CMS' Medicare Shared Savings Program proposal would make substantial changes to the benchmarking methodology for accountable care organizations.
Cowritten by Sean Cavanaugh, chief administrative officer of Aledade; Hayden Rooke-Ley of Aledade; and Farzad Mostashari, MD, cofounder and CEO of Aledade.
On August 9, CMS issued the first update to the Medicare Shared Savings Program (MSSP) of the current administration. This is the second of a series of blogs that explores the proposed regulation. In this blog, we explore the proposed changes to how MSSP benchmarks are set.
We have been known to say “it is all about the benchmark,” and this is doubly true when performance-based risk is involved. Our first post identified 3 types of risk an accountable care organization (ACO) must consider: medical risk, insurance risk, and regulatory risk. Medical risk is changes in costs driven by the delivery of healthcare services, such as how well diabetic patients are managed. Insurance risk is changes in costs that are not driven by the delivery of healthcare services, such as changes in the rates of accidents. Regulatory risk is the risk of changing the rules and/or rules that affect risk in nontransparent ways. The benchmark methodology is crucial to determining how much medical and insurance risk an ACO assumes, and whether ACOs are appropriately rewarded for reducing costs. CMS is proposing several substantial changes to this benchmark methodology.
In today’s value-based landscape, physicians are becoming more amenable to accepting responsibility for medical risk. But insurance risk remains daunting—and beyond the purview of most providers. Therefore, a benchmark methodology should minimize the transfer of insurance risk to healthcare providers. The primary mechanism for transferring medical risk, but not insurance risk, is risk adjustment. CMS uses the Hierarchical Condition Categories, or the “HCC mode,” in both Medicare Advantage (MA) and in MSSP. In MA, payments are adjusted up or down for changes in risk score, without caps, and based on a normalized risk score. In MSSP, the ACO population is divided into continuously assigned beneficiaries (those who saw an ACO provider both last year and this year) and newly assigned beneficiaries (those who did not see an ACO provider last year). For the continuously assigned beneficiaries (75% to 90% of the ACO population), the population-level risk score is asymmetric: it cannot increase, but it can decrease. This means that essentially all insurance risk that this population gets sicker is transferred to the ACO. For the newly assigned population, the risk score floats freely like in MA.
CMS is proposing to remove the distinction between the 2 ACO populations; instead, it would calculate a single risk score for the entire assigned population to adjust the benchmark. However, unlike MA, CMS is proposing a 3% cap, up or down, on changes in risk score compared with the most recent benchmark year. This limit could hold for as long as 5 years, although an ACO can shorten this timeframe by ending its agreement early and going to the Enhanced track. CMS proposes to use the cap instead of a MA-style coding intensity adjustment. CMS is asking for comment on whether the caps should be higher, whether the caps should be replaced with a coding intensity adjustment, or whether the risk score should just freely float with no adjustment until CMS has more data in such an environment.
CMS next explored how best to balance past performance in managing medical risk with future improvements in managing medical risk. As MSSP is inserted into our existing system, no one is starting from zero. Every possible mix of ACOs will include some providers who are better than average at controlling costs on day 1 and some who are worse than average on day 1. How do you simultaneously encourage inefficient providers to achieve efficiency while encouraging efficient providers to fight against reversion to the mean when your program is voluntary? Currently, MSSP does this by transitioning from a historical benchmark to a regional benchmark over time. In the first 3 years an ACO gets rewarded only for improvement over historical costs. In subsequent years, ACOs are progressively awarded bonuses or penalties at the start of each agreement based on how regionally efficient or inefficient they are. CMS is proposing to move regional benchmarking into the first agreement period. This will remove the waiting period that efficient providers must survive to be rewarded for their relative performance which is expected to increase both the number and longevity of regionally efficient ACOs. The tradeoff is that the program will become less attractive to regionally inefficient ACOs than it is today. As the perfect balance is unknowable, CMS seeks comment on this tradeoff.
Percent of ACO Benchmark based on Regional Comparison
(Years 1 to 5)
(As soon as year 2, as late at year 6)
Regionally Efficient ACO
Regionally Inefficient ACO
It was not all good news for regionally efficient ACOs, though. The ultimate percentage in MSSP of regional benchmark is reduced from 70% to 50% (MA is 100%). Furthermore, CMS is proposing a 5% cap on the bonus a regionally efficient ACO can receive at the start of an agreement period. The stated purpose of the cap is to prevent excessively high savings for past performance. By using the national average costs to calculate the cap, CMS is rewarding ACOs in low-cost regions with a higher cap than 5% and penalizing ACOs in high-cost regions with a cap that is lower than 5%. This rewards ACOs that accomplish the harder task of regional separation in a low-cost area. CMS asks for comments on whether the cap should be lower, higher, or whether CMS should have a cap at all.
Regional Trend Factors
The next proposed change to the benchmarking methodology is also moving regional trend factors into the first contract. Currently, CMS moves from a 100% national trend (or annual inflation) update to a 100% regional trend in the second contract. CMS is proposing to move to a blended regional and national trend for all years. The proposed blend is based on the penetration of the ACO in its region. If 20% of the Medicare fee-for-service beneficiaries in the ACO’s region are assigned to the ACO, then the trend update would be 80% regional trend and 20% national trend. CMS gives 2 reason for the blend. First, CMS has long sought to find ways to reward ACOs in regions where trend increases less than the national rate. This continues that attempt. Second, CMS is trying to solve the circularity problem of including ACOs in its own region trend. As ACOs drive down costs for their aligned beneficiaries, they are also driving down costs in their region and thus lowering their own regional adjustment. This effect can be large for rural ACOs that serve a high proportion of a county’s beneficiaries.
Unfortunately, we do not believe the blended trend addresses this problem. For example, consider a hypothetical ACO that has 20% market penetration and reduces costs by 5% in a region where trend went up 5% in the 80% of beneficiaries not assigned to the ACO. Under current methodology, the regional trend would be 4% for the region (5% x 80% + 0% x 20%). Under the proposed rule the regional trend would still be 4%, but it would only account for 80% of the trend (3.2%). National trend would make up the other 20%. If national trend is 3%, then we are actually at a lower trend than current policy for the ACO (4% x 80% + 3% x 20% = 3.8%). CMS captures 100% of the 1.2% of the savings generated by the ACO (the difference between the 5% cost reductions achieved by the ACO and the proposed blended trend of 3.8%) in addition to the share rate CMS receives on the rest of the savings. However, it is not the reduction in the actual share rate that is the most concerning part of this policy. An ACO must have 5000 members. This effectively means that rural ACOs will have a greater percentage of the savings they generate captured by CMS than urban ACOs. CMS has not proposed to do the straightforward thing of removing an ACO’s assigned beneficiaries from regional trend because of a concern about small numbers in some counties. This concern is easily addressed by expanding the region to contiguous counties.
Taken together, these proposals do make risk more appealing to an ACO by better measuring medical risk, reducing insurance risk, and reducing regulatory risk. We look forward to working with our partner physicians, the ACO community, and CMS to make these policies simpler and more predictable to create the best ACO program possible.