ABSTRACT
Since 1984 the French Communicable Disease Network (FCDN) collects and analyses epidemiological information obtained online from a team of "Sentinel General Practitioners" (SGPs). It redistributes this information in the form of standardised weekly incidence estimates. These weekly estimates now appear on the Internet and are the basis for issuing alerts of influenza epidemics. We postulate that day to day estimations would be highly desirable to achieve timely detection of the actual onset of the epidemic, a need dramatically underscored by the emergence of bioterrorism. The present paper suggests the feasibility of reconstructing daily epidemiological information using local smoothing with a suitable spline function to obtain short latency alert messages.