Estimating incidence of bacterial meningitis with capture-recapture method, Lazio Region, Italy.

To estimate the incidence of bacterial meningitis in the Lazio Region, including the city of Rome, and to assess the quality of the surveillance systems, we adopted a multiple-capture model by merging cases from three sources available in 1995-1996: the Notifiable Disease Surveillance (NDS) system, the Special Hospital Surveillance (SHS) system and the Hospital Discharge (HD) registry. A medical record revision was carried out to confirm the cases of bacterial meningitis. A total of 199 individuals was classified as probable or confirmed cases of bacterial meningitis in 1995-1996. In this period, the incidence of reported meningitis was 3.8/100,000 (population = 5,209,633). The log-linear model yielded a total estimated number of 236 cases (95% confidence interval (CI): 206-306), the estimate of incidence reaching the value of 4.5/100,000. Hospital Discharge registry showed the highest sensitivity (77%), the SHS system the highest positive predictive value (83%). In 1997-1998, the meningitis surveillance was integrated with an additional laboratory-based source and yielded 326 cases, with an incidence of reported cases of 6.3/100,000. Laboratory surveillance, involving 115 (92%) public hospitals and 84 (57%) private clinics, contributed 35 (27%) cases in addition to those notified to NDS (n = 130). Multiple-capture models, in our experience could estimate the bacterial meningitis incidence with a very good approximation. In order to improve both sensitivity and positive predictive value of surveillance, hospital and public health sources should be integrated with laboratory-based system.