CMS finalized the CY2027 Rate Announcement on April 6, 2026, with a net average payment increase of 2.48 percent, over $13 billion, rising to roughly 4.98 percent (about $26 billion) once projected risk score trend is included. [1][4]
But the headline hides the exposure. CMS finalized two diagnosis-source exclusions that reduce risk scores: diagnoses from unlinked chart review records and diagnoses from audio-only encounters no longer count. [1] The unlinked chart review exclusion alone cuts MA payments by an average of about 1.53 percent. [4]
That 1.53 percent is an average. Your number depends entirely on how much of your risk score was riding on those two sources, and most plans have never measured them separately.
This post is about the money. For how the underlying model works, see our CMS-HCC Model V28 guide. For audit mechanics, see RADV audits in 2026.
Key Takeaways
- Net CY2027 payment increase: 2.48 percent, over $13 billion. With risk score trend: roughly 4.98 percent, about $26 billion. [1][4]
- The increase is real, but it exists largely because CMS declined to recalibrate V28. It is a deferral, not a windfall. [1]
- Two diagnosis-source exclusions were finalized: unlinked chart review records and audio-only encounters (modifiers 93 and FQ). [1]
- The unlinked CRR exclusion reduces payments by about 1.53 percent on average. Without the MA-to-MA switcher exception CMS added, it would have been roughly 1.78 percent. [4]
- These are averages. Plans with heavy reliance on unlinked chart review or telephonic outreach will land well above them.
- The MA coding pattern difference adjustment stays at the statutory minimum of 5.90 percent, unchanged from CY2026. [5]
- Add-only retrospective coding is now both unpaid and legally exposed. The March 2026 Aetna settlement, $117.7 million, turned on exactly that design. [3]
What CMS finalized, and what it costs
The two exclusions
Unlinked chart review records. Diagnoses drawn from chart review records not associated with a specific beneficiary encounter are excluded from risk score calculation. Plans may still submit them. Those diagnoses simply will not generate payment. [1]
One exception: diagnoses from unlinked CRRs still count for beneficiaries switching from one MA organization to another. CMS added this after comment, and it materially softened the blow. Without it, the payment reduction would have been approximately 1.78 percent instead of the 1.53 percent finalized. [4]
Audio-only encounters. Diagnoses coded from audio-only services, identified by modifiers 93 and FQ, are also excluded from risk score calculation. [1][5]
This one gets overlooked, and it is a distinct exposure. A plan that leaned on telephonic annual wellness visits or telephonic chronic-condition outreach carries risk here even if its chart review program was clean. Both exclusions also apply to the Part D RxHCC models. [1]
What did not change
CMS did not finalize the proposed V28 recalibration. The 2024 CMS-HCC model, calibrated on 2018 diagnoses and 2019 expenditures, continues into CY2027. CMS cited a need to give the market more time to adjust after the phase-in completed in CY2026. [1]
This deferral is the single largest reason the final rate came in at 2.48 percent against a proposed 0.09 percent. [4]
Two things follow. The coding value map is stable for one more year, so documentation priorities do not have to shift around a new model. And this is temporary. CMS has signalled recalibration returns, and the longer the deferral runs, the sharper the eventual adjustment.
Is Your Risk Adjustment CY2027-Ready?
A self-assessment for Medicare Advantage and value-based care organizations
CY2027: proposed vs finalized
| Metric | Advance Notice (Jan 2026) | Rate Announcement (Apr 6, 2026) |
| Net average payment increase | 0.09% (~$700M) | 2.48% (over $13B) |
| Including risk score trend | 2.54% (~$13B) | ~4.98% (~$26B) |
| V28 recalibration | Proposed | Not finalized |
| Unlinked CRR exclusion | Proposed | Finalized, with switcher exception |
| Audio-only exclusion | Proposed | Finalized (modifiers 93, FQ) |
| Payment impact of CRR exclusion | ~1.78% reduction | ~1.53% reduction |
| Coding pattern difference adjustment | 5.90% | 5.90% (unchanged) |
Sources: CMS CY2027 Rate Announcement [1], Georgetown CHIR [4], HealthEdge [5].
The number that matters is not 1.53 percent. It is yours.
Every summary of this rate cycle quotes the 1.53 percent average. Almost none of them say the obvious next thing: an average is a distribution, and somebody is on the wrong end of it.
Two plans with identical membership can land in completely different places depending on how their risk scores were built. Most organizations have never separated these exposures, because until CY2027 there was no payment reason to.
Here is how to quantify yours before bid season closes.
Step 1: Isolate diagnoses that exist only in unlinked chart reviews
This is the load-bearing question, and it is narrower than “how much chart review do we do.”
For each HCC contributing to your CY2026 risk scores, ask: is there a qualifying encounter that independently supports this diagnosis? If the answer is yes, the CRR exclusion does not touch it. The diagnosis survives on the encounter data.
The exposure is the residual: HCCs whose only supporting source is an unlinked chart review record. In most programs that residual is smaller than leadership fears and larger than the coding team expects. Measure it, do not estimate it.
Step 2: Isolate diagnoses sourced from audio-only encounters
Filter for modifiers 93 and FQ. Then run the same test: does another qualifying encounter independently support the HCC?
Plans that pushed hard on telephonic outreach during and after the pandemic often carry a concentrated exposure here, and it is completely invisible in a chart-review-focused analysis. This is the number most teams have not looked at.
Step 3: Net out the switcher population
Diagnoses from unlinked CRRs still count for beneficiaries who switched into your plan from another MA organization. Identify that cohort and remove it from your Step 1 exposure.
For plans with heavy MA-to-MA switching, this exception is worth real money. For plans growing mostly from age-ins and Original Medicare, it is worth close to nothing. Know which you are.
Step 4: Compare against the 1.53 percent benchmark
Now you have a defensible internal figure. Set it against the industry average.
- Materially below 1.53 percent: your risk scores were already substantially encounter-linked. Your CY2027 problem is smaller than the market’s, and that is a competitive position worth understanding.
- At or above 1.53 percent: you carry above-average exposure, and it will show up in bids and in margin. It is also, separately, an audit signal.
Step 5: Separate the recoverable from the unrecoverable
Not all of the exposure is lost. Some diagnoses currently supported only by an unlinked chart review are clinically real and were simply never captured at an encounter. Those are recoverable through prospective capture: get the condition documented at the visit, and it counts again.
The rest are not recoverable, and they were never defensible. Those codes were going to fail a RADV review whether or not CMS changed the payment rule. The exclusion did not create that liability. It revealed it.
Splitting your exposure into these two buckets is the single most useful output of this exercise, because it tells you what to fix with workflow and what to write off.
Why the Aetna settlement changes how you read all of this
In March 2026, the Department of Justice reached a $117.7 million False Claims Act settlement with Aetna over its risk adjustment practices. [3]
The allegation was not that the codes were merely wrong. It was that the chart review program only ran one way: it added diagnoses that increased payment, while failing to remove unsupported diagnoses that the same review had already identified. Of the total, $106.2 million resolved the chart review allegations, and $11.5 million resolved unsupported morbid obesity codes for payment years 2018 through 2023. [3]
Read that alongside the CY2027 exclusions and the picture is unambiguous. CMS has stopped paying for unlinked retrospective volume. DOJ has demonstrated what it costs when that volume cannot be defended.
Which means the exposure analysis above is not only a bid exercise. Step 5 is a compliance exercise. The unrecoverable bucket, the diagnoses that exist only in an unlinked chart review and cannot be supported by a real encounter, is precisely the population an auditor would ask about. Finding them and removing them is not a revenue loss. It is the removal of a liability you were already carrying.
An add-only program cannot do this. By construction, it has no delete path.
What to do before bid season closes
- Run the five-step exposure analysis. Quantify unlinked CRR and audio-only reliance separately. Do not accept the 1.53 percent average as a proxy for your own number.
- Model both exclusions into CY2027 bids. Apply them alongside the normalization update and the unchanged 5.90 percent coding pattern adjustment. [5]
- Convert the recoverable bucket to prospective capture. Conditions that are clinically real but only ever appeared in a chart review need to be surfaced at the point of care, where they can be documented, evaluated, and coded defensibly.
- Build a delete path. If your retrospective program cannot remove an unsupported code as readily as it adds a missed one, you are running the design that produced a nine-figure settlement.
- Use the V28 freeze. The coefficients are stable for one more year. That is a window to fix documentation quality, not a reason to coast. When recalibration returns, it will land on whatever foundation you built this year.
How RAAPID approaches this
RAAPID’s Clinical AI Platform, built on Neuro-Symbolic AI, links every suggested HCC to MEAT-based evidence in the clinical note, so each diagnosis carries an explainable, auditable trail rather than a black-box score.
RAAPID’s retrospective risk adjustment solution is two-way by design. It surfaces missed diagnoses that carry clinical support and flags unsupported codes for removal, which is exactly the delete path the CY2027 exclusions and the Aetna settlement both require. Prospective risk adjustment closes the other half, moving capture to the encounter, where CY2027 now requires it to live.
Frequently Asked Questions
CMS finalized a net average increase of 2.48 percent, over $13 billion. Including projected risk score trend, the figure is roughly 4.98 percent, about $26 billion. [1][4]
Plans may still submit them, but diagnoses from unlinked chart review records no longer count toward risk score calculation. One exception: they still count for beneficiaries switching between MA organizations. [1]
Yes. Diagnoses coded from audio-only services, modifiers 93 and FQ, are excluded from risk score calculation for CY2027. This applies to Part C and to the Part D RxHCC models. [1][5]
The industry average is about 1.53 percent. [4] Your figure depends on what share of your HCCs are supported only by an unlinked chart review, and on your MA-to-MA switcher population. Both need to be measured, not estimated.
No. CMS did not finalize the proposed recalibration. The 2024 CMS-HCC model continues into CY2027, and CMS has signalled that recalibration will return in a future year. [1]
Primarily because CMS declined to recalibrate the V28 model, and secondarily because it added the switcher exception and incorporated more complete fee-for-service spending data. [4]
Autonomous Retrospective Risk Adjustment Solution
One platform. Every HCC validated. Revenue secured.
