OBJECTIVES: Long-term continuous use of HMG-CoA inhibitors (statins) has been shown to be beneficial to patients with coronary artery disease in several large clinical trials. This study demonstrates a promising model to assess the relationship between medication adherence with statins and the risk of adverse event (hospitalization or emergency visit) in a managed care population. METHODS: Conventional outcomes assessments for medication compliance often encounter the problem of identifying the causal relationship between adherence and outcomes (e.g., adverse event). It is not unusual that medication discontinuation occurs after and possibly due to a hospitalization event. To determine the strength of the relationship between statin adherence and risk of hospitalization, a Cox proportional hazard model was developed with the time-varying variable defined as proportion of days covered by statins (PDC) as of the date of first adverse event. For censored subjects, PDC was chosen as of the end of study period or the date of disenrollment, whichever occurred first. 68,974 adult patients were identified as new statin starters during a 2-year period from June 1998 to June 2000. Other covariates included demographics, payer types, previous drug, medical and procedure uses, and comorbidities. RESULTS: The time varying covariate (PDC) was the second most significant predictor of time to adverse event (chi-square<1610.9, p<0.0001). The most significant predictor of an adverse event was the existence of a previous adverse event (chi-square=2043.1, p <0.0001). Patients with higher age, comorbidities and previous high utilization have higher risk. In addition, patients in HMO and POS have significant lower risk than patients with FFS and PPO. CONCLUSIONS: Time-varying PDC is one of the strongest predictors of the risk of an adverse event. Persistence is significantly associated with lower risk of occurrence of hospitalization or emergency visit.