Abstract P201: Use of Medicare Claims Data for the Identification of Myocardial Infarction. The REasons for Geographic And Racial Differences in Stroke (REGARDS) Study

Abstract

Background: Administrative claims are used to conduct epidemiology research. However, few studies have compared results using administrative claims versus primary data collection. Objective: To compare myocardial infarction (MI) rates using primary data collection and Medicare claims in the REasons for Geographic and Racial Differences in Stroke (REGARDS) study. Methods: We included 9,951 REGARDS study participants ≥65 years of age with Medicare Part A (inpatient) fee-for-service coverage at baseline in 2003-2007. Participants were asked every 6 months to report heart-related hospitalizations which were subsequently adjudicated to detect definite or probable MIs (MI Definition 1, see Table footnotes). Events detected through surveillance in REGARDS were supplemented with adjudicated definite or probable MIs detected through Medicare inpatient claims with a diagnosis code for MI (i.e., ICD-9 code 410.x0 or 410.x1) in any position (Definition 2) and in the primary position (Definition 3). MIs were also defined by a Medicare inpatient claim with a code for MI in any position (Definition 4) and in the primary position (Definition 5), without further adjudication. MIs were ascertained through December 2012. Results: REGARDS study procedures detected and adjudicated 669 definite or probable MIs over a mean follow-up of 6.3 years, representing 10.7 MIs per 1,000 person-years (Definition 1, Table ). Supplementing adjudicated MIs with Medicare inpatient claims resulted in a 12% (any diagnosis position, Definition 2) and 6% (primary diagnosis position, Definition 3) higher rate for MI. Using only Medicare claims without adjudication underestimated the rate for MI by 8% (any diagnosis position, Definition 4) and 32% (primary diagnosis position, Definition 5), compared with REGARDS study procedures. Conclusion: Supplementing MIs detected through participant self-report with those identified in claims could improve event detection. Using only Medicare claims to identify events may underestimate MI rates.