Study of Predictive Models for Payment Parity and Population Health Management

• Activity Number: 269
• Type: Contributed
• Date/Time: Monday, August 10, 2015 : 2:00 PM to 3:50 PM
• Sponsor: Health Policy Statistics Section
• Author(s): Donghui Wu*
• Companies: LexisNexis Health Care
• Keywords: Predictive Modeling ; Risk Adjustment ; High Risk Identification ; Claims Data Modeling ; Statistical Learning ; Payment Parity

Abstract

Claims data based predictive models have been studied for more than 30 years. They have been widely adopted by CMS, states and insurance plans for risk-adjusted payments to providers in lieu of fee-for-services payments. Most of these predictive models, retrospective, concurrent or prospective, are group-based models, e.g. CMS-HCC Medicare, CDPS Medicaid, ACG, CRG, DxCG, etc. The goal is to pay all providers fairly regardless patient case-mix or severity. The main purpose of Predictive Models for Population Health Management (PHM) is to identify high risk members for proper intervention, i.e. assign them to different type of population health management programs according to their diagnoses and predicted risks. Group-based models are often also used to identify high risk members for PHM. We study the group-based payment parity model and individual attributes based predictive models developed solely for high risk member identification. In particular we compare the CMS-HCC model produced risk scores, and the risk scores produced by predictive models for PHM. The PHM model accuracy is much higher, and member risk stratification is much more accurate.