On April 2, 2026, CMS published its Contract Year (CY) 2027 Medicare Advantage Rate Announcement — and buried inside a payment policy document that usually draws payer actuaries rather than medical coders was a change that reshapes how HCC risk adjustment works at the point of documentation.
Starting January 1, 2027, diagnoses discovered through unlinked chart review records can no longer be used to calculate Medicare Advantage risk scores. For the hundreds of health systems and MA plans that have spent years sending retrospective chart reviewers through prior-year records to capture HCC-eligible diagnoses, the economics just changed.
What CMS Actually Finalized
The CY2027 Rate Announcement finalized two specific diagnosis source exclusions. First, it bars diagnoses from unlinked chart review records (CRRs) — a chart review record is “unlinked” when the diagnosis it captures is not associated with a specific clinical encounter for that beneficiary during the measurement year. Second, it excludes diagnoses from audio-only services coded with modifiers 93 or FQ, which CMS has increasingly scrutinized as a risk adjustment inflation vector.
There is one narrow exception: beneficiaries who switch from one MA organization to another mid-year. CMS designed this carve-out to avoid penalizing receiving plans for their inability to link diagnoses to a prior plan’s encounter data.
The rule also maintained the 2024 CMS-HCC risk adjustment model — no recalibration this cycle — and held the statutory minimum coding pattern difference adjustment at 5.90%. But those constants do little to soften the structural shift: any plan or health system with a historically high dependence on unlinked retrospective chart reviews will see risk score compression starting in contract year 2027. The key source is the CMS CY2027 MA finalization press release (April 2, 2026).
Why the Old Model Is Going Away
Retrospective chart review has been the dominant risk adjustment “recovery” strategy for MA plans for over a decade. The workflow is straightforward: after the measurement year closes, a vendor or internal team reviews beneficiary medical records looking for HCC-eligible diagnoses — conditions that may have been documented by a treating provider but were never coded on a submitted claim. When found, those diagnoses are coded and submitted as chart review records, raising the beneficiary’s risk score and triggering higher per-member per-month payments from CMS.
CMS has viewed this practice with increasing suspicion. The agency’s retrospective data showed that chart review submissions were disproportionately adding diagnoses that had no corresponding encounter — meaning a plan could receive higher payments for comorbidities that may have been historical, resolved, or otherwise not reflective of current-year clinical status. The Elevance Health case — in which CMS forced a $342 million repayment in May 2026 for Medicare Advantage overpayments tied to risk coding practices — put a face on the systemic scale of the problem.
The unlinked chart review ban is CMS’s structural response: not just enforcement after the fact, but a prospective rule change that removes the payment incentive entirely.
What Agentic AI Changes in This New Environment
The chart review ban does not eliminate the need for thorough diagnosis capture — it changes when and how that capture must happen. The diagnosis still has to be documented; now it must be tied to a real clinical encounter during the measurement year.
This is where prospective CDI and agentic AI tools matter. Rather than mining charts after the year closes, these systems work inside the encounter workflow in real time, doing several things that retrospective chart reviewers cannot:
- Flagging HCC-relevant chronic conditions from prior years that have not yet appeared in the current year’s encounter documentation
- Prompting providers to document or confirm the continued presence of conditions such as CHF, COPD, or diabetes with complications before the encounter note is closed
- Surfacing coding specificity gaps — for example, “diabetic nephropathy” coded at a non-HCC level when a more specific ICD-10-CM code would qualify for risk adjustment
The result is a diagnosis that is both clinically valid and encounter-linked — exactly what the CY2027 rule requires.
Why Specificity Matters More Now
Under the old model, a chart reviewer could find a prior-year record noting “type 2 diabetes with CKD” and submit that as an HCC capture for the current year. Under the new model, the provider must document and encounter must occur in the current measurement year. That creates a hard documentation quality imperative: if the provider sees the patient but codes only “type 2 diabetes” (non-HCC) instead of “type 2 diabetes with chronic kidney disease, stage 3” (HCC 136), the capture is permanently missed for that measurement year — there is no retrospective fix available.
Higher specificity at the point of care is no longer just a coding best practice. Under CY2027 rules, it is the only remaining path to complete, audit-defensible risk adjustment.
The Elevance Precedent and What Comes Next
The Elevance repayment was a landmark enforcement action — the first time CMS used its enrollment authority, rather than the False Claims Act or a whistleblower suit, to compel direct repayment of $342,209,085 in Medicare Advantage overpayments. The mechanism was unusual; the implications were not. CMS has made clear that it has both the data tools and the regulatory willingness to pursue MA plans whose coding patterns diverge systematically from fee-for-service benchmarks.
The CY2027 chart review exclusion closes the regulatory loop. Plans that continue submitting unlinked chart review records after January 1, 2027 will find those diagnoses excluded from risk score calculation automatically — and any claims-based recouping that follows will be on weaker footing than before, since the diagnosis source is now explicitly non-compliant.
Three Things Coding Teams Must Do Before January 2027
Providers, coding directors, and MA plan CDI teams have roughly six months to adjust their workflows. Here is where to start.
First, audit your current RAF improvement program. Identify what percentage of your annual risk score uplift is derived from unlinked CRR submissions. Those risk points are at risk under the new rule, and plans should quantify the exposure now — not in Q4 2026.
Second, retool HCC training around encounter-based capture. Coding education that focuses on retrospective recognition must shift to real-time documentation — helping providers understand what to document, at what specificity level, and why the encounter note is now the only valid HCC source.
Third, evaluate prospective CDI and AI-assisted coding tools. Platforms that integrate into the EHR workflow, surface HCC gaps before the note is closed, and provide provider-facing specificity prompts are now a compliance requirement, not just an efficiency play.
The Documentation Standard CMS Is Enforcing
The unlinked chart review ban is not a technical footnote in a payment update — it is the regulatory conclusion of a decade of CMS concern about Medicare Advantage risk score inflation. For HCC coders and CDI specialists, the message is unambiguous: a diagnosis must be documented in a real clinical encounter during the measurement year to count toward risk adjustment.
The upstream implication is that documentation quality now belongs in every provider-coder interaction, every CDI workflow, and every EHR design decision — before the measurement year closes, not after it ends. MA plans and health systems that make that shift in the second half of 2026 will enter 2027 with risk adjustment programs that are both financially intact and audit-ready. Those that do not will be running a program CMS has effectively made illegal.
Medikode’s automated medical coding platform helps health systems and MA plans capture HCC-eligible diagnoses at the point of encounter, ensuring every code is encounter-linked, clinically specific, and audit-ready under the CY2027 rules.