CMS Bans Unlinked Chart Reviews: $7B Hit to MA Plans Starting CY 2027

CMS published the Contract Year 2027 Medicare Advantage and Part D final rule on April 2, 2026, confirming that diagnoses submitted on unlinked chart review records will no longer count toward CMS-HCC risk scores. The agency puts the payment effect at $7.12 billion, a 1.53% aggregate reduction.

The average is the least useful number in the rule. Reliance on unlinked records was concentrated, so a single-digit national figure sits on top of plan-level exposure that ranges from nothing to several points of revenue, with bids due June 1.

Key Takeaways

• $7.12 billion, or 1.53% of aggregate MA payments, comes out of CY 2027 risk scores, the largest single-year risk adjustment methodology change since the V28 phase-in began in CY 2024.
• MA organizations submitted 88.8 million unlinked chart review records in 2023 for 2024 payment, and CMS found 85% could not be matched to any encounter record even allowing a three-day window on beneficiary, billing provider, and service dates.
• The switcher exception is worth about $2.5 billion. CMS scored the change at negative 1.24% with it against negative 1.78% without, a 0.54 percentage point spread.
• Exposure runs from 0% to 4% of revenue by plan archetype, so the 1.53% average describes almost no individual plan's bid.
• The 2.48% final rate is not a clean offset. A plan with 3% unlinked-record exposure still nets a 0.5% revenue reduction against its CY 2026 baseline before medical trend.

What Was Excluded, and What Was Not

The rule cuts one submission type, not chart review. A linked chart review record ties back to a previously submitted encounter data record: the plan reviews a chart, finds a diagnosis present at a documented face-to-face visit but absent from the original claim, and corrects the gap against a specific date of service. Those remain fully eligible in CY 2027.

An unlinked record references no encounter. It introduces a diagnosis into the risk adjustment data without connecting it to a billable service, so a condition documented somewhere in a beneficiary's chart could raise the risk score whether or not the beneficiary was treated for it that year.

The volume is what made this a payment-policy question rather than a coding question. MA organizations submitted 88.8 million unlinked records in 2023 for 2024 payment, and CMS analysis found 85% could not be matched to any encounter record even when matching on beneficiary, billing provider, and dates of service within a three-day window.

The HHS Office of Inspector General had quantified the same pattern on payment year 2017 data six years earlier, and Georgetown's analysis of the final rule reads it as the first direct administrative strike at coding intensity.

The Average Describes Almost No One's Bid

Exposure tracks how a plan built its documentation function, which is why the distribution is bimodal rather than spread around 1.53%.

National plans with centralized retrospective coding operations, running vendor chart review across 80% or more of membership, sit at an estimated 2.5% to 4.0% revenue reduction. Regional plans that used unlinked records selectively on high-acuity members or specific hierarchies run 1.0% to 2.5%. Provider-sponsored and staff-model organizations, where documentation improvement lives in the clinical workflow and diagnoses arrive through face-to-face encounters and linked records, sit between 0% and 1.0%.

The rate announcement does not neutralize that. Plans that began building CY 2027 bids against the January Advance Notice's 0.09% baseline got a 2.48% final rate, which helps, but a plan carrying 3% unlinked-record exposure still lands at a net 0.5% reduction against its CY 2026 baseline before medical trend runs through.

The switcher exception is the one lever CMS added on comment, and its value is plan-specific in the same way. A receiving MA organization may submit unlinked records for an incoming member's pre-existing conditions, because it has no encounter history with that beneficiary and would otherwise carry the cost of managing a condition like stage IV chronic kidney disease for months before capturing it.

CMS scored the finalized policy at negative 1.24% with the exception against negative 1.78% without, roughly $2.5 billion preserved. It does not cover transitions from original Medicare, so its value is a direct function of MA-to-MA churn: plans in markets churned by Star Ratings moves or competitor service-area exits recover more of the loss than plans with stable enrollment.

That leaves the bid itself. With the final rule published April 6 and bids due June 1, the net reduction has to flow through the A/B revenue projection in about eight weeks, and for plans at the top of the exposure range it can exceed the margin built into the CY 2026 bid, which puts supplemental benefits (dental, vision, hearing, fitness, OTC allowances) into the reconciliation.

The Records Stop Paying and Stay Auditable

Closing the prospective mechanism does not close the retrospective one, and the two run on different clocks.

From CY 2027 unlinked diagnoses do not contribute to risk scores, so they cannot generate a Risk Adjustment Data Validation overpayment finding going forward. Payment years before CY 2027 are a different matter, because in those years the same diagnoses did drive payment.

The OIG work plan carries an active project, "Audits of Medicare Part C Unlinked Chart Review Diagnosis Codes" (SRS-A-25-018), aimed at exactly that window. A plan that ran a large vendor-driven unlinked program is therefore losing the revenue prospectively while retaining the audit exposure on the years that revenue was earned, which belongs in a contingent liability estimate rather than a bid memo.

Sizing it is genuinely uncertain, and the disagreement is on the record. OIG estimated $2.7 billion of payments from unlinked records for a single payment year; CMS put the figure nearer $675 million after applying blending adjustments between encounter data and RAPS submissions. Both parties agreed on the direction and differ by a factor of four on the amount, which is the range a plan's own exposure estimate has to live inside.

Nor is the correction complete. MedPAC's March 2026 report puts MA payments $76 billion above what traditional Medicare would spend on the same beneficiaries, 14% above FFS equivalent, with coding intensity running 10.3% higher. Unlinked records are the component least connected to clinical activity, which is why they went first, and CMS has deferred the risk adjustment model recalibration using 2023 diagnoses and 2024 expenditures to a later payment year.