Health Stream Article - Issue 48 - December 2007

Distribution System Health Study

Researchers in Norway have conducted an epidemiological study suggesting that repairs and maintenance works in water distribution systems are associated with increased risks of gastroenteritis among residents in the area affected by the work (1). The outcome of the study once again raises questions about the role of tap water as a contributor to endemic gastrointestinal disease in developed nations. This study differed from previous investigations of endemic disease because it specifically targeted areas of the distribution system affected by low pressure events. Previous research has highlighted how even transient low pressure events may permit ingress of groundwater, potentially containing enteric pathogens, into water distribution systems (2).

The study was conducted over a 12 month interval in urban areas of Norway with a total population of about 1.1 million. Tap water was provided by seven waterworks which individually served between 35,000 and 460,000 people. During the study the water companies were asked to notify the researchers of low-pressure episodes affecting their service area. Such episodes were defined as events were a section of the water distribution system was closed off, with a presumed loss of water pressure. Episodes could be planned (eg scheduled maintenance) or unplanned (eg spontaneous pipe breaks or damage during construction work).

Each waterworks was asked to identify up to two episodes per month where at least 10 households were affected by low pressure; the first planned episode and the first unplanned episode during the month. When notifiable episodes occurred, the researchers were contacted and supplied with information on the time and place of the event, the climatic conditions, the reason for the event, any measures taken to prevent contamination, and the location of sewage pipes in relation to water distribution pipes. The waterworks personnel also provided their assessment on the risk of contamination associated with the episode (low, medium or high risk). The waterworks company randomly selected 10 “exposed” households (affected by the low pressure episode) and 10 “unexposed” households (in the same general area but unaffected by the low pressure episode) from its customer register and provided their contact details to researchers. A letter was sent to these households advising that they would be contacted by telephone and asked to participate in the study. The letter requested that one person aged over 16 years be prepared to answer questions on behalf of all household members.

Telephone contact was made with households between 8 and 14 days after the low pressure episode, and a standard questionnaire was administered. Data collected included demographic details of household members, average tap water intake at home, any overseas travel in the last month, whether children attended a daycare centre, employment of household members in kindergartens, the presence of pets or regular contact with other animals. People were asked about gastroenteritis episodes in household members during the week following the low pressure event, and whether they had noticed discolouration or a strange taste in tap water in the last 14 days, or if they thought any work had been done on the water pipes. In order to avoid bias in data collection, the same preliminary letter was sent to both exposed and unexposed households, and interviewers were unaware of the exposure status of the households.

Data from 88 low pressure episodes were included in the statistical analysis, with the number of episodes for individual waterworks ranging from 2 to 24 during the 12 month study. It was estimated that 5,935 households had been exposed to low pressure conditions from these 88 episodes. Some reported low pressure episodes were not able to be included in the study due to lack of interviewing capacity. A total of 612 exposed households and 547 unexposed households were interviewed from a target of 880 households (88 episodes x 10 households in each group). Reasons for missing interviews included inability to establish telephone contact (37%), relocation or no phone number obtainable (21%), refusal to participate (20%) or unknown reasons (15%). Six households (4 exposed and 2 unexposed) were excluded as they were unable to specify the time of gastrointestinal illness in relation to the low pressure episode.

The majority (63%) of low pressure episodes were due to mains breaks or leaks, 26% were due to changes of equipment (valves, pipes), while 11% were due to various other causes (cleaning of pipes, construction near pipes, defective valves etc). Waterworks companies reported flushing pipes to eliminate contamination after 87% of episodes, but only one of the seven companies chlorinated after repair work (performed in 12 of the 14 episodes reported by this company). Water samples were obtained for only 18 low pressure episodes and one was positive for E. coli . Householders were not advised to boil water following any of the episodes.

There were no significant differences between exposed and unexposed households in terms of demographics or most non-water related risk factors (eg children were in childcare, overseas travel, animal contact etc). More unexposed households had a member who worked in a kindergarten (7% compared to 4% in the exposed group) and this was statistically significant. In both groups, 83% of households reported average water consumption was more than one glass per person per day. The analysis of gastrointestinal illness was performed at the household level with a case household defined as having at least one person experiencing gastrointestinal illness during the observation period (the week following the low pressure episode). Gastrointestinal illness was defined as an episode of vomiting and/or diarrhoea with at least 3 loose stools during a 24 hour period.

There was a significant difference in the reported rate of gastrointestinal illness by exposed households (12.7%) compared to unexposed households (8%). The crude relative risk (RR) was 1.58 (95% CI 1.1-2.3). Stratified analysis for foreign travel or employment at a kindergarten did not change the RR estimate. Exposed households were significantly more likely to say they believed there had been work/repair of water mains in the last two weeks compared to unexposed households (75% vs 25%, p less than 0.001). Exposed households were also more likely to report they had noticed water discolouration (29% vs 7.3%), but reporting of bad tasting water was similar in both groups (3.8% vs 4.2%). To test the potential effect of bias due to participants being aware of some problem with the water supply, stratified analysis was carried out. Relative risks for “thought work/repair had been done”, “noticed discolouration”, or “bad taste of water” were 1.38 (95% CI 0.9-2.1), 1.37 (95% CI 0.9-2.0) and 1.54 (95% CI 1.1-2.2) respectively. These relatively small changes in RR suggest that awareness of a water problem did not have a large influence on the results of the analysis.

Among exposed households there was a significant difference in reported illness rates between households with higher average daily water consumption (more than 1 glass per person) and those with lower water consumption (RR=4.9 95% CI 1.6-15.2). However in unexposed households, illness risks did not differ significantly with reported average water consumption.

When individuals (n=3020) were considered instead of households, the attack rate for gastrointestinal illness was 7.5% in the exposed group compared to 3.9% in the unexposed group (Odds Ratio 2.0, 95% CI 1.3-3.2). Illness rates were similar in males and females. The highest illness rates were seen in children aged 5 years and less (17.7% in exposed households, 10.0% in unexposed households), however the greatest difference between exposed and unexposed groups occurred in adults aged 20-39 years (OR=7.2, 95% CI 2.8-18.7). Symptoms of illness were similar in the exposed and unexposed groups, and most episodes of illness were of short duration (median 2 days, range 1-14 days). Details of individual water consumption were not collected, so it was not possible to assess individual risk in relation to water intake.

When the data for each waterworks were separately analysed, relative risks varied from 0.9 to 2.2. The waterworks with the lowest RR reported only two low pressure episodes during the study so the RR estimate was very imprecise. The next lowest RR of 1.1 was for the waterworks which hyperchlorinated after most low pressure episodes. The remaining five waterworks had RRs ranging from 1.3 to 2.2.

Relative risks were also assessed for different kinds of work or repairs on the distribution system. On univariate analysis a significantly increased RR was observed for swabbing (4 episodes, RR=2.2, 95% CI 1.1-4.2), and a significantly decreased RR for hyperchlorination (12 episodes from one waterworks, RR=0.4, 95% CI 0.2-1.0). Non-significant increases in RR were seen for rain during repair work, prolonged water shut off (more than 6 hours), and water and sewage pipes being in the same ditch. A non-significant decrease in RR was seen for flushing of pipes after repairs. Analysis using a multivariate logistic regression model showed a significantly increased risk for prolonged water shut off (RR=1.9 95% CI 1.0-3.4) and a significantly decreased risk for flushing (RR=0.4 95% CI 0.2-0.8). Of the 88 low pressure episodes examined, none were considered high risk by waterworks personnel, seven were considered medium risk, and the remaining 81 episodes were considered low risk. Medium risk episodes were significantly associated with a higher rate of illness in affected households (RR=1.8 95% CI 1.0-3.2) compared to low risk episodes.

Although all of the water supplies were disinfected with chlorine at their respective water treatment plants, chlorine residual levels in the distribution systems were low and would have had little protective effect against significant contamination ingress. Previous research has shown that even when measurable free chlorine residual is present, inactivation of viral or protozoan pathogens is slow, although bacterial pathogens and indicators may be rapidly inactivated (3). This may explain why E. coli was detected only once among water samples from 18 low pressure events in the Norwegian study despite the increased risk of gastrointestinal illness observed in exposed households.

Overall, the results of the study are consistent with the hypothesis that low pressure episodes permit ingress of enteric pathogens into the distribution system, and that this may result in a subsequent increase in the risk of gastrointestinal infections in consumers in the affected area. The dose-response relationship observed with water consumption among exposed households, and the reduction in relative risk seen with decontamination procedures (hyper-chlorination or flushing after repair work) are also consistent with this model.

The authors note a number of limitations with the study. Firstly, it was carried out in urban areas and may not be applicable to other settings. However rural systems may be even more vulnerable than urban systems due to longer distribution systems. Secondly, water exposures other than drinking were not assessed. For example, the authors suggest that use of tap water to fill small swimming/paddling pools may represent another exposure route contributing to illness rates. The observation period in the study was limited to one week after the low pressure events, so illnesses caused by pathogens with longer incubation periods (such as Cryptosporidium and Giardia ) would have been missed, however these organisms are less common causes of gastroenteritis than bacterial or viral pathogens. Households and interviewing staff were not informed of the exposure status in order to minimise the potential for recall bias (over-reporting of illness by exposed households). However, the nature of low pressure events means that many households in the exposed group would have been aware of a recent interruption or problem with the water supply. Stratified analysis suggested this had a relatively minor impact on risk estimates.

One previous study has reported an association between low pressure events and gastrointestinal illness among residents of affected areas (4). Data from the control group in a UK case-control study of cryptosporidiosis was analysed and it was found that households affected by low pressure events in the two weeks before interview had a significantly increased risk of illness in the same period compared to households not affected by such events. However due to the limitations of the study design, the low pressure events relied on reporting by participants without confirmation by water companies, and the time relationship between events and the onset of illness could not be established. Therefore the association between low pressure events and illness could only be regarded as tentative. Other studies of endemic waterborne illness in developed nations have used point-of-use interventions (such as comparison of households with real and sham water treatment devices) to assess the contribution of waterborne pathogens to community gastroenteritis, however the design of such studies means that they measure rates of illness in a relatively large area over a prolonged time period rather than targeting short potential exposure periods in localised areas. This new study establishes a novel approach for assessing the role of transient distribution system contamination in endemic gastrointestinal illness.

The results of the Norwegian study suggest that individuals exposed to low pressure events have an increased risk of illness (7.5% in the week following exposure vs 3.9% for individuals in unexposed households), however it is not possible to extrapolate this to an overall estimate of disease burden as the prevalence of exposure to such events across the whole population is not known. The authors speculate that if 20% of households are affected once per year, then an estimated 33,000 cases of acute gastrointestinal illness may be occurring among the 4.5 million residents of Norway due low pressure events in water distribution systems. The illness burden from distribution system contamination may be higher if transient low pressure events rather than the more prolonged events examined in this study also contribute significantly to ingress of contamination. Conversely, risks may be lowered if water companies routinely use decontamination practices such as hyperchlorination and flushing after repair and maintenance work. If the illness rate in unexposed households represents the “background” gastroenteritis rate from all causes in the general population then it can be estimated that the average rate would be about 2 episodes per person per year or a total of 9 million episodes for the whole population. This is reasonably similar to previous estimates for other developed nations (around 1 case of gastroenteritis per person per year).

(1) Breaks and maintenance work in the water distribution systems and gastrointestinal illness: a cohort study. Nygard K, Wahl E, Krogh T et al. (2007) International Journal of Epidemiology 36 :873–880.

Health Stream thanks Dr Nygard for additional information on disinfection practices in the study area .

(2) The potential for health risks from intrusion of contaminants into the distribution system from pressure transients. LeChevallier MW, Gullick RW, et al. (2003). Journal of Water and Health 1 (1): 3-14.

(3) Poor efficacy of residual chlorine disinfectant in drinking water systems to inactivate waterborne pathogens in distribution systems. Payment P (1999) Canadian J Microbiology 45 709-715. (Reviewed in Health Stream Issue 16).

(4) Self-reported diarrhea in a control group: a strong association with reporting of low-pressure events in tap water.  Hunter PR, Chalmers RM, Hughes S and Syed Q. (2005) Clinical Infectious Diseases 40 :e32-4. (Reviewed in Health Stream Issue 37).