The Medicare program represents a sacred promise to America’s elderly and disabled citizens. The Medicare Fee-For-Service (FFS) program, introduced as part of the Great Society in 1965 to mirror a Blue Cross Blue Shield commercial market standard, covered hospital and physician benefits separately. As the private health care market evolved, policymakers introduced private plans into Medicare, followed by the Medicare Risk Program and the integration of health maintenance organizations, eventually leading to today’s Medicare Advantage (MA) program.

This article addresses the history of the Medicare program and the current debate around the recent MedPAC estimate of MA overpayments. We propose nuanced analytical considerations to ensure accurate coding, address favorable selection, and foster a holistic policy future for Medicare.

A Program In Evolution Since 1965

With the intent to create a program serving the changing needs of beneficiaries, various administrations and Congresses oversaw the design of benefits, addressing evolving medical practice and patient complexity. In the 1950s, hypertension was only newly being treated, with national treatment guidelines first issued in the 1970s. Powered by epidemiological research and pharmaceutical innovation, prescription drugs became a norm within medical practice, and, by 1986, 84.7 percent of Medicare HMO enrollees received prescription drug benefits. Driven by an increasingly diverse population, policymakers and analysts recognized the programmatic agility of MA, which, for example, allowed benefit design that better served beneficiaries in long-term care and provided home and team-based care for multi-morbid individuals.

Recognizing that customization—not standardization—coupled with plan quality transparency was critical to caring for an increasingly diverse population, policymakers created special needs plans and later added a quality star rating system to grade MA plans as part of the Patient Protection and Affordable Care Act. Subsequently, regulators have worked to improve ratings accuracy in response to analytical concerns. Given that the star rating quality bonus system does not apply to FFS Medicare, quality bonus incentives thus financially favor the MA program, the FFS Medicare plan lacks a quality rating, and beneficiaries cannot easily compare the two formulations of Medicare.

Simultaneous fiscal pressures drove commercial insurers to implement cost-control strategies via networks, utilization review, and employee cost shifting. States and the federal government faced similar pressures, motivating the evolution of FFS payment to risk-adjusted capitation as part of the volume-to-value transition supported across Democratic and Republican administrations.

Facing the pressures of fixed incomes and limited finances, just over half of all seniors today elect MA with its financial protections and enhanced benefits. The FFS program has become increasingly expensive for beneficiaries via premiums for Part B, Part D, and Medigap (exceeding $183 monthly); MA has become an appealing alternative by serving as an affordable Medigap plan coupled with no-cost or low-cost Part D coverage.

Recent Policy Questions And The Need For Analytical Rigor

With over 30 million elderly, disabled, and impoverished Americans turning to MA, analytical rigor in assessing MA’s value and performance compared to FFS Medicare is an absolute requirement. Analyses must integrate insights and data across academia, the policy community, government, and industry, all aimed at equalizing the evaluation and treatment of MA and FFS Medicare. This section responds to MedPAC’s recent analysis claiming $83 billion in MA overpayments as a result of greater coding intensity in, and favorable selection into, MA.

Coding Intensity Differences

While MA plans are likely overfunded, approaches to developing estimates of coding intensity and favorable selection must validate policy models by connecting them to real-world business and clinical operations. Coding intensity differences, a longstanding MedPAC concern, derive from different payment methodologies in FFS Medicare versus MA. In FFS Medicare, specific providers are paid based on relative value units while hospitals are paid based on diagnosis-related groups (DRGs). In FFS, providers are not incented to capture all diagnosis codes that accurately reflect a member’s health condition. In MA, health plans are paid on a per-member, per-month base capitation rate, risk-adjusted for health status. This incents MA plans to more accurately code disease prevalence, incidence, acuity, and complications.

Such differing incentives create problems when policy experts seek to do programmatic comparisons between FFS Medicare and MA. These coding intensity differentials stem from three potential components. First, outright fraudulent coding remains a concern given ongoing cases of plans adding diagnoses unsupported by medical documentation. A second, related concern is diagnostic upcoding, wherein complexity is falsely enhanced to drive payment. The third component is clinically appropriate diagnostic coding intensity, the reciprocal of FFS undercoding.

Regarding the third component, consider the case of a Medicare beneficiary with diabetes and cardiac, renal, and ophthalmic complications. In FFS Medicare, physicians and hospitals may capture only a portion of these diagnoses to justify the provision of a specific service. In contrast, an MA plan that bears full underwriting risk has an incentive to capture all of these diagnoses as a means of assessing the full cost to cover all beneficiary benefits. In this setting, policymakers must address chronic FFS undercoding, a problem acknowledged but not yet measured and accounted for in the MedPAC analysis of FFS versus MA spending.

While the first two examples of coding intensity differences represent MA plan “overpayment” in the sense that people normally use that word, the third—clinically appropriate coding—does not. Stringent CMS regulations and Medicare Risk Adjustment Data Validation (RADV) audits would greatly minimize the first two components. The recent MedPAC analysis of coding intensity in MA does not differentiate between, measure, or account for these three factors driving coding intensity. Thus, with FFS undercoding unaccounted for and the three components of coding intensity all lumped together as overpayment, the MedPAC analysis likely overestimates coding intensity effects.

With an eye towards solving problems and improving measurement, policymakers and regulators should measure all components of coding intensity via chart audits comparing large samples of FFS and MA beneficiaries, using them to address overpayment. Other drivers of policy consternation, like in-home health risk assessments, pose a regulatory opportunity to require two years of data as MedPAC recommended and go a step further by transforming what is functionally a data-harvesting visit into a meaningful, more convenient clinical encounter for elderly or disabled beneficiaries. CMS has begun to address challenges in MA coding through an appropriately muscular approach to implementing RADV audits, updating Medical Loss Ratio guidance, and attempting to target problematic coding practices.

Favorable Selection

Finally, favorable selection within the MA program has been a longstanding concern. Historically, bad actors have engaged in tactics ranging from deterring sicker beneficiaries through third-story sales seminars to targeted advertising designed to drive healthier enrollment. This has inspired enhanced oversight and a series of regulator- and policymaker-driven reforms: CMS designed policy interventions to address real-world market problems and now reviews all plan marketing materials, operationalizing existing regulations regularly updated through marketing guidelines.

MedPAC’s recent updated estimate of MA favorable selection violates several analytical norms, including the use of a non-peer-reviewed comparator model and the inclusion of beneficiaries enrolled in employer-group waiver plans (EGWPs). EGWPs are not available to the general public and are not subject to CMS plan bidding requirements for MA and prescription drug benefits.

In addition, for unclear reasons, MedPAC’s recent model excludes beneficiaries with end-stage renal disease (ESRD), despite MA penetration of ESRD Medicare beneficiaries rising from 27 percent in 2020 to 47 percent in 2022, nearly approaching the MA general market share of 51 percent. While there is undoubtedly health status variation among ESRD beneficiaries, caring for this population is generally costly, with evidence suggesting that the MA’s required maximum out-of-pocket benefit is highly attractive to many ESRD Medicare beneficiaries—resulting in likely negative selection for MA plans. Furthermore, growing MA enrollment in D-SNPs—special needs plans for beneficiaries who are dually eligible for Medicare and Medicaid, another high-cost population—and the Medicare trustees report denoting decreasing FFS costs per beneficiary due to this trend raise questions about the real-world validity of the recent MedPAC favorable selection model.

While invariably there is some favorable selection in MA, the program’s integrated benefits likely result in negative selection into the program. For example, a Wakely report concludes that, if the mandatory MA maximum out-of-pocket (MOOP) limit of $7,550 were included in FFS Medicare, FFS spending would be 2.8 percent higher. A desire for a MOOP limit likely drives some high-cost beneficiaries into MA, but this factor is unaccounted for in MedPAC’s recent program comparison methodologies, which thus likely overestimate the effect of favorable selection.

To ensure accurate measurement of favorable selection across FFS Medicare and MA, selection effects must be examined bidirectionally. As the example of ESRD beneficiaries demonstrates, models, including MedPAC’s, must undergo stress-testing with subsets of Medicare populations in order to ensure internal validity. Finally, MedPAC’s model and other models must be externally validated by connecting them to real-world business practices. As the historical and recent actions of CMS’ work to address favorable selection demonstrate, externally validating models in this fashion guides policymakers and regulators to focus regulatory policy on issues that harm Medicare beneficiaries.

Considerations For Future Analytical Work And Program Policy

While it is clear that MedPAC’s new model is incomplete and likely overestimates coding intensity and favorable selection effects through unaddressed analytical questions, the model also fails to distinguish overpayment from differential payment. Given a purported overpayment of $83 billion for beneficiaries enrolled in MA versus FFS, if that entire amount represented plan profit, UnitedHealthcare and Humana, representing nearly 47 percent of enrollees, would presumably reap $39 billion/year in excess spending. However, Humana reported $102.6 billion in annual revenues and $4.3 billion in earnings before interest, taxes, depreciation, and amortization (EBITDA) across all insurance lines for 2023; UnitedHealth Group reported annual revenue of $372 billion with $32 billion in EBITDA. Thus, the $39 billion in excess spending would represent more than the two companies’ combined pre-tax income across all lines of insurance business, suggesting that the health plans are losing money on their remaining lines of business—an unlikely scenario.

Just as not all differential payment is overpayment, neither is all differential payment contributing to plan profits. In fact, MA uses $2,328 in rebates per beneficiary annually to deliver additional benefits, meaning $69.8 billion (84 percent) of purported overpayments go toward reduced A/B cost-sharing, premium reductions, a prescription drug plan, and other supplemental benefits. MA thus attracts beneficiaries who are unwilling or, more worrisome, unable to purchase Medigap plans. Consequently, blind cuts in MA, versus targeted policy improvements that equalize the treatment of MA and FFS, would hurt the many poor and minority beneficiaries in the program and damage long-standing efforts to improve health equity.

Instead, future policy analyses of MA and FFS spending must be holistic. In addition to analyzing statutory program spending, analysts must consider component-by-component costs and the cost to both taxpayers and beneficiaries for the construction of a holistic health benefit package in both the FFS and MA programs. This could include analysis of taxpayer/beneficiary costs and induced demand as MA plans buy down Parts A and B beneficiary cost-sharing versus the much greater induced demand from FFS beneficiaries with Medigap, as nearly three-quarters of beneficiaries with Medigap are without any cost sharing.

Policymakers should also look to equalize the treatment of MA and FFS Medicare, promote value-based care, and differentiate between good and bad actors. If policymakers are concerned that MA is coded differently than FFS, regulators should look to improve coding accuracy for both programs. Solutions could include using artificial intelligence to crawl charts as a means of equalizing payment. In conjunction, policymakers could budget for thorough chart audits of FFS and MA populations to better measure differential coding. Other policies could include promoting site-neutral payments, competitive telehealth pricing, and tech-assisted and tech-driven solutions that reduce cost and expand access.

If MA program spending driven by inflated, formulaic benchmarks is a primary concern, policymakers could also consider staged reforms to Medigap plans, long a focus for programmatic reform; they could address the previously mentioned induced demand in FFS Medicare, driven by a lack of cost sharing created by Medigap, through the implementation of value-based insurance design in Medigap plans. Other policy alternatives worth exploring include larger geographies for plan bidding or benchmark reform through competitive bidding inclusive of FFS Medicare, with rigorous consumer protections such as grandfathering and risk corridors to protect vulnerable beneficiaries. To combat overpayments due to star ratings, policymakers and regulators could eliminate double-bonus counties (counties with high MA penetration and low FFS Medicare spending receive double bonuses) and steward the creation of a uniform quality ratings program for both FFS and MA.

To address high drug costs, regulators could mirror best practices in Medicaid and use managed care tools to implement value-based contracting for revolutionary new therapies. To better support consumers, regulators should improve the plan finder to help provide beneficiaries with more transparent cost and benefits information. Finally, with an increasingly complex Medicare population and changing delivery system, policymakers should support a diversity of beneficiaries through modernization and mass customization versus product or benefit standardization.

Indiscriminate, across-the-board program cuts would harm the millions of elderly and poor beneficiaries enrolled in MA. Policymakers should undertake a “measure twice, cut once” ethos, internalizing consumer protection as a core principle. Creative, contemporary policy solutions that preserve the strengths and minimize the weaknesses of the MA program should be coupled with simultaneous improvements to FFS Medicare, as the two programs are inextricably linked. While MA is imperfect, population-based payment provides the flexibility to meet the varying needs of America’s increasingly diverse population while creating a long-term framework for continuing the transition from volume to value.

Authors' Note

We would like to acknowledge the years of hard work on implementing and improving the Medicare Advantage and Part D Prescription drug programs by Jeffrey Kelman, MD, MMSc, the former Chief Medical Officer of CMS whose curiosity and skepticism inspired us all.

Kenny Kan is Chief Actuary of Horizon BCBS, which operates MA plans. The views expressed are the authors’ own and not necessarily those of their employers or affiliations.