HCC / Risk Adjustment - Engine V28

CY2026 is the first year your RAF runs on 100% V28, with zero V24 weight left to cushion it.

The three year phase-in is over. For payment year 2026, CMS is scoring your Medicare Advantage population entirely on the CMS-HCC model Version 28, and the softening blend that carried the old V24 coefficients is gone.

The short answer: most groups do not lose members, they lose documented risk that no longer maps to a payment HCC. Chronic, well managed populations coded to the old V24 logic tend to give back RAF under V28, while complex, actively treated cohorts hold or gain. The delta shows up as a lower average risk score for the same panel, coded the same way. Below is where it moved and what to do about it.

V24 to V28 Blend, Per CMS Rate Announcements

source: CY2024 Rate Announcement three-year transition

CY2024

transition begins

V24 - 67%V28 - 33%

CY2025

majority V28

V24 - 33%V28 - 67%

CY2026

fully phased in

V24 - 0%V28 - 100%

The recalibration, in numbers

What actually changed between the two models

These are structural facts from the CMS-HCC V28 model documentation, not scenario estimates. They explain why the same chart can carry less RAF in 2026 than it did in 2023.

115 HCCs

Payment HCCs in V28, restructured and renumbered from the 86 in V24. More categories, but with reweighted and constrained coefficients.

~2,000 fewer

ICD-10-CM diagnosis codes that no longer map to any payment HCC under V28. Codes that risk-adjusted in V24 now score zero.

100%

Share of the CY2026 risk score set by V28 alone. The V24 blend weight is fully retired per the CY2026 Rate Announcement.

5.9% min

The statutory MA coding pattern adjustment floor under the ACA, applied on top of the V28 recalibration. It compounds the RAF pressure.

Counts reflect CMS-HCC Version 28 model documentation and the CY2024 through CY2026 Rate Announcements. Category and code figures are directional summaries of the published model files.

Where the score moved

Which populations gave back RAF, and which held

Same panel, same documentation habits, two models. The pattern below is what a mixed Medicare Advantage book looks like when you re-score V24-coded members under V28. Read it as direction and relative magnitude, not a promise about your specific membership.

Cohorts that lost score

Diabetes without complicationsRAF down

Protein-calorie malnutrition capturesRAF down

Stable vascular / atherosclerosis dxRAF down

Lower-acuity behavioral codingRAF down

Cohorts that held or firmed

Diabetes with chronic complicationsheld

CKD stage 4, ESRD, dialysis statusheld

CHF with active managementheld

Serious mental illness, treatedheld

Illustrative modeling of a mixed MA panel, not client data. Bar lengths show relative movement between cohorts, not a guaranteed percentage change for any organization. The directions track the published V28 coefficient and mapping changes.

Under the hood

Why the same diagnosis pays less

V28 was not a tweak. CMS rebuilt the condition categories on ICD-10-CM from the ground up, pruned diagnoses it judged discretionary or loosely specified, and reweighted the coefficients. The RxHCC model was recalibrated on the same clinical logic.

V24 model - retired

payment HCCs

Broader ICD-10 mappings, more codes counted
Higher weight on some stable chronic conditions
Carried 33% weight in CY2025, zero in CY2026

recalibrated

V28 model - CY2026 at 100%

115

payment HCCs, reweighted

~2,000 fewer ICD-10-CM codes map to a payment HCC
Constrained coefficients within disease hierarchies
Updated normalization factor and renumbered categories
Paired RxHCC model recalibrated on the same base

Cited authorities

The 2026 rules this is built on

CMS-HCC Model V28

The risk-adjustment model fully phased in for CY2026, replacing V24 for 100% of the payment-year risk score.

CY2026 Advance Notice

CMS proposal confirming the final year of the V28 transition and the effective growth and normalization assumptions.

CY2026 Rate Announcement

The final policy that locks V28 at 100% blend and publishes the applicable normalization factor and coding pattern adjustment.

RxHCC Model

The Part D risk model recalibrated alongside V28, which shifts prescription drug risk scores on the same clinical basis.

Operator playbook

What to actually do before the 2026 sweep closes

Re-score last year's panel on V28, not V24

Run your closed 2024 and 2025 charts back through the V28 mappings. The gap between the two scores is your real exposure, and it tells you which cohorts to prioritize.

Rebuild suspecting logic around the surviving HCCs

Retire suspects tied to the ~2,000 dropped codes. Redirect coder and provider attention to conditions that still risk-adjust and are clinically present but under-documented, especially complications and status codes.

Chase specificity where V28 rewards it

Diabetes with named chronic complications, CKD by stage, and treated severe conditions hold under V28. Loose, unspecified coding is exactly what the model pruned, so specificity is now the difference between a paid HCC and a zero.

Model the RxHCC hit separately

Do not assume Part D moves with Part C. The RxHCC recalibration lands on its own coefficients, and drug risk revenue deserves its own re-score before you set budget.

We built the V28 re-score into our risk-adjustment engine, so you see the RAF delta before it hits the check.

Our HCC engine (V28) runs your historical charts against both models, flags the cohorts that gave back score, and surfaces the surviving, documentable HCCs your coders should be chasing right now. De-identified before anything leaves your environment, tied to the 2026 CMS Rate Announcement math, and reconciled to your own membership. No fabricated lift, just where the risk really sits.

Model your V28 RAF exposure → Talk to our risk-adjustment team