Estimation of Disease-specific Health Care Costs Using Causal Inference Framework

• Activity Number: 82
• Type: Contributed
• Date/Time: Sunday, July 29, 2012 : 4:00 PM to 5:50 PM
• Sponsor: Health Policy Statistics Section
• Author(s): Irina Bondarenko*+ and Trivellore Raghunathan
• Companies: University of Michigan and University of Michigan
• Keywords: causal inference ; Tukey GH-distributions ; potential outcomes ; health care expenditures ; semi-parametric imputations ; cost attribution

Abstract

Estimation of the costs attributable to various diseases is needed to understand the structure of health care spending and to develop strategies to improve efficiency of health care. This estimation problem can be framed as a causal inference from an observational study. This article will develop and compare various methods for causal inference. We investigate four methods--propensity score stratification, and three approaches based on multiple imputations of counterfactuals, for estimating disease-specific costs. For multiple imputation inference, we use three approaches: (1) Parametric approach using log-normal distribution; (2) Tukey's gh-distribution(GH) on the original scale; and (3) Approximate Bayesian Bootstrap (ABB). Data from the Medicare Current Beneficiary Survey (MCBS) is used to illustrate the methodology. We also evaluate the repeated sampling properties of the estimates through a simulation study.