From V24 to V28 to 2027: How CMS Risk Adjustment Policy Keeps Getting It Wrong

In early 2024, when CMS released its Advance Notice proposing sweeping changes to the CMS-HCC risk adjustment model, I published an analysis warning that the proposed revisions were contradictory to the agency's own stated goals. I argued that flattening coefficients for chronic conditions, removing thousands of diagnosis codes, and misreading the causation behind coding variation between Medicare Advantage and traditional FFS Medicare would reverse a decade of progress — harming the very populations and provider organizations that had made value-based care work.

At the time, some of those warnings may have seemed like advocacy. The data since has made them look more like forecasting.

Now, with the 2027 rate notice finalized, it's worth stepping back for a full accounting: what was proposed, what CMS adjusted between the Advance Notice and the Final Rule (and critically, why those adjustments matter), what actually happened in the market, and what the trajectory of these decisions signals for ACOs, risk-bearing providers, and the future of value-based care. The Final Rule stepped back from the most damaging proposals — but CMS has been clear that further model changes are coming. The direction of travel has not changed; only the pace.

This is that accounting.

Part One
What We Said in 2024 — and Why

The core argument in my 2024 analysis was straightforward: CMS's proposed HCC reclassification was built on a flawed assessment. The agency identified higher rates of certain diagnoses in Medicare Advantage populations compared to traditional FFS Medicare and concluded this represented discretionary or inappropriate coding. The proposed remedy was to remove or flatten those diagnoses in the risk model.

The problem is that correlation is not causation. There are at least three structural reasons why MA populations carry more documented diagnoses that have nothing to do with inappropriate coding:

Population differences. MA plans disproportionately serve lower-income, socioeconomically disadvantaged beneficiaries — populations with higher burdens of chronic illness. You would expect to see more documented diagnoses in these populations regardless of coding behavior.

Provider network differences. MA plans contract with high-performing, value-oriented provider networks that practice thorough documentation as part of coordinated care. Traditional FFS providers have no financial incentive to document comprehensively. The difference in coded diagnoses reflects a difference in care quality, not fraud.

Care model differences. Value-based MA arrangements incentivize early detection, preventive care, and ongoing chronic disease management. Patients who are actively managed will have more documented conditions than patients who only seek acute care. Again, this reflects the system working as intended.

By treating these documented differences as evidence of fraud rather than evidence of better care, CMS proposed to punish the organizations doing the most rigorous, preventive, patient-centered work. I further noted that the proposed changes selectively applied Principle 10 of the CMS model development guidelines — which excludes discretionary coding categories — while ignoring Principle 5, which specifically calls for encouraging specific and detailed coding. The proposals were internally inconsistent with CMS's own framework.

The variation in coding can be explained by sample biases. Any inappropriate coding of these conditions should be addressed through the RADV audit and oversight responsibilities of the plans — not by restructuring the risk model in ways that harm every compliant provider in the country.

The specific concern was the downstream incentive structure these changes would create. By flattening or reducing coefficients for chronic conditions like diabetes and depression while leaving high coefficients for severe acute complications like necrotic wounds and vascular disease, the model would financially reward providers who let conditions deteriorate to crisis level and penalize those who prevented the crisis from occurring. This is the opposite of value.

Part Two
What the V28 Advance Notice Proposed vs. What the Final Rule Delivered

When the original Advance Notice was released, the proposed risk model impact was a -3.12% reduction in risk scores across the MA population. The outcry from plans, provider groups, and patient advocates was significant. America's Physician Groups, the American College of Physicians, and numerous health equity organizations submitted comments arguing — consistent with my analysis — that the changes would harm disadvantaged populations and undermine value-based care models.

CMS made modifications between the Advance Notice and the Final Rule. The net revenue impact was adjusted to -2.16% after those modifications, with expected average revenue rising modestly to 3.32%. A phased implementation was adopted: 33% of the new model in 2024, blended up to 100% by 2026.

These were meaningful concessions. But they did not address the structural problem — the coefficient flattening and hierarchy changes that rewarded acute-focused, facility-centric care over prevention-oriented, community-based care remained intact. The phase-in softened the financial blow; it did not change the direction of travel.

The result, as the market has now demonstrated, is that the phase-in bought time but not correction. The incentive misalignment was baked into the model regardless of the pace of implementation.

Part Three
What Actually Happened — The Market Verdict

Three years later, the outcomes are measurable. A Milliman study commissioned by America's Health Insurance Plans found that approximately 10% of Medicare Advantage enrollees — roughly 3.4 million people — were forced to switch plans for 2026 due to exits, mergers, or drastic benefit reductions.¹

Premiums rose an average of 5.2%, out-of-pocket maximums climbed 3–4%, and networks narrowed — with the most pronounced impacts on low-income, rural, and minority beneficiaries. This reverses the equity gains that CMS's own 2021 study documented in the MA program from 2009 to 2018.

High-performing providers in value-based arrangements — precisely the organizations CMS claims to want to grow — saw revenue drops of 20–35% as the flattened coefficients failed to capture the complexity of their patient populations. These are the providers doing the work CMS's ACO REACH and MSSP programs were designed to expand. The risk model changes pulled the financial rug out from under them.

The one element I underestimated in my 2024 analysis was timing. I predicted the Trust Fund savings would prove illusory over time as higher utilization offset reduced payments. In reality, the cost increases materialized more quickly than anticipated — higher utilization is already pushing HI Trust Fund insolvency projections to 2033, three years earlier than pre-2024 estimates under some analyses.³

This is the textbook definition of a policy that achieved its short-term accounting goal while accelerating the long-term problem it was meant to solve.

Part Four
The 2027 Notice — Building on a Flawed Foundation

Against this backdrop, CMS released the 2027 Advance Notice in January 2026 and finalized its approach in the subsequent Call Letter. The 2027 model sticks with V28's structure of 86 HCCs and its fundamental hierarchy, recalibrating coefficients using more recent data — 2023 diagnoses predicting 2024 expenditures — resulting in an overall risk score impact of -3.32%.

Some elements of the 2027 proposals were genuinely positive — and these are the reforms that made it into the Final Rule. Excluding unlinked chart reviews and audio-only encounters from risk score calculation improves data integrity and ties payments more directly to actual documented care. These are good-faith improvements and deserve acknowledgment.

But CMS's initial Advance Notice proposals would have created a deeper structural problem: a coefficient hierarchy that continues to undervalue chronic disease management while assigning outsized weights to advanced, often hospital-treated complications. The Final Rule stepped back from the most damaging of those coefficient changes — though the agency has signaled this is a pause, not a reversal, and that further model recalibration is forthcoming.

The Coefficient Story

Consider how the model values common conditions in the community non-dual aged segment:

While the most significant coefficient changes were struck from the Final Rule, the pattern the Advance Notice revealed is clear and consistent: conditions that respond to proactive primary care management — depression, diabetes — lose value under the direction CMS is signaling. Conditions that represent the failure of that management — vascular disease, necrotic wounds — remain highly compensated. The model, as CMS has proposed it, rewards escalation over stabilization, and the agency appears set on continuing down that path in future rulemaking.

If a provider optimizes for their risk score under this model, the rational behavior is to under-document early-stage chronic conditions (low coefficient) and over-invest in identifying and documenting advanced complications (high coefficient). This is precisely backwards from what value-based care is designed to achieve. Expert commentary from the American College of Physicians, the American Hospital Association, and Brookings Institution analysts has reached similar conclusions.^6,7,8

What the 2027 Notice Got Right

It is worth being precise about where CMS's direction is sound. The exclusion of unlinked chart reviews — documentation not tied to an actual clinical encounter — removes a vector for gaming the risk score without providing care. Audio-only encounter exclusions similarly tighten the connection between payment and documented in-person care. These changes should be preserved and built upon. The problem is not that CMS is trying to improve data integrity; the problem is that the underlying model architecture they are refining continues to incentivize the wrong outcomes.

Part Five
The Broader Context: A Bipartisan Policy Failure

It would be convenient to frame this as a story about one administration's mistakes, but the record does not support that framing. The V24-to-V28 transition began under one administration and the 2027 proposals compound those changes under another. The structural pressure to reduce MA payment rates — driven by trust fund concerns and documented fraud in corners of the market — crosses political lines. So does the failure to adequately account for the collateral damage to the legitimate, high-performing providers who are doing exactly what the program should incentivize.

This is a policy design problem as much as a political one. The risk adjustment model is a technical instrument that shapes clinical and financial incentives for millions of providers and patients. When it is calibrated incorrectly, the consequences are not abstract — they show up in plan exits, benefit reductions, revenue losses for community-based providers, and ultimately in patients who receive less coordinated, less preventive care.

The solution is also bipartisan in nature. Improving data integrity, addressing genuine fraud through robust RADV auditing, and redesigning the model to appropriately reward prevention rather than reward its failure — these are goals that any administration should be able to advance. The question is whether the agency will engage seriously with the evidence that its current approach is producing the opposite of its stated outcomes.

Part Six
What ACOs and Risk-Bearing Providers Should Do Now

The comment period for the 2027 Final Notice has closed — and it's worth acknowledging that it worked. The advocacy by groups including America's Physician Groups, the American College of Physicians, and the American Hospital Association, alongside the broader provider and plan community, contributed to CMS stepping back from the most harmful coefficient proposals in the Final Rule. That's a meaningful win and a demonstration that organized, evidence-based advocacy can move policy.

CMS has been clear, however, that further risk model recalibration is coming in future years. The structural direction has not changed — only the timeline. As that future rulemaking takes shape, continued provider advocacy will be critical to ensuring the model evolves toward accurately capturing risk adjustment in ways that encourage value-based care rather than punish it. The comment periods ahead matter. Engaging through your trade associations, contributing real-world performance data to the public record, and building relationships with Congressional representatives who oversee CMS are all levers that remain open.

Conclusion

Three years ago, I argued that CMS's proposed risk adjustment changes were built on a misreading of evidence and would produce outcomes contrary to the agency's stated goals. The market data since has been consistent with that warning: millions of beneficiaries disrupted, revenue losses for high-performing VBC providers, and a trust fund depletion trajectory that accelerated rather than improved.

The 2027 finalized notice makes targeted improvements at the edges — better data integrity provisions are genuinely welcome. But it does not correct the fundamental architecture of a model that financially rewards the failure of preventive care over its success. Until that architecture is revisited, every recalibration will refine a structure that points in the wrong direction.

The evidence for a better approach exists. SDoH-integrated coefficients that actually predict cost for the populations most burdened by social risk. Nuanced differentiation between early-stage and advanced chronic conditions that rewards the management preventing progression. Continued investment in RADV auditing to address actual fraud without penalizing compliant, high-performing providers across the board.

These are not radical proposals. They are the logical extension of CMS's own stated commitments to health equity, value, and innovation. The question is whether the agency will follow the evidence — or continue to refine a model that contradicts it.