The 80,000 residents of the lower town of Le Havre (France) receive their water supply from the karst springs of Radicatel and Saint-Laurent. Every year, turbidity events affect these springs, leading to microbiological degradation of the water. The study period ran from March 1997 to July 2000. At that time, the water was chlorinated prior to distribution, and the water from Radicatel was pre-clarified. The distributed water complied with European regulations. The study aimed to establish and characterize the link between water supply conditions and the incidence of gastroenteritis. Five pharmacists provided sales records for medications used to treat acute gastroenteritis (AEG). Turbidity, free chlorine, precipitation, and flow rates were tested as exposure proxies. Daily medication sales were modeled using a "GAM" framework. We systematically explored the data to identify significant covariates and lags. To be considered, these had to be both significant and plausible. Decreases in air temperature, chlorination failures in Saint-Laurent, and the turbidity of raw and treated water were all significantly correlated with an increase in daily sales. Most risks peaked for delays ranging from 5 to 10 days, a latency consistent with the incubation period of GEA. Our results suggest that, in the field of "GEA and turbidity" time-series studies, the key conditions for accurate risk quantification lie in (i) the sensitivity of the morbidity indicator, (ii) the plausibility of turbidity as a proxy for exposure to pathogens, and (iii) the temporal variability of the exposure signal. We plan to study 10 additional sites to model the meta-risk associated with turbidity in drinking water. (R.A.)