Medical costs are confounded with differences in patients' demographic and clinical characteristics, and there is a need to adjust for heterogeneity in the population. In addition, healthcare costs present special modeling challenges as they often have a large fraction of observations with zeros, and distributions of positive values that are skewed, non-normal, and heteroskedastic.
This module provides a general view of simple regression analysis with an explanation of why a standard simple approach fails in the estimation of medical costs. Furthermore, an overview of quantitative methods for modeling health care costs will be presented, including propensity score matching, multivariate analysis, and selection bias models. The module includes proposed approaches as well as guidelines to choose among these possible approaches.
The module consists of a presentation and a discussion part. The presentation includes tests of the underlying assumptions, a unified framework for computing marginal effects, and links to recent research. The discussion will revolve around how to choose among alternative estimation strategies that are sensitive to different types of health costs applications. Examples with SAS and STATA codes will be presented during the module.
By the end of the Modeling Health Care Costs Module, you will be able to:
- Understand the basic principles of quantitative methods for modeling health care costs;
- Understand the modeling challenges related to modeling health care costs; and
- Be able to choose among alternative estimation strategies.