A time series study of drug sales and turbidity of tap water in Le Havre, France.

The 80,000 inhabitants of the lower part of Le Havre obtain their water supply from two karstic springs, Radicatel and Saint-Laurent. Until 2000, the Radicatel water was settled when turbidity exceeded 3 NTU, then filtered and chlorinated, whereas the Saint-Laurent water was simply chlorinated. Our study aimed to characterize the link between water turbidity and the incidence of acute gastroenteritis (AGE). Records on drug sales used for the treatment of AGE were collected from January 1994 to June 1996 (period 1) and from March 1997 to July 2000 (period 2). Daily counts of drug sales were modeled using a Poisson Regression. We used data set 2 as a discovery set, identifying relevant (i.e. both significant and plausible) exposure covariates and lags. We then tested this model on period 1 as a replication dataset. In period 2, the daily drug sales correlated with finished water turbidity at both resources. Settling substantially modified the risk related to turbidity of both raw and finished waters at Radicatel. Correlations were reproducible in period 1 for water from the Radicatel spring. Timeliness of treatment adaptation to turbidity conditions appears to be crucial for reducing the infectious risk due to karstic waters.