Approximate Bayesian Bootstrap Procedures to Estimate Multilevel Treatment in Observational Studies with Application to Type 2 Diabetes Treatment Regimens

• Activity Number: 447 - Recent Advances in Propensity Score Methods for Observational Studies with Multiple Treatments
• Type: Invited
• Date/Time: Wednesday, July 31, 2019 : 8:30 AM to 10:20 AM
• Sponsor: Health Policy Statistics Section
• Author(s): Roee Gutman* and Anthony D. Scotina and Robert J Smith and Andrew R Zullo
• Companies: Brown University and Simmons University and Brown University and Brown University
• Keywords: Approximate Bayesian Bootstrap; Causal Inference; Multiple Imputation; Multiple Treatments; Sensitivity Analysis

Abstract

Randomized clinical trials are considered the goal standard for estimating causal effects. Nevertheless, in studies that are aimed at examining adverse effects of interventions these are often impractical because of ethical and financial considerations. In observational studies, matching on the generalized propensity scores was proposed as a possible solution to estimate the treatment effects of multiple interventions. However, the derivation of point and interval estimates for these matching procedures can become complex with non-continuous or censored outcomes. We propose novel approximate Bayesian bootstrap algorithms that result in statistically valid point and interval estimates of the treatment effects with dichotomous outcomes. The procedures rely on the estimated generalized propensity scores and multiply impute the unobserved potential outcomes for each unit. In addition, we describe a corresponding easily interpretable sensitivity analysis to examine the unconfoundedness assumption. The procedures are motivated and illustrated using an observational study that examines the cardiovascular safety of common, real-world anti-diabetic treatment regimens for Type 2 diabetes.