What changed between V24 and V28
V24 and V28 are two versions of the same CMS-HCC model, the formula that converts a member's documented diagnoses into the risk score that sets Medicare Advantage payment. V28 is not a small revision of V24. The Centers for Medicare and Medicaid Services (CMS) rebuilt the category structure, cut the set of diagnosis codes that map to a payment condition, and recalculated the coefficients on newer claims data.
The table below shows the headline differences between the two models.
| Attribute | V24 (2020 model) | V28 (2024 model) |
|---|---|---|
| Model year | 2020 | 2024 |
| Payment HCC categories | 86 | 115 |
| Mapped diagnosis codes | 9,797 | 7,770 |
| Timing / phase-in | Sole model through PY2023; blended down from PY2024 | Phased in PY2024 to PY2026; 100% in PY2026 |
| Net RAF effect | Reference baseline | Generally lower RAF for the same population |
The renumbering matters in practice. V28 raised the count of payment HCC categories from 86 to 115 and reassigned the numbers, so an HCC number under V24 does not point to the same condition under V28. Crosswalks built for the old model give wrong answers against the new one, and any internal report that still keys off V24 category numbers needs to be rebuilt before it can be trusted.
The coefficients changed too. CMS refit the model weights on a more recent set of Medicare Advantage claims, so even conditions that survived the transition can carry a different value than they did under V24. Two plans with identical documentation can see different RAF movement depending on which conditions dominate their panel, because the cuts and the new weights land unevenly across disease groups.
The phase-in timeline
CMS did not switch models in a single year. It blended V24 and V28 across three payment years so plans had time to adjust their documentation and coding. Each year the V28 share grew and the V24 share shrank, and for payment year 2026 the transition is complete.
- Payment year 2024: 67% V24 plus 33% V28.
- Payment year 2025: 33% V24 plus 67% V28.
- Payment year 2026: 100% V28. Risk scores are now calculated entirely on the new model.
The blend is why a plan could watch RAF erode gradually rather than drop in one step. A member's score was a weighted mix of the two models, and the weight shifted toward V28 every year. For payment year 2026 there is no V24 cushion left, so any condition that V28 stopped recognizing now contributes nothing to payment.
The phase-in also hid the size of the change from plans that read only the blended number. A panel that looked stable in 2024 and 2025 can see a sharper correction in 2026 when the V24 weight finally goes to zero. Plans that modeled the full-V28 score early had time to fix documentation; plans that watched the blend got the news late.
What V28 removed and why
V28 reduced the diagnosis codes that map to a payment HCC from 9,797 under V24 to 7,770 under V28. That is roughly 2,236 codes removed and 209 added. CMS constrained or dropped several high-frequency, lower-cost categories, including many diabetes-without-complication codes and certain vascular conditions, and rebuilt the coefficients on more recent data.
The stated aim was a model that predicts cost more accurately and responds less to coding intensity. The HHS Office of Inspector General (OIG) has tracked the differences between the 2020 V24 model and the 2024 V28 model as part of its risk-adjustment oversight work. For most plans the practical effect is downward pressure on RAF: conditions that used to earn a category no longer do, so the same patient documented the same way scores lower than under V24.
The model change does not alter the clinical truth of the patient. What it alters is which documented, codable conditions count toward payment. A member with stable diabetes is exactly as sick as before, but the code that captured it may no longer map. That shift makes accurate annual recapture and specific coding worth more under V28, because there are fewer codes left carrying the score.
Specificity is where most of the lost value hides. Under V24 a less specific code often still landed in a payment category, so documentation that stopped at a general diagnosis was usually enough. V28 expects the coding to reflect the actual severity and complications, and the gap between a general code and the specific one is now the gap between a category that pays and one that does not. The clinical encounter has to capture that detail, and the chart has to carry it through to the claim.
What it means operationally
Under V28 the margin sits in complete, specific documentation of the conditions that still map, recaptured every calendar year. Every chronic condition has to be re-documented each year to count, and codes that used to earn a category now need the higher specificity V28 expects. Missing a recapture costs more than it did under V24 because fewer codes carry the risk score.
This is operations work, not analytics. A gap list tells you which members have a suspected or lapsed condition that still maps under V28. Closing the gap means getting the member in front of a provider, getting the condition documented to the right specificity, and confirming it landed in the submission against the 277CA and MOR. The list is the easy part. The follow-up across providers and payer portals is where RAF is won or lost.
The work compounds across the year. Suspected conditions have to be surfaced before the visit, documented during it, and verified afterward, and a single member can carry several conditions that each need their own recapture. A small team owning a large panel cannot do that by working a spreadsheet. The volume of members times conditions times follow-ups is larger than manual outreach can cover, which is why most of the lost RAF under V28 is documentation that was clinically true but never closed.
For the year-one workflow, see the V28 readiness guide. For the recapture and specificity playbook, see how to improve a RAF score under V28.
How Pelica handles the move to V28
Most platforms show you what needs to happen. Pelica actually does it. Pelica is the AI-native execution layer for value-based care: one live member record with full ICD-to-HCC mapping across V24 and V28, and a Risk Adjustment copilot that works the recapture queue instead of just ranking it.
The copilot prioritizes the conditions that still map under V28, prepares the documentation a provider needs, and follows up across charts and payer portals until each gap is closed and confirmed in the submission. Because the mapping is current for both models, the team is never working a V24 crosswalk against a V28 payment year, and a recapture is tracked through the 277CA and MOR so a closed gap is one that actually paid, not one that was merely worked.
That cuts risk-adjustment coordination from 30 minutes to 3 minutes per member and lets the same team cover two to three times more members. At HealthCare Partners, the largest IPA in the country, that approach added 0.4 to RAF in two quarters with no new headcount, across more than 175,000 patients managed live.
The execution frame is the point. Knowing which conditions V28 cut is a reporting question that any analytics tool can answer. Recapturing the conditions that remain, to the specificity V28 requires, across a full panel before the submission deadline, is operations. That is the part Pelica owns.
Related terms
V28 is the current CMS-HCC model. See V28 for the model itself, HCC for how condition categories work, and RAF for the risk score the model produces.