OBJECTIVE: To estimate the one-year cumulative incidence of discontinuation of post-MI beta-blocker therapy after hospital discharge and the one-year cumulative incidence of restarting therapy following discontinuation. METHODS: We conducted a retrospective, population-based, inception cohort study among all enrollees (n=1334) of Group Health Cooperative (GHC, a health maintenance organization) aged 30-79 years who survived a first hospitalized MI during 1986-1996 (mean follow up 1.4 years) and were discharged from the hospital on beta-blocker therapy. Using the GHC computerized pharmacy database, we calculated the duration of therapy by assuming subjects were on average 80% compliant with each beta-blocker prescription fill. We considered subjects to have discontinued therapy on the 60th consecutive day they were without medication and to have restarted therapy on the first day they obtained a prescription fill following discontinuation. In sensitivity analyses, we varied when we considered subjects to have discontinued therapy (i.e., on the 30th and the 90th consecutive day without medication). We estimated the cumulative incidence using failure time analyses. RESULTS: By one year post-discharge, 511 subjects discontinued therapy; the cumulative incidence of discontinuation was 39.0% (95% CI: 36.3, 41.6). The cumulative incidence increased to 50.7% (95% CI: 47.9, 53.3) and decreased to 33.4% (95% CI: 30.9, 36.0) when our discontinuation definition required 30 and 90 days without medication, respectively. By one year post-discontinuation, the cumulative incidence of restarting was 42.7% (95% CI: 39.1, 46.2). The cumulative incidence of restarting increased to 52.9% (95% CI: 49.6, 56.1) and decreased to 32.0% (95% CI: 28.4, 35.6) when our discontinuation definition required 30 and 90 days without medication, respectively. CONCLUSIONS: Our results indicate that patients are likely to restart post-MI beta-blocker therapy after having discontinued it. Studies considering only time to first discontinuation may be oversimplifying the long-term utilization patterns of therapy by misclassifying patients who subsequently restart.