CMS-HCC V28 Transition Guide: What Changed, What It Means for Your RAF

The CMS-HCC V28 risk adjustment model is the most significant change to Medicare Advantage risk scoring since the V21 model was introduced in 2014. CMS finalized V28 in the 2024 Advance Notice and is phasing it in over three payment years, with the authoritative software and mapping files published on the CMS risk-adjustment model page. This guide explains what actually changed between V24 and V28, the phase-in schedule, the disease families that lost the most RAF, and the documentation areas where coders need to focus to protect risk scores. For a plain-English industry summary of the same material, see AAPC's CMS-HCC model V28 overview.

The Phase-In Schedule

CMS is blending V24 and V28 over three payment years. For payment year (PY) 2024, plans are paid 67% V24 / 33% V28. For PY 2025, the blend shifts to 33% V24 / 67% V28. By PY 2026, plans are paid 100% on V28. This staggered transition was intended to give plans time to adjust documentation and coding workflows before the full revenue impact lands.

It is important to remember that the dates of service that drive each payment year are 12 to 24 months earlier than the payment year itself. PY 2026 RAF, which is 100% V28, is built primarily on dates of service from January 1 through December 31 of 2025. This is why V28 documentation work cannot wait until 2026, the encounters that drive 2026 revenue have already happened in many cases.

What Actually Changed

V28 is not just a re-weighting of the existing model. CMS made structural changes that affect both which conditions count and how they roll up. The four most consequential changes:

1. The HCC numbering system was rebuilt. The HCCs in V28 use a new numbering system that does not match V24. HCC 19 (Diabetes without Complication) in V24 is not the same as HCC 19 in V28. Coders who memorized V24 numbering need to relearn the mappings, a chart that says "this is an HCC 18" tells you nothing about V28.

2. The total number of HCCs grew from 86 to 115. Despite the larger count, the model is more selective. CMS removed some condition categories entirely and split others into more specific groups, with the explicit goal of reducing what they describe as "discretionary coding", diagnoses where small documentation choices have outsized payment impact.

3. The number of ICD-10-CM codes that map to any HCC dropped significantly. CMS removed roughly 2,200 ICD-10-CM codes from HCC mapping in V28 compared to V24, as laid out in the CMS ICD-10-CM code set reference. The removed codes are concentrated in five disease areas that we will cover in detail below.

4. RAF coefficients were recalibrated using more recent data. CMS rebuilt the model coefficients on FFS Medicare data from 2018 to give a more current view of the relative cost of each condition. Some conditions (CHF, vascular disease) saw their coefficients increase. Others (diabetes with complications, several depression codes) saw their coefficients drop, and a number of lower-acuity conditions were removed from HCC mapping entirely.

The Five Disease Families That Lost the Most RAF

If you only have time to study a handful of areas, focus on these five. They account for the majority of the V24-to-V28 RAF gap that most plans are seeing.

1. The Diabetes Complication Differential Collapsed

In V24, uncomplicated type 2 diabetes (E11.9) mapped to HCC 19 (Diabetes without Complication) with a community non-dual aged (CNA) RAF weight of 0.105. In V28, E11.9 maps to HCC 38 (Diabetes with No, Glycemic, or Unspecified Complications) at a CNA weight of 0.166. The practical effect is less dramatic than it first appears: CMS constrained HCC 37 (diabetes with chronic complications) and HCC 38 to the same 0.166 CNA weight in V28, so the RAF contribution is identical regardless of complication status. The coding imperative is clinical accuracy and RADV defensibility, not a weight difference. That said, capturing the correct complication code still matters because hierarchy rules require the highest-supported HCC, and audit reviewers expect documentation to match the coded specificity.

The real revenue change is the loss of the complication premium. Under V24, a chronic-complication code (HCC 18, 0.302) paid nearly 0.20 RAF more than uncomplicated diabetes (HCC 19, 0.105); under V28 that gap is gone, HCC 37 and HCC 38 both pay 0.166. For a Medicare Advantage population where 25–30% of members have diabetes, losing that differential is a meaningful revenue shift, and it is recovered not from the base diabetes code but by capturing the separate manifestation HCCs (diabetic CKD, retinopathy, foot ulcer) that stack on top of the diabetes category.

The fix is not "code every diabetic as complicated." The fix is provider documentation. If the patient has CKD and diabetes, the provider must explicitly link them: "diabetes mellitus type 2 with diabetic chronic kidney disease, stage 3a." A note that lists "diabetes" and "CKD stage 3a" as separate problems in the assessment does not establish the causal link required for the combination code. The combination requires the provider to document that the kidney disease is a complication of the diabetes.

2. Vascular Disease

V24 had a single broad HCC 108 (Vascular Disease) that captured codes like I70.0 (Atherosclerosis of aorta), I70.90 (Unspecified atherosclerosis), and many peripheral vascular disease codes. V28 split vascular disease into more specific HCCs and removed the unspecified atherosclerosis code from the mapping entirely.

Coders who relied on I70.90 in V24 to capture vascular disease without strong documentation will lose that HCC under V28. Intermittent-claudication codes (I70.21x) also carry no payment HCC in V28. To capture a vascular HCC, the documentation has to support a complication: atherosclerosis of the extremities with rest pain (I70.221, I70.222, and so on) maps to HCC 264, and with ulceration or gangrene (I70.23x through I70.26x) maps to the higher-weighted HCC 263. If the chart does not specify the vessel and that level of complication, the HCC is gone.

3. Major Depression

V24 had HCC 59 (Major Depressive, Bipolar, and Paranoid Disorders) that captured a broad range of mood disorder codes, including F33.9 (Major depressive disorder, recurrent, unspecified). V28 narrowed the mapping: neither F33.9 nor F32.9 (Major depressive disorder, single episode, unspecified) carries an HCC in V28. To capture depression, the provider must document severity: mild, moderate, severe without psychotic features, severe with psychotic features, or in partial / full remission.

The pattern that loses RAF is the recurrent depression chart with no severity documented. In V24, that chart could be coded F33.9 and capture HCC 59. In V28, that same chart captures nothing; severity (moderate, severe, with or without psychosis) is now required for any depression HCC. Provider documentation training in this area produces some of the largest measurable RAF recovery.

4. Protein-Calorie Malnutrition

V24 had a Protein-Calorie Malnutrition HCC that included a range of malnutrition codes. V28 removed malnutrition from payment mapping entirely. None of the malnutrition codes (E40 through E46), including severe forms such as E43, map to a payment HCC under V28.

This is a significant change for SNF and post-acute populations where malnutrition is commonly documented. Continue to document and code malnutrition accurately for clinical and quality purposes, but do not count on it for risk adjustment under V28. Verify the current mapping for any code in the CMS 2026 model software.

5. Drug and Alcohol Use Disorders

V24 had HCC 55 (Substance Use Disorder, Moderate/Severe, or Substance Use with Complications) that captured a relatively broad set of F10–F19 codes. V28 narrowed the mapping to require explicit moderate or severe documentation, and many of the "in remission" codes were removed from HCC mapping entirely.

For chronic-care populations where prior substance use is common but not currently active, this means the provider must document either current moderate/severe use or active dependence, not just history. The "in remission" codes that captured HCC value under V24 are gone in V28.

Documentation Patterns That Protect RAF

The plans that are losing the least RAF in the V24-to-V28 transition share three documentation habits:

They train providers on linking statements. "Diabetes with CKD" is not the same as "diabetes mellitus type 2 with diabetic chronic kidney disease." The first is two separate diagnoses; the second is a combination code that drives an HCC. Training providers to write the linking phrase is the single highest-RAF-recovery intervention available under V28.

They build CDI workflows around severity. Depression severity, malnutrition severity, substance use severity, CHF type and acuity, V28 punishes unspecified codes harder than V24 did. CDI specialists who query providers for severity on these conditions recover meaningful RAF.

They use V28 mappings during prospective review, not just retrospective audit. Prospective coding under the V24 mindset captures a different set of HCCs than V28 requires. Coders running prospective chart reviews need to be looking at V28 mappings, not the mappings they learned three years ago.

What Comes Next

CMS has signaled that V28 is not the end of the road. The agency has discussed further model recalibration in future Advance Notices, and there is ongoing pressure from MedPAC and from Congress to continue reducing what is described as "coding intensity" in Medicare Advantage -- a position MedPAC made explicit in its February 2026 comment letter on the CY2027 Advance Notice. The OIG is moving in parallel: the OIG work plan tracking trends across the V24 and V28 CMS-HCC models signals continued audit focus on RAF drift between the two generations. Plans should expect that the patterns established by V28 (less reward for unspecified codes, more reward for genuine clinical specificity) will continue or accelerate in future model updates, most of which land first in the Federal Register as proposed and final rules.

The practical implication is that documentation improvement is no longer a one-time project. Plans that want to maintain stable RAF need provider documentation programs as part of their standard operating model, not as a temporary response to a model change.