Finding Next Year's High-Cost Cases: What Information Matters Most?

• Activity Number: 345
• Type: Topic Contributed
• Date/Time: Wednesday, August 14, 2002 : 2:00 PM to 3:50 PM
• Sponsor: Section on Health Policy Statistics*
• Author(s): Arlene Ash*+
• Affiliation(s): Boston University

Abstract

Consider a group of under-age-65, commercially-insured people receiving health care during a given year who survive at least into the beginning of the next year. The 1% most costly people this year use 32% of this year's health care dollars, and they remain costly (although much less so) in the following year (using 12% of year two's dollars).

We used a nationally representative data base tracking nearly three million people in 1997 and 1998 to address the following questions.: How well does prior cost predict cost next year? Can diagnostic profiles extracted from administrative data bases predict as well (or better) than prior cost? Do the two methods identify substantially different expected high-cost people? What features of a clinical profile lead to the highest expected costs next year?