John Kent Lin, MD
Peter Veazie, PhD
Seminar date: 5/15/2019
Description: John Kent Lin and Peter Veazie share the spotlight this month.
Risk adjustment is a foundational component of health services research. It has often been assumed that incorporating clinical information from previous years would improve predictive capability. This seminar describes work that evaluates whether incorporating clinical information from previous years improves a risk score’s predictive capability.
The GLM with Gamma family and Log link function is common for modeling costs. However, it is typically specified such that the variance is proportional to the square of the mean (e.g. STATA and SAS glm models). This seminar provides an alternative specification for the GLM with Gamma family and Log link function that can work provide better estimates, standard errors, and R2. The underlying maximum likelihood function will be shown as programmed in STATA and used with STATA’s ml commands. VA cost data are used as an example.
Intended Audience: Health services researchers who use risk adjusted cost models in their research.