主页
/ 市场洞察
/ White Papers
/ Reviewing Methodological Issues in MedPAC Analysis

Reviewing Methodological Issues in MedPAC Analysis

Favorable Selection in Medicare Advantage

Abstract

We assessed the findings contained in recent reports by the Medicare Payment Advisory Commission (“MedPAC”).^{1, 2} MedPAC’s top line conclusion is that MA plans were paid 22% more in 2021 than what Medicare would have paid for those individuals under Fee-For-Service Medicare (“FFS”).³ However, MedPAC’s conclusions are based on data which limit the scope of their findings and methodologies that affect the accuracy of their estimates. MedPAC also elides significant countervailing facts and fails to account for plan bids, which affects the interpretation of their results.

Introduction

Because CMS payments to Medicare Advantage plans are capitated and based on the average costs of beneficiaries in Fee For Service, Medicare Advantage plans can financially benefit from enrolling beneficiaries who are healthier than the average FFS beneficiary. The healthier beneficiary utilizes health care services less frequently, and therefore costs less to the plan than CMS is paying them. This phenomenon is known as favorable selection. Because of concerns this could create incentives for MA plans to target healthier seniors, CMS risk adjusts payments to account for these differences in costs, but despite these adjustments, favorable selection is still a concern.

Researchers have been claiming different levels of favorable selection through each iteration of the MA program over the past 25 years despite attempts to address it. Analyses show that from 2003 to 2009, private plans were paid more than FFS practically by design, to attract more competition.⁴ The payment difference peaked at 14% in 2009 before the ACA reduced payments to private plans.⁵

In 2017, according to MedPAC, FFS and MA were at parity.⁶ Now, 7 years later, despite no changes in legislation or how the program is administered, MedPAC is claiming that private plans are paid substantially more.⁷ According to their new analysis, MedPAC is claiming that favorable selection results in plans being overpaid up to 22%.⁸ However, MedPAC’s analyses of favorable selection make several strong and problematic assumptions, that, along with the lack of relevant data should call into question their conclusions.

Absence of Data Limits Legitimacy of Conclusions

Direct to Medicare Advantage Beneficiaries
Attempts to estimate the magnitude of favorable selection have always been challenged by the substantial differences between Medicare Advantage and FFS Medicare. FFS Medicare is an all-inclusive network, government-administered plan while Medicare Advantage comprises multiple private insurance companies offering different plans with various benefit designs, cost-sharing structures and complex networks of providers.

In addition to these structural complications, information on Medicare Advantage beneficiaries is difficult to obtain. Consequently, many studies, including the MedPAC studies, try to estimate favorable selection using only FFS data and extrapolating, from beneficiaries’ experience in FFS, what their costs would be in Medicare Advantage. Determining effects through extrapolation alone is always fraught and specific problems are detailed below. It is also important to distinguish two general categories of Medicare Advantage enrollees—those who were originally in FFS and switched, and those who never participated in FFS at all but joined MA when they turned 65. Research based on FFS data, like the MedPAC reports have some data on the first group, which they are using to speculate as to the experience for both groups, despite having no information on the experience of the second group.

While MA enrolled only 25% of people in 2010, it has grown rapidly since then. In 2024, half of Medicare beneficiaries are enrolled in Medicare Advantage plans.⁹ As Medicare Advantage displaces FFS, the number of people who never participated in FFS and whose experience is therefore excluded from these analyses grows. In 2015, approximately 950,000 people enrolled directly into MA without participating in FFS, representing 27% of enrollees.¹⁰ In 2022, 1.6 million, or 44% of new beneficiaries enrolled directly into MA.¹¹

Beneficiaries who first enrolled in MA (and subsequently stayed in MA) may have a different health care cost profile than MA beneficiaries who started in FFS. This represents a material, and growing segment of membership that were entirely omitted from this analysis. MedPAC’s methodology similarly omits millions of direct enrollees from previous years. Any conclusions made by the MedPAC reports are incomplete without understanding the healthcare expense trends of a significant group of MA beneficiaries who were never included in the analysis. Without further analysis of this population, one cannot draw definitive conclusions.

Fee For Service to MA Switchers
While there is no information on which to estimate favorable selection for beneficiaries who enrolled directly in Medicare Advantage, there is at least some information on the beneficiaries who switched to MA after first participating in FFS. However, even for these beneficiaries, the MedPAC analyses could misconstrue the available data and generate estimates that are flawed.

First, research has shown that costs vary from year to year, and, importantly, while a person or group may have a single low-cost or high-cost year, that year may be an aberration and costs revert to the mean across everyone over time.^{12, 13} MedPAC attempts to account for this by tracking beneficiaries within FFS across several years and assuming that the pattern will be reflected for beneficiaries who switched. Their approach, however, will not account for differences in utilization in MA caused by differences in plan design.

Medicare Advantage plans can offer lower out-of-pocket costs than FFS.¹⁴ MA plans often provide more comprehensive coverage (e.g., typical MA plans include prescription drug coverage and supplemental benefits such as vision and dental), and they also cap a beneficiary’s annual out-of-pocket costs.¹⁵ Beneficiaries who switch may forego care while in FFS and delay it until they’re in Medicare Advantage when their out-of-pocket costs will be lower. This could distort the analyses, making switchers’ health care costs appear lower while they are enrolled in FFS even though their costs are just delayed until they join Medicare Advantage. This will lead to overestimates of the magnitude of favorable selection.

This experience, in reverse, has been documented in private plans. As plans periodically adjust their structures and benefits, beneficiaries react by accelerating their utilization to incur any treatment costs while benefits are more generous. This then causes a decline in healthcare once the new benefits package goes into effect.^{16, 17}

Methodological Decisions Leads to Flawed Estimates For Switchers

The previous section explained how lack of information for beneficiaries while enrolled in Medicare Advantage raises doubts as to the applicability of MedPAC’s estimates. In this section we discuss how even with the available data, methodological decisions will also lead to flawed results.

In its analyses, MedPAC excludes certain groups or ignores mitigating facts that could impact its conclusions. Examples of excluded information that could impact results:

Existence of and implications of Dual Eligibles and Special Needs Plans
Beneficiaries enrolled only in Medicare Part A
Beneficiaries who have ESRD
Beneficiaries who don’t maintain the same residence throughout the year of analysis
Beneficiaries who have Medicare as a secondary payer
Beneficiaries enrolled only in Medicare Part A¹⁸

Below, we provide detail as to how these omissions might affect their estimates and conclusions.

Part A-Only Beneficiaries
Some Medicare FFS beneficiaries are enrolled only in Part A of Medicare and do not receive Part B benefits. These beneficiaries are not eligible to participate in Medicare Advantage though their costs are included in CMS’s benchmark calculations which affect MA plan payments.¹⁹ Notably, MedPAC chose to exclude these beneficiaries from its analyses and so the favorable selection estimates that they generate will not reflect real world experience.²⁰ Because Part A-only beneficiaries are less costly on average to Medicare than beneficiaries who are eligible for Medicare Advantage, they reduce the benchmarks that CMS uses to determine MA plan payments below what they would be if CMS only used MA-eligible beneficiaries.²¹

By excluding Part A-only beneficiaries from their analysis, MedPAC may have produced a more accurate comparison of costs between eligible beneficiaries, but their estimate is less applicable to actual experience because this is not the population that CMS bases its benchmarks on, and, consequently, MA plans are not paid according to the estimate determined based on this approach.

Multi-Home Seniors
MedPAC analyses also require beneficiaries to live in the same county during the reference year.²² Excluding these beneficiaries removes people who could materially impact cost estimates such as those who are housing-insecure (e.g., a low-income dual population) and those who are more affluent and migrate between homes seasonally.²³

Dual Eligibles
The Medicare dual-eligible population has grown from 8.6 million beneficiaries in 2006 to 12.5 million in 2023.²⁴ Concurrently, dual-eligible members have migrated away from FFS, and Special Needs Plans (SNPs) have grown in popularity, with dual-eligible SNPs (“D-SNPs”) being the predominant type of SNP.²⁵ MA plans have risen in popularity among dual eligibles in line with their popularity among the rest of seniors. In January 2013, 23.6% of dual eligibles were enrolled in an MA plan²⁶ and by the end of 2021, that figure exceeded 50%.²⁷

MedPAC’s methodology assumes dual-eligible beneficiaries can be assessed in the same manner as other Medicare-eligible beneficiaries. Dual eligibles are included in the MedPAC risk cohorts, which categorizes beneficiaries according to their characteristics such as risk score, age, gender, etc. However, this approach does not accurately reflect the administrative costs to MA plans associated with developing and operating a D-SNP.

The Centers for Medicare & Medicaid Services (“CMS”) requires all SNPs to establish rigorous processes around a Model of Care that clearly defines the unique health requirements associated with what are often a plan’s most vulnerable enrollees, such as creating integrated care plans for each beneficiary, developing specialized networks targeted at the disease states present in the population, and building quality programs that monitor a plan’s success in serving the population’s care requirements. This level of intense care coordination can create better outcomes for beneficiaries and result in lower long-term costs of care. However, it requires that plans incur additional administrative expenses, which are not accounted for in MedPAC’s analysis.

Payments to Plans Not Considered

Importantly, policymakers must understand that the MedPAC reports’ estimates are not descriptive of any effect of favorable selection on the payments to MA plans, which are distinct from the costs of care. MedPAC does not clarify the differences, but their analysis is focused on the “benchmarks” or the average costs of care, in FFS, for beneficiaries. Some readers may infer that the benchmarks and the payments to plans are equivalent or the distinction between them is irrelevant, but this is not the case.

Payments to plans are a function of the plan’s bid, the benchmark, and the quality of the plan. If the plan’s bid is lower than the benchmark, then the payment CMS makes to the plan will be lower than the benchmark, and the lower the bid, the smaller the payment, and the smaller the effect of any favorable selection.

Given competition to attract consumers, plans incorporate savings they can generate into lowering their bids to make these plans more attractive to members. This is particularly true in highly competitive markets. Previous research cited in the MedPAC report acknowledges this point.²⁸ Further, MedPAC notes that higher levels of MA penetration, as has been the trend over the past ten years, may be correlated with less favorable selection.

The National Library of Medicine confirmed that competition incentivizes plans to bid low to attract consumers.²⁹ Low bids produce larger rebates which are channeled to and attract the beneficiary.³⁰ MedPAC reported that 95% of beneficiaries live in counties with at least four MA organizations, “which appears large enough to ensure a sufficient level of competition”.³¹ MedPAC acknowledges that bids are nearly always lower than the benchmarks.³²

By not incorporating the bids into the analysis, the estimates of favorable selection MedPAC generates will apply only to the benchmarks, which are not the only factor in what the plans are paid. Favorable selection based exclusively on benchmarks will overstate any difference in payments. Analysis that focuses entirely on FFS costs and ignores the bidding process would only inform policymakers on the differences in the costs of treatment if Medicare FFS treated both groups of people; it would not be an accurate estimate of the payments to the plans or the costs to CMS or taxpayers.

Conclusion

The MedPAC reports’ exclusion of plan bids and Medicare Advantage beneficiaries significantly limit the validity and implication of their conclusions. Further, the approach taken to estimate the difference for FFS beneficiaries who switch includes several methodological decisions that will lead to inaccurate estimates.

Further investigation that accounts for these issues should be pursued before any conclusions about the level of selection into MA are drawn, or any changes to Medicare are pursued. Modifying a system that is poorly understood can lead to unintended consequences that could adversely impact millions of American seniors who currently rely on the program.

This article reflects a summary of our thoughts reviewing methodological issues in MedPAC’s analysis of favorable selection in Medicare Advantage.

Download the full Reviewing Methodological Issues in MedPAC’s Analysis white paper here

Footnotes:

1: “Chapter 12: The Medicare Advantage program: Status report,” MedPAC (March 2024), Page 364.

2: “Chapter 13: Estimating Medicare Advantage coding intensity and favorable selection,” MedPAC (March 2024), Page 417.

3: “Chapter 12: The Medicare Advantage program: Status report,” MedPAC (March 2024), Page 372.

4: Patel Y, Guterman S, “The Evolution of Private Plans in Medicare,” The Commonwealth Fund (December 8, 2017).

5: Patel Y, Guterman S, “The Evolution of Private Plans in Medicare,” The Commonwealth Fund (December 8, 2017).

6: Patel Y, Guterman S, “The Evolution of Private Plans in Medicare,” The Commonwealth Fund (December 8, 2017).

7: Patel Y, Guterman S, “The Evolution of Private Plans in Medicare,” The Commonwealth Fund (December 8, 2017).

8: “Chapter 12: The Medicare Advantage program: Status report,” MedPAC (March 2024), Page 364.

9: Freed, M., Fuglesten Biniek, J., Damico, A., Neuman, T., “Medicare Advantage in 2024: Enrollment Update and Key Trends” KFF.org (August 8, 2024).

10: Xu L, Welch WP, Sheingold S, De Lew N, Sommers BD. Medicare switching: patterns of enrollment growth in Medicare Advantage, 2006–22. Health Aff (Millwood). 2023;42(9), Link.

11: Xu L, Welch WP, Sheingold S, De Lew N, Sommers BD. Medicare switching: patterns of enrollment growth in Medicare Advantage, 2006–22. Health Aff (Millwood). 2023;42(9), Link.

12: Joshi S, Nuckols T, Escarce J, Huckfeldt P, Popescu I, Sood N, “Regression to the Mean in the Medicare Hospital Readmissions Reduction Program,” JAMA Intern Med. (September 1, 2019);179(9):1167–1173. doi:10.1001/jamainternmed.2019.1004

13: Newhouse JP, Price M, McWilliams JM, Hsu J, Souza J, Landon BE, “Adjusted Mortality Rates Are Lower For Medicare Advantage Than Traditional Medicare, But The Rates Converge Over Time,” Health Aff. (April 2019);38(4):554-560. doi: 10.1377/hlthaff.2018.05390. PMID: 30933606; PMCID: PMC6555557.

14: USC Schaeffer Center, “Out-of-Pocket Costs Are Substantially Lower in Medicare Advantage Than Traditional Medicare”, (November 4, 2024).

15: USC Schaeffer Center, “Out-of-Pocket Costs Are Substantially Lower in Medicare Advantage Than Traditional Medicare”, (November 4, 2024).

16: Balasinkam A, “Benefit design and the rush, hush, and crush cycles,” Milliman (April 8, 2015).

17: Barrett J, “Timing’s Everything: The Impact of Benefit Rush,” Health Watch (May 2008).

18: “Chapter 13: Estimating Medicare Advantage coding intensity and favorable selection,” MedPAC (March 2024), p440.

19: “The Use of Benchmarks for Payment in Medicare Advantage and Necessary Adjustments”, Better Medicare Alliance Fact Sheet (February 2021).

20: “Chapter 13: Estimating Medicare Advantage coding intensity and favorable selection,” MedPAC (March 2024), Page 416.

21: In an analysis, Wakely calculated that Part A spending was 14.2% higher for beneficiaries who were enrolled in Part A and B compared to all beneficiaries (including those enrolled in only Part A). Courtney, T FSA,MAAA, Stewart, R ASA,MAAA, “Value of Medicare Advantage Compared with Fee for Service – Response to MedPAC,” (January 18, 2024), Page 7.

22: “Chapter 13: Estimating Medicare Advantage coding intensity and favorable selection,” MedPAC (March 2024), Page 441.

23: “Chapter 13: Estimating Medicare Advantage coding intensity and favorable selection,” MedPAC (March 2024), Page 441.

24: “Data Analysis Brief: Medicare-Medicaid Dual Enrollment 2006 through 2019,” CMS Medicare-Medicaid Coordination Office.

25: “What We Know About Medicare Enrollment for Dual Enrollees”, Urban Instutute, May 30, 2024.

26: “What We Know About Medicare Enrollment for Dual Enrollees”, Urban Instutute, May 30, 2024.

27: “What We Know About Medicare Enrollment for Dual Enrollees”, Urban Instutute, May 30, 2024.

28: Pelech D, Song Z. “Pricing and pass-through in response to subsidies and competition: Evidence from Medicare Advantage before and after the Affordable Care Act,” (January 12, 2023).

29: Song Z., Landrum, M.B., Chernew, M.,“COMPETITIVE BIDDING IN MEDICARE ADVANTAGE: EFFECT OF BENCHMARK CHANGES ON PLAN BIDS”, National Library of Medicine (December 1, 2014).

30: Song Z., Landrum, M.B., Chernew, M.,“COMPETITIVE BIDDING IN MEDICARE ADVANTAGE: EFFECT OF BENCHMARK CHANGES ON PLAN BIDS”, National Library of Medicine (December 1, 2014).

31: “Chapter 4: Favorable selection and future directions for Medicare Advantage payment policy,” MedPAC (June 2023), Page 186.

32: “Chapter 12: The Medicare Advantage program: Status report,” MedPAC (March 2024), Page 364.