Seasonal patterns of gastrointestinal illness and streamflow along the Ohio River

Abstract

Waterborne gastrointestinal (GI) illnesses demonstrate seasonal increases associated with water quality and meteorological characteristics. However, few studies have been conducted on the association of hydrological parameters, such as streamflow, and seasonality of GI illnesses. Streamflow is correlated with biological contamination and can be used as proxy for drinking water contamination. We compare seasonal patterns of GI illnesses in the elderly (65 years and older) along the Ohio River for a 14-year period (1991-2004) to seasonal patterns of streamflow. Focusing on six counties in close proximity to the river, we compiled weekly time series of hospitalizations for GI illnesses and streamflow data. Seasonal patterns were explored using Poisson annual harmonic regression with and without adjustment for streamflow. GI illnesses demonstrated significant seasonal patterns with peak timing preceding peak timing of streamflow for all six counties. Seasonal patterns of illness remain consistent after adjusting for streamflow. This study found that the time of peak GI illness precedes the peak of streamflow, suggesting either an indirect relationship or a more direct path whereby pathogens enter water supplies prior to the peak in streamflow. Such findings call for interdisciplinary research to better understand associations among streamflow, pathogen loading, and rates of gastrointestinal illnesses.

Keywords: drinking water quality; gastrointestinal infections; hydrology; pathogens; seasonality; streamflow.

Figure 1

Selected cities and characteristics of the Ohio River watershed. Cities (populations >50,000 persons) which fall within a 10-mile buffer of the Ohio River (Panel A). Average annual temperature (°C) by county (Panel B) and annual cumulative precipitation (mm) by county in the Ohio River watershed (PRISM Data - http://www.prism.oregonstate.edu/) (Panel C). Percent of county public water supply from surface water sources in the Ohio River watershed (Panel D) (USGS Data, http://water.usgs.gov/watuse/data/2000/index.html).

Figure 2

Cross correlations between streamflow and outcomes rates for GI symptoms (Panel A), Ill-defined GI infections (Panel B), and All GI infections (Panel C) with lags up to 25 weeks for each of the six selected counties. ( Allegheny Co, PA, Cabell Co, WV, Hamilton Co, OH, Jefferson Co, KY, Daviess Co, KY, Vanderburgh Co, IN).

Figure A

Details of the derivation of the seasonal curve properties based on the values predicted by the Poisson harmonic regression.