Under CMS rules, a chronic condition only counts in the year it is documented from a face-to-face encounter. It does not roll forward. If diabetes with complications was captured in 2025 and never re-documented in 2026, the risk score drops to zero on that condition. This is the checklist our V28 engine runs on every chart to make sure that does not happen quietly.
Every chronic HCC must be re-documented at least once each calendar year. Our engine takes last year's captured conditions, cross-checks 2026 encounters, and surfaces the ones that are suspected but not yet coded for provider review. It flags the gap. It never invents the diagnosis.
CMS-HCC risk scores are rebuilt from scratch every payment year off the diagnoses documented in the prior calendar year. A condition documented in 2025 does not persist into the 2026 model run. Miss the re-documentation window and the member's risk score, and the care they qualify for, silently deflates.
The risk-adjustment year runs on the calendar. On January 1, 2026, every chronic HCC captured in 2025 dropped back to unproven. It has to be re-documented through a qualifying 2026 encounter to count for the payment year.
V28 is the model doing the counting. The CMS-HCC V28 model, finalized in the CY2024 Rate Announcement, completes its three-year transition at PY2026, when payment is fully weighted to V28.
Source: CMS CY2024 & CY2026 Rate AnnouncementsPY2026 payment is fully weighted to CMS-HCC V28. The V24 blend that softened the transition is gone.
V28 recalibrated the crosswalk and constrained several categories. Codes that mapped to an HCC under V24 may no longer, so last year's capture logic cannot be trusted blind.
The CMS RADV Final Rule (2023) enables extrapolated recovery of unsupported diagnoses. Every recaptured condition needs encounter-level documentation behind it.
These are archetype conditions, not real patients. For each one the engine watches the prior-year capture plus the clinical signals in this year's chart, then hands the provider a prompt to confirm and document with MEAT (Monitor, Evaluate, Assess, Treat). The checkmark is earned only when the 2026 encounter closes the gap.
Active A1c orders, insulin or GLP-1 on the med list, standing ophthalmology or podiatry referrals.
Link the complication, not just "diabetes." The generic code and the complication map to different HCCs under V28.
Loop diuretics, a recent echo or BNP result, cardiology follow-up on the calendar.
Document systolic vs diastolic and acuity. Unspecified heart failure leaves specificity, and score, on the table.
An eGFR trend line in the labs, a nephrology referral, dialysis or transplant status codes.
V28 keys on stage. Stage 1–2 no longer carries the same weight, so the specific stage has to be in the note.
Inhaled bronchodilators or steroids, a prior exacerbation, pulmonary function testing.
Note whether an exacerbation is active. Acuity changes the code and the documentation the auditor expects.
Antidepressants on the med list, a documented PHQ-9, behavioral health referral.
V28 reorganized behavioral health categories. Capture severity and single vs recurrent episode, not "depression, unspecified."
Antiplatelet therapy, a prior revascularization or amputation status, an ABI result.
Tie atherosclerosis to the vessel and any complication. Bare "history of" language does not support the recapture.
This is the line that keeps recapture compliant. Suspect analytics point a clinician at a condition that looks likely from the record. They never write the diagnosis. Confirmation is a clinical act, documented at a face-to-face encounter, every time.
Every chronic HCC documented in 2025 becomes a candidate that must be re-earned in 2026.
Labs, med lists, referrals and problem lists in the 2026 chart, aligned to CMS chart-review and HRA guidance.
Where a signal is strong but no 2026 diagnosis exists yet, the engine raises a suspect prompt for the provider.
The clinician evaluates, confirms or rules out, and documents. Only then does the checkmark turn green.
The guardrail, stated plainly. The engine never fabricates data and never auto-adds an unconfirmed condition. It surfaces a documented signal and hands the decision to the clinician. That is what keeps a recaptured diagnosis defensible under RADV, and it is a hard line we do not cross.
Recapture is a volume problem with a compliance edge. Doing it chart by chart in a spreadsheet leaves score, and defensibility, on the table. Our V28 engine runs the gap-closure checklist across every member, routes suspect prompts to the right clinician, and keeps the documentation trail an auditor will accept. You keep clinical judgment where it belongs. We handle the surfacing, the tracking, and the proof.
Guideline references: CMS-HCC V28 model (CY2024 Rate Announcement); CY2026 Rate Announcement (PY2026 fully weighted to V28); CMS annual diagnosis-recapture requirement (conditions documented each calendar year from a qualifying face-to-face encounter, no carryover); CMS RADV Final Rule (2023); CMS chart-review and Health Risk Assessment guidance. MEAT is a widely used clinical-documentation framework. Condition examples are archetypes and do not represent any real patient. This page is educational and is not coding or legal advice.
Related reading
A RADV audit HCC documentation MEAT playbook: why a diagnosis without Monitor, Evaluate, Assess or Treat evide
Read →InsightNorth Carolina's draft Clinical Coverage Policy 8F rewrites Medicaid autism treatment as RB-BHT. See the dated
Read →BriefingA RADV auditor does not read your intent. They read your chart. Here is how the V28 risk-adjustment engine fla
Read →