Last updated: March 2026 | By Wynda Clayton, MS, RHIT, CRC, Director of Risk Adjustment Coding & Compliance, RAAPID
Quick Answer: The CMS-HCC Model V28 is fully operative as of January 1, 2026. It replaces version 24 entirely, reduces valid ICD-10-CM diagnostic codes from 9,797 to 7,770, expands hierarchical condition categories (HCCs) from 86 to 115, and is projected to reduce average Medicare Advantage risk scores by 3.12 percent. Health plans and providers that have not updated their coding practices and documentation workflows face real payment shortfalls and audit exposure, in an enforcement environment that has never been more active.
Key Takeaways
- The CMS-HCC Model V28 is fully operative as of payment year 2026.
- Valid diagnostic codes dropped by over 2,000: conditions that drove risk adjustment factor scores under V24 may no longer count.
- Severity of illness now matters more than the number of diagnoses.
- Documentation specificity is the core of defensible risk adjustment under V28.
- Healthcare organizations running add-only retrospective programs face payment reductions and regulatory scrutiny backed by active DOJ enforcement.
- The CMS 2027 Advance Notice continues tightening the model, with 65 HCCs decreasing in coefficient values.
- AI-powered platforms with explainable, MEAT-based evidence trails are the most defensible path forward.
What Is the CMS-HCC Model V28?
The CMS-HCC Model V28 (Hierarchical Condition Category risk adjustment model, version 28) is CMS’s updated methodology for calculating risk adjustment factor (RAF) scores for Medicare Advantage beneficiaries, calibrated on more recent fee-for-service utilization, cost, and diagnostic data.
V28 replaces V24, the risk adjustment model that governed Medicare Advantage for the prior decade. Where V24 mapped patient diagnoses to 86 HCC categories using 9,797 ICD-10-CM codes, V28 uses 115 categories but applies only 7,770 diagnostic codes — a net reduction of more than 2,000 valid codes. [1] That difference changes which conditions generate risk scores, how much those scores are worth, and which coding practices hold up under audit.
The transition reflects CMS’s stated goal of aligning the HCC model more accurately with actual Medicare FFS spending patterns using more current data. The proposed 2027 model update is calibrated on 2023 diagnosis dates and 2024 expenditures, compared to the 2018 and 2019 data that underpinned the 2024 CMS-HCC model. [2] CMS’s signal is consistent: pay for improved patient care, not improved coding volume.
Autonomous Retrospective Risk Adjustment Solution
One platform. Every HCC validated. Revenue secured.
Why Does V28 Matter for Medicare Advantage Plans Right Now?
As of January 1, 2026, the CMS-HCC Model V28 governs 100 percent of Medicare Advantage risk score calculations. The phased implementation that started in payment year 2024 is complete: 33 percent V28 in PY 2024, 67 percent V28 in PY 2025, 100 percent V28 in PY 2026. [1]
CMS projected the V28 transition would reduce average Medicare Advantage risk scores by 3.12 percent, representing approximately $11 billion in net savings to the Medicare Trust Fund. [1] The CMS 2027 Advance Notice tightens this further: for every 10,000 patients in a Medicare Advantage risk plan, the 2027 proposed rule would cut approximately $1.2 million in risk score care funding, with coefficient weight shifts for the top ten chronic conditions accounting for a $2.3 million decrease in care funding per 10,000 patients. [2]
The enforcement context compounds this pressure. CMS’s January 2026 RADV memorandum confirmed that the agency has resumed active audits covering payment years 2018 through 2024, estimates approximately $17 billion in annual MA overpayments from unsupported diagnosis data, and is committed to completing outstanding audits on an expedited schedule. [4] The DOJ settled with Aetna/CVS for $117.7 million in March 2026, specifically because Aetna’s PY 2015 retrospective chart review program submitted additional diagnosis codes but failed to delete unsupported codes the same reviews identified. [3]
V28’s specificity requirements make that risk concrete. Codes that previously generated risk scores may now be invalid, and if plans continue to submit them, they aren’t just missing a revenue adjustment: they may be submitting unsupported diagnoses into an active audit environment.
The Essential Guide to CMS-HCC Model V28
Access the Full List of Chronic Conditions & Key Updates Now
What Changed from CMS-HCC V24 to V28?
Understanding the V24 to V28 shift requires looking at four areas: code volume, HCC category expansion, the constraining process, and severity weighting.
Code Volume and HCC Categories
The total number of valid ICD-10-CM diagnostic codes dropped from 9,797 in V24 to 7,770 in V28, a net reduction of 2,027 codes, reflecting 2,236 codes removed and 209 added. [1] HCC categories expanded from 86 to 115. The net effect: the risk adjustment model covers more clinical conditions at a higher level of specificity but excludes a large portion of codes that previously contributed to risk scores. Healthcare organizations running on V24-era coding logic are almost certainly submitting some diagnostic codes that no longer count.
The Constraining Process
V28 introduces constraining, where related HCCs are assigned the same coefficient values. This prevents double-counting clinically related conditions. Under V24, a patient with diabetes and diabetic peripheral neuropathy received separate, additive HCC contributions. Under V28, both conditions map to the same coefficient. [1] The resulting RAF score for that patient is lower, with no change in their health status.
Severity Over Volume
V28 explicitly prioritizes severity of illness over the number of diagnoses. A patient with two severe conditions may generate a higher risk score than a patient with five moderate ones. This is a structural reversal of the incentive that drove add-only retrospective chart review programs.
Specific Category Changes
Depression coding changed significantly: codes for mild, unspecified, or in-remission depression are no longer valid. Only moderate and severe major depression generate HCC risk scores. Morbid obesity coding came under direct enforcement scrutiny in the Aetna settlement: $11.5 million of the $117.7 million resolved specifically alleged false morbid obesity diagnoses for PY 2018 through 2023. [3] Protein-calorie malnutrition is no longer a valid HCC code under V28.
How Does V28 Impact Risk Adjustment Factor Scores?
RAF scores are lower under V28 for most Medicare Advantage patients. CMS projected a 3.12 percent average decline. [1] For members with multiple related chronic conditions, the actual decline can be steeper. The shift in coefficient weights for the top ten chronic conditions alone accounts for a $2.3 million decrease in care funding per 10,000 patients under the proposed 2027 model. [2]
Expert Perspective
“V28 isn’t a technical update to a model. It’s CMS signaling that they want documentation to reflect actual patient complexity, not a list of codes. If you can’t link a diagnosis to a current encounter with MEAT-based evidence — Management, Evaluation, Assessment, and Treatment — you shouldn’t be submitting it.”
— Wynda Clayton, MS, RHIT, CRC, Director of Risk Adjustment Coding & Compliance, RAAPID | Former CMS RADV Auditor
CMS 2027 Advance Notice: Proposed Coefficient Changes
Condition | Change (Proposed 2027) |
CKD Stage 3A | -50.4% |
COPD | -18.8% |
Morbid Obesity | -19.4% |
Rheumatoid Arthritis | -17.3% |
Major Depression | -13.7% |
Heart Failure | -10.8% |
Diabetes | -6.6% |
Drug Use Disorder (moderate/severe) | -24.3% |
Septicemia/Sepsis/SIRS/Shock (HCC 2) | +15.6% |
Bladder/Colorectal/Other Cancers (HCC 22) | +12.1% |
Source: CMS 2027 Advance Notice, February 2026 [2]
What Does the OIG Audit Series Tell Us About High-Risk Diagnoses?
The OIG’s March 2026 compliance audit (Report A-07-22-01207) is one of the most instructive data points available on what V28-era audits actually find in practice. OIG audited nine high-risk diagnosis categories, reviewing 271 sampled enrollee-years.
Source: OIG Report A-07-22-01207, March 2026 [5]
The most common error pattern across all nine categories was history-of conditions coded as active diagnoses: a past stroke coded as acute stroke, resolved cancer coded as active malignancy, prior embolism coded as current vascular disease. This is the documentation distinction that V28 makes critical.
The CMS 2027 Advance Notice proposes higher coefficient weights for several of these same conditions, including sepsis and certain cancers, precisely because they represent acute, high-cost events warranting increased reimbursement when genuinely documented. [2] But as the OIG audit confirms, they are also the categories most likely to attract RADV scrutiny when documentation does not hold up.
What Is the Current Status of RADV Audits?
The January 2026 CMS memorandum from the Audits and Vulnerabilities Group resolved any ambiguity about whether RADV audits would resume: they have, on an accelerated schedule. [4]
CMS confirmed that despite federal estimates suggesting approximately $17 billion in annual overpayments from unsupported diagnosis data, the last significant recovery of MA overpayments under the RADV program occurred for the PY 2007 audit. Audits for PYs 2011 through 2013 found overpayment rates between five and eight percent, and recoveries on those amounts will begin soon.
Key RADV Facts (CMS, January 2026)
- Audits cover PYs 2018 through 2024 on an expedited timeframe
- Variable sample sizes: 35 to 200 enrollees per contract depending on size
- Medical record submission window: five months
- New audits initiated approximately every three months
- Maximum two medical records per audited HCC
- All overpayment coding decisions made by human certified medical coders
- AI used as coder support tool only, not decision-maker
With audits now covering every eligible contract across seven consecutive payment years, the question is not whether a plan will be audited but how its documentation holds up when it is.
What Is MEAT Criteria and Why Does It Matter Under V28?
MEAT criteria (Management, Evaluation, Assessment, and Treatment) is the documentation standard CMS uses to validate HCC diagnoses during RADV audits. A diagnosis must be supported by at least one of these four elements in the medical record, tied to a current face-to-face encounter, to be considered valid.
Under V24, the relatively broader code set meant some diagnoses could persist in annual submissions without airtight MEAT-based evidence. V28 closes that window. Fewer valid codes and a constraining process that eliminates duplicative credit mean every code submitted needs to earn its place in the record.
The OIG audit makes this operational. In every one of the nine high-risk categories reviewed, the independent medical review contractor’s standard was whether the medical record contained documentation that supported the coded condition as a current, active, appropriately treated diagnosis: not just a historical reference. Plans whose documentation practices do not meet that bar are carrying audit exposure on every submission cycle.
How Should Health Plans Adapt to CMS-HCC Model V28?
The compliance path under V28 follows a clear logic: document what is real, link it to evidence, review it in both directions, and make sure your tools can show their work.
- Audit current coding practices against V28 mapping. Identify which ICD-10-CM codes in your current submission workflow are no longer valid under V28. The removal of 2,027 diagnostic codes means existing workflows almost certainly include obsolete codes that will not generate risk scores and may generate audit flags.
- Shift from volume to clinical specificity. Under V28, the imperative is to document each diagnosis precisely enough to survive scrutiny. Invest in clinical documentation improvement (CDI) workflows that support clinical precision at the point of care, not retrospective code addition.
- Run two-way retrospective reviews. The Aetna enforcement action establishes that retrospective programs which add codes but never remove unsupported ones are now a legal liability. Every review program should systematically identify both missed diagnoses and unsupported diagnoses. See RAAPID’s retrospective risk adjustment solution.
- Flag high-risk diagnosis categories proactively. Apply specific validation logic to OIG target categories before submission: acute stroke, acute MI, embolism, sepsis, and cancer diagnoses where history-of conditions may have been coded as active.
- Use AI with an auditable evidence trail. CMS’s January 2026 RADV memo confirms it plans to use AI as a coder support tool in audits. [4] Plans that rely on AI tools that cannot produce an auditable evidence trail are at a structural disadvantage. RAAPID’s Novel Clinical AI Platform uses Neuro-Symbolic AI, combining large language models with clinical knowledge graphs, to validate every HCC suggestion against MEAT-based evidence extracted directly from the medical record. Charts take 8 to 12 minutes to review.
What Does the CMS 2027 Advance Notice Mean for Risk Adjustment Strategy?
The CMS 2027 Advance Notice, published February 2026, makes clear that the trajectory established by V28 continues. [2]
The most consequential proposed change for risk adjustment programs is the exclusion of unlinked chart review records. CMS specifically stated it is proposing to exclude diagnosis information from unlinked Chart Review Records, meaning diagnosis information not associated with a specific beneficiary encounter, from risk score calculation starting in CY 2027. CMS estimates this change alone would reduce payments by 1.53 percent, representing more than $7 billion. This is a direct regulatory prohibition on the add-only chart mining model.
The implication for healthcare organizations: the era of retrospective volume is structurally ending. The path forward runs through encounter-based documentation, prospective clinical decision support at the point of care, and coding programs that remove unsupported diagnoses as readily as they add missed ones.
How Does V28 Affect Dementia and Cognitive Impairment Coding?
Cognitive impairment and dementia coding requires special attention under V28. Dementia remains a condition with high-RAF HCC categories, and V28 expands those categories to better capture severity distinctions. [1] However, dementia is estimated to be underdiagnosed in approximately 60 percent of cases by current standards, creating both missed revenue and documentation risk. [6]
Mild cognitive impairment sits at a classification threshold where clinical precision directly determines whether a patient’s condition generates a valid HCC. Under V28, that threshold is higher than it was under V24. Healthcare organizations with significant Medicare Advantage populations of older, higher-acuity members need to ensure that cognitive impairment documentation specifically supports the severity level being coded, with clear MEAT-based clinical evidence, not just a historical mention in the record.
What Role Does AI Play in V28 Compliance?
AI-driven platforms are now a practical necessity for V28 compliance across healthcare organizations. The volume of code changes, the clinical precision required in ICD-10-CM to HCC mapping, and the dual obligation to both add and remove diagnostic codes are beyond what manual coding workflows can reliably handle at scale.
The most defensible AI platforms for accurate risk adjustment under V28 do four things: they analyze both structured and unstructured data to identify conditions that qualify under V28’s more specific requirements; they automatically cross-walk V24 codes to V28, flagging diagnostic codes that have been removed or reclassified; they provide documentation guidance during patient encounters, not just retrospectively; and they generate evidence trails that link every suggested HCC to the specific clinical language in the medical record.
CMS’s own RADV expansion plan calls for AI as a coder support tool, with human certified medical coders retaining all decision authority over overpayment determinations. [4] That’s the same human-in-the-loop model RAAPID uses: Neuro-Symbolic AI surfaces the clinical evidence and validates the HCC; the clinician or coder makes the final call. Generic NLP tools built on pattern matching can identify that a term like “diabetes” appears in a chart. What they cannot reliably do is distinguish between a current, actively managed diabetic condition with documented complications and a historical mention of resolved glycemic issues. V28’s severity-first logic requires that distinction at scale.
Conclusion
The CMS-HCC Model V28 has fundamentally changed what defensible risk adjustment looks like. Fewer valid codes. A severity-first weighting structure. A constraining process that eliminates duplicate credit. And an enforcement environment backed by active RADV audits, OIG compliance reviews finding 91 percent error rates on high-risk diagnoses, and DOJ settlements specifically targeting add-only retrospective programs.
This is not a documentation problem. It is a clinical accountability problem. The plans that navigate V28 successfully link every diagnosis to real, encounter-grounded clinical evidence, run two-way reviews that capture missed complexity and remove unsupported codes, and use AI that can show its work to a RADV auditor.
Risk adjustment has shifted from a revenue function to a clinical, compliance, and enterprise AI discipline. RAAPID’s Novel Clinical AI Platform was built for exactly this environment: explainable, evidence-based, two-way, and audit-ready. To learn how RAAPID supports Medicare Advantage plans under CMS-HCC Model V28, Request a demo.
Frequently Asked Questions About CMS-HCC Model V28
CMS-HCC Model V28 became fully operative for payment year 2026, meaning 100 percent of Medicare Advantage risk scores are now calculated using V28. The phased implementation (33% V28 in 2024, 67% V28 in 2025) is complete. [1] Voice answer (20-30 words): CMS-HCC Model V28 took full effect on January 1, 2026. All Medicare Advantage risk scores are now calculated exclusively under V28.
V28 reduced valid diagnostic codes from 9,797 in V24 to 7,770, a net removal of 2,027 codes. Of that total, 2,236 codes were removed and 209 new codes were added. [1] Voice answer: V28 removed a net 2,027 diagnostic codes compared to V24, dropping from 9,797 to 7,770 valid ICD-10-CM codes.
CMS requires risk adjustment programs to both add missed diagnoses and remove unsupported ones. The Aetna/CVS $117.7 million DOJ settlement in March 2026 specifically penalized an add-only retrospective program that submitted additional codes without deleting unsupported diagnoses identified in the same review. [3]
RADV audits are active and expanding. CMS’s January 2026 memo confirmed audits covering PYs 2018 through 2024, with sample sizes from 35 to 200 enrollees per contract, a five-month medical record submission window, and new audits initiated approximately every three months. [4]
Based on OIG’s March 2026 audit series (A-07-22-01207), the highest-risk categories are acute stroke, acute myocardial infarction, embolism, lung cancer, breast cancer, colon cancer, prostate cancer, sepsis, and pressure ulcer. Acute stroke and acute MI had 100 percent error rates in the sampled population. [5]
The 2027 Advance Notice proposes to exclude unlinked chart review records from risk score calculations starting CY 2027, which CMS estimates will reduce payments by 1.53 percent. It also continues reducing coefficient weights for high-prevalence chronic conditions: CKD Stage 3A down 50.4 percent, COPD down 18.8 percent, morbid obesity down 19.4 percent. [2]
MEAT criteria (Management, Evaluation, Assessment, and Treatment) is the documentation standard CMS uses to validate HCC diagnoses in RADV audits. Every submitted diagnosis must be supported by at least one of these four elements in the medical record, tied to a current face-to-face encounter, not just a historical reference.
V28 expands high-RAF dementia HCC categories but requires more specific clinical documentation to qualify. Mild cognitive impairment documentation must reflect the specific severity level coded, supported by MEAT-based clinical evidence. Dementia is estimated to be underdiagnosed in approximately 60 percent of cases, creating documentation risk alongside missed revenue. [6]
Source
[1] CMS. 2024 Announcement of Calendar Year (CY) 2024 Medicare Advantage Capitation Rates and Medicare Advantage and Part D Payment Policies. cms.gov/files/document/2024-announcement-pdf
[2] CMS. 2027 Advance Notice for Medicare Advantage Capitation Rates and Part C and Part D Payment Policies. cms.gov/files/document/2027-advance-notice.pdf (February 2026)
[3] U.S. Department of Justice. Aetna Agrees to Pay $117.7 Million to Resolve False Claims Act Allegations. justice.gov/opa/pr/aetna-agrees-pay-1177-million-resolve-false-claims-act-allegations (March 11, 2026)
[4] CMS, Center for Program Integrity. Update on the Status of Medicare Advantage Risk Adjustment Data Validation Audits. Memorandum from Steven Ferraina, Acting Director, Audits and Vulnerabilities Group. January 27, 2026. cms.gov/medicare/health-plans/medicareadvantageapps/radv-audit-resources
[5] OIG, HHS. Medicare Advantage Compliance Audit of Specific Diagnosis Codes (Contract H0104). Report A-07-22-01207. oig.hhs.gov (March 2026)
[6] Alzheimer’s Association. 2024 Alzheimer’s Disease Facts and Figures. alz.org
About the author
Wynda Clayton
Director of Risk Adjustment Coding & Compliance
Wynda is a recognized leader with over 20 years of experience in risk adjustment, coding, and compliance. A seasoned RADV auditor and educator, she focuses on maximizing coding accuracy and maintaining regulatory standards. At RAAPID, Wynda spearheads AI-driven initiatives to enhance value-based care delivery and reimbursement accuracy.
