All ETDs from UAB

Advisory Committee Chair

Paul Muntner

Advisory Committee Members

Monika M Safford

Emily B Levitan

Meredith L Kilgore

Huifeng Yun

Document Type

Dissertation

Date of Award

2016

Degree Name by School

Doctor of Philosophy (PhD) School of Public Health

Abstract

Claims are being used for the identification of myocardial infarctions (MI) in research studies, either as a complement of primary data collection or as the sole data source. However, few data are available to support these approaches. The current dissertation analyzes the use of claims for the identification of MIs in a large population-based cohort linked to Medicare data, the REasons for Geographic and Racial Differences in Stroke study. Analyses included 9,951 study participants ≥65 years of age with Medicare Part A (inpatient) fee-for-service coverage at baseline in 2003-2007. Participant self-report and Medicare inpatient claims with a diagnosis code for MI (i.e., international classification of diseases, ninth revision codes 410.x0-410.x1) were used to identify hospitalizations through December 31, 2012 which were subsequently classified as MIs or no MIs by expert adjudication. Using expert adjudication as the reference standard, Medicare claims with a code for MI in any and in the primary position had a sensitivity of 49.0% and 40.1%, respectively, and a positive predictive value (PPV) of 84.3% and 89.7%, respectively. Very small non-ST-segment elevation MIs and type 2 and type 4-5 MIs were more likely to be identified through participant self-report but not present in claims. The rate for MI was 10.7 per 1,000 person-years using MIs identified through participant self-report. Adding MIs identified through claims with a code for MI in any and in the primary position increased the rate for MI by 12% and 6%, respectively. Using claims with a code for MI in any and in the primary position without adjudication underestimated the rate for MI by 8% and 32%, respectively compared with using MIs identified through participant self-report. Associations between participant characteristics and MI were similar using MIs identified through participant self-report, MIs identified through participant self-report or Medicare claims, and Medicare claims with a code for MI without adjudication. In conclusion, Medicare claims have high PPV but moderate sensitivity for MI identification. The improvement in MI identification using Medicare claims as a complement of participant self-report is small. Using claims data as a complement of participant self-report for MI identification may not be needed.

Included in

Public Health Commons

Share

COinS