Inpatient rehabilitation facilities (IRFs) provide rehabilitation services to certain Medicare beneficiaries who need intensive rehabilitation care following an acute hospital stay. IRF patients must be able to tolerate and benefit from intensive therapy. In 2022, enrollees in traditional Medicare had 383,000 stays provided by 1,181 IRFs, for which Medicare spent $8.8 billion. Medicare pays IRFs on a per-stay basis, paying a base amount for each stay that may be adjusted higher or lower depending on a patient's condition. In this report prepared for the Medicare Payment Advisory Commission, we use Medicare claims data for fiscal year 2019 to examine the correspondence between IRF payments and costs. We find substantial variation in payment-to-cost ratios (PCRs) across stays by case-mix group (CMG). The pattern of findings shows that PCRs were generally higher for patients with lower motor and cognitive functioning levels who typically receive higher payments. The systematic variability in profitability for different groups of patients raises concerns about the performance of the IRF payment system. The Medicare program could be paying substantially more than justified by the average cost for some groups of patients while paying less than warranted for others. Average payment weights and CMIs have risen over time, increasing Medicare payments. Differences in profitability could provide an incentive for facilities to selectively admit some types of patients and adversely affect access to or quality of care for others.

We also investigate the reasons for variation in PCRs across IRF stays. Ruling out other potential explanations, we attribute the observed pattern of rising PCRs with CMGs with lower functioning levels to the IRF payment system’s use of the hospital-specific relative value (HSRV) method of computing relative payment weights. We show that the problem with using HSRV weights was not evident in the early years of the IRF payment system but emerged over time. Further, with an alternative method of computing relative payment weights based on average cost, such as is used in Medicare’s acute inpatient hospital prospective payment system, PCRs are nearly equal across patient groups and produce a much better correspondence between IRF payments and costs at the facility level. We examine why the performance of HSRV weights changed over time—simulations show the HSRV weights seem to have undesirable properties in that they resist adapting to provider or system efficiency changes. Finally, we simulate the financial impact of replacing the current HSRV weights with average-cost-based payment weights in the IRF payment system by type of provider and find relatively modest changes across all types. A shift to cost-based weights may benefit the IRF prospective payment system and better meet Medicare’s objectives.