Title: A Bayesian nonparametric approach for missing data for causal inference in EHRs that uses auxiliary information
Authors: Michael Daniels - University of Florida (United States) [presenting]
Sebastien Haneuse - Harvard TH Chan School of Public Health (United States)
David Lindberg - University of Florida (United States)
Abstract: A Bayesian nonparametric using enriched Dirichlet process mixtures is proposed to model the observed data in EHRs with an ultimate goal of causal inference. Missing data (in confounders and the outcome) is allowed to be nonignorable and auxiliary information in the EHR can be exploited to move the missingness closer to MAR. We illustrate the approach via simulations and a data example.