For many Medicare Advantage organizations, the retrospective chart sweep has been the backbone of risk adjustment. Pull records after the year closes, have coders find documented conditions that never reached a claim, and submit them as chart review records. The CY2027 Advance Notice, released January 26, proposes to stop counting diagnoses from unlinked chart review records in risk scores beginning in 2027. If CMS finalizes it in April, the economics of that operating model change overnight.

The proposal is not yet final, and comments close February 25. But organizations that wait for the Rate Announcement to begin redesigning their programs will lose most of a dates-of-service year. Diagnoses that count for 2027 payment come largely from 2026 encounters, which are happening now.

What Is Actually at Risk

The distinction that matters is linkage. A linked chart review record is tied to a specific encounter data record that the plan has already submitted. An unlinked record reports a diagnosis without connecting it to any submitted service. CMS's position is that a chart review should not be the only record of a healthcare service, and that the diagnosis should be supported by a documented encounter.

The HHS Office of Inspector General raised this issue in 2019, finding that chart reviews were used almost entirely to add diagnoses and recommending that CMS reassess the use of unlinked reviews. Georgetown University's Center on Health Insurance Reforms reports that nearly 58% of MA contracts submitted unlinked chart reviews in 2022. CMS estimates the proposed exclusion would reduce average MA payments by 1.53%, with a larger effect for organizations that rely more on unlinked records.

Two things are not proposed to change. Linked chart reviews would still count. And diagnoses still have to be supported by medical record documentation from an acceptable face-to-face encounter. The proposal does not eliminate retrospective review. It eliminates retrospective review as a substitute for encounter-level documentation.

Step One: Measure Your Exposure Precisely

Before redesigning anything, risk adjustment leaders need a clear inventory. For each contract, determine:

  • The share of RAF attributable to HCCs that appear only on unlinked chart review records.
  • Which provider groups, specialties and settings generate those diagnoses.
  • Whether a matching encounter exists that the diagnosis could have been linked to, and why it was not.
  • Which HCCs are most affected. Chronic conditions that are rarely addressed in primary care visits, such as certain vascular, neurologic and behavioral health conditions, often surface only in retrospective review.

The last two questions separate the problem into two categories. Some unlinked diagnoses had a real encounter behind them and were unlinked because of workflow or data issues. Others were found in records with no corresponding submitted service. The first category is largely fixable. The second requires capturing the condition during a future encounter.

Step Two: Shift Capture Earlier

The durable answer is to document conditions when the patient is in front of the provider. That means rebalancing the program across three modes:

  1. Prospective review. Before a scheduled visit, review the chart, prior claims and clinical data to identify suspected or previously documented conditions that need to be assessed this year. Deliver those to the provider as evidence-based prompts, not as diagnosis lists.
  2. Concurrent review. After the visit is documented but before the claim is final, review the note for specificity, MEAT support (monitor, evaluate, assess, treat) and missed conditions. Problems caught here are corrected in the record that generates the encounter, so linkage is automatic.
  3. Retrospective review. Still valuable, but redirected toward validation, linked corrections and deletions, compliance auditing and identifying documentation patterns to fix upstream.

Many organizations have historically spent most of their coding budget retrospectively. A reasonable target for 2026 is to move the majority of new HCC capture into prospective and concurrent workflows, with retrospective review increasingly focused on accuracy rather than volume.

Step Three: Fix Linkage at the Data Level

Where real encounters exist, the work is operational. Organizations should confirm that their chart review submissions reference the correct encounter identifiers, that encounter data from delegated providers is complete and timely, and that rejected encounters are resolved quickly. Linkage failures often trace back to provider data that never reached the plan cleanly in the first place. Improving encounter submission from provider partners protects both risk adjustment and quality reporting.

Step Four: Use Provider Queries and Addenda Correctly

When a reviewer finds a condition that appears clinically supported but is incompletely documented, the compliant path is a provider query, followed where appropriate by an addendum to the original note. Done well, this corrects the source record, keeps the diagnosis tied to the encounter and creates an auditable trail. Done poorly, it creates compliance risk.

A sound query program has a few consistent features:

  • Queries are non-leading and present the clinical evidence rather than suggesting a diagnosis.
  • Addenda are signed and dated by the treating provider, within a defined time window after the encounter.
  • Response rates and agreement rates are tracked by provider, and outliers are reviewed.
  • Conditions not supported by the record are deleted, not just left unsubmitted.

Step Five: Rework Incentives and Contracts

Retrospective programs paid on a per-chart or contingency basis were built around the volume of HCCs found. If those HCCs no longer count when unlinked, the incentives in those agreements need revisiting. Similarly, provider incentive programs should reward accurate documentation at the visit, completion of annual assessments for chronic conditions and timely query responses, not year-end retrospective yield.

A Short Readiness Checklist

  • Exposure quantified by contract, provider group and HCC.
  • Prospective pre-visit review in place for high-risk and rising-risk members.
  • Concurrent review running before claims are finalized.
  • Encounter linkage and rejection rates monitored monthly.
  • Query and addendum policy documented and audited.
  • Vendor and provider incentive contracts reviewed against the proposed policy.

Where a Platform Approach Helps

Moving capture upstream is mostly a workflow problem. Curitics Health supports prospective, concurrent and retrospective coding review in one place, with AI chart classification that separates documented from inferred HCCs for coder review, AI summaries, provider queries and signed addenda, and an EHR point-of-care overlay that brings AI chart review into the visit. Patient-level CMS-HCC V28 RAF scoring makes it possible to see the effect of each documented condition as it is captured.

The Bottom Line

Whether CMS finalizes the exclusion exactly as proposed or modifies it, the policy direction is clear: diagnoses should come from documented encounters. Organizations that rebuild around prospective and concurrent capture will be more accurate, more defensible in audits and less exposed to the next change in diagnosis sourcing rules.