CMS-HCC V28 2026: 115 HCCs, Full Non-PACE Use

What changed from V24 to V28

Five changes matter to a coder. The categories were renumbered, so a V24 number tells you nothing about V28. Diabetes, heart failure, and dementia were split into severity tiers. Whole code families lost their payment category (unspecified angina, unspecified peripheral vascular disease, protein-calorie malnutrition, unspecified recurrent depression). Most categories that kept their label got a different coefficient. And the model is larger: more categories, more hierarchies, more interactions. The counts below come straight from the CMS model software and mapping files.

Ten largest coefficient moves where the category label did not change

Community non-dual aged (CNA) coefficients, matched on the exact category label because the number is never the join key. Categories that were split, merged, or relabeled are not in this table; the sections below cover those by code family.

A coefficient is a category reference, not a member total. Source: CMS-HCC V28 PY2026 model software (V28_CE_Relative_Factors.csv) and the PY2025 V24 model files, both registered in the source manifest and regenerated with each CMS release.

The V24/V28 blend schedule (2024 to 2027)

CMS introduced V28 as a three-year blend to smooth revenue impact on MA plans. Each payment year's RAF score is computed twice (once under V24, once under V28) and the results are weighted:

Coders auditing claims for a specific payment year should apply the weighted model in effect for that year. The CMS-HCC V28 PY2026 calculator scores a complete member context, with every line cited to the CMS file. Use the model that matches the payment year in your organization's approved scoring system.

One detail in that table catches teams out every year: the dates of service that drive a payment year run 12 to 24 months earlier than the payment year itself. PY2026 RAF, the first year scored on 100% V28, is built largely on encounters from January through December 2025. Those visits have already happened. V28 documentation work was never something that could wait for 2026 to start, and the same lag applies to every year that follows.

Code families to recheck: valid codes that stopped paying

A code being valid in the ICD-10-CM code set does not mean it maps to a payment HCC. The registered current CY2026 V28 mapping and separate historical PY2025 V24 reference have 7,660 unique codes mapped in both models, 2,267 only in historical V24, and359 only in current V28. The current 2026 Final mapping file does not contain a generic CMS-HCC V24 column. Use the comparison to find code families that need a fresh lookup, not as a substitute for checking the exact code and payment year.

Angina is the clearest example

I20.9 (angina pectoris, unspecified) mapped to V24 HCC 88, Angina Pectoris. It carries no payment HCC under V28. The same holds for I20.1 (with documented spasm), I20.89 (other forms, which is where stable angina lands), and I25.119 (atherosclerotic heart disease with unspecified angina). V24 HCC 88 has no V28 successor.

Do not widen that to the whole family, because the exception is the one worth catching. Unstable angina kept its category: I20.0 and I25.110 both map to V28 HCC 229, Unstable Angina and Other Acute Ischemic Heart Disease, at 0.240, up from V24 HCC 87 at 0.195. So the split inside I20 and I25.11 is acuity: unstable pays, stable and unspecified do not. Two codes that pay in neither model, and are often assumed to: I25.10 (atherosclerotic heart disease without angina) and I10.

Unspecified PVD and claudication got much less forgiving

V24 rewarded a broad vascular bucket. V28 does not. I73.9 mapped to V24 HCC 108, Vascular Disease, and has no V28 payment HCC. Same for I70.213 (atherosclerosis with intermittent claudication, bilateral legs). The family did not disappear, though. Rest pain moves the code into HCC 264, Vascular Disease with Complications (0.455), and ulceration or gangrene reaches HCC 263 (1.118). The difference between I70.213 and I70.233 is documented severity, not the disease label. If the note stays at claudication, there is nothing to capture.

Unspecified depression lost the automatic path

F33.9 (recurrent, unspecified) mapped to V24 HCC 59 and carries no V28 payment HCC. F32.9 (single episode, unspecified) is the harsher case: it maps to nothing under either model, so it was never the safe fallback some worklists treat it as. F33.1 (recurrent, moderate) maps to HCC 155, Major Depression, Moderate or Severe, without Psychosis. F33.3 (severe with psychotic symptoms) maps to HCC 152, Psychosis, Except Schizophrenia. When you look up HCC 152 and see "Psychosis," that is correct; the psychotic features are what move the code there, not the depression severity alone. The chart that loses RAF is recurrent depression with no severity documented.

Malnutrition came off the model entirely

V24 rolled the whole malnutrition family into HCC 21, Protein-Calorie Malnutrition. Under V28 that category has no successor for these codes: E40 through E46, R64 (cachexia), and E88.A all carry no CMS-HCC V28 category. Severity never tiered this HCC in the first place, so there is no "keep the severe one" carve-out to hunt for. Keep coding it when the record supports it; it remains an audit-scrutinized diagnosis and it still drives acuity and quality measures. It just does not move the Part C RAF.

Substance use: the rule is not "moderate or severe only"

This one is widely mis-stated. For alcohol, only dependence maps: F10.20 reaches HCC 139 (0.242) while F10.10 (alcohol abuse, mild) carries no HCC at all. For other drugs, even mild uncomplicated use disorder maps: F11.10 (opioid abuse) reaches HCC 138, Drug Use Disorder, Mild, Uncomplicated, Except Cannabis, at 0.423, which outweighs the alcohol category and sits above it in the hierarchy. Check the substance before you assume the severity threshold.

And the trap that runs the other way

Not every V28 story is a loss. E11.65 (type 2 diabetes with hyperglycemia) started in the V24 chronic-complications tier (HCC 18, the same tier as diabetic CKD and neuropathy) and lands in V28 HCC 38, the no-or-unspecified tier. It dropped a full tier, not just a number. Meanwhile E11.9 went the other way, from 0.105 under V24 to 0.166 under V28. The honest summary is that the diabetes tier itself stopped being where the value sits, since V28 sets HCC 36, 37, and 38 to the same 0.166; the manifestation codes may map to separate HCC categories, with the member contribution still subject to hierarchy, cleanup, interactions, and member context. Before you write off any familiar code, check it against the current mapping file rather than last year's deck. A code that paid under V24 but not V28 is still a valid code, and an auditor does not care what the old training said.

The 2026 HCC list, chronic conditions by category

The 2026 V28 payment HCC list is organized by clinical body system. The categories below are the highest-frequency HCCs across the Medicare Advantage population, the ones most MA plans see on 60%+ of risk-adjustable enrollees. A full category-to-category mapping against specific ICD-10 codes lives on the ICD-10 to HCC mapping hub, and every individual code can be looked up via the HCC Buddy encoder.

The top 10 HCC codes across the MA population and the top 10 miscoded HCCs reports are the quickest way to prioritize chart review capacity for 2026. Both lists are HCC Buddy-maintained and updated against the current V28 coefficient set.

MEAT documentation under V28

For each diagnosis, confirm that the documentation supports the diagnosis and meets the applicable coding, encounter, program and payer requirements. MEAT (Monitor, Evaluate, Assess, or Treat) is a review mnemonic, not a universal CMS coding rule. Document conditions according to the record and applicable rules, regardless of whether they map to an HCC. V28 mapping changes do not determine which conditions deserve documentation. See the MEAT criteria hub for the framework, the provider query templates for AHIMA-compliant prompts, and the RADV audit prep guide for how V28 documentation stands up under audit.

The parallel 2023 RADV final rule removed the fee-for-service adjuster and allowed CMS to extrapolate audit findings across contracts, which means unsupported V28 HCCs now produce larger payment recoveries than unsupported V24 HCCs would have under the prior rule. Compliance teams treating V28 as "just a coefficient change" are misreading the risk.

Practically, the MEAT bar under V28 is higher for the conditions that still pay. A diabetes encounter that previously cleared audit with "DM, on metformin, stable" now needs to name the complication (neuropathy, CKD stage, retinopathy) to land on a V28-mapped code like E11.22 , and the MEAT narrative has to support that specificity. Heart-failure notes that said "CHF, continue Lasix" need to document systolic vs diastolic, acute vs chronic, and the clinical monitoring that justified the encounter. Coders who learned V24-era shortcuts, accepting a "history of" phrase as MEAT, or letting ambiguous laterality ride, will see those charts flagged on the first RADV pull under the 2023 rule. Retraining the chart-review rubric is not optional; it is the difference between keeping and refunding V28 dollars.

The shift also changes what provider-education sessions should emphasize. Under V24, "code to the highest specificity" worked as a generic mantra because almost every chronic condition carried some RAF weight. Under V28, the three diabetes tiers (HCC 36, 37, and 38) all carry the same 0.166 base weight, so moving E11.9 to E11.22 does not, by itself, lift the diabetes coefficient. What specificity unlocks is the separate manifestation HCC: documenting the CKD behind E11.22 lets you also capture the kidney-disease category on top of diabetes, and documenting a foot ulcer (E11.621) captures HCC 383 in its own right. The uncomplicated code is not unmapped; the RAF lift comes from the additional mapped condition the specificity lets you code, not from a higher diabetes tier. Education content should lead with the map, not with the alphabet, and coder-QA dashboards should surface under-specified diabetes and heart-failure encounters as rework candidates, not as pass-through clean claims. Teams that rebuild their provider-facing cheat sheets to the V28 hierarchy code to the categories V28 actually pays; decks still written to V24 categories can steer coders toward HCCs that no longer map under V28, under-capturing where the model now rewards specificity. The same logic applies to retrospective chart review: V24-trained reviewers will flag dozens of "missed" HCCs per chart that no longer map, burning reviewer hours on zero-yield rework. Re-scoping the review queue to V28-mapped categories is a one-time project that pays back in the first quarter.

Related coding tools