Seasonal variation of acute gastro-intestinal illness by hydroclimatic regime and drinking water source: a retrospective population-based study

Uncorrected Proof 1 © IWA Publishing 2013 Journal of Water and Health | in press | 2013 Seasonal variation of acute gastro-intestinal illness by hydroclimatic regime and drinking water source: a retrospective population-based study Lindsay P. Galway, Diana M. Allen, Margot W. Parkes and Tim K. Takaro ABSTRACT Acute gastro-intestinal illness (AGI) is a major cause of mortality and morbidity worldwide and an important public health problem. Despite the fact that AGI is currently responsible for a huge burden of disease throughout the world, important knowledge gaps exist in terms of its epidemiology. Specifically, an understanding of seasonality and those factors driving seasonal variation remain elusive. This paper aims to assess variation in the incidence of AGI in British Columbia (BC), Canada over an 11-year study period. We assess variation in AGI dynamics in general, and disaggregated by hydroclimatic regime and drinking water source. We use several different visual and statistical techniques to describe and characterize seasonal and annual patterns in AGI incidence over time. Our results consistently illustrate marked seasonal patterns; seasonality remains when the dataset is disaggregated by hydroclimatic regime and drinking water source; however, differences in the magnitude and timing of the peaks and troughs are noted. We conclude that systematic descriptions of infectious illness dynamics over time is a valuable tool for informing disease prevention strategies and generating hypotheses to guide future research in an era of global environmental change. Key words | acute gastro-intestinal illness, climate change, ecological determinants, seasonality Lindsay P. Galway (corresponding author) Faculty of Health Sciences, Simon Fraser University, 11830 Blusson Hall, 8888 University Drive, Burnaby, BC, Canada E-mail: lpg@sfu.ca Diana M. Allen Earth Sciences, Faculty of Science, Simon Fraser University, 8888 University Drive, Burnaby, BC Margot W. Parkes Canada Research Chair in Health, Ecosystems and Society, School of Health Sciences/Cross-appointed, Northern Medical Program, University of Northern British Columbia, 3333 University Way, Prince George, BC., Canada Tim K. Takaro Faculty of Health Sciences, Simon Fraser University, 8888 University Drive, Burnaby, BC., Canada INTRODUCTION ‘Whoever wishes to investigate medicine properly, to the health care system remain high (Payment & should proceed thus: in the first place to consider the sea- Riley-Buckley ; Majowicz et al. ; Fleury et al. sons of the year, and what effects each of them produces, ). Additionally, major waterborne outbreaks such for they are not all alike, but differ much from themselves as in regard to their changes’ reminders of the potentially devastating population Hippocrates (Lloyd et al. ) the Walkerton, Ontario example in 2001 are health impacts of waterborne AGI (Ali ; Eggertson ). Despite the fact that AGI is currently responsible Acute gastro-intestinal illness (AGI) is a major cause of for a huge burden of disease throughout the world mortality and morbidity worldwide (Prüss et al. ). and is a major public health problem, important Estimates suggest that AGI infections cause four million knowledge gaps exist in terms of its epidemiology. cases of diarrhea worldwide each year (WHO ). Specifically, an understanding of seasonality, the sys- In developing nations, diarrhea is the third leading tematic periodic occurrence of illness, and those factors Q1 cause of death (WHO 2008). Although mortality due driving seasonal variation remain poorly understood to AGI in developed nations is relatively low, the (Grassly & Fraser ; Naumova ; Fisman ; morbidity, associated socio-economic costs, and burden Lal et al. ). doi: 10.2166/wh.2013.105 Uncorrected Proof 2 L. P. Galway et al. | Seasonal variation of AGI by hydroclimatic regime and drinking water source The causes and consequences of seasonal variation in Journal of Water and Health | in press | 2013 control strategies, and improve the accuracy of early warn- infectious disease occurrence have intrigued medical pro- ing systems (Pascual & Dobson ; Altizer et al. ). fessionals and epidemiologists for more than a century (i.e. Infectious diseases that vary seasonally demonstrate some Ransome (); Lloyd et al. ()). More recently, increas- form of climatic dependence and are most likely to be influ- ing concerns regarding global environmental change, enced by climate change and variability. Hence, a greater globalization and ‘an apparent surge in infectious disease understanding of seasonal variation can play an important emergence’ and re-emergence have inspired a renewed inter- role in the climate change adaptation process (McMichael est in the seasonality of infectious illness (Fisman ; Lal ; Fleury et al. ; WHO ). Furthermore, an in- et al. ). The dynamics of AGI is surely influenced by a depth understanding of past and current seasonal variation complex interaction of ecological, social, and biological in infectious illness epidemiology can provide a baseline determinants; however, waterborne AGI in particular is for monitoring the early impacts of climate change on mediated by ecological factors that influence drinking health (Martens & McMichael ). Lastly, exploring dis- water quality and quantity (Eisenberg et al. ; Institute ease dynamics, and specifically how these vary by relevant of Medicine (US) ; Lal et al. ). Ecological drivers factors and across different settings, can generate hypoth- have not been adequately explored due to the dominant eses and highlight future research priorities. public health paradigm that is largely focused on individual This paper analyses variations in the incidence of AGI in level risk factors (McMichael ; March & Susser ; British Columbia (BC), Canada over the period of 1999–2010 Ruiz-Moreno et al. ; Eisenberg et al. ; Lal et al. ). in order to answer the following questions: (1) Does the inci- The role of hydroclimatology, a term commonly used to dence of AGI in BC, Canada illustrate a seasonal pattern?; describe the dominant climatic drivers for watershed (2) Does the seasonal pattern of AGI in BC differ according responses (Allen et al. ), is one such ecological factor to hydroclimatic regime and drinking water source?; and that has not been adequately explored to date. Recent (3) Has the incidence of AGI in BC changed over time? Sev- advances have been made with regards to our understanding eral different methods are applied in order to provide a of cholera epidemiology, in part due to increased interest and complete picture and robust understanding of seasonality in research pertaining to environmental drivers of seasonality, AGI incidence. To the best of our knowledge, this is the including hydroclimatology (Bertuzzo et al. ). The work first study to describe and compare seasonal patterns of Q2; Q3 of Akanda et al. (2011) and Bertuzzo et al. (2011) has high- AGI by hydroclimatic regime and drinking water source. lighted ways in which distinct hydroclimatic regimes can influence pathogen transmission and disease occurrence in different ways, resulting in distinct seasonal patterns of illness METHODS at the population level. Hydroclimatology may play a role in driving the seasonality of AGI. Other potentially important Design ecological drivers of seasonality include drinking water source, agricultural and other land use activities, and vari- We conducted a retrospective, population-based study to ations in wild animal populations. Herein, we examine the assess the seasonality of reported and laboratory confirmed dynamics of AGI over time and across different hydrocli- AGI cases from January 1, 1999 to January 1, 2010 in tar- matic regimes and drinking water sources to explore geted communities in the province of BC. This study was whether and how these factors may drive seasonality of AGI. approved by the SFU Research Ethics Board for the use of A better understanding of AGI seasonality, those factors secondary data. driving disease dynamics, and differences in seasonality across settings is key to disease prevention at the population Setting level. More specifically, this knowledge can facilitate the prediction and forecasting of longer-term changes in the BC, Canada was selected as the setting for this work for several risk of illness, inform effective disease prevention and reasons. Firstly, the province offers a rich diversity of Uncorrected Proof 3 L. P. Galway et al. | Journal of Water and Health Seasonal variation of AGI by hydroclimatic regime and drinking water source | in press | 2013 hydroclimatology and drinking water sources, thus acting as a unique case study to examine the seasonal patterns of gastrointestinal illness across these factors (Allen et al. ). Second, a province wide surveillance system, the Integrated Public Health Information System (iPHIS), operational since the late 1990s, offered access to laboratory confirmed illness data of high quality and adequate time span for the analysis for seasonal disease dynamics. Third, despite the fact that AGI rates are higher in this province than the rest of the country, a knowledge gap exists with regards to seasonality of infectious AGI in this setting (Davies & Mazumder ). The hydroclimatology of BC can be broadly classified as rainfall-dominated (pluvial) and Figure 1 | Hydroclimatic regimes and drinking water sources across the study communities. snowmelt-dominated (nival) (Allen et al. ). Rainfall-dominated regimes are found primarily in coastal lowland areas and snowmelt- in turn report cases to the BC Centre for Disease Control dominated regimes occur in the interior plateau and moun- through iPHIS. Five potentially waterborne AGI pathogens tain regions (Pike et al. ). The hydrology of rainfall- are included in this study: Campylobacter, Salmonella, vero- dominated regimes is characterized by seasonal changes in toxigenic Escherichia coli, Giardia, and Cryptosporidium. rainfall, with peak streamflow and groundwater recharge The first three are bacterial pathogens while the latter two occurring during the rainy winter months (November–Feb- are protozoan. Reported cases along with information on ruary) and age at onset of illness, sex, disease type, serotype, episode groundwater levels occurring the late summer and early date, report date, address, and a unique identifier were fall (July–September) (Pike et al. ). On the other hand, extracted from the iPHIS system for the selected study hydrological processes in snowmelt-dominated regimes are communities. and the lowest monthly streamflows controlled primarily by melting snowpack and glaciers. A total of 89 cases (3.9% of all reported cases) were Thus, the hydrology of snowmelt-dominated regimes is excluded. Cases were excluded if they were identified as characterized by the highest flows occurring in the spring travel cases (identified by disease subtypes known to be and early summer, and the lowest flows in the late only travel related), or were identified as clearly outside of summer and throughout the winter (Pike et al. ). the study community after geocoding the cases (for example, Drinking water sources in BC include groundwater, sur- on a nearby island or more than 75 km from the center of face water, or a mixture of groundwater and surface water the study community). We expect that travel cases remain (hereafter referred to as mixed water). Approximately 30% in our dataset as the travel variable was poorly populated. of BC residents rely on groundwater for drinking water (Government of Canada ). Linking cases to a drinking water source Eight study communities were selected from across the province of BC for this analysis. These eight communities Cases were geocoded by street address and postal code of were selected to represent different combinations of hydro- residence using a geographic information system (GIS) in climatic regimes and drinking water sources relevant ArcMap10.0 (ESRI, Redlands, CA, USA). A spatial join across the province, as illustrated in Figure 1. function was then performed to assign each case to a specific drinking water system based on land parcel infor- Case data mation, ultimately linking each case to a drinking water source. More specifically, we gathered spatial data indicat- Physicians in BC are mandated to report cases of notifiable ing water service areas for the various drinking water waterborne illness to regional public health authorities, who systems within each study community, and province-wide Uncorrected Proof 4 L. P. Galway et al. | Seasonal variation of AGI by hydroclimatic regime and drinking water source layers indicating the location of groundwater wells and sur- Journal of Water and Health | in press | 2013 groundwater and surface water. In these cases, information face water extraction points. Information about small from local community officials and water managers regard- community systems was gathered from community water ing service populations for all community water systems managers and manually added to the GIS. Using these was used to estimate the proportion of the total population spatial datasets, we assumed that all cases linked to a resi- serviced with each type of water source. This proportion dential land parcel located within a drinking water system was then applied to the total population of communities service area were serviced by that system, and thus we D and H to estimate the population in these two commu- assigned the case the associated water source unless there nities supplied with groundwater, surface water, and was an active surface water extraction license, groundwater mixed water. The calculations were then used to generate well, or small community system linked to the land parcel. total estimated population at risk for each drinking water This method of linking case data to a drinking water source. system and source is adapted from the work of Jones et al. Analysis Q4 (2005). We used time-series plots, monthly plots, negative binomial Generating time-series models, and spectral analysis as visual and statistical tools to We generated monthly times-series with rates expressed as characterize the seasonal and secular trends in the AGI the number of cases per month per 100,000 population for time-series. Analysis involved the use of several graphical the entire dataset and for cases disaggregated by hydrocli- and statistical techniques to triangulate findings and provide matic regime and drinking water source. a more robust understanding of disease dynamics. We The variable ‘episode date’ (defined as the date of symp- applied each of these methods to the full dataset and to tom onset) was used to derive the day, month, and year of the dataset disaggregated by cases occurring in each hydro- illness occurrence, and then, to generate monthly time- climatic regime and linked to each drinking water source. series representing the number of cases per month over By examining seasonal variation in the times-series disaggre- the 132-month study period. Monthly rates were then gener- gated by these selected factors, seasonal patterns under ated using total population at risk derived from census data different conditions were evaluated. as the denominator. Population at risk within each commu- (1) Time-series plots: To reflect trends in AGI data, we nity of plotted the 11-year time-series of monthly AGI incidence. dissemination area population totals – the smallest adminis- was estimated spatially using aggregates The time-series was smoothed using the three-month trative unit with census derived data. Population at risk was moving average, a basic smoothing technique, to enhance calculated at two points during the 11-year study period, interpretation of trends (Zeger et al. ). By superim- 2001 and 2006, and an average population at risk over the posing the time-series disaggregated by hydroclimatic study period was generated. regime and drinking water source, similarities and differ- We also estimated population at risk totals for the ences across these factors could be explored. different hydroclimatic regimes and water sources of inter- (2) Monthly plots: Total monthly incidence over the 11-year est. To generate population totals for each hydroclimatic study period was calculated and plotted. Again, we super- regime, we pooled the population counts for those commu- imposed the total monthly incidence over the 11-year nities located in snowmelt-dominated regimes and for study period disaggregated by hydroclimatic regime and those communities located in rainfall-dominated regimes. drinking water source. Also, negative binomial regression This pooling approach could not be used to generate popu- models using PROC GLM were used to test statistical sig- lation at risk estimates for each drinking water source, nificance of monthly variation. A negative binomial was because communities D and H (Figure 1) are characterized used rather than a Poisson model due to over-dispersion by some entirely groundwater systems, some entirely sur- in the data as indicated by deviance factors greater than face water systems, and others that are a mix of both 1 (Osgood ). The month with the lowest number of Uncorrected Proof 5 L. P. Galway et al. | Seasonal variation of AGI by hydroclimatic regime and drinking water source Journal of Water and Health | in press | 2013 cases was used as the reference month as per convention 11-years in the study period. Additionally, a negative (Naumova et al. ). binomial regression model was used to test the statisti- (3) Spectral analysis: Spectral analysis is useful for detecting periodicities (dominant cyclical patterns) in time-series and to test data for seasonality (Cryer & Chan ). We constructed a periodogram to identify influential periodicities present in the AGI data. We also tested the statistical significance of seasonal patterns using two cal significance of variation across years relative to 1999. Statistical analyses were carried out using SAS software, version 9 (SAS Institute, Inc., Cary, North Carolina) and graphs were created using Microsoft Excel. formal tests: the Fisher–Kappa (FK) test and the Bartlett Kolmogorov–Smirnov (BKS). The FK tests the hypothesis that the series is white noise against the alternative hypothesis that the series contains a periodic component of unspecified frequency. The BKS tests the null hypothesis that the series is white noise, or that there is no periodicity. (4) Annual plots: Finally, we examined secular trends in the data using simple plots of illness rates across the RESULTS During the study period from January 1, 1999 to January 1, 2010 2308 cases of AGI were reported to iPHIS after excluding known travel cases and cases outside the study communities. Of the total 2,308 cases, 1,805 cases (78%) were caused by bacterial pathogens and 458 cases (22%) were caused by protozoan pathogens (Table 1). The inci- Table 1 | dence of AGI across age categories is bimodal with one Characteristics of all AGI cases, 1999–2010 peak among children less than 5 and another between the Characteristic No. of Cases (n ¼ 2308) % of Total 1,805 78.2 458 21.8 Snowmelt-dominated 1,081 47.0 Rainfall-dominated 1,227 53.1 Groundwater 700 30.3 Surface water 1,011 43.8 597 25.9 Pathogen type Bacterial Protozoan females (Figure 2). Hydroclimatic regime Water source Mixed water Figure 2 | ages of 20–30. A similar pattern is seen for males and AGI rates per 100,000 by sex and age, 1999–2010. Time-series plots Visual examination of the time-series plots for all cases illustrates a clear pattern in the data characterized by annual cyclical peaks and troughs (Figure 3(a)). The timing and amplitude of the peak rates clearly vary from year to year. In some years, the time-series illustrates a bimodal peak, a large peak in the summer followed by a second smaller Uncorrected Proof 6 L. P. Galway et al. Figure 3 | | Seasonal variation of AGI by hydroclimatic regime and drinking water source Journal of Water and Health | in press | 2013 Time-series of monthly AGI rates (1999–2010); (a) for all cases, (b) by hydroclimatic regime, and (c) by water source. peak in the fall, while in other years there is a single sum- drinking water source (Figure 3(c)). However, it is difficult mertime peak. There is a range of peaks across the years; to identify any consistent patterns in the time-series plots however, more than half of the years in the study period across these factors. For example, in some years the seasonal illustrate a peak incidence in July and all years show peak cycles in Figure 3(b) show an earlier peak among those cases incidence in the summer or fall months. occurring in snow-dominated regimes compared to those Cyclical patterns remain, although the amplitude and cases occurring in the rainfall-dominated regimes, while in timing of peaks and troughs differ when the cases are disag- other years the opposite is the case. Aside from the consist- gregated by a hydroclimatic regime (Figure 3(b)) and ently higher peaks among those cases linked to a mixed Uncorrected Proof 7 L. P. Galway et al. | Journal of Water and Health Seasonal variation of AGI by hydroclimatic regime and drinking water source | in press | 2013 water source, there are no clear or consistent trends looking at the time-series disaggregated by drinking water source. Monthly plots Figure 4 shows the monthly plots of AGI incidence. The monthly incidence for all cases (Figure 4(a)) illustrates a seasonal pattern with the trough occurring in April and a peak occurring from July to September (summertime). Statistically significant peak incidences were found in July (IRR ¼ 2.21, 95% CI: 1.79–2.73), August (IRR ¼ 2.00, 95% CI: 1.62–2.48) and September (IRR ¼ 2.13, 95% CI: 1.72– 2.64) (Table 2). Among those cases occurring in rainfall-dominated regimes, the trough incidence occurs in April and the peak incidence occurs in September (IRR ¼ 2.37, 95% CI: 1.78– 3.17) (Table 2). In contrast, the trough among those cases in the snowmelt-dominated regime occurs in March and the peak incidence occurs in July (IRR ¼ 2.36, 95% CI: 1.75–3.20) (Table 2) with a steady decline through the fall and winter (Figure 4(b)). Monthly plots by drinking water source show that the timing of the peak occurs during different months of the summer and early fall across different drinking water sources (Figure 4(c)). Among those cases linked to a surface water source, the seasonal peak occurs in July–September (IRR-July ¼ 2.44, 95% CI: 1.77–3.38; IRR-Aug ¼ 2.16 95% CI: 1.55– 3.01; IRR-Sept ¼ 2.07 95% CI: 1.49–2.90) (Table 2). Similar patterns are seen in those cases linked to a groundwater source where the peak incidence also occurs from July through to September (IRR-July ¼ 2.03, 95% CI: 1.40–2.93; IRR-Aug ¼ 2.01, 95% CI: 1.39–2.90; IRR-Sept ¼ 2.00, 95% CI: 1.38–2.9) (Table 2). For those cases linked to a mixed water source the peak incidence occurs in September (IRR ¼ 2.41, 95% CI: 1.59–3.63), with another smaller peak in July (IRR ¼ 2.09 95% CI: 1.37–3.17) (Table 2). The magnitude of the peak of incidence is greatest among cases occurring in a rainfall-dominated hydrocli- Figure 4 | Total monthly AGI rates (1999–2010); (a) for all cases, (b) by hydroclimatic regime, and (c) by water source. matic regime and linked to mixed water cases. spectral density at 12 months in the periodogram (Figure 5). Spectral analysis The same annual periodicity is seen in the time-series data when disaggregated by hydroclimatic regime and drinking The periodogram for all cases shows an annual (12-month) water source (results not shown). The Fisher’s κ and periodicity in the time-series as indicated by the large Bartlett–Kolmogorov–Smirnov tests confirm a statistical Uncorrected Proof 8 L. P. Galway et al. Table 2 | | Table 3 Results of negative binomial for years, all cases, 1999–2010 95% Cl | IRR Upper Lower 2000 1.53a 1.26 1.85 2001 1.41a 1.16 1.72 2002 1.34a 1.10 1.64 2003 1.25a 1.02 1.53 2004 1.17 0.96 1.44 2005 1.20 0.98 1.47 2006 1.03 0.83 1.27 2007 1.05 0.85 1.30 2008 a 1.31 1.08 1.60 2009 1.27a 1.04 1.55 Indicates statistical significance, p < 0.05. | in press | 2013 FK test and BKS test results Time-series Year a Journal of Water and Health Seasonal variation of AGI by hydroclimatic regime and drinking water source FK BKS All GI Illness cases b 18.02 0.370b Snowmelt-dominated cases 1.041b 0.193a Rainfall-dominated cases b 15.35 0.403b Groundwater cases 10.72b 0.284b Surface water cases b 10.73 0.245b Mixed water cases 10.78b 0.239b a p < 0.05. b p < 0.01. FK test: the 5 and 1% critical values for the test are 5.93 and 6.56, respectively. another increase in risk during 2008 and 2009 (Figure 5(a)). The years 2004 to 2007 are not statistically significantly IRR is incidence risk ratio, CI is confidence interval. different from the 1999 reference period. Annual trends 1999 is the reference year. differ somewhat across the hydrological regimes and drinking water sources (see Figure 5(b) and (c); Table 2). DISCUSSION This study examined the seasonality of AGI in BC over an 11-year period using laboratory confirmed case data from eight communities across the province. To our knowledge, this is the most comprehensive examination of AGI seasonality in BC and the first study anywhere to describe trends in AGI by hydroclimatic regime and drinking water source. Additionally, the majority of research examining the seasonality of AGI, and also Q7 Figure 5 | Total annual AGI rates (1999–2010); (a) for all cases, (b) by hydroclimatic regime, and (c) by water source. research examining those factors that may drive these trends, originates from the US, England, and Australia. There has been little research on the subject in BC. Tar- significant seasonal pattern (p < 0.01) for all AGI overall geted ecological studies at the regional level are needed (Table 3). This statistical significance remains when the to inform local decision-making, policy and action (Lal cases are disaggregated by hydroclimatic regime and drink- et al. ). ing water source (Table 3). We used several different visual and statistical techniques to characterize seasonal patterns and our results Annual plots consistently illustrate seasonality. These results are perhaps not surprising given that seasonal variation of AGI has been A non-linear trend is evident over the 11-year study period. documented in numerous other settings (Naumova et al. Looking at the full dataset, there is a slight increase in risk , ; Hall et al. ; Thomas et al. ; Zhang relative to the reference year between 2001 and 2003, fol- et al. ; White et al. ; Febriani et al. ; Harper lowed by a decrease between 2004 and 2007, and then et al. ; Lal et al. ). What is novel is consideration Uncorrected Proof 9 L. P. Galway et al. | Seasonal variation of AGI by hydroclimatic regime and drinking water source Journal of Water and Health | in press | 2013 Examining the time-series data disaggregated by the hydroclimatic regime and drinking water source enables a better understanding of the seasonal dynamics of AGI and can help generate hypotheses about possible mechanisms driving seasonal variation. Differences across the hydroclimatic regime and drinking water source are most easily identified in the monthly plots. Examining monthly plots disaggregated by the hydroclimatic regime demonstrates that the illness peak for those cases occurring in a rainfall-dominated regime occurs in September, compared to July for those cases occurring in a snowmelt-dominated regime. This may reflect differences in temperatures and hydrological patterns contributing to source water contamination at different times of the year in the two distinct regimes. Research indicates that source water microbial loads (both surface water and groundwater sources) are positively correlated with rainfall and may be influenced by the timing of precipitation and runoff events as well as temperature patterns (Atherholt et al. ; Kistemann et al. ). The snowmelt-dominated hydroclimatic regime in BC is characterized by major runoff and groundwater recharge in the spring and early summer caused by warming temperatures that lead to rapid snowmelt; this is known as the spring freshet. At this time, the rivers run high and groundwater levels reach their maximum. Flooding is also a common occurrence. Runoff and groundwater recharge in the rainfall-dominated regimes is driven by precipitation, with streamflow and groundwater levels peaking through the late fall and early winter (Allen et al. ). Although the summertime peak in incidence is likely in part explained by higher temperatures which influence both human behavior and pathogen survival and replication, it is also plausible that the spring freshet in the snow-dominated watersheds causes runoff events and significant source water contamination contributing to the early summertime peak among Q6 Figure 6 | Periodogram of monthly AGI. The peak in the periodogram indicates the dominant periodicity in the data (Fisman 2007). those cases occurring in snowmelt-dominated watersheds. Rapid snowmelt may also overwhelm drinking water treatment systems (Naumova ). ‘A rapid snowmelt, of the seasonal patterns by different hydroclimatic regimes resultant runoff, and filtration system failure at the over- and water source relevant in the context of BC. The season- loaded drinking water treatment plant were implicated ality of these diseases suggests that hydroclimatic factors with the largest known waterborne outbreak of cryptospor- such as temperature, precipitation, and runoff may mediate idiosis, which occurred in Milwaukee, Wisconsin in 1993’ influence disease occurrence. (Naumova ). Thomas et al. () examined the Uncorrected Proof 10 L. P. Galway et al. | Seasonal variation of AGI by hydroclimatic regime and drinking water source association between spring freshet events and AGI out- Journal of Water and Health | in press | 2013 We expect environmental factors that vary across years are breaks across Canada, but found no statistical significance more relevant in this case as the reporting of notifiable dis- (Thomas et al. ). Our results suggest that any associ- eases is required by law in BC (Hill ). ation may have been masked by the inclusion of cases in both snowmelt-dominated and rainfall-dominated watersheds. When interpreting these results, one must note that any effects of weather or hydrological variables on AGI occurrence would be lagged in time, likely on the scale of Fall precipitation may play a role in source water con- several weeks to months. Furthermore, it is a challenge to tamination and the fall incidence peak among those cases separate the effects of various factors driving seasonal illness occurring in rainfall-dominated watersheds (Patz et al. patterns. More advanced statistical analysis are needed to ). Fall rains following the summertime drought are tease out the effects of different potential environmental dri- characteristic of rainfall-dominated watersheds and can vers of seasonality. Poisson time-series analysis using GLM lead to more severe runoff that is highly contaminated, and GAM models, which have been used extensively in air thus posing a greater risk of contamination and illness pollution research, are good candidates for further research. (Charron et al. ). Additionally, high levels of rainfall Case-crossover analysis may prove useful in examining the onto saturated soil may facilitate the movement of patho- influence of more extreme events such as rapid snowmelt, gens over land and into drinking water sources (Lake extreme rainfall events, and flooding. et al. ; Britton et al. ). More research is needed to There are no consistent discernible patterns in seasonal- test and better understand the possible role of the spring fre- ity across those cases linked to different drinking water shet in snowmelt-dominated regimes, and precipitation in sources. Although we might expect to see some differences rainfall-dominated regimes in driving AGI transmission in patterns given that research has shown that water and seasonality. source mediates risk of illness in BC (Uhlmann et al. ; Our results indicate that seasonal patterns vary across Teschke et al. ), our results may be limited by misclassi- years. Inter-annual variability in seasonal patterns may fication of drinking water source exposure. We are confident suggest that ecological factors, such as weather and the with the accuracy of our approach linking cases to their resi- hydrologic response of the watershed, which are known to dential drinking water source because of high predictive vary across years, play a role in driving seasonal patterns. values documented in a sensitivity analysis, but it may be Interestingly, some evidence in settings where weather that using residential drinking water source as a surrogate variability is limited both within years and across years for all source water exposure is problematic. Individuals points to limited variation in AGI rates (Araj et al. ). are likely to drink water from a different source at work or It is also possible that longer scale climatic processes such when they are away from their homes in the evening or as the El Nino-Southern Oscillation (ENSO), a periodic fluc- weekends, such that the residential drinking water source tuation of atmospheric pressure and sea surface temperature may not be a good surrogate measure for exposure. This in the equatorial Pacific Ocean, could influence inter-annual issue has been highlighted in the literature, but no solutions variation. Past research has found evidence for a relation- have been identified (Jones et al. ). This issue may be ship between ENSO and multiannual cycles of cholera particularly relevant in our dataset as 64% of our cases orig- (Pascual et al. ; Koelle et al. ; Hashizume et al. inate from communities where there are drinking water ). Time-series data longer than 11 years are needed to systems supplied by groundwater, surface water, and a mix adequately explore the possible role of longer scale Earth of the two within the community itself. For these commu- system processes. nities, it is likely that individuals are exposed to different Other explanations are also possible, such as changes in drinking water sources when drinking water in their resi- reporting behavior from year to year (Zeger et al. ). Sev- dences and when drinking water in other settings in the eral studies have shown that reporting of infectious illness community. Furthermore, trends in the data across different can be affected by factors such as publicized outbreaks in water sources may be influenced by foodborne cases which other settings or policy changes (Naumova et al. ). we would not expect to vary by water source. Although we Uncorrected Proof 11 L. P. Galway et al. | Seasonal variation of AGI by hydroclimatic regime and drinking water source Journal of Water and Health | in press | 2013 had originally hoped to focus on the protozoan cases, which freshet is an example. This testing could be used to inform are very commonly waterborne pathogens, low case counts early warning systems for the community and could also did not allow this approach. Future research should con- serve as data for future research. sider an analysis focused on specific pathogen groups There are both strengths and weaknesses in this study. (bacterial and protozoan) or pathogens whenever possible. The main strength is the use of different methods for charac- Explicit attention to cryptosporidiosis and giardiasis, recog- terizing seasonality. Often, temporal trends are examined nized by the WHO as ‘neglected diseases’, is called for using visual analysis of plotted time-series only; we have (Savioli et al. ). Another limitation of this study, and taken this approach but have used additional methods to which may warrant attention in future studies, is the need generate richer and more robust findings. This study is for attention to recreational exposures. unique because we have characterized seasonality by In addition to the knowledge generation and methodo- selected hydroclimatic regime and drinking water source. logical contributions of this study, our efforts to compare We have used a GIS to link cases to the hydroclimatic seasonal patterns of AGI by hydroclimatic regime and drink- regime and drinking water source, allowing for analysis of ing water source also contribute to an enhanced conceptual seasonality across these factors. Finally, this study answers understanding of AGI as well as informing of practical the increasingly common call for an interdisciplinary recommendations. Although the combination of hydrocli- approach to infectious disease research. The study team is matology and drinking water source is unlikely to an interdisciplinary group bringing together knowledge represent the full complexity of factors driving AGI in BC, from medicine, human ecology, public health, and earth our explicit attention to interrelated ecological determinants sciences. indicates an approach grounded in systems thinking and There are also important limitations to consider when complexity. The ecological factors that we have examined interpreting these results. Although a diverse group of com- in this work may explain some seasonal variation, but munities with different drinking water sources and clearly other factors also contribute to the seasonal geographic locations across the province was selected, it is dynamics of AGI in BC. Future research in BC and possible that there are important differences between our beyond may consider eco-bio-social approaches that have study communities and other settings limiting the generaliz- been applied to other re-emerging infectious diseases (see ability of this work. Another limitation is that the AGI case for example Arunachalam et al. () and Kittayapong data used in these analyses likely under-represent the true et al. ()) and should examine the important linkages number of illness cases. Under-reporting could introduce between ecological and social factors when data permit. bias into the study if those cases captured in the data are sys- With regards to the practical implications arising from tematically different than those cases not captured in the these results, we suggest two specific practice and policy data. Additionally, our case data do not include information oriented recommendations. First, water managers, environ- about foodborne versus waterborne transmission; such data- mental health officers, and researchers should work sets do not exist in BC and are generally uncommon. together to identify locally-relevant conditions and times of Unfortunately, AGI data from surveillance systems rarely, the year when the risk of AGI illness may be heightened. if ever, include information about foodborne versus water- In communities located in snow-dominated watersheds, borne transmission pathways. Additionally, we have used early summer and the spring freshet are likely to be high- the variable ‘episode date’ from the iPHIS reporting risk times. For those communities located in rain-dominated system to represent the date of onset of the GI illness case. watersheds, late summer and early fall precipitation events It is possible, however, that the ‘episode date’ in the report- are likely candidates. Second, microbiological and/or tur- ing system is not reflective of the true date of onset which bidity monitoring programs could be adapted to account may introduce lag-time between case reporting and case for high-risk conditions and times, taking into account occurrence. Finally, although we have excluded those local conditions and water system characteristics. Increas- cases known to be travel related, we suspect that travel- ing water supply and source testing during the spring related cases remain in the dataset. Uncorrected Proof 12 L. P. Galway et al. | Seasonal variation of AGI by hydroclimatic regime and drinking water source CONCLUSIONS Seasonality represents a rich area for future research, particularly given widespread environmental degradation and our changing climate (Fisman ). These findings provide strong evidence for seasonality in general and differences in seasonality across selected ecological factors. This knowledge can provide insight into disease etiology and can contribute to public health policy-making and water resource management. Knowledge of the timing of disease peaks, for example, could allow public health programs to focus resources and preventative actions at certain times of the year. Furthermore, ‘disease forecasting and warning systems could allow public health officials to alert the populace when specific meteorological conditions pose considerable risk to health’ (Naumova ). We conclude that systematic descriptions of infectious disease dynamics over time is a valuable tool for generating hypotheses for future research, establishing climate change sensitivity and potentially informing of disease prevention strategies. ACKNOWLEDGEMENTS Parkes was supported by the Canada Research Chair Program during her involvement in this study. Galway was supported by the Canadian health Research Institute during her involvement in this study. The authors acknowledge the support of the Simon Fraser University, Community Trust Endowment Fund and appreciate Sunny Mak for his editing of the paper. REFERENCES Ali, S. H.  A socio-ecological autopsy of the E. coli O157:H7 outbreak in Walkerton, Ontario, Canada. Soc. Sci. Med. 58 (12), 2601–2612. Allen, D. M., Whitfield, P. H. & Werner, A.  Groundwater level responses in temperate mountainous terrain: regime classification, and linkages to climate and streamflow. Hydrol. Process. 24 (23), 3392–3412. Altizer, S., Dobson, A., Hosseini, P., Hudson, P., Pascual, M. & Rohani, P.  Seasonality and the dynamics of infectious diseases. Ecol. Lett. 9 (4), 467–484. Journal of Water and Health | in press | 2013 Araj, G. F., Abdul-Baki, N. Y., Hamze, M. M., Alami, S. Y., Nassif, R. E. & Naboulsi, M. S.  Prevalence and etiology of intestinal parasites in Lebanon. J. Med. Liban. 44 (3), 129. Arunachalam, N., Tana, S., Espino, F., Kittayapong, P., Abeyewickrem, W., Wai, K. T., Tyagi, B. K., Kroeger, A., Sommerfeld, J. & Petzold, M.  Eco-bio-social determinants of dengue vector breeding: a multicountry study in urban and periurban Asia. Bull. WHO 88 (3), 173–184. Atherholt, T. B., LeChevallier, M. W., Norton, W. D. & Rosen, J. S.  Effect of rainfall on Giardia and Cryptosporidium. J. Am. Water Works Assoc. 90 (9), 66–88. Bertuzzo, E., Mari, L., Righetto, L., Gatto, M., Casagrandi, R., Rodriguez-Iturbe, I. & Rinaldo, A.  Hydroclimatology of dual-peak annual cholera incidence: Insights from a spatially explicit model. Geophys. Res. Lett. 39 (5), pL05403. Britton, E., Hales, S., Venugopal, K. & Baker, M. G.  The impact of climate variability and change on cryptosporidiosis and giardiasis rates in New Zealand. J. Water Health 8, 561–571. Charron, D. F., Thomas, M. K., Waltner-Toews, D., Aramini, J. J., Edge, T., Kent, R. A., Maarouf, A. R. & Wilson, J.  Vulnerability of waterborne diseases to climate change in Canada: a review. J. Toxicol. Environ. Health A 67 (20–22), 1667–1677. Cryer, J. D. & Chan, K.-S.  Time Series Analysis: With Applications in R. Springer, New York. Davies, J.-M. & Mazumder, A.  Health and environmental policy issues in Canada: the role of watershed management in sustaining clean drinking water quality at surface sources. J. Environ. Manag. 68 (3), 273–286. Eggertson, L.  Investigative report: 1766 boil-water advisories now in place across Canada. Can. Med. Assoc. J. 178 (10), 1261–1263. Eisenberg, J. N. S., Desai, M. A., Levy, K., Bates, S. J., Liang, S., Naumoff, K. & Scott, J. C.  Environmental determinants of infectious disease: a framework for tracking causal links and guiding public health research. Environ. Health Perspect. 115 (8), 1216–1223. Eisenberg, J. N. S., Trostle, J., Sorensen, R. J. D. & Shields, K. F.  Toward a systems approach to enteric pathogen transmission: from individual independence to community interdependence. Annu. Rev. Public Health 33 (1), 239–257. Febriani, Y., Levallois, P., Gingras, S., Gosselin, P., Majowicz, S. & Fleury, M.  The association between farming activities, precipitation, and the risk of acute gastrointestinal illness in rural municipalities of Quebec, Canada: a cross-sectional study. BMC Public Health 10 (1), 48. Fisman, D.  Seasonality of viral infections: mechanisms and unknowns. Clin. Microbiol. Infect. 18 (10), 946–954. Fisman, D. N.  Seasonality of infectious diseases. Annu. Rev. Public Health 28, 127–143. Fleury, M., Charron, D. F., Holt, J. D., Allen, O. B. & Maarouf, A. R.  A time series analysis of the relationship of ambient temperature and common bacterial enteric infections in two Canadian provinces. Int. J. Biometeorol. 50 (6), 385–391. Uncorrected Proof 13 L. P. Galway et al. | Seasonal variation of AGI by hydroclimatic regime and drinking water source Fleury, M. D., Stratton, J., Tinga, C., Charron, D. F. & Aramini, J.  A descriptive analysis of hospitalization due to acute gastrointestinal illness in Canada, 1995–2004. Can. J. Public Health 99 (6), 489–493. Government of Canada, E.C.  Environment Canada – Water – Groundwater [WWW Document]. Available from: www.ec. gc.ca/eau-water/default.asp?lang=En&n=300688DC-1#use. Grassly, N. C. & Fraser, C.  Seasonal infectious disease epidemiology. Proc. R. Soc. B Biol. Sci. 273 (1600), 2541–2550. Hall, G. V., Kirk, M. D., Ashbolt, R., Stafford, R. & Lalor, K.  Frequency of infectious gastrointestinal illness in Australia, 2002: regional, seasonal and demographic variation. Epidemiol. Infect. 134 (1), 111–118. Harper, S. L., Edge, V. L., Schuster-Wallace, C. J., Berke, O. & McEwen, S. A.  Weather, water quality and infectious gastrointestinal illness in two Inuit communities in Nunatsiavut, Canada: potential implications for climate change. EcoHealth 8 (1), 93–108. Hashizume, M., Armstrong, B., Hajat, S., Wagatsuma, Y., Faruque, A. S. G., Hayashi, T. & Sack, D. A.  The effect of rainfall on the incidence of cholera in Bangladesh. Epidemiology 19 (1), 103. Hill, A.  Indifferent reporting of notifiable diseases. Can. Med. Assoc. J. 184 (10), 513–514. Institute of Medicine (US) Forum on Microbial  Global issues in water, sanitation, and health: Workshop summary. In Vulnerable Infrastructure and Waterborne Disease Risk. National Academies Press, Washington, D.C. Jones, A. Q., Dewey, C. E., Doré, K., Majowicz, S. E., McEwen, S. A. & Waltner-Toews, D.  Exposure assessment in investigations of waterborne illness: a quantitative estimate of measurement error. Epidemiol. Perspect. Innov. 3 (1), 6–13. Kistemann, T., Classen, T., Koch, C., Dangendorf, F., Fischeder, R., Gebel, J., Vacata, V. & Exner, M.  Microbial load of drinking water reservoir tributaries during extreme rainfall and runoff. Appl. Environ. Microbiol. 68 (5), 2188–2197. Kittayapong, P., Thongyuan, S., Olanratmanee, P., Aumchareoun, W., Koyadun, S., Kittayapong, R. & Butraporn, P.  Application of eco-friendly tools and eco-bio-social strategies to control dengue vectors in urban and peri-urban settings in Thailand. Pathog. Glob. Health 106 (8), 446–454. Koelle, K., Rodó, X., Pascual, M., Yunus, M. & Mostafa, G.  Refractory periods and climate forcing in cholera dynamics. Nature 436 (7051), 696–700. Lake, I. R., Bentham, G., Kovats, R. S. & Nichols, G. L.  Effects of weather and river flow on cryptosporidiosis. J. Water Health 3, 469–474. Lal, A., Hales, S., French, N. & Baker, M. G.  Seasonality in human zoonotic enteric diseases: a systematic review. PLoS ONE 7 (4), pe31883. Lloyd, G. E., Chadwick, J. & Mann, W. N.  Hippocratic Writings. Penguin (Non-Classics). Majowicz, S. E., McNab, W. B., Sockett, P., Henson, S., Doré, K., Edge, V. L., Buffett, M. C., Fazil, A., Read, S., McEwen, S., Journal of Water and Health | in press | 2013 Stacey, D. & Wilson, J. B.  Burden and cost of gastroenteritis in a canadian community. J. Food Protect. 69 (3), 651–659. March, D. & Susser, E.  The eco- in eco-epidemiology. Int. J. Epidemiol. 35 (6), 1379–1383. Martens, W. J. M. & McMichael, A. J.  Environmental Change, Climate, and Health: Issues and Research Methods. Cambridge University Press, London. McMichael, A. J.  Prisoners of the proximate: loosening the constraints on epidemiology in an age of change. Am. J. Epidemiol. 149 (10), 887–897. McMichael, A. J.  Climate Change and Human Health: Risks and Responses. World Health Organization, Geneva. Naumova, E.  Mystery of seasonality: getting the rhythm of nature. J. Public Health Pol. 27 (1), 2–12. Naumova, E., Chen, J. T., Griffiths, J. K., Matyas, B. T., Estes-Smargiassi, S. A. & Morris, R. D.  Use of passive surveillance data to study temporal and spatial variation in the incidence of giardiasis and cryptosporidiosis. Public Health Rep. 115 (5), 436–447. Naumova, E., Jagai, J. S., Matyas, B., DeMaria Jr, A., MacNeill, I. B. & Griffiths, J. K.  Seasonality in six enterically transmitted diseases and ambient temperature. Epidemiol. Infect. 135 (2), 281–292. Osgood, D. W.  Poisson-based regression analysis of aggregate crime rates. J. Quant. Criminol. 16 (1), 21–43. Pascual, M. & Dobson, A.  Seasonal patterns of infectious diseases. PLoS Med 2 (1), pe5. Pascual, M., Rodó, X., Ellner, S. P., Colwell, R. & Bouma, M. J.  Cholera dynamics and El Niño-Southern oscillation. Science 289 (5485), 1766–1769. Patz, J. A., Vavrus, S. J., Uejio, C. K. & McLellan, S. L.  Climate change and waterborne disease risk in the Great Lakes region of the US. Am. J. Prevent. Med. 35 (5), 451–458. Payment, P. & Riley-Buckley, M. S.  Resolving the Global Burden of Gastrointestinal Illness: A Call to Action. American Academy of Microbiology, Washington, DC. Pike, R., Redding, T., Wilford, D., Moore, R. D., Ice, G., Reiter, M. & Toews, D. A.  Chapter 4 – regional hydrology. In: Compendium of Forest Hydrology and Geomorphology in British Columbia. B.C. Ministry of Forests and Range Research Branch, Victoria, B.C. Prüss, A., Kay, D., Fewtrell, L. & Bartram, J.  Estimating the burden of disease from water, sanitation, and hygiene at a global level. Environ. Health Perspect. 110 (5), 537–542. Ransome, A.  On epidemic cycles. Proc. Manchester Lit. Phill. Soc. 19, 75–96. Ruiz-Moreno, D., Pascual, M., Bouma, M., Dobson, A. & Cash, B.  Cholera seasonality in madras (1901–1940): dual role for rainfall in endemic and epidemic regions. EcoHealth 4 (1), 52–62. Savioli, L., Smith, H. & Thompson, A.  Giardia and Cryptosporidium join the ‘neglected diseases initiative’. Trends Parasitol. 22, 203–208. Q5 Uncorrected Proof 14 L. P. Galway et al. | Seasonal variation of AGI by hydroclimatic regime and drinking water source Teschke, K., Bellack, N., Shen, H., Atwater, J., Chu, R., Koehoorn, M., MacNab, Y., Schreier, H. & Isaac-Renton, J.  Water and sewage systems, socio-demographics, and duration of residence associated with endemic intestinal infectious diseases: A cohort study. BMC Public Health 10 (1), 767. Thomas, K., Majowicz, S. E., Pollari, F. & Sockett, P. N.  Burden of acute gastrointestinal illness in Canada, 1999– 2007: interim summary of NSAGI activities. Can. Commun. Dis. Rep. 34 (5), 8–15. Thomas, K. M., Charron, D. F., Waltner-Toews, D., Schuster, C., Maarouf, A. R. & Holt, J. D.  A role of high impact weather events in waterborne disease outbreaks in Canada, 1975–2001. Int. J. Environ. Health Res. 16 (3), 167–180. Uhlmann, S., Galanis, E., Takaro, T., Mak, S., Gustafson, L., Embree, G., Bellack, N., Corbett, K. & Isaac-Renton, J.  Where’s the pump? Associating sporadic enteric disease with drinking water using a geographic information system, in British Columbia, Canada, 1996–2005. J. Water Health 7 (4), 692. Journal of Water and Health | in press | 2013 White, A. N. J., Kinlin, L. M., Johnson, C., Spain, C. V., Ng, V. & Fisman, D. N.  Environmental determinants of campylobacteriosis risk in Philadelphia from 1994 to 2007. EcoHealth 6 (2), 200–208. WHO (World Health Organization)  Global Water Supply and Sanitation Assessment 2000 Report. Available from: www.who.int/water_sanitation_health/monitoring/ globalassess/en/ (accessed January 8 2013). WHO (World Health Organization)  WHO | Vision 2030: The Resilience of Water Supply and Sanitation in the Face of Climate Change. Available from: www.who.int/ water_sanitation_health/publications/9789241598422/en/ (accessed January 8 2013). Zeger, S. L., Irizarry, R. A. & Peng, R. D.  On time series analysis of public health and biomedical data. Dept. of Biostatistics Working Papers 54, Johns Hopkins University. Zhang, Y., Bi, P. & Hiller, J.  Climate variations and salmonellosis transmission in Adelaide, South Australia: a comparison between regression models. Int. J. Biometeorol. 52 (3), 179–187. First received 17 May 2013; accepted in revised form 16 September 2013. Available online 28 October 2013 Author Queries Journal: Journal of Water and Health Manuscript: JWH-D-13-00105 Q1 WHO (2008) is not listed in the reference list. Please provide publication details to insert in the list. Q2 Q3 Akanda et al. (2011) is not listed in the reference list. Please provide publication details to insert in the list. Bertuzzo et al. (2011) is not listed in the reference list. Please provide publication details to insert in the list. Q4 Q5 Jones et al. (2005) is not listed in the reference list. Please provide publication details to insert in the list. Please provide editor names for Pike et al. (2007). Q6 Q7 Please indicate where in the main text Figure 6 should be mentioned. In Figure 5, part figures ‘a’, ‘b’ and ‘c’ are mentioned in caption, but part figures not provide. Please check.