CY2027 Rate Announcement: Audio-Only and Unlinked CRR Diagnoses Out of MA Risk Scores

The era of encounter-disconnected retrospective diagnosis capture is over for 2027

If your retrospective coding pools rely on audio-only visits or CRRs that aren't tied to a matching encounter record, those diagnoses are out of the risk score calculation starting January 1, 2027. CMS published the CY2027 Rate Announcement on April 6, 2026, and both exclusions are final.

This isn't a surprise policy direction. CMS has signaled for years that it would tighten what counts. CY2027 is where the technical capability caught up with the intent.

What changed — and what didn't

V28 is fully implemented and continues for CY2027. No V29 transition is happening for 2027.

Audio-only exclusion: modifier 93 and FQ

CMS framed the audio-only exclusion as enforcing a longstanding policy that its systems can now technically apply. Diagnoses from encounters coded with modifier 93 or modifier FQ are out of the risk score calculation. Diagnoses from telehealth that includes a real-time audio-and-video component still count — it's the audio-only encounter, with no video, that gets excluded.

To be clear on the mechanics: 93 and FQ are service modifiers on the encounter claim, not diagnosis codes. The diagnosis itself may be valid and codeable. The exclusion is based on the encounter type — audio-only — not the dx code itself. A diagnosis from an audio-only visit needs a face-to-face encounter to carry risk score weight.

CMS estimated the aggregate payment impact of the audio-only exclusion at approximately 0% on average, but plans with concentrated audio-only telehealth volume will see larger effects at the contract level.

If you're doing retrospective sweeps and pulling diagnoses from telehealth visits, check the encounter modifier. An audio-only visit with modifier 93 or FQ doesn't support a risk-adjustable dx in CY2027 without a separate face-to-face record for the same condition.

Unlinked CRR exclusion: what "unlinked" means

A Chart Review Record is a document — typically a retrospective review — submitted as a diagnosis source. An unlinked CRR is one where the diagnoses captured in that review don't also appear in a separately submitted encounter data record tied to a specific service.

Starting January 1, 2027, unlinked CRR diagnoses don't count. The diagnosis needs to be anchored to an actual encounter that also appears in encounter data.

The finalized payment impact is -1.53%. Georgetown puts that at a reduction in MA payments of over $7 billion in 2027 relative to 2026 — the dollar scale of what unlinked CRRs are currently adding to risk scores.

The switcher exception

There is one carve-out: unlinked CRR diagnoses still count for beneficiaries who switch from one MA organization to another. The logic is that a member switching plans may not have encounter data available from the prior plan, so the CRR provides the only capture path.

This exception does not apply to members who stay in the same plan. For members who remain, the rule is firm: diagnoses from unlinked CRRs don't count starting CY2027.

The -1.78% figure you may have seen elsewhere is the counterfactual impact without the switcher exception — the number CMS used to illustrate how much the exception reduces the exclusion's overall effect. The finalized, adopted figure is -1.53%.

What didn't move: V28 recalibration and the coding pattern adjustment

Two items matter for how you project CY2027 scores: the proposed V28 recalibration, which CMS did not finalize, and the coding pattern adjustment, which CMS held flat.

V28 recalibration: CMS proposed recalibrating the V28 model for CY2027 using more recent Medicare FFS data, which would have changed HCC coefficients. It did not finalize that recalibration. Instead, CMS retained the existing 2024 model calibration — using 2018 diagnosis data and 2019 expenditure data — for CY2027.

V28 is still fully in effect. No code that didn't map under V28 suddenly maps again. If you've already worked through the V28 model changes, that mapping set is what applies.

Coding pattern adjustment: This is the one item that didn't change. CMS held the MA coding pattern difference adjustment at the 5.9% statutory minimum for CY2027 — the same level as CY2026, and the same figure proposed in the Advance Notice. It was not increased. So the downward pressure on effective risk scores from the coding adjustment is steady year over year; don't model a higher coding intensity haircut than the 5.9% you already have.

What the +2.48% overall rate change means

The net CY2027 payment rate increase is 2.48% over CY2026. The Advance Notice had proposed 0.09%. The jump is driven substantially by the decision not to recalibrate V28 — retaining the older calibration preserves higher coefficients that a recalibration would have adjusted downward.

The two new exclusions (audio-only and unlinked CRR) create headwinds at the plan level, but the overall rate environment is better than what the Advance Notice signaled.

What this means operationally for retrospective coding

The practical effect is a higher bar for any diagnosis captured outside of a face-to-face encounter. For conditions identified in retrospective chart reviews, the diagnosis needs to appear in both the CRR and an encounter data record for the same service. For conditions identified from telehealth, an audio-only visit isn't enough — you need a face-to-face confirmation.

Neither of these exclusions affects prospective coding at a face-to-face visit with full documentation. That pathway is unchanged.

If you're unsure whether a specific dx has the MEAT and encounter linkage to survive these new exclusion rules, that's the place to start.