Thursday, February 15 | |
PS1 Poster Session 1 and Opening Mixer |
Thu, Feb 15, 5:30 PM - 7:00 PM
Salons F-I |
Combining Historical Data and Propensity Score Methods in Observational Studies to Improve Internal Validity (303636)Heather Angier, Oregon Health & Science UniversityDavid Ezekiel-Herrera, Oregon Health & Science University Nathalie Huguet, Oregon Health & Science University *Miguel Marino, Oregon Health & Science University Jean P. O'Malley, Oregon Health & Science University Lewis Raynor, OCHIN Teresa Schmidt, OCHIN Rachel Springer, Oregon Health & Science University Steele Valenzuela, Oregon Health & Science University Keywords: Observational Studies, Propensity Score Methods, Longitudinal Data Analysis, Electronic Health Records Randomized experiments are the gold standard for establishing causality between a treatment and an outcome. For many research studies (e.g. studying the impact of gaining Medicaid public health insurance on health outcomes), ethics and/or feasibility prevent conduct of randomized experiments. In lieu of randomization, observational studies provide a rich opportunity to answer many questions. To produce reliable results, however, observational studies must appropriately control for confounders. We will describe how combining historical data and propensity score methods appropriately controls for confounders and thus increases the internal validity of observational studies. We demonstrate this easy-to-implement and widely applicable methodology in a study using electronic health records to evaluate the impact of gaining Medicaid through the Affordable Care Act (a.k.a. Obamacare) on diabetes-related biomarkers (i.e., hemoglobin A1c, low density lipoprotein). Specifically, we compare changes in biomarkers 24 months pre- to 24 months post-ACA among four insurance cohorts: 1) continuously insured, 2) continuously uninsured, 3) discontinuously insured, and 4) gained insurance post-ACA.
|