• John S Hughes, Richard F Averill, Jon Eisenhandler, Norbert I Goldfield, John Muldoon, John M Neff, and James C Gay.
• Department of Medicine, Yale University School of Medicine, New Haven, Connecticut, USA. jshughes@mmm.com
• Med Care. 2004 Jan 1; 42 (1): 81-90.

ObjectiveTo develop Clinical Risk Groups (CRGs), a claims-based classification system for risk adjustment that assigns each individual to a single mutually exclusive risk group based on historical clinical and demographic characteristics to predict future use of healthcare resources. STUDY DESIGN/DATA SOURCES: We developed CRGs through a highly iterative process of extensive clinical hypothesis generation followed by evaluation and verification with computerized claims-based databases containing inpatient and ambulatory information from 3 sources: a 5% sample of Medicare enrollees for years 1991-1994, a privately insured population enrolled during the same time period, and a Medicaid population with 2 years of data.ResultsWe created a system of 269 hierarchically ranked, mutually exclusive base-risk groups (Base CRGs) based on the presence of chronic diseases and combinations of chronic diseases. We subdivided Base CRGs by levels of severity of illness to yield a total of 1075 groups. We evaluated the predictive performance of the full CRG model with R2 calculations and obtained values of 11.88 for a Medicare validation data set without adjusting predicted payments for persons who died in the prediction year, and 10.88 with a death adjustment. A concurrent analysis, using diagnostic information from the same year as expenditures, yielded an R2 of 42.75 for 1994.ConclusionCRGs performance is comparable to other risk adjustment systems. CRGs have the potential to provide risk adjustment for capitated payment systems and management systems that support care pathways and case management.

29 keywords...

hide…