HCC Coding Under V28: What Changed and How AI Helps

The CMS-HCC risk adjustment model is the financial engine behind Medicare Advantage. It determines how much plans are paid for each enrollee based on the diagnoses documented and coded during the measurement year. When CMS announced the V28 model overhaul in the 2024 Medicare Advantage Rate Announcement and Final Call Letter, it set off the most significant restructuring of hierarchical condition categories in over a decade.

Now that 2026 marks the first year of full V28 implementation, every risk adjustment coder, CDI specialist, and revenue cycle leader needs to understand exactly what changed, what it means financially, and how to adapt workflows accordingly. This guide breaks down the V28 model in detail and explains where AI-assisted coding fits into the picture.

What Is the CMS-HCC Risk Adjustment Model?

The CMS Hierarchical Condition Category model assigns risk scores to Medicare Advantage beneficiaries based on their demographic profile and coded diagnoses. Higher risk scores correspond to sicker patients and result in higher capitated payments to the plan. The model uses ICD-10-CM diagnosis codes mapped to HCC categories, which are then fed into payment algorithms that produce a Risk Adjustment Factor (RAF) score for each member.

The CMS-HCC model serves two purposes: it compensates plans fairly for enrollees with complex medical needs, and it discourages cherry-picking healthier populations. Accurate HCC coding is therefore not just a compliance requirement — it is the primary mechanism through which Medicare Advantage organizations capture appropriate revenue.

Why Did CMS Replace V24 with V28?

CMS periodically recalibrates the risk adjustment model to reflect updated clinical evidence, coding patterns, and expenditure data. The V24 model, which had been in use since payment year 2020, was built on older fee-for-service claims data and diagnostic classification logic that CMS determined no longer reflected current medical practice.

In the 2024 Rate Announcement, CMS finalized the transition to V28, citing several objectives: improving payment accuracy, reducing the impact of coding intensity on risk scores, updating clinical classifications to align with current ICD-10-CM coding standards, and better distinguishing between severity levels within disease categories. CMS also noted that the V24 model's reliance on certain high-prevalence, low-cost conditions was inflating risk scores beyond what actual enrollee expenditures justified.

What Is the V24-to-V28 Transition Timeline?

CMS implemented V28 through a three-year phased blend to give plans time to adjust coding operations and financial forecasts. The transition schedule, as published in the CMS Rate Announcements for payment years 2024 through 2026, is as follows:

| Payment Year | V24 Weight | V28 Weight | Status | |---|---|---|---| | 2024 | 67% | 33% | Complete | | 2025 | 33% | 67% | Complete | | 2026 | 0% | 100% | Current year — full V28 |

Payment year 2026 is the first year in which risk scores are calculated entirely under V28 with no V24 blending. This means that any HCC category removed from the model no longer generates any risk adjustment revenue, and any newly added category is now fully weighted in the payment calculation. Organizations that delayed V28 preparation during the blended years face the full financial impact this year.

How Many HCC Categories Changed Between V24 and V28?

The V28 model restructured the HCC category set substantially. The V24 model used 86 HCC categories that mapped to risk-adjusted payments. The V28 model expanded this to 115 HCC categories, reflecting more granular clinical distinctions and new severity hierarchies.

However, the expansion in categories does not mean more conditions are captured overall. The V24 model accepted 9,797 ICD-10-CM diagnosis codes as valid inputs for risk adjustment. The V28 model accepts only 7,770 ICD-10-CM codes — a reduction of approximately 2,027 codes. This means that many diagnoses which previously contributed to a patient's risk score under V24 no longer map to any HCC under V28.

This combination — more categories but fewer accepted codes — is the central tension of V28. The model is more clinically precise but also more restrictive in what it pays for.

What Are the Key HCC Category Changes in V28?

Dropped HCC Categories

Several high-prevalence conditions that generated significant risk adjustment revenue under V24 were removed or substantially narrowed in V28. Notable eliminations and reductions include:

• HCC 48 (Morbid Obesity) — Removed entirely. Under V24, morbid obesity (BMI 40+) was a standalone HCC. V28 no longer assigns a risk score to obesity alone, though it may still serve as a supporting condition in hierarchies.
• HCC 59 (Major Depressive Disorder and Bipolar Disorders) — Restructured. Several mental health conditions were reclassified, with lower-severity depression codes dropped from risk adjustment.
• HCC 40 (Specified Rheumatoid Arthritis) — Narrowed. Many rheumatoid arthritis codes no longer map to a payment-eligible HCC.
• Peripheral vascular disease categories — Consolidated and severity-tiered, with lower-acuity PVD codes removed.
• Certain chronic kidney disease stages — CKD Stage 3 codes were dropped from risk adjustment, with only Stage 4 and above remaining payment-eligible.

Added and Reclassified Categories

V28 introduced new categories to capture conditions that were previously underweighted or grouped too broadly:

• Dementia with and without complications — Split into severity-tiered HCCs, allowing higher payment for dementia patients with documented behavioral disturbances or complications.
• Heart failure severity tiers — Expanded from a single HCC to multiple tiers distinguishing reduced ejection fraction, preserved ejection fraction, and unspecified heart failure.
• Substance use disorders — New HCCs added for alcohol and opioid use disorders, reflecting CMS recognition of the cost burden these conditions impose.
• Chronic pain categories — Select chronic pain conditions now map to risk adjustment for the first time.

Hierarchy Restructuring

V28 also introduced new hierarchies and modified existing ones. The model now includes dual-model processing, running separate risk adjustment calculations for community and institutional enrollees with updated coefficient structures. Conditions that previously existed in flat payment groups were reorganized into severity ladders, meaning that documentation specificity — not just the presence of a diagnosis — determines payment.

What Is the Financial Impact of V28?

CMS estimated in the 2024 Rate Announcement that the V28 model would reduce Medicare Advantage risk adjustment payments by approximately 3% in the first year of full implementation relative to V24. This reduction reflects the removal of high-prevalence, lower-cost conditions from the model and the recalibration of coefficients using more recent claims data.

For a mid-sized Medicare Advantage plan covering 100,000 members, a 3% reduction in risk-adjusted revenue can translate to $30 million to $50 million in annual payment reductions, depending on the plan's population acuity mix. Plans with large populations of members previously coded for conditions like morbid obesity, lower-severity depression, or early-stage CKD face disproportionately higher losses.

However, the financial impact is not uniformly negative. Plans that improve documentation specificity for newly granular categories — such as heart failure severity or dementia complications — can partially offset losses by capturing higher-weighted HCCs that were not available under V24. The net financial outcome depends heavily on how quickly coding operations adapt.

How Do Documentation Requirements Change Under V28?

MEAT Criteria Remain Central

The MEAT documentation standard — Monitor, Evaluate, Assess/Address, and Treat — remains the foundation for supporting HCC code capture under V28. Every condition submitted for risk adjustment must be documented with clinical evidence that the provider monitored, evaluated, assessed, or treated the condition during the encounter. This requirement is unchanged from V24 and is enforced through CMS Risk Adjustment Data Validation (RADV) audits.

Specificity Requirements Are Higher

What has changed is the level of specificity required to map to a payment-eligible HCC. Under V24, a general diagnosis of heart failure might map to a single HCC regardless of type or severity. Under V28, coders must capture whether the heart failure involves reduced ejection fraction (HFrEF), preserved ejection fraction (HFpEF), or is unspecified — and the HCC assignment and payment weight differ accordingly.

This means providers must document not just the presence of a condition but its clinical characteristics. For conditions like dementia, the documentation must distinguish between dementia with behavioral disturbance and dementia without. For chronic kidney disease, stage-specific documentation is mandatory because only stages 4 and 5 remain risk-adjustable.

What Should Coders Focus On?

Risk adjustment coders working under V28 should prioritize:

• Verifying code validity — Confirm that every ICD-10-CM code selected still maps to an HCC under V28. Codes that were valid for risk adjustment under V24 may now be unmapped.
• Capturing severity indicators — Ensure documentation supports the highest appropriate severity tier within each hierarchy.
• Reviewing dropped conditions — Identify members whose risk scores relied heavily on conditions removed from V28 and flag them for clinical documentation improvement (CDI) outreach.
• Dual-model awareness — Understand that community and institutional models have different coefficients and that a member's enrollment status affects which model applies.

How Does AI Help With the V28 Transition?

The V28 model's increased complexity — more categories, fewer accepted codes, severity-tiered hierarchies, and dual-model processing — creates a significant operational challenge for coding teams. Manual coders must now cross-reference every diagnosis against an updated mapping table, verify severity documentation, and track which model version applies to each member. This is precisely the type of high-volume, rule-dense work where AI-assisted coding delivers the most value.

Dual-Model Processing

AI coding platforms can run V24 and V28 models simultaneously on the same clinical document, allowing organizations to see exactly how each chart's risk score changes between model versions. This dual-model capability was critical during the 2024 and 2025 blended years and remains valuable in 2026 for retrospective analysis and financial forecasting.

Jivica's ANICA engine, for example, supports dual-model processing natively, running both V24 and V28 scoring on every chart so that risk adjustment teams can quantify the revenue impact of the transition at the individual member level.

Automated Code Validity Checks

One of the most common V28 errors is submitting ICD-10-CM codes that mapped to HCCs under V24 but are unmapped under V28. AI systems maintain updated mapping tables and flag invalid submissions automatically, preventing revenue leakage from codes that no longer count and avoiding compliance risk from submitting unsupported conditions.

Evidence Extraction and MEAT Compliance

AI-powered coding tools extract clinical evidence directly from provider notes and map it to MEAT criteria for each captured condition. Under V28, where specificity requirements are higher, this evidence extraction capability ensures that every HCC submission is backed by documentation that would withstand RADV audit scrutiny.

ANICA's architecture deploys 9 specialized AI agents and 24 MCP (Model Context Protocol) tools to handle different aspects of the coding workflow — from clinical NLP and code assignment to MEAT validation and RADV readiness scoring. Each agent focuses on a specific task, and the multi-agent design allows the system to process the increased complexity of V28 without sacrificing speed or accuracy.

Identifying Revenue Recovery Opportunities

AI can systematically scan charts for conditions that are newly risk-adjustable under V28 but were not previously captured. Substance use disorders, chronic pain conditions, and severity-tiered heart failure classifications represent new revenue opportunities that manual coders may overlook, especially during the transition period when coding habits are still calibrated to V24 logic.

V24 vs. V28 Comparison at a Glance

| Feature | V24 | V28 | |---|---|---| | HCC categories | 86 | 115 | | Accepted ICD-10-CM codes | 9,797 | 7,770 | | Severity tiering | Limited | Extensive | | Dual community/institutional models | Yes | Yes (updated coefficients) | | Morbid obesity as standalone HCC | Yes | No | | Substance use disorder HCCs | Limited | Expanded | | Heart failure granularity | Single tier | Multi-tier (HFrEF, HFpEF, unspecified) | | Estimated payment impact | Baseline | Approximately -3% vs. V24 | | CKD Stage 3 risk-adjustable | Yes | No |

Key Takeaways

1. 2026 is the first year of 100% V28 implementation. There is no longer any V24 blending to cushion the financial impact. Every risk adjustment submission is scored entirely under V28 rules.

2. V28 has more HCC categories (115 vs. 86) but accepts fewer ICD-10-CM codes (7,770 vs. 9,797). The model is more granular but more restrictive, requiring higher documentation specificity to capture risk-adjustable conditions.

3. High-prevalence conditions like morbid obesity and lower-severity depression were dropped. Plans that relied on these categories for risk score volume face measurable revenue reductions.

4. Documentation specificity is now a direct revenue driver. Under V28's severity-tiered hierarchies, the difference between a general and a specific diagnosis can mean the difference between a mapped and unmapped HCC.

5. CMS estimates a roughly 3% payment reduction under full V28 implementation. For large plans, this translates to tens of millions of dollars in annual revenue impact.

6. AI-assisted coding tools mitigate V28 complexity by automating dual-model processing, code validity checks, evidence extraction, and RADV readiness scoring — enabling coding teams to maintain accuracy and throughput despite the model's increased demands.

What Should Risk Adjustment Teams Do Now?

The organizations that navigate V28 successfully will be those that treat the transition as a documentation and technology challenge, not just a coding update. Clinical documentation improvement programs must be recalibrated to target the conditions and specificity levels that V28 rewards. Coding teams need updated mapping references and real-time validation tools. And leadership needs accurate financial models that reflect V28 economics.

If your organization is evaluating AI-assisted coding to address V28 complexity, ANICA is purpose-built for this transition — with dual-model HCC processing, automated MEAT compliance validation, and RADV audit readiness built into every chart. Schedule a demo to see how it handles your V28 workflow, or explore ANICA's capabilities to learn more about the platform's multi-agent architecture.

---

References: CMS 2024 Rate Announcement and Final Call Letter, CMS Risk Adjustment Data Validation (RADV), AHIMA HCC Coding and Risk Adjustment Guidelines, AAPC Risk Adjustment Coding Resources.