REPORTABLE ENTERIC ILLNESS DUE TO DRINKING WATER SOURCE IN PENNSYLVANIA

Abstract

Background: An estimated 1.27 million people per year are experiencing acute gastrointestinal illness (AGI) due to private water sources, which are federally unregulated drinking water sources in the US (Murphy et al., 2017). Previous studies have found a relationship between drinking water source (domestic well or publicly supplied water) and waterborne disease. Although some work has been done to characterize the quality of water in karstic terrain, and karst is considered to be a risk factor for pathogenic contamination of domestic wells, little work has been done to assess the relationship between karst and incidence of enteric disease. There is also little known about the socio-demographic factors of populations in the US using domestic wells. Although studies have also shown an association between rain, antecedent rain, and waterborne illness, the relationship is not well defined. Objective: The dissertation uses spatiotemporal and epidemiologic methods to characterize domestic well users in Pennsylvania, determine the relationship between use of domestic wells and cases of illness, and investigate an interactive effect between rain and drought on county level occurrence of reporting of illness. This work includes illness due to Salmonella, Campylobacter, Giardia, and Cryptosporidium as reported to the PA department of health. The population of interest is the estimated 12.8 million people who lived in PA from 2010-2019, of which an estimated 3.5 million used a domestic well. Aim 1: Determine whether populations of Pennsylvania that rely on domestic wells are more socially vulnerable than those with access to public water supply or wastewater treatment. Hypothesis: Census tracts that are generally more socially vulnerable are also more likely to lack access to public water supply. Aim 2: Determine presence of spatial clusters of reportable enteric diseases (Salmonella, Campylobacter, Giardia, or Cryptosporidium) between 2010 and 2019 in PA, and whether those clusters are associated with drinking water source (whether a domestic well or publicly supplied water). Hypothesis 1: Enteric disease in PA for the 2014 - 2018 time period cluster in time and space. Hypothesis 2: The incidence risk ratio of reported enteric disease will be positively associated with the proportion of households utilizing domestic wells. Aim 3: Determine the effects of rain and drought on weekly counts of reportable enteric illness (Salmonella, Campylobacter, Giardia, or Cryptosporidium) in PA, 2010-2019. Hypothesis 1: County level weekly enteric illness incidence rates are positively associated with total rain (cm) in prior week(s). Hypothesis 2: The incidence rate of reportable enteric illness by county is associated with an interaction between total rain (cm) in the prior week by county and drought conditions as measured by the Palmer Drought Severity Index in the same week the rain is measured preceding disease onset. Methods: Using county and zip code tabulation area incidence data on enteric illnesses made available from the Pennsylvania Department of Health for the years 2010 to 2019, this thesis investigated the potential relationships between rain, drought conditions, use of private wells, and the occurrence of karstic terrain and illness. The social characteristics of those who do not have access to public water supply were also explored. This was done in three parts, all utilizing ecologic study designs. Aim 1: We assessed how social vulnerability measures co-occur with domestic well use and also how those patterns change over space using a profile regression method. Aim 2: We examined the spatial patterns, by Zip Code Tabulation Area using a Poisson Spatial only SaTScan analysis with 5% of the state population without Philadelphia allowing for hierarchical clusters, and temporal patterns, using the R software trending package to develop a season adjusted negative binomial model to assess outbreaks, of weekly reported enteric illnesses. The weekly county incidence rates for each illness were also assessed by percentage of a county that did not have access to public water supply and presence of karst using a zero-inflated negative binomial model with random intercept for county. Philadelphia was excluded because they did not share their data. To assess the second hypothesis, that the incidence risk ratio of reported enteric disease is positively associated with the proportion of households utilizing domestic wells, a zero-inflated negative binomial model with a random intercept for county was utilized to determine if there was an association. This same model was expanded as assessment of an association between illness rates and percentage of a county underlain by karst. Aim 3: We assessed multiple time lags using a zero-inflated negative binomial model with random intercept for county to determine if there was an relationship between county level weekly rainfall (cm) and drought, as measured by the Palmer Drought Severity Index, on the weekly incidence rates of these four reportable enteric illnesses. Philadelphia was excluded, because they did not share their data. If both the amount of rain and the level of drought were found to be related to the number of cases of illnesses, an interaction was assessed for that lag time.Results: Aim 1: Measures of social vulnerability have 15 distinct profiles or clusters in the state of PA. These clusters occur spatially heterogeneously across the state. Five distinct population-profiles are more likely to rely on a domestic well. Two of these profiles are also more likely to experience social vulnerability when measured at the census tract level. In general, census tracts with higher proportions of homeowners, lower proportions of those without a high school diploma, lower median per capita income, and higher proportions of children under the age of 5 and 17 are more likely to rely on an unregulated drinking water source than populations with greater proportions of their population having a high school diploma, higher median per capita income and lower proportions of their population consisting of people under the age of 5 and 17. Aim 2: All four pathogens, which were assessed, have spatial heterogeneity in the state of Pennsylvania. All pathogens had higher incidence rates in the summer and lowest rates in the winter. However, most outbreaks occurred in the wintertime. A positive association was found between quartile of area of a county not served by public water supply and incidence rates of campylobacteriosis [IRR = 1.35 (95%CI 1.21, 1.51 p < 0.001)], cryptosporidiosis [IRR = 1.33 (95%CI 1.07, 1.66, p < 0.05)], and giardiasis [IRR = 1.25 (95%CI 1.11, 1.42, p < 0.01)]. An association was also observed between the quartile of area of a underlain by karst and incidence rates of campylobacteriosis [IRR = 1.21 (95%CI 1.08, 1.35, p <0.01)] and cryptosporidiosis [IRR = 1.36 (95%CI 1.09, 1.69, p < 0.01)]. Aim 3: There was at least one lag time (weeks between weekly measure of rain and specimen collection) where the amount of rain was positively associated with counts of campylobacteriosis, salmonellosis, and giardiasis. We also evaluated potential associations between enteric disease and rain as well as previous drought conditions (wetness or lack thereof) with various time-lags for 66 counties of PA for a 10-year time, Philadelphia is not included in this analysis because their data was not shared. In the case of Salmonella, there were four different lag times with an increase in incidence (weeks 4, 5, 11 and 12) and Giardia had two different lag times that showed an increase in incidence (weeks 12, and 14 ). At week 11 for cases due to Campylobacter, there was also an assessed interaction, while both rain and drought, as measured by the weekly Palmer Drought Index (PDI) had a positive relationship with risk, the relationship changed direction and was protective [IRR = 0.9997 (95%CI 0.9997, 0.9999, p < 0.01)]. Conclusions: The key findings are: 1. There are distinct profiles of domestic well users that differ by measures of social vulnerability. Two types of census tracts in PA are likely to have a high number of domestic wells and experience high social vulnerability traits. 2. Enteric illnesses due to Salmonella, Campylobacter, Giardia and Cryptosporidium, have patterns of clustering spatially across the state, and have a strong relationship with season. 3. There is a positive association between the area of a county not served by public water supply and incidence rates of campylobacteriosis, cryptosporidiosis, and giardiasis. 4. There is a positive association between the area of a county underlain by karst and incidence rates of campylobacteriosis and cryptosporidiosis. 5. That rain and drought (wetness) conditions weeks to several weeks prior, affect the incidence rate ratio of cases of Giardia, Cryptosporidium, Campylobacter, and Salmonella in Pennsylvania. The relationship between rain and drought conditions and illness varies by organism type which could be attributed to incubation period, reporting time, transport time in the environment and exposure sources (drinking water, food, recreation). There was evidence of an interactive effect between rain and drought conditions for cases of campylobacteriosis, with an 11-week lag time.