B1105
Title: Causal framework for subgroup treatment evaluation using multivariate generalized mixed effect models
Authors: Yizhen Xu - Johns Hopkins University (United States) [presenting]
Jisoo Kim - Johns Hopkins University (United States)
Ami Shah - Johns Hopkins University (United States)
Scott Zeger - Department of Biostatistics-- Johns Hopkins Bloomberg School of Public Health (United States)
Abstract: Dynamic prediction of causal effects under different treatment regimes conditional on individual's characteristics and longitudinal history is an essential problem in precision medicine. This is a challenging problem in practice because outcomes and treatment assignment mechanisms are unknown in observational studies, individual's treatment efficacy is a counterfactual, and the existence of selection bias is empirically untestable. We propose a framework for identifying the long-term individualized treatment effect adjusting for unobserved stable trait factors, using Bayesian G computation with multivariate generalized mixed effect models. Existing methods mostly focus on balancing the confounder distributions of observables between different treatments, while our proposal also accounts for a latent tendency towards each treatment due to unobserved time-invariant factors. We assume sequential ignorability conditional on unobserved stable trait factor in treatment assignment, and dynamically updates stable unobserved factors in outcomes progression as an individual's history data increases over time. Our framework naturally incorporates sensitivity analysis, providing an alternative to defining an additional sensitivity parameter for quantifying the impact of unmeasured confounding.
