Insights into a large waterborne Campylobacter outbreak from a cross-sectional telephone survey

Campylobacter is the most common cause of bacterial gastroenteritis around the world.¹ Symptoms of infection include diarrhoea (which may be bloody), fever, abdominal pain/cramps, nausea and vomiting.² Transmission may occur through consumption of contaminated food or water, contact with animals or their faeces or by person-to-person faecal-oral transmission.¹ Infections via consumption of undercooked or contaminated food, particularly poultry products and raw milk, are well documented.^3–6 In contrast to foodborne outbreaks which generally only affect a small number of individuals, contamination of drinking water supplies with Campylobacter has led to numerous large outbreaks.^2,7–11

The largest waterborne Campylobacter outbreak to date occurred in the town of Havelock North, New Zealand in 2016. It was caused by contamination of the town’s unchlorinated, groundwater-derived drinking water supply with sheep faeces following heavy rainfall, with the contamination lasting for 5 days.¹² There were 953 notified Campylobacter cases and four deaths.¹² A series of landline telephone surveys were undertaken during and after the outbreak which provided the basis for an age adjusted estimate of 5,540 illnesses among the 14,118 residents of Havelock North, and up to another 2,230 illnesses among those living outside of Havelock North.¹² In this paper we describe insights from the fourth of these telephone surveys regarding the epidemic curve, attack rates, dose response, access to healthcare, economic impact and awareness or compliance with public health messaging.

Methods

Data collection

Four landline telephone surveys were conducted by UMR Research Ltd (ISO20252 accredited) on behalf of the Hawke’s Bay District Health Board using a call centre based in Auckland, New Zealand. The surveys were conducted on 16 August 2016, 18 August 2016, 22 August 2016 and on the 27/28 September 2016 (days 8, 10, 14 and 50 after the outbreak started).

Households were selected for the initial three survey waves using a combination of landline telephone numbers sourced from a local directory and random digit dialling using numbers generated by a computer-assisted telephone interviewing (CATI) system to include numbers not listed in the local directory. The locations of the selected households were overlayed with the distribution grid for the Havelock North water supply to exclude households not supplied by Havelock North water. Survey four participants were a subset of those contacted during three initial survey waves selected based on willingness to participate in the follow-up survey. For all survey waves, questions were usually answered by a single person on behalf of the household. The questions asked in this survey are provided in the Appendix.

Confirmed and notified cases of campylobacteriosis were extracted from EpiSurv notifiable disease database as previously described.¹²

This study was undertaken under the direct supervision and mandate of the Hawke’s Bay Medical Officer of Health Dr Nicholas Jones under the New Zealand Health Act 1956.

Statistical analyses

Attack rates were calculated from notified cases relative to the Havelock North population, from the proportion of telephone respondents with diarrhoea, and from the proportion of telephone respondents who drank the Havelock North water and had diarrhoea. Results were adjusted by age group.

P-values for demographic groups were calculated using Pearson’s Chi-squared test. Poisson confidence intervals (95% CI) for attack rates were calculated using binomial distribution for proportions. To assess potential differences in the number of cups of water drunk by symptom status, a t-test to compare the difference between the observed means in two independent samples was performed. All calculations including odds ratios were performed using MedCalc Version 20 (MedCalc Software, Ostend, Belgium).

Results

Participation and representation

Not all the questions were answered by all the respondents, so denominators differ for some questions. Of the 195 households included in surveys one to three, 169 participated in the fourth survey, representing 473 residents. The surveyed population aligned reasonably well with the age distribution of Havelock North residents, although those aged 30–39 years old were slightly under-represented in the sample and those aged 60–69 years old were slightly over-represented (p<0.01) (Table 1).

Drinking water exposure

During the 5 days that it is likely that the water in Havelock North was contaminated (7–12 August 2016), 361 of the 421 people for whom this question was answered indicated that they had consumed unboiled Havelock North tap water, with 125 (35%) developing diarrhoea (Table 1). Among the 60 who stated that they did not drink the water, only two (3%) developed diarrhoea (odds ratio 15.4, 95% CI 3.7–63.9, p<0.01). While in every age group those who developed diarrhoea drank more water than those that did not, this was only statistically significant for those 15–19 years of age, those over 50 years of age and for all cases combined (Table 1). On average, individuals who developed diarrhoeal symptoms drank 5.3 glasses of unboiled tap water per day, whereas those who did not develop diarrhoea drank an average of 3.9 glasses per day (p<0.01).

Reported diarrhoeal symptoms

Within the 169 households surveyed, 81 households contained at least one member who suffered diarrhoea, with a total of 144/473 individuals with diarrhoea. Of these, 45 (31%) reported a relapse after improvement, with two individuals reporting that their diarrhoea had not stopped after more than 6 weeks. The attack rate was higher in those under 30 years of age (44%) compared with those over 50 years of age (26%) (p<0.01). Among telephone survey respondents, those over 70 years old had a lower attack rate than most other age groups. This contrasted with notified cases where those over 70 years old had the highest attack rate (Table 1). Among the 144 individuals with diarrhoea, 34 visited a general practitioner (GP), one visited a hospital and three visited both a GP and a hospital. In households of two or more where at least one person had diarrhoea, 86% (66/77) had at least one person who did not report diarrhoea (Table 2).

The onset dates of cases in the household survey aligned well with confirmed cases, with a peak in cases on 11 August (Figure 1). The apparent second peak in probable cases (symptomatic but not confirmed by isolation of Campylobacter from a patient sample), on 15 and 16 August was not observed among household survey cases.

View Figure 1, Table 1–2.

Impact of outbreak and household symptoms on daily life

Of the 473 people surveyed, 131 symptomatic individuals (28% of the total) indicated they had to take an average of 7 days off work or school (range of 1 to 21 days, total of 936 days). Another 41 days were taken off by 11 individuals that had no symptoms of campylobacteriosis.

Of the 81 households with at least one person suffering from diarrhoea, 55 (68%) indicated that they spent money at pharmacies, with an average spend of NZ$58 (range from NZ$10 to NZ$200, total of NZ$3,180). Within these 55 households, 26 (47%) contained individuals who sought treatment by a GP or hospital, while for the other 29 the pharmacy was their only contact with the medical profession. Another 10 households with no cases of diarrhoea indicated they spent a total of NZ$230 at pharmacies due to the outbreak (average of NZ$23, range from NZ$4 to NZ$44).

On 12 August 2016 a boil water notice was posted and not withdrawn until 3 September 2016. Among surveyed households, 135 (80% of total) indicated that they boiled their water during the outbreak, with 11 of these indicating they routinely boil their tap water and an additional 51 indicating they were still boiling their tap water at the time of the fourth survey (27/28 September). Two-thirds of households (n=113) also chose to buy bottled water, with households using bottled water for a median of 24 days (range from 1 to 46 days).

Discussion

The 2016 Havelock North waterborne Campylobacter outbreak resulted in an estimated 6,260 to 8,320 illnesses.¹² Core to this estimate was the use of cross-sectional telephone questionnaires administered during and after the outbreak to households within Havelock North. During the last of these telephone questionnaires a range of additional questions beyond disease incidence were asked, and this paper provides an analysis of the data, providing insights into campylobacteriosis, disease surveillance, outbreak investigation and community response.

Attack rates

The observation that those who drank more water were more likely to have become ill supports a dose-dependent relationship, which was previously reported for a waterborne Campylobacter outbreak in Tune, Denmark.¹³ Notified cases were distributed across the reticulated network,¹² and the water was contaminated for potentially 5 days, suggesting that all of those who reported drinking the water were likely to have ingested Campylobacter. However, only 35% of people who consumed the water reported symptoms, and within a household it was uncommon for all members to be sick. This attack rate of 35% is important for the epidemiological investigation of campylobacteriosis, highlighting that in similar situations only a limited number of those exposed will have symptoms.

Among those who claimed not to have drunk the water, only two reported symptoms. These symptoms could be due to campylobacteriosis acquired from another source such as exposure to poultry.³ However they could still be connected to the outbreak: exposure to contaminated water during brushing of teeth, consumption of food prepared using unboiled tap water, swallowing of water during bathing/showering or from person-to-person transmission. Brushing of teeth has previously been implicated in development of campylobacteriosis during waterborne outbreaks in Norway⁹ and in Darfield, New Zealand.⁷

Two differences between notified cases and telephone survey respondents were in disease incidence in over-70-year-olds and in the potential second peak seen for notified cases. While 35.5% of notified campylobacteriosis cases in this outbreak were over 70 years old,¹² in this telephone survey over-70-year-olds had the lowest attack rates. Older community members were proactively contacted during the outbreak to determine their health status and if they needed any support. This appears to have artificially increased the apparent incidence among the older population in notified data. This proactive identification of cases may also partly explain what appeared to be a possible second peak in notified cases on 15 and 16 August (Figure 1), which was not supported by the telephone questionnaire data. Lower attack rates with increasing age could reflect increased immunity obtained from previous Campylobacter exposures.^14,15

Pharmacy surveillance

A survey of local pharmacists’ activities during this outbreak¹⁶ highlighted the important role of pharmacists in public wellbeing, pharmaceutical distribution and medicine therapy management. Drug sales data have been previously identified as a useful surveillance tool,^17,18 with this study quantifying the potential sensitivity of this approach. Seventy-two percent of the households with diarrhoea spent money at a pharmacy due to the outbreak, with less than half of those engaging with other medical care. This suggests that pharmacy-based surveillance could detect twice as many cases as may be possible via disease notification. The ease of access and lower cost relative to GPs increases the potential for sick individuals to visit pharmacies. However, without diagnostic tests this would not be particularly specific and, as highlighted by purchases from households without illness, there would be a false-positive rate that would need to be accounted for.

Impact

The telephone surveys also highlight the considerable loss in productivity caused by gastrointestinal outbreaks such as this, with individuals having an average of 7 days off school or work. A number of non-ill individuals were also unable to work or attend school due to the need to care for sick household members or as a consequence of the closure of schools or early learning centres. An age group–adjusted extrapolation to the whole population of Havelock North would be over 30,000 absentee days during this outbreak. This is important not only for the public health response, but also for evaluating the cost–benefits of future interventions and preventative actions. Households also spent an average of NZ$58 at pharmacies, which through extrapolation to the population of Havelock North would suggest expenditure at pharmacies of over NZ$100,000 due to this outbreak.

A relapse of symptoms after recovery was reported by 31% of cases, which is higher than previously reported relapse rates of up to 20% of cases.¹⁹

Boiled water

On 12 August 2016, in response to the outbreak, the Hastings District Council flushed the reticulation network and began chlorinating the water supply.¹² A boil water notice was released to media and posted on social media the same day. In this study, we found that most households boiled their water or obtained water from other sources. While the boil water notice was lifted on 3 September 2016,²⁰ 51 households (38% of the total who boiled their water) were still boiling their tap water 3 weeks later. Boiling of drinking water is known to pose an inherent risk of burns or scalds, particularly to the elderly,²¹ which is concerning when residents of at least 26 of these 51 households were over 65 years old. This highlights the importance of conveying not only the boil water notice but also the recommendation to stop boiling water to minimise the inherent risk that boiling water poses. For this communication to be effective, it may require additional supporting evidence of the safety of the water. There were 11 households who reported routinely boiling their water, but despite this seven still had cases of diarrhoea. Previous studies have noted that many of those who boil their water for drinking will still use unboiled water for washing food and brushing teeth.^22–25

Conclusions

Not all individuals who develop gastrointestinal symptoms seek medical treatment, which means it is often difficult to assess the scale and impact of outbreaks. Telephone surveys or online questionnaires can be used to understand more about disease outbreaks, the disease itself and the behaviours of the public in response. Intense public interest in this outbreak may have increased the willingness to participate, which may not be as high for less prominent outbreaks.

The telephone surveys in this Havelock North outbreak were able to be undertaken remotely by individuals from outside the region (the interviewers were based in Auckland). This required minimal input from the local public health service who were both overwhelmed with responding to the actual outbreak and hampered by illness among public health staff who were also patients in the outbreak. This type of telephone survey does not require trade-offs in local resources and can be done in addition to any other response activities. Provided the exposed population can be reached, surveys can be faster than traditional approaches as they can be administered on the date of symptom onset, rather than waiting for clinical appointments, laboratory testing and notifications. While self-reported symptoms do not confirm the causative organism, it can, as illustrated, support laboratory confirmed surveillance systems. It would also be useful in places without effective surveillance systems or for organisms not tested for clinically, which would include many viruses and new emerging pathogens. The decline in landline telephone usage means that future surveys will need to make use of cellphone and/or online tools to administer surveys. These tools are not geographically bound in the same way as landline phones, so additional steps will need to be taken to identify individuals within an area of interest. This could be achieved by making use of GP patient lists, school roles or other groups. As part of preparedness activities, pre-prepared questionnaires and established relationships with professional survey companies would ensure that these types of surveys can be rapidly deployed.

View Appendix.