Influenza vaccine effectiveness evaluations were carried out among the elderly, as part of a demonstration established to estimate the value of including influenza vaccination as a covered Medicare benefit. Cases hospitalized with pneumonia and influenza-related diagnoses during November through April were identified and group matched to randomly selected community controls. Data were collected from cases and controls on influenza vaccination status and other factors which could have confounded the association between vaccination and hospitalization. A community-based influenza surveillance programme was conducted each year to determine the timing and aetiology of influenza activity. Logistic regression analyses were carried out to evaluate the association of influenza vaccination with the likelihood of hospitalization. In 1990-1991, during the peak of the influenza type B outbreak, influenza vaccination was estimated to be 31% (95% CI: 4-51%) effective in reducing the likelihood of hospitalization. In 1991-1992, during the peak of the influenza type A(H3N2) epidemic, a nearly identical point estimate for vaccine effectiveness was demonstrated (32%, 95% CI: 7-50%). Identical analyses carried out each year during the periods of low or absent influenza activity failed to demonstrate a significant effect for vaccination in preventing hospitalization. Results indicated that a significant benefit for vaccination could be expected during both type A and type B influenza seasons.