Online Program Home
My Program

Abstract Details

Activity Number: 167 - SPEED: Missing Data and Causal Inference Methods, Part 1
Type: Contributed
Date/Time: Monday, July 29, 2019 : 10:30 AM to 12:20 PM
Sponsor: Health Policy Statistics Section
Abstract #304890
Title: Heterogeneous Treatment Effects with Subgroups via the Overlap Weights
Author(s): Elizabeth Lorenzi*
Companies:
Keywords: subgroup analysis; causal inference; propensity scores; heterogeneous treatment effects; machine learning
Abstract:

Learning heterogeneous treatment effects across subgroups is becoming the norm rather than the exception in most comparative effectiveness research. However, guidance for observational comparisons within subgroups is lacking as is a connection to classic design principles for propensity score (PS) analyses. We address these shortcomings by proposing a novel PS method for subgroup analysis that seeks to balance existing strategies in an automatic and efficient way. With the use of overlap weights, we prove that a propensity model including interactions between subgroups and all covariates results in exact covariate balance within subgroups. We estimate the propensity score model using machine-learning methods to better select the appropriate interaction terms and to address the inherent bias-variance tradeoff. We explore these ideas through simulations, showing that overlap weights estimated using a post-LASSO propensity model (GLM with LASSO selected variables) provide better subgroup balance and therefore lower bias, compared to IPW and other model choices. We also discuss results in the context of procedural comparisons in the COMPARE-UF registry of women with uterine fibroids.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2019 program