant in the environment and also to chlorine, therefore transmission through chlorinated water is not exceptional [9, 12, 26]. Drinking water from municipal supplies fed by lake or groundwater has already been identified in the past as the source of waterborne NoV outbreaks in Italy [27–29] and NoV is now considered one of the pathogens causing the largest number of affected consumers in drinking waterborne outbreaks [12]. NoV can persist in water longer than 15 days [30] and it is resistant to chlorine disinfection when free chlorine levels or pre-treatment processes are inadequate [31, 32]. Moreover, microfiltration filters labelled as certified by NSF Standards 53 or 58 (with a pore size ranging from 0.05 to 5 µm) are not effective in removing viruses and can even decrease chlorine concentration in filtered water [33]. Actually, the owner of the resort involved in our outbreak was encouraged to serve tap water as drinking water at the restaurant relying on the effectiveness of such a filter purifier. Unfortunately, we did not have a tap water sample from the restaurant of the resort taken before correction measures and whose volume was appropriate for virus detection. Therefore, we could not demonstrate the failure to eliminate viral pollution of the microfiltration system. In Mascali, correction measures on the water network consisted in the maintenance of the old chlorination device and in the addition of a new one at a key point located upstream and close to suspected illegal sewage dumping (Fig. 1), but as poor protection of the pipeline from sewage leakage risked to reproduce faecal pollution episodes, renovation of the municipal water system was also advised, as well as systematic control of illegal sewage dumping. The additional chlorination device was apparently the most effective intervention, since right after its activation faecal bacterial pollution indicators disappeared. However, NoV genome was detected in a water sample taken 6 weeks after the last detection of bacterial indicators. Unfortunately, detection of faecal contamination of drinking water is only based on the use of bacterial indicators despite no correlation between levels of enteric bacteria and enteric viruses has been found [34–36]. According to ISGEV enteric virus circulation surveillance data, in 2016 the circulation of GII.P2/GII.2 NoV strains in Palermo, home town of the travellers involved in the outbreak, was the highest ever observed from the starting of local NoV surveillance in 2004. GII.P2/GII.2 was the fourth most common genotype in 2016 (12.2% prevalence), following GII.P4/GII.4 Sydney 2012 (28.4%), GII.P16/GII.2 (18.9%) and GII.Pe/GII.4 Sydney 2012 (17.6%). GII.P2/GII.2 strains were mostly observed starting from July 2016, leaving space for the possibility that their increased circulation was linked to the introduction of the ‘Mascali clone’. However, a couple of GII.P2/GII.2 strains (including PA36/16) genetically close to the ‘Mascali clone’ had been already detected in January 2016 in Palermo and GII.P2/GII.2 NoV had sporadically been observed also in 2011 and 2014. Moreover, the circulation of GII.P2/GII.2 NoVs has been increasing worldwide over the last period [37]. Therefore, the unusually high circulation observed in Palermo might just reflect the global epidemiology of GII.P2/GII.2 strains. Noteworthy, the older GII.P2/GII.2 NoV strains detected in Palermo were closely related in the ORF2 (98.9% nt id) to the GII.P16/GII.2 Nashville strain (KY865307) which is supposed to have generated the recombinant GII.P16/GII.4 clone recently circulating epidemically worldwide [37, 38] and were more distantly associated (95.3% nt id) to the GII.2 capsid sequence (AB662868) of another putative GII.P16/GII.4 ancestor from Japan [39]. The increasing frequency of detection of waterborne NoV outbreaks calls for the introduction of routine monitoring for viruses of drinking waters and for the definition of affordable and reliable indicators for the presence of NoV in drinking waters. The use of ultrafiltration systems as a concentration method in order to recover viruses from environmental waters combined with RT-PCR to detect viral genomes has been demonstrated to grant a recovery rate of >50% of multiple viruses [40] and is at present one of the most sensitive approaches for the detection of NoVs, although the discrimination between viable and nonviable viruses remains a problem. The burden of NoV outbreaks, either food associated or waterborne, is still poorly defined due to the lack of specific diagnostic tools in many clinical laboratories. NoV epidemics should always be suspected when the main symptom is vomiting and the duration of disease is short (<48 h). Waterborne transmission is suggested by the rapid spread of disease and the high attack rate. Care should be taken to perform sensitive diagnostic tests on a sufficient number of stool samples in order to identify NoV as the causative agent of enteritis outbreaks, possibly preferring genome screening over antigen detection. Only the systematic detection of NoV outbreaks can guide in the choice of the safety procedures that could be most appropriate in containing such microbiological hazard.

What were the infrastructure complaints?