Institutional members access full text with Ovid®

Share this article on:

Assessing the Impact of Propensity Score Estimation and Implementation on Covariate Balance and Confounding Control Within and Across Important Subgroups in Comparative Effectiveness Research

Girman, Cynthia J. DrPH*,†; Gokhale, Mugdha MS; Kou, Tzuyung Doug PhD; Brodovicz, Kimberly G. PhD*; Wyss, Richard MS; Stürmer, Til MD, PhD

doi: 10.1097/MLR.0000000000000064
Original Articles

Purpose: Researchers are often interested in estimating treatment effects in subgroups controlling for confounding based on a propensity score (PS) estimated in the overall study population.

Objective: To evaluate covariate balance and confounding control in sulfonylurea versus metformin initiators within subgroups defined by cardiovascular disease (CVD) history comparing an overall PS with subgroup-specific PSs implemented by 1:1 matching and stratification.

Methods: We analyzed younger patients from a US insurance claims database and older patients from 2 Medicare (Humana Medicare Advantage, fee-for-service Medicare Parts A, B, and D) datasets. Confounders and risk factors for acute myocardial infarction were included in an overall PS and subgroup PSs with and without CVD. Covariate balance was assessed using the average standardized absolute mean difference (ASAMD).

Results: Compared with crude estimates, ASAMD across covariates was improved 70%–94% for stratification for Medicare cohorts and 44%–99% for the younger cohort, with minimal differences between overall and subgroup-specific PSs. With matching, 75%–99% balance improvement was achieved regardless of cohort and PS, but with smaller sample size. Hazard ratios within each CVD subgroup differed minimally among PS and cohorts.

Conclusions: Both overall PSs and CVD subgroup-specific PSs achieved good balance on measured covariates when assessing the relative association of diabetes monotherapy with nonfatal myocardial infarction. PS matching generally led to better balance than stratification, but with smaller sample size. Our study is limited insofar as crude differences were minimal, suggesting that the new user, active comparator design identified patients with some equipoise between treatments.

*Department of Epidemiology, Merck Sharp & Dohme, North Wales, PA

Department of Epidemiology, University of North Carolina, Chapel Hill, NC

Department of Global Pharmacovigilance & Epidemiology, Bristol Meyers Squibb, Hopewell, NJ

T.D.K. was at Merck Sharp & Dohme when this work was performed.

Supported by grants from Merck & Co. Inc., and NIA R01 AG023178 “Propensity Scores & Preventive Drug Use in the Elderly.”

Parts of this research were presented as an oral presentation at the International Conference on Pharmacoepidemiology and Therapeutic Risk Management, August 24, 2012, Barcelona, and in poster form at the DEcIDE Methods workshop, Agency for Healthcare Research and Quality, June 12–13, 2012.

Commercial employment: Merck: C.J.G. and K.G.B.; BMS: T.D.K. Advisory Board: ASSESS Steering Comm (RTI, GSK): T.S. Industry-sponsored grants: GSK: T.S.; Sanofi: T.S. Stock ownership/options: Merck: C.J.G., and K.G.B.; BMS: T.D.K. The remaining authors declare no conflict of interest.

Reprints: Cynthia J. Girman, DrPH, Department of Epidemiology, UG1D-60, Merck Sharp & Dohme, 351 N Sumneytown Pike, North Wales, PA 19454. E-mail:

© 2014 by Lippincott Williams & Wilkins.