Heavy precipitation, drinking water source, and acute gastrointestinal illness in Philadelphia, 2015-2017

Abstract

Runoff from heavy precipitation events can lead to microbiological contamination of source waters for public drinking water supplies. Philadelphia is a city of interest for a study of waterborne acute gastrointestinal illness (AGI) because of frequent heavy precipitation, extensive impervious landcover, and combined sewer systems that lead to overflows. We conducted a time-series analysis of the association between heavy precipitation and AGI incidence in Philadelphia, served by drinking water from Delaware River and Schuylkill River source waters. AGI cases on each day during the study period (2015-2017) were captured through syndromic surveillance of patients' chief complaint upon presentation at local emergency departments. Daily precipitation was represented by measurements at the Philadelphia International Airport and by modeled precipitation within the watershed boundaries, and we also evaluated stream flowrate as a proxy of precipitation. We estimated the association using distributed lag nonlinear models, assuming a quasi-Poisson distribution of the outcome variable and with adjustment for potential confounding by seasonal and long-term time trends, ambient temperature, day-of-week, and major holidays. We observed an association between heavy precipitation and AGI incidence in Philadelphia that was primarily limited to the spring season, with significant increases in AGI that peaked from 8 to 16 days following a heavy precipitation event. For example, the increase in AGI incidence related to airport precipitation above the 95th percentile (vs no precipitation) during spring reached statistical significance on lag day 7, peaked on day 16 (102% increase, 95% confidence interval: 16%, 252%), and declined while remaining significantly elevated through day 28. Similar associations were observed in analyses of watershed-specific precipitation in relation to AGI cases within the populations served by drinking water from each river. Our results suggest that heavy precipitation events in Philadelphia result in detectable local increases in waterborne AGI.

Fig 1. City of Philadelphia study region showing the Schuylkill River and Delaware River watershed areas upstream of Philadelphia Water Department intakes.

Fig 2

Timeseries plots of daily values during the study period, for: a) AGI counts identified by syndromic surveillance; b) total precipitation recorded at the Philadelphia International Airport.

Fig 3. Association between daily combined streamflow (average of Schuylkill and Delaware River streamflow values) and AGI incidence in Philadelphia across lag days, comparing AGI at streamflow levels >95th percentile to ≤median (RRs and 95% confidence interval shown), with adjustment for temperature, day-of-week, holidays, season, and temporal trends (natural spline variables for consecutive day of the study and day of the calendar year).

Fig 4. Association of daily precipitation (PHL) with AGI incidence, comparing AGI at exposure values >95^th percentile to ≤median (first RR peak after lag 0 shown), across lags and by season (smoothed RRs in red, smoothed 95% CIs in grey) with adjustment for temperature, day-of-week, holidays, and temporal trends (natural spline variables for consecutive day of the study and day of the calendar year).

Fig 5

Association between daily precipitation and AGI incidence in Philadelphia during the spring season across lag days, fit using natural spline with comparison of AGI at each precipitation level to the median level (exposure estimated as mean precipitation in 50-mile watershed region), with adjustment for temperature, day-of-week, holidays, and temporal trends (natural spline variables for consecutive day of the study and day of the calendar year), for: a) Philadelphia, overall; b) Population with Delaware River source water; c) Population with Schuylkill River source water.