Specialty Payment Model Opportunities and Assessment

This article describes the results of a simulation analysis of a payment model for specialty oncology services that is being developed for possible testing by the Center for Medicare and Medicaid Innovation at the Centers for Medicare & Medicaid Services (CMS). CMS asked MITRE and RAND to conduct simulation analyses to preview some of the possible impacts of the payment model and to inform design decisions related to the model.

Data and Methods Used in the Simulation Analysis

The simulation analysis used an episode-level dataset based on Medicare fee-for-service (FFS) claims for historical oncology episodes provided to Medicare FFS beneficiaries in 2010. Spending on those historical episodes was inflated to 2016, which is the first year in which the proposed payment model could be implemented. Under the proposed model, participating practices would continue to receive FFS payments, would also receive per-beneficiary per-month (PBPM) care management payments for episodes lasting up to six months, and would be eligible for performance-based payments based on per-episode spending for attributed episodes relative to a per-episode spending target. For the simulation analysis, the care management payments were set at $160 PBPM; that amount has not yet been finalized by CMS. RAND assumed that only practices providing 50 or more episodes of chemotherapy treatment per year would consider participating and would be eligible to participate in the payment model—these medium- and high-volume oncology practices account for 73 percent of all chemotherapy episodes and 80 percent of total spending on chemotherapy episodes. Among those practices providing 50 or more episodes of treatment per year, 10 percent were assumed to participate in the payment model. For participating practices, the simulation model was used to calculate care management payments, a spending target, and actual spending, taking into account the fact that participating practices have an incentive to reduce spending on attributed episodes. The simulation model was then used to calculate performance-based payments, taking into account the practices' behavioral response.

Conclusion

The simulation described in this report offers several insights into the proposed payment model for oncology:

1. The care management payments used in the simulation analysis—$960 total per six-month episode—represent only 4 percent of projected average total spending per episode (around $27,000 in 2016), but they are large relative to the FFS revenues of participating oncology practices, which are projected to be around $2,000 per oncology episode. By themselves, the care management payments would increase physician practices' Medicare revenues by roughly 50 percent on average. This represents a substantial new outlay for the Medicare program and a substantial new source of revenues for oncology practices.
2. The use of performance-based payments requires the assignment of spending benchmarks, but those benchmarks cannot reflect the desired counterfactual—i.e., spending per episode in the absence of the payment model. Inaccuracies in benchmarks arise due to variation in individuals' health care spending that cannot be explained by case-mix adjusters. That unexplained variation affects both the historical claims data used to set benchmarks and actual spending in the performance period. As a result, the Medicare program will in some cases pay “noise bonuses”: performance-based payments that are due to inaccurately low benchmarks rather than to providers' behavioral responses. The likelihood of noise bonuses would be reduced if the payment model were limited to larger practices (e.g., 100 or more episodes).
3. For the Medicare program to break even, participating oncology practices would have to reduce utilization and intensity by roughly 4 percent. (“Breaking even” in this case means that Medicare spending per episode would be the same, with or without the payment model.) Based on research on other similar payment models, a behavioral response of that magnitude may be possible, but it is by no means certain.
4. The break-even point can be reduced—i.e., less of a behavioral response would be required for the Medicare program to achieve savings—if the care management payments are reduced or if the performance-based payments are reduced.

The research addressed in this article was conducted in RAND Health, a division of the RAND Corporation, under a subcontract to MITRE.