Background: Data collected as part of routine clinical practice could be used to detect cardiovascular outcomes in pragmatic clinical trials or clinical registry studies. The reliability of claims data for documenting outcomes is unknown.
Methods and results: We linked records of Women's Health Initiative (WHI) participants aged ≥65 years to Medicare claims data and compared hospitalizations that had diagnosis codes for acute myocardial infarction or coronary revascularization with WHI outcomes adjudicated by study physicians. We then compared the hazard ratios for active versus placebo hormone therapy based solely on WHI-adjudicated events with corresponding hazard ratios based solely on claims data for the same hormone trial participants. Agreement between WHI-adjudicated outcomes and Medicare claims was good for the diagnosis of myocardial infarction (κ, 0.71-0.74) and excellent for coronary revascularization (κ, 0.88-0.91). The hormone:placebo hazard ratio for clinical myocardial infarction was 1.31 (95% confidence interval, 1.03-1.67) based on WHI outcomes and 1.29 (95% confidence interval, 1.00-1.68) based on Medicare data. The hazard ratio for coronary revascularization was 1.09 (95% confidence interval, 0.88-1.35) based on WHI outcomes and 1.10 (95% confidence interval, 0.89-1.35) based on Medicare data. The differences between hazard ratios derived from WHI and Medicare data were not significant in 1000 bootstrap replications.
Conclusions: Medicare claims may provide useful data on coronary heart disease outcomes among patients aged ≥65 years in clinical research studies.
Clinical trials registration information: URL: www.clinicaltrials.gov. Unique identifier: NCT00000611.
Keywords: myocardial infarction; patient outcome assessment; pragmatic clinical trials; randomized controlled trials; validation studies.