Abstract 26: Comparison of Medicare Claims with Clinical Trial Outcomes for Follow-up of Older Individuals with Acute Coronary Syndrome

Background: Medicare administrative data has been proposed as an efficient alternative for long-term event detection. The accuracy of these data relative to investigator reported and event adjudicated results are unknown. Methods: We linked 3217 older patients (65+ years) with acute coronary syndromes (ACS) from 4 ACS clinical trials to Medicare inpatient standard analytic files (SAFs) using a deterministic linkage and indirect patient identifiers (e.g., sex, date of birth, date of index hospitalization). We compared event ascertainment for death, all-cause rehospitalization, myocardial infarction (MI), heart failure and revascularization to 180 days after hospital discharge. Up to four Medicare ICD-9-CM coding algorithms were evaluated for each outcome. Sensitivity, specificity and positive and negative predictive values were assessed for Medicare events considering trial records as the ‘gold standard’. Results: The 180 day incidences of adverse events following discharge were: death 4.6%, all-cause rehospitalization 50.9%, MI 5.6%, heart failure 15.1%, and any revascularization 13.2%. For most outcomes, Medicare claims displayed similar or higher event rates as those reported for trials. In general, Medicare claims had high sensitivity, specificity, and positive and negative predictive value when compared to trial outcomes; however, these differed depending on the Medicare coding algorithm used (Table). Conclusions: Medicare claims detect similar or greater numbers of events as those identified by existing clinical trial mechanisms. These data demonstrate the potential utility of administrative claims for efficient long-term follow-up of older patients enrolled in clinical trials or post-market surveillance efforts. Further investigation is needed to both optimize Medicare coding algorithms and determine which mechanism is most accurate among discordant reported events.