HCC Coding & Risk Adjustment Explained (2026): RAF Scores and the CMS-HCC V28 Model

What HCC coding actually is, in one paragraph

HCC coding is the practice of translating a patient's documented chronic and serious diagnoses into the Hierarchical Condition Categories that Medicare and other payers use to predict next year's cost of care. Each qualifying ICD-10-CM diagnosis maps to an HCC, each HCC carries a numeric weight, and the sum of those weights plus a demographic component becomes the patient's Risk Adjustment Factor (RAF) score. A sicker, well-documented patient produces a higher RAF, and the health plan receives a higher monthly payment to care for them. That is the whole loop: document the disease, map it to an HCC, the HCC drives the RAF, the RAF drives the payment. This article explains every step of that loop the way the AAPC Certified Risk Adjustment Coder (CRC) exam expects you to understand it — and then shows you exactly what changed now that the CMS-HCC V28 model is 100% live for payment year 2026.

free CRC practice questionsPractice questions with detailed explanations

The four moving parts of risk adjustment

Every risk adjustment scenario — and every CRC exam question — sits on four parts. Learn them as a chain, not as isolated facts.

The critical mental shift for new risk adjustment coders: not every codeable diagnosis is risk-adjustable. Roughly 70,000+ ICD-10-CM codes exist, but only a fraction map to a payment HCC. A sprained ankle is codeable and not risk-adjustable. Diabetes with chronic kidney disease is both. The CRC exam is largely a test of telling those apart and defending the call from the chart.

How a RAF score is built (with a worked example)

A RAF score is additive, then adjusted. Start with the patient's demographic coefficient (driven by age, sex, Medicaid status, and whether they are community or institutional, originally disabled, etc.). Then add the coefficient for each unique HCC the patient's documentation supports for the year. Disease interactions can add more.

Here is a simplified community-patient example using illustrative coefficients:

That raw score is then divided by the annual normalization factor (which CMS sets to keep the average beneficiary at 1.0) and reduced by the statutory coding-intensity adjustment, which has been held at the minimum 5.9% since 2018. So the plan does not get paid on 1.065 directly — it gets paid on a normalized, intensity-reduced version of it. The exact coefficients change each year; the structure does not, and that structure is what CRC tests. The official coefficients live in the CMS Risk Adjustment program documentation.

The coder's job is the middle rows of that table: making sure every chronic condition that is monitored, evaluated, assessed, or treated during the year is captured once, with the specificity the HCC requires — and that nothing unsupported sneaks in.

Hierarchies: why "hierarchical" is in the name

The "H" in HCC is the rule new coders trip over. Within a disease family, a more severe condition trumps a less severe one, and only the highest-weighted HCC in that hierarchy counts. If a patient has both diabetes without complications and diabetes with chronic complications documented, you do not get paid for both — the more severe HCC supersedes the milder one in the same hierarchy. Capturing the same diabetes twice does not raise the RAF; capturing it at the wrong (lower) specificity lowers the legitimate score. CRC questions love this trap: they hand you two conditions in one hierarchy and ask which one drives the score.

MEAT: the documentation filter that decides everything

A diagnosis on a problem list is not automatically reportable for risk adjustment. The condition has to be addressed during a face-to-face encounter in the data-collection year, and the standard evidence test coders apply is MEAT — Monitor, Evaluate, Assess, Treat:

• Monitor — signs, symptoms, disease progression noted
• Evaluate — test results, response to treatment reviewed
• Assess — ordering tests, discussion, records review
• Treat — medications, therapies, referrals, plan of care

If at least one MEAT element supports the diagnosis in the note, it is defensible. If the chart only carries a stale problem-list mention, a history-of code, or a rule-out, it is not. This is why risk adjustment is a documentation-abstraction discipline before it is a code-lookup discipline, and why a CRC drills MEAT until it is automatic.

The CMS-HCC V28 model: what changed for 2026

This is the single biggest reason to learn risk adjustment in 2026 rather than from a 2022 textbook. CMS replaced the long-running V24 model with V28, phasing it in over three payment years so plans could adjust:

As of payment year 2026, risk scores are calculated 100% on V28 — the blend is over. Any prep material that still teaches V24 mappings as current is wrong for this year. CMS documents the transition in its Advance Notice and Rate Announcement materials.

V28 by the numbers

V28 is not a cosmetic update. The structure changed:

• The model expanded from 86 HCCs (V24) to 115 payment HCCs (V28) and renumbered them, so the HCC numbers you memorized under V24 no longer line up.
• The set of ICD-10-CM codes that map to a payment HCC shrank from about 9,797 to roughly 7,770 — about 2,236 codes removed and 209 added. Thousands of diagnoses that used to risk-adjust no longer do.
• CMS constrained coefficients across many disease families (giving related HCCs the same weight) to reduce the payment swing from minor coding differences.

Concrete conditions that lost or changed value

The CRC exam and your daily work both hinge on knowing which conditions moved. High-profile V28 changes include:

• Diabetes coefficients were consolidated and constrained — the various diabetes presentations now carry a single, roughly halved weight rather than escalating by complication.
• Protein-calorie malnutrition was removed as a payment HCC.
• Atherosclerosis of the extremities with intermittent claudication and stable (chronic) angina were dropped or moved to non-payment status.

CMS targeted exactly the diagnoses it found most subject to coding variation. For a coder, the lesson is blunt: the documentation still matters clinically, but the payment impact of these specific conditions has fallen, and capturing them at V24 logic in 2026 is simply wrong. The American Academy of Family Physicians summarized the clinical-side fallout in its HCC change briefing.

Prospective vs retrospective review

Risk adjustment coders work in two modes, and CRC tests both:

• Prospective review happens before or during the encounter — flagging suspected, undocumented chronic conditions so the provider can address and document them at the visit. It improves accuracy at the source.
• Retrospective review happens after — auditing closed charts to find supported diagnoses that were missed or to remove ones that were not supported.

A mature program does both. The exam may ask which review type fixes a given problem, so anchor each to its timing.

RADV: why compliance is not optional

Everything above runs under the shadow of RADV — Risk Adjustment Data Validation, CMS's audit program that verifies submitted diagnoses against the actual medical record. If a plan submitted an HCC that the chart does not support under MEAT, RADV can claw the payment back. CMS finalized an expanded RADV rule in February 2023 that removed the fee-for-service adjuster and allowed CMS to extrapolate audit findings across a contract — but note the currency trap that almost every competing page misses: a federal court vacated that 2023 rule on procedural grounds in September 2025, so the exact extrapolation framework is unsettled and may be re-promulgated or appealed. What is not unsettled is the underlying principle the CRC tests: a submitted diagnosis must be supported by the record. For a CRC, that is the whole point of the credential — code what the record supports, at the right specificity, and nothing else. "Defensible" is the operative word.

What the CRC exam tests from all this

The AAPC CRC exam — 100 multiple-choice questions, 4 hours, 70% to pass — is essentially a graded version of this article. Expect questions that ask you to:

• map an ICD-10-CM diagnosis to its correct HCC (under V28),
• decide whether documentation meets MEAT,
• resolve a hierarchy (which condition supersedes),
• distinguish CMS-HCC from HHS-ACA, CDPS, and commercial models,
• and identify a compliance/RADV risk in a chart scenario.

Membership in AAPC is required and the exam is open-book with an approved ICD-10-CM code book, but "open-book" rewards speed, not browsing. The full format and current fees are on the official AAPC CRC certification page, and our free CRC exam guide walks the format, fee, and study-plan side in depth.

A fast path to actually learning this

Reading about HCCs builds recognition; answering questions builds the skill the exam scores. Use this sequence:

1. Drill the chain — diagnosis to HCC to RAF to payment — until you can explain it without notes.
2. Memorize the V28 deltas: 86 to 115 HCCs, the shrunken code set, constrained diabetes, removed malnutrition/angina.
3. Practice MEAT on chart excerpts until "is this supported?" is instant.
4. Run timed mixed sets with your code book beside you.

free CRC practice questionsPractice questions with detailed explanations

---

Official sources used

• CMS Risk Adjustment program — https://www.cms.gov/medicare/payment/medicare-advantage-rates-statistics/risk-adjustment
• CMS Advance Notice and Rate Announcement documents (V28 phase-in) — https://www.cms.gov/medicare/payment/medicare-advantage-rates-statistics/announcements-and-documents
• AAFP, HCC model V28 change briefing — https://www.aafp.org/pubs/fpm/issues/2023/1100/hcc-update.html
• AAPC CRC certification page — https://www.aapc.com/certifications/crc