An exploration and case study of population classification for managed healthcare within a state-based modelling framework

Abstract

The healthcare industry currently accounts for more than 12% of the US Gross Domestic Product (GDP). The industry is struggling with a shift from the fee for service compensation model prevalent during the 20th Century to a delivery model based on some form of managed care. Efforts at implementing capitation rate contracts with physicians have been significantly hampered by the inability to accurately forecast and adjust for risk differences among various practice populations. We approach the problem within the framework of state-based modelling to develop state definitions that represent homogeneous sub-populations. We use the state definitions to develop a First Order Markov Model for healthcare expense prediction and compare performance to the current state of the art expense prediction models on actual data from a family practice in Calgary.