The V28 migration playbook for risk-bearing health plans.
CMS-HCC moves to V28 at full weight in payment year 2026. Two thousand three hundred codes are re-weighted. Over two hundred are removed entirely. The average modeled RAF drop sits near 9.3 percent. This is the eight-step playbook to land the year inside the new model.
Executive summaryFive things to know about V28 before 2026.
V28 is the largest CMS-HCC change since the model was introduced. The arithmetic of the change is settled. The operational discipline to absorb it is not. This playbook is a working document for revenue-cycle leaders, coding directors, and medical economics teams who own the V28 transition inside a risk-bearing organization.
The headline arithmetic is hard to argue with. For a Medicare Advantage plan with an average V24 RAF near 1.04, the modeled V28 RAF on the same population is closer to 0.94 to 0.96. On a benchmark of twelve thousand dollars per member per year, that is roughly twelve hundred dollars per member of revenue at risk. Multiplied by panel size, the dollars compound quickly. The playbook is not about avoiding the model. It is about converting documentation discipline into recovered RAF inside the model.
The transition is phased, which matters operationally. Payment year 2024 blends sixty-seven percent V24 and thirty-three percent V28. Payment year 2025 blends thirty-three percent V24 and sixty-seven percent V28. Payment year 2026 runs at one hundred percent V28. Most plans saw a partial signal in 2024, a larger signal in 2025, and the full impact in 2026. Plans that anchored their training, QA, and chart-review programs to V24 specificity will feel a discontinuity. Plans that rebuilt the operating cadence around V28 specificity over the last twenty-four months are absorbing the change inside the model rather than reacting to it after the close.
This paper exists because the V28 conversation, as it shows up in trade press and in vendor decks, has been reduced to two numbers: the count of re-weighted codes and the average modeled RAF drop. Both numbers are real and both are quoted accurately here. Neither number tells an operator what to do on Monday morning. The leaders responsible for landing the year inside V28 need a sequenced operating plan, a condition-level priority list, a documentation rubric tuned to V28 specificity, a coding QA framework that catches the new failure modes, and a way to read RAF trajectory each month instead of waiting for the CMS final reconciliation. That is what the eight steps cover, and that is what the worked example illustrates.
The intended reader is the operator. Chief medical officers, chief financial officers, coding directors, medical economics leads, and ACO REACH program directors will recognize the cadence. The paper assumes familiarity with CMS-HCC at the working level, including the difference between prospective risk adjustment, retrospective chart chase, RADV defensibility, and the coding intensity adjustment that sits between submitted RAF and paid RAF. Where a concept needs a one-line refresher, the glossary at the back of the paper carries it. Where a concept needs full treatment, the body of the paper expands it inline. The goal is a working document, not a primer, and not a sales artifact.
The landscapeWhere the weight actually moved.
A condition-by-condition heatmap of V24 to V28 weight movement for the highest-prevalence categories inside a typical Medicare Advantage panel. Green means the category gained weight or held steady. Red means the category lost weight. Intensity scales with the size of the move on a representative population.
Two patterns matter most. First, the diabetes split. V24 paid the same weight for diabetes regardless of which specified complication was documented. V28 distinguishes the categories and assigns materially different weights. Without specificity in the chart, the V28 capture collapses to the lowest-weight category. Second, the CKD acceleration. CKD stage 4 and CKD stage 5 carry meaningfully higher weight in V28 than V24, and they reward the clinical discipline of staging clearly in the note. CKD is a category where V28 rewards documentation that was already best practice.
The chronic conditions that lose weight are the categories where the playbook focuses operational attention. Major depression, morbid obesity, and the diabetes-with-complication category all lose meaningful weight. The right response is not to chase the lost weight inside V24 thinking. The right response is to re-anchor the documentation guidance around V28 specificity so that the conditions that still carry weight are captured at full weight, every encounter, without exception.
The regulatory context is worth restating in plain terms because it is often muddled in operator conversations. CMS finalized the V28 model in the Calendar Year 2024 Rate Announcement and confirmed the three-year phase-in at the same time. Payment year 2024 ran on a blend of two-thirds V24 plus one-third V28. Payment year 2025 inverted the blend to one-third V24 plus two-thirds V28. Payment year 2026 runs on pure V28. The Medicare Advantage coding intensity adjustment of 5.9 percent under section 1853 sub-paragraph (c)(1)(C)(ii)(IV) continues to apply on top of the V28 calculation, so the submitted RAF and the paid RAF are not the same number. ACO REACH participants run on a separate normalization pathway with a REACH-specific coding intensity adjustment and a discount factor that is not symmetric to the MA adjustment. Confusing the two pathways is a common source of forecasting error.
The most common operational mistake in the field is to treat V28 as a coding problem. Coding teams cannot code what the documentation does not support, and the V28 model rewards specificity that V24 did not require. A note that read "diabetes with complications" carried a single HCC under V24. The same note under V28 collapses to the unspecified diabetes bucket because the model needs the specific complication named. The fix is upstream of the coder. It is in the encounter, in the problem list, in the assessment, and in the plan. The second common mistake is to confuse RAF drift with RAF drop. Drift is the year-over-year decay that happens when chronic conditions are not recaptured each calendar year. Drop is the model change. The two compound. A plan that does not recapture is losing seven to ten percent each year on the chronic base, on top of the model change. The third common mistake is to chase one-off chart retrieval campaigns instead of building the prospective discipline. Retrospective chases lift the current year. Prospective discipline lifts every year.
Top 10 moversThe conditions that drive the drop.
Ranked by aggregate impact on a representative Medicare Advantage panel, the top ten condition categories explain roughly seventy-eight percent of the modeled V28 RAF drop. Operational focus on the top three categories alone is enough to recover roughly half of the drop with documentation discipline.
The diabetes category alone explains roughly thirty-five percent of the modeled drop on a typical panel. The major depression category explains another twenty percent. Together with morbid obesity and substance use disorder, four condition families account for the bulk of the operational target. The playbook concentrates the documentation training, the prospective chart review, and the QA rubric on these categories first, before broadening to the long tail.
Inside the diabetes family, the consolidation is the operational story. V24 mapped most diabetes-with-complication codes into a single HCC that paid one weight regardless of which complication was named. V28 splits the family into a hierarchy where the specified vascular complication, the specified neurological complication, and the specified renal complication each carry distinct weights, and where unspecified diabetes collapses to a noticeably lower category. An encounter coded as E11.22 (type 2 diabetes mellitus with diabetic chronic kidney disease) now anchors both the diabetes line and the CKD line in the appropriate stage, and the linkage matters for the renal complication weight to attach. An encounter coded as E11.9 (type 2 diabetes mellitus without complications) anchors neither, and the chart loses everything the patient actually carries.
Mental health and substance use disorders show the second consolidation. Several unspecified major depression codes and several unspecified substance use codes were removed from V28 HCC mapping outright. The codes that remain require the severity, the episode (single versus recurrent), and the remission status to be documented. A note that reads "depression" carries nothing under V28. A note that reads "major depressive disorder, recurrent, severe, without psychotic features" carries HCC-155 and the matching weight. The substance use family follows the same pattern, with remission status as the pivot. The documentation discipline is the same on both: the diagnosis needs the qualifier the model is asking for, every encounter, in the assessment and plan, not in the problem list alone.
Cardiovascular and renal categories tell the other side of the story. CHF gained weight in V28, with the specified systolic and diastolic categories carrying meaningfully more than the unspecified category. A note that reads "CHF" loses to a note that reads "chronic systolic congestive heart failure, NYHA class III" anchored to I50.32 or its peer codes. CKD stages 4 and 5 carry the largest single category gain on most panels. The clinical staging discipline that nephrology has always followed is now financially reinforced. Specified arrhythmias, specified polyneuropathy, and specified vascular disease all reward the same upgrade from generic to specific. The pattern across the entire delta list is consistent: V28 rewards specificity and penalizes generality.
Worked example5,000-life MA plan. Real arithmetic.
An anonymized Medicare Advantage panel of five thousand members. The pre-program RAF on V24 sat at 1.04. The modeled V28 RAF on the same panel landed at 0.94, a 9.6 percent drop. The playbook recovered roughly half of the drop in twelve months. The dashboard below shows the condition-level capture rate the playbook drove.
The panel was a mature MA HMO book in a regional plan with a payer mix tilted toward dual-eligible and chronic care management members. Average age was 74. Diabetes prevalence ran at 41 percent of the panel. CHF prevalence ran at 17 percent. CKD stage 3 or higher ran at 12 percent. Major depression ran at 14 percent. The pre-program documentation footprint was typical: the problem list carried the conditions, the assessment carried two or three of the top three, and the plan rarely re-stated the chronic conditions that were not the focus of the visit. The coding team was experienced and the audit agreement rate was high. The gap was upstream.
The recommendation was sequenced. First, baseline V24 and modeled V28 on the same twelve months of claims. Second, rank conditions by panel-level RAF dollar impact. Third, build a one-page documentation cheat sheet for each of the top fifteen V28 categories with the specificity language the model rewards. Fourth, rebuild the coding QA rubric around V28 categories with a structured kick-back rule when a chart supports a higher-specificity category than the code submitted. Fifth, schedule prospective gap-closure visits for panel members whose prior-year HCCs had not refreshed in the current calendar year, with the visits structured to address the chronic problem list in full. Sixth, update the EHR HCC mapping logic and the operational dashboard so the team could see V28 capture by category in real time rather than waiting for the CMS reconciliation.
Projected outcome at twelve months was a V28 RAF lift from 0.94 to between 0.98 and 1.00, recovering roughly half of the modeled drop. Actual landed RAF was 0.99, inside the modeled band. Recovered premium at the twelve-thousand-dollar PMPY benchmark was approximately three million dollars at the panel level. The largest single contribution was the diabetes recapture, which alone accounted for nearly forty percent of the lift. The CHF and CKD lifts were second and third largest. The major depression and morbid obesity categories continued to lose weight inside V28 even after the recapture, which was expected and modeled. The work was discipline, not magic.
RAF and dollars before vs after the playbook.
V28 is not a coding problem. It is a documentation discipline problem with a coding consequence. The plans that landed inside the new model are the ones that rebuilt the documentation guide around V28 specificity twelve months before the close. The rest are still trying to recover lost ground.
Implementation checklistLand the year inside V28.
The checklist is sequenced so the first six items deliver the bulk of the recovery. The remaining items lock in the discipline and feed the next year's program.
A few operational notes ride alongside the checklist. The baseline run in item one is best done on the same claim extract twice, once with the V24 mapping and once with the V28 mapping, so the comparison is apples to apples. The kick-back rule in item four needs a defined turnaround clock and a paired training loop so the same gap does not recur. Prospective gap closure in item five works only when the visit type is structured to address the chronic problem list, not just the chief complaint, and only when the visit is scheduled before the close of the calendar year so the recapture lands in the right RAF year. The dashboard work in item six should use a single source of truth for HCC mapping, ideally inside the EHR rather than in a downstream spreadsheet, so coders and clinicians see the same picture. Common pitfalls include suspect-list governance that lacks a clinical validator, retrospective chart chase campaigns that crowd out the prospective work, and RADV exposure from suspect-driven adds that lack clean MEAT support. Escalation criteria are simple: any condition with a panel-level RAF dollar impact above fifty thousand dollars that drifts more than two percentage points below target in a given month gets escalated to the medical director and the coding lead together, not separately.
Capability stackThe HCC Coding AI under the playbook.
The playbook is the operating cadence. The HCC Coding AI is the assist. It reads the chart, returns a deterministic V28 condition list, cites the supporting documentation, and flags the specificity gap when the note will not support the highest-weight category. The coder still owns the code. The AI runs against every chart so no specificity gap reaches submission unseen.
The MEAT framework anchors the documentation review. Every HCC needs evidence in the encounter for the date of service that the condition was Monitored, Evaluated, Assessed, or Treated. The AI surfaces the MEAT element it found and flags any condition where the encounter carries the diagnosis on the problem list but lacks the supporting evidence in the body of the note. That gap is the most common driver of RADV downgrades, and it is also the most common driver of submitted RAF that does not survive audit. The audit defense work begins at the encounter, not at the chart retrieval stage.
RADV defensibility is the second discipline layered on top. CMS allows roughly 25 weeks for medical record retrieval in a RADV cycle, and the best-one record selection rules under 42 CFR 422.310(e) mean that the strongest supporting record per HCC is the one that defends the diagnosis. The playbook builds the chart selection logic into the same pipeline that does the prospective gap work, so the record that supports each submitted HCC is identified and indexed at submission time, not reconstructed two years later under audit pressure. The attestation rules require a credentialed provider signature and a date of service that aligns with the diagnosis. Misalignment on either dimension is a downgrade waiting to happen.
ACO REACH participants run on a parallel but distinct calculation. The REACH model uses normalized CMS-HCC with a REACH-specific coding intensity adjustment and a coding intensity ratio (CIR) that benchmarks the participant against its peer cohort. A participant that codes more intensely than the cohort sees a CIR adjustment that pulls the submitted RAF back toward the peer average. The MA Advantage 5.9 percent coding intensity adjustment is symmetric across plans. The REACH CIR is not symmetric and can move year over year. Forecasting that confuses the two pathways underestimates the variance the REACH participant will see at reconciliation.
GlossaryThe vocabulary of V28.
Common questionsFrequently asked: V28 migration.
What is V28 and how does it differ from V24?
Who is affected by the V28 transition?
Why do diabetes complications matter so much in V28?
Is the V28 transition phased or a cliff?
What is the 8-step migration timeline?
How does the HCC Coding AI help?
What does the worked example show?
Does ASP-RCM replace our coders?
Want this playbook applied to your panel?
Send twelve months of de-identified claims. We run V24 and V28 on the same population, deliver a written condition-by-condition delta, and a recovery estimate in dollars. Yours to keep.