OBJECTIVE: To highlight data limitations, the need to improve data collection, the need to develop better analytic methods, and the need to use alternative data sources to conduct research related to the Medicare program. Objectives were achieved by reviewing existing studies on risk selection in Medicare HMOs, examining their data limitations, and introducing a new approach that circumvents many of these shortcomings. DATA SOURCES: Data for years 1995-97 for five states (Arizona, Florida, Massachusetts, New York, and Pennsylvania) from the Healthcare Cost and Utilization Project (HCUP) State Inpatient Databases (SIDs), maintained by the Agency for Healthcare Research and Quality; and the Health Care Financing Administration's Medicare Managed Care Market Penetration Data Files and Medicare Provider Analysis and Review Files. STUDY DESIGN: Analysis of hospital utilization rates for Medicare beneficiaries in the traditional fee-for-service (FFS) Medicare and Medicare HMO sectors and examination of the relationship between these rates and the Medicare HMO penetration rates. PRINCIPAL FINDINGS: Medicare HMOs have lower hospital utilization rates than their FFS counterparts, differences in utilization rates vary across states, and HMO penetration rates are inversely related to our rough measure of favorable selection. CONCLUSIONS: Substantial growth in Medicare HMO enrollment and the implementation of a new risk-adjusted payment system have led to an increasing need for research on the Medicare program. Improved data collection, better methods, new creative approaches, and alternative data sources are needed to address these issues in a timely and suitable manner.
