The purpose of this study is to investigate the relationship between the number of influenza-like illness (ILI), weekly registered by the general practitioners (sentinel stations), and the monthly overall influenza mortality in people over 60 years of age, provided by the Dutch Statistical Bureau during the period July 1970 to June 1989. The quantitative impact of influenza-morbidity is expressed by three summary parameters, calculated from the 52(53) weekly ILI-f igures per season-year, (i) their sum (i.e. global extent of an epidemic), (ii) their standard deviation, and (iii) their maximum (i.e. peak number of ILI during an epidemic). In the analysis influenza A subtype is also included. These four parameters are mutually compared with respect to their predictability for yearly total influenza mortality in the 19 season-years available. In most cases, the standard deviation and the peak number of ILI are more powerful for prediction of mortality, thanthe global extent of the epidemic. The peak number of ILI is of special interest. It is particularly useful for estimating the effect on current influenza mortality during an ongoing epidemic. From the model it is possible to calculate a threshold (of weekly ILI) beyond which mortality increases proportionately more than the number of illness episodes. By using the peak value of morbidity it is possible to calculate the minimal impact of epidemic mortality. This study indicates that the weekly number of influenza-like illnesses has a certain prognostic value for the real impact of influenza. An electronic surveillance system could detect immediately the threshold above which influenza mortality increases more than proportionally. When this level is reached electronic bulletins could stress the importance of prophylactic measures or, especially for the high-risk patient, the use of amantadine to reduce excess mortality. The benefits of such an electronic surveillance system are not restricted to the epidemic. We believe that sucha system could improve the ...