A time series construction of an alert threshold with application to S. bovismorbificans in France.

Epidemiological surveillance can be used to initiate preventive measures and Salmonella infection is a prime example of application. It is important that any unusual increase of reported cases should be detected as rapidly as possible. Reliable forecasts based on a suitable model of the temporal structure of a pathological indicator are necessary for the elaboration of an alert threshold. We first discuss the specific problems associated with the identification and the estimation of the temporal structure of indicators of infectious diseases when they present both epidemic peaks and an underlying stationary structure. Then we consider how an alert threshold can be defined. Our application concerns infections caused by S. bovismorbificans in France for which an alert threshold is constructed and tested both retrospectively and prospectively.