Risk-Adjustment Approach to Compensating Health Care Plans More Appropriately for Serving Chronically Ill People

Improving access to care by restructuring provider payments

From 1997 to 2001, researchers at the Institute for Health Policy Studies at the University of California, San Francisco, conducted a two-phase study to test the hypothesis that risk adjustment could be improved by removing patients with high-cost chronic conditions from the general risk pool and assessing their risk by analysis of detailed clinical information.

In the project's first phase, researchers identified 100 high-cost chronic conditions—termed the VEP 100 (for verifiable, expensive and predictable)—and applied various risk adjustment models to a subset of study data to determine how the predictive power changed based on exclusion of VEP 100 conditions.

For the project's second phase, the team sought patient records from four collaborating health systems to test the project's risk adjustment hypothesis on four VEP 100 conditions: HIV/AIDS, depression, cystic fibrosis and post-transplant follow-up. For such patients, the team also proposed to adjust risks concurrently by using this year's data to predict this year's costs instead of prospectively by using last year's data to predict this year's costs. Concurrent models have demonstrated a higher predictive power but necessitate more auditing.

Key Findings

The following were among key findings reported to RWJF and/or included in an article ("The Best of Both Worlds? The Potential of Hybrid Prospective/Concurrent Risk Adjustment") published in Medical Care:

  • The project team reported that 9.3 percent of patients with one or more of the VEP 100 conditions accounted for 49 percent of total expenses and 84 percent of the variation from the average cost for the entire population.
  • Almost all of the benefit of concurrent risk adjustment models is due to their predictive power for patients with a VEP 100 condition and that applying prospective adjustment for the remaining patients entailed little loss in predictive accuracy.
  • The principal investigator reported that the project demonstrated that researchers face a number of potential difficulties when they rely on data from ongoing health care businesses.