reports

Minimal-burden risk adjusters for the Medicare risk program

Medical College of Virginia Associated Physicians, Mathematica Policy Research. Report prepared for the Health Care Financing Review (HCFA Contract No. 17C-90366/3/01). Washington, DC, USA: Mathematica Policy Research.

The Medicare risk program, often touted as a solution to the high rates of increase in Medicare costs, will not save the government money if payments to risk plans exceed the cost that HCFA would have incurred under traditional fee-for-service (FFS) Medicare. Most research to date (Brown et al. 1993; Riley et al. 1996; and U.S. General Accounting Office 1997) has shown that, in fact, this is what occurs. Beneficiaries with serious health problems are less likely than other beneficiaries to enroll in a risk plan. Estimates suggest that HCFA pays 6 or 7 percent more to plans than it would have spent had these enrollees been in traditional FFS Medicare, even though the payments are set at 95 percent of the average FFS costs that HCFA expects to incur for beneficiaries with similar demographic characteristics. The overpayment occurs because the mechanism for setting the capitation rates that risk plans are paid for providing coverage of Medicare services fails to reflect health status adequately. The payment method, based on the Adjusted Average Per Capita Cost (AAPCC), adjusts for some factors that are associated with Medicare costs, including age, sex, whether enrolled in Medicaid, whether residing in a nursing home, and county of residence. However, studies have repeatedly shown that this measure explains only about one percent of the variance in beneficiary cost.

This concern prompted HCFA to fund several studies, including this one, to develop more effective risk adjusters for the general Medicare population. HCFA has decided to phase out the AAPCC-based adjuster by blending it with one of these adjusters, the Principal Inpatient Diagnostic Cost Group (PIPDCG), beginning in January 2000. The PIPDCG model sets payments to plans for a given enrollee based on the principal diagnoses for hospital admissions the enrollee had during the prior year, plus the beneficiary’s age, sex, and Medicaid status. The weight on the PIPDCG component in this blended rate will gradually increase over time. In 2004, HCFA intends to move to a risk adjuster that includes diagnoses from other Medicare-covered services (physician visits, skilled nursing facility, and home health claims). Although this study was completed too late to influence the choice of risk adjusters, we believe that it may provide some useful guidance and possible alternatives as risk adjusters, and the HMO data systems on which they depend, evolve over the next few years. While the simple adjuster that we had originally set out to test does not predict costs with sufficient accuracy, the one that we ultimately derived performs much better than the AAPCC and still requires substantially less data than the best-known diagnosis-based adjusters.