Title: Bayesian nonparametrics for comparative effectiveness research in EHRs
Authors: Michael Daniels - University of Florida (United States) [presenting]
Abstract: A Bayesian nonparametric approach is proposed to address both confounding and selection bias for inference using electronic health records (EHRs). Data provenance, the collection of decisions that give rise to the observed data, is modularized and modeled to properly adjust for the selection bias due to missing data. The approach is motivated by a study to assess the long terms effects of bariatric surgery on weight gain.