[General practice and surveillance: the example of influenza in France]

Data registration by the GROG national network (Regional Group for the Surveillance of Influenza) since 1984 has helped to identify methodological problems. The choice of sentinels and the selection of indicators depend on the analysis of actual influenza consequences. Various sentinels may be concerned: health insurance companies, private companies, schools, physicians, pharmacists. Health care organization modifies the validity of indicators. In France, for instance, home visits were an excellent indicator for early warning before 1995 but this indicator is no longer as efficient as before. The virological detection of Influenza depends on the organization of transportation (samples, results). The predictive value of cases definitions depends on the incidence of infection. The level of choice between specificity and sensibility modifies the perception of outbreaks. Sentinel participation rate influences the sample representativity. The farther this rate is from 100%, the more result validity decreases (in the same way that patients lost to follow-up compromise the validity of results in clinical trials). The publication of results can modify health expenses and behaviors.

The GROG network stresses the important role that general practitioners play in health surveillance; it also raises questions in the field of mathematics, statistics, professional organization, training, education, and politics.