This serial cohort study assessed the risk of hospitalization or death associated with influenza and the effectiveness of influenza vaccination among subgroups of elderly members of 3 managed-care organizations in the United States. Data on baseline characteristics and outcomes were obtained from computerized databases. A total of 122,974 (1996-1997 season) and 158,454 (1997-1998 season) persons were included in the cohorts. Among unvaccinated persons, hospitalizations for pneumonia/influenza or death occurred in 8.2 of 1000 healthy and 38.4 of 1000 high-risk persons in year 1, and in 8.2 of 1000 healthy and 29.3 of 1000 high-risk persons in year 2. After adjustments, vaccination was associated with a 48% reduction in the incidence of hospitalization or death (95% confidence interval [CI], 42-52) in year 1 and 31% (95% CI, 26-37) in year 2. Effectiveness estimates were statistically significant and generally consistent across the healthy and high-risk subgroups. The absolute risk reduction, however, was 2.4- to 4.7-fold higher among high-risk than among healthy elderly persons. All elderly individuals may substantially benefit from vaccination. However, the impact of influenza is greater in persons with high-risk medical conditions.