Exploring the Sensitivity of Propensity Score Matching Analyses to Unobserved Covariates in the Context of an Intervention to Reduce Hospitalization Rates

• Activity Number: 550
• Type: Contributed
• Date/Time: Wednesday, August 7, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Sitaram Vangala*+
• Companies: UCLA Department of Medicine Statistics Core
• Keywords: Propensity score ; Observational data ; Causal inference ; Sensitivity analysis

Abstract

The method of propensity score matching (PSM), often used in analyses of observational data, can correct for imbalanced pre-treatment variable distributions between experimental groups, thereby facilitating causal inference. PSM may not succeed, however, if unobserved variables markedly influence treatment assignment. In a study measuring the efficacy of a patient-centered medical home intervention in reducing hospitalization rates, PSM was used to pair patients in a treatment clinic who utilized the intervention with patients in a control clinic who did not have access to it. A sensitivity analysis based upon Rosenbaum and Rubin (1983)[1] was performed to identify the extent to which unobserved confounders (e.g., patient motivation) render uncertain the PSM-based estimate of the intervention effect.

[1] P. R. Rosenbaum and D. B. Rubin. "Assessing Sensitivity to an Unobserved Binary Covariate in an Observational Study with Binary Outcome." J. R. Statist. Soc. B (1983), 45, No. 2, pp. 212-218.