Executive Summary
The benchmarks used to determine payments to Medicare Advantage (MA) plans are based on the expected cost of covering a Traditional Medicare (TM) enrollee in a plan’s local area. TM now enrolls less than half of all eligible Medicare beneficiaries, following the large and ongoing shift of beneficiaries into MA in recent years. This shift has spurred questions about whether TM enrollment is or soon will be too small to allow the Centers for Medicare and Medicaid Services (CMS) to construct reliable estimates of local TM costs. If so, this would undermine the logic of basing MA payments on estimates of local TM costs since these estimates would no longer be a reliable guide to the costs that MA enrollees would incur if they instead enrolled in TM.
In this paper, we estimate the amount of error in CMS’ estimates of county-level TM costs attributable to the fact that CMS must rely on data for a finite population of TM enrollees. In addition to estimating sampling error, we estimate the non-sampling error introduced by the “credibility adjustments” that CMS applies to mitigate sampling error; these adjustments blend estimates for counties with fewer than 1,000 TM enrollees with estimates for the counties’ regions and introduce error if the affected counties differ from their regions. We do not consider other reasons that observed TM spending may be a poor guide to what MA enrollees would cost if enrolled in TM, notably the possibility that TM and MA enrollees systematically differ from each other; these related, but distinct, issues have been studied in prior research.
We have four main findings:
- The absolute number of TM enrollees, the main determinant of the amount of sampling error in CMS’ estimates of county TM costs, is projected to be roughly stable over the coming decade. Both the Medicare Trustees and the Congressional Budget Office expect declines in TM’s market share to be offset by increases in the total number of Medicare beneficiaries over the next ten years. These projections do assume that TM’s market share will decline more slowly over the next ten years than it did over the past decade. Nevertheless, they suggest that statistical challenges posed by low TM enrollment may not meaningfully worsen in the near-term.
- Sampling error in CMS’ estimates of TM costs is currently modest and will likely remain so even if TM’s market share falls faster than expected. We estimate that, in 2022, the median Medicare beneficiary lived in a county where the standard error of CMS’ estimate of county TM costs was 0.6% of the underlying estimate. The 95th percentile beneficiary lived in a county where this value was 1.9%. Even in a scenario where TM’s market share continues to fall rapidly, the median standard error would be 0.7% of the underlying estimate and the 95th percentile value would be 2.4% in 2033.
- The credibility adjustments applied in counties with low TM enrollment introduce considerable non-sampling error, but the affected counties contain a small fraction of beneficiaries. For counties subject to a credibility adjustment, we estimate that the total root mean squared error of CMS’ estimates of county TM costs—accounting for both sampling error and the non-sampling error introduced by the credibility adjustment process—averaged 6.5% in 2022. By contrast, this estimate was just 3.5% when accounting for sampling error alone. However, only 1.2% of beneficiaries lived in counties subject to a credibility adjustment in 2022, and only 2.2% of beneficiaries will live in such counties by 2033 even in the lowest-enrollment scenario we consider.
- Changes to the credibility adjustment process could increase the accuracy of CMS’ estimates for counties with very low TM enrollment. We estimate that the current credibility adjustment process increases the root mean squared error of CMS’ estimates for many affected counties relative to applying no adjustment. CMS could improve the accuracy of its estimates for these counties by changing how it ensures that the credibility adjustment is budget neutral and how it determines the weight placed on the regional estimate versus the county’s own estimate.
We conclude that CMS can construct fairly precise estimates of local TM costs at both current TM enrollment levels and those likely to prevail over the next decade. Thus, concerns about the precision of CMS’ estimates offer, at best, a weak basis for proposals to “decouple” MA payment from local TM costs, although there may be other rationales for such changes (e.g., concerns about the performance of the MA risk adjustment system). Our results do, however, suggest that CMS should refine how it estimates TM costs in counties with very low TM enrollment.
Authors
Related Content
2024
-
Acknowledgements and disclosures
The authors thank Richard Frank for helpful comments, Chloe Zilkha for research assistance, Rasa Siniakovas for editorial assistance, and Chris Miller for assistance with web posting. All errors are our own. This work was supported by a grant from Arnold Ventures.
The Brookings Institution is financed through the support of a diverse array of foundations, corporations, governments, individuals, as well as an endowment. A list of donors can be found in our annual reports published online here. The findings, interpretations, and conclusions in this report are solely those of its author(s) and are not influenced by any donation.
The Brookings Institution is committed to quality, independence, and impact.
We are supported by a diverse array of funders. In line with our values and policies, each Brookings publication represents the sole views of its author(s).
