Online Program Home
  My Program

Abstract Details

Activity Number: 597 - New Development for Causal Inference in Health Policy Statistics: A Bayesian Perspective
Type: Invited
Date/Time: Thursday, August 3, 2017 : 8:30 AM to 10:20 AM
Sponsor: Health Policy Statistics Section
Abstract #322182
Title: Does Hospice Reduce End-Of-Life Medical Costs? Evidence from a Bayesian Analysis
Author(s): Fan Li* and Jerry Reiter and David Klemish and Don Taylor
Companies: Duke University and Department of Statistical Science, Duke University and Duke University and Duke University
Keywords: hospice care ; causal inference ; electronic health record ; Principal Stratification

This project aims to evaluate the effect of utilization of hospice care on medical cost among end-of-life patients. The existing literature has focused on comparing patients who utilized hospice versus who did not, adjusting for (e.g. through matching) patient characteristics. However, arguably such an approach fails to take into account of several important sources of confounding, including physicians' attitude towards hospice and the timing of patients' entry to hospice. Additionally, the number of days a patient lives after assignment is a post-treatment outcome and hence, ideally, should not be used in matching. Here, we provide a framework to draw valid causal inferences that address these complications. We will define the treatment as physician's recommendation; we treat patient's hospice care utilization as a confounded post-treatment variable, and use Principal Stratification to draw causal inference. We will also develop a Bayesian statistical model/algorithm to estimate a summary score of patients' health status and thus "eligibility" to hospice based on electronic health record data. We will apply the methodology to the Medicare claims records with

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

Back to the full JSM 2017 program

Copyright © American Statistical Association