Identifying Future High Cost Users of Medical Care
by
Arlene S. Ash
Boston University School of Medicine and DxCG, Inc.
Coauthors: Yang Zhao, DxCG, Inc.

A fundamental need in managing health care is the abiity to prospectively identify people who, without management, are likely to incur high medical costs "next year." We introduce the idea of a fixed-size (small) "Top Group" of patients who are good candidates for case management - as a natural and practical way to compare the value of various methods for case selection. We apply these ideas to data for over a million members of employer-sponsored benefit plans from Medstat's MarketScan Research Database (1997-1999). In comparing a wide-range of methods (based on previous diagnoses and/or costs) for identifying 0.5-sized Top Groups, we identify trade-offs in achieving either the highest average costs versus the highest prevalence of manageable disease. The best models for forming Top Groups have a surprising structure, and 3/4 of their members go on to cost

Date received: August 29, 2003