Evaluating the performance of medicare fee-for-service providers using the health outcomes survey: a comparison of two methods.

Background: Health status measures are now being used for evaluating the performance of health care organizations. Trends in SF-36 component scores have previously been examined for Medicare-managed care plans but not for providers serving Medicare fee-for-service (FFS) beneficiaries. We compared 2 methods for evaluating the performance of Medicare FFS providers, the Research Triangle Institute (RTI) and Health Assessment Laboratory (HAL) methods.

Methods: Data were collected from 6547 Medicare FFS beneficiaries in 10 cohorts. SF-36 Physical Health (PCS) and Mental Health (MCS) component scores were computed at baseline and after a 2-year follow-up. The RTI approach predicts follow-up scores based on a standard care regression model. The HAL approach determines the percentage of beneficiaries whose status is the "same or better" at follow-up. Both approaches then compare observed to expected scores for each cohort.

Results: The HAL method did not detect any statistically significant differences for the PCS; the RTI method detected a small PCS difference for one cohort. The HAL method identified 4 cohorts that had significantly higher MCS scores; the RTI approach identified one cohort with significantly lower scores.

Conclusions: The 2 approaches provided consistent assessments of provider performance for the PCS but not for the MCS. The differences in the MCS results may have been affected by differing treatment of deaths during follow-up. The HAL approach disregards deaths for the MCS, whereas the RTI method imputes values for death. Implications of using self-reported health status for monitoring provider performance are discussed.