HCC Risk Adjustment · Engine V28

RADV-Ready: Building MEAT-Standard Documentation Into the Risk Workflow

The short version: a RADV audit does not fail at the audit letter. It fails months earlier, when a coded HCC was submitted without documentation that a reviewer can defend.

The answer up front: you become RADV-ready by moving the medical-record standard upstream. Every diagnosis your engine codes should pass a MEAT check (Monitor, Evaluate, Assess, Treat) before it is ever submitted, so the chart CMS pulls for a Risk Adjustment Data Validation review already contains the evidence a validator is looking for. Readiness is a workflow gate, not an after-the-fact scramble.

See it run on your charts →

Why the bar moved

RADV stopped being a paperwork exercise

CMS finalized the policies that made unsupported HCCs expensive. These are the rules that reshape how a risk workflow has to behave, cited by name.

Extrapolation

The RADV Final Rule (CMS-4185-F2, 2023) applies extrapolated recovery to audit findings.

SOURCE // CMS RADV FINAL RULE

PY2018+

Extrapolation applies beginning with payment year 2018 audits; findings on a sample project across the contract.

SOURCE // CMS-4185-F2

No FFS Adjuster

CMS finalized RADV audits without a Fee-For-Service Adjuster, so unsupported codes are not cushioned.

SOURCE // CMS-4185-F2

V28 by 2026

The 2024 CMS-HCC model (V28) is phased in over three years and fully weights risk scores in payment year 2026.

SOURCE // CY2024 RATE ANNOUNCEMENT

The evidence standard

MEAT is what a validator reads for

A diagnosis on a claim is a code. A diagnosis that survives RADV is a code with a story in the note. MEAT is the shorthand for that story: the four ways a record shows a condition was actively addressed on the date of service.

Monitor

Signs, symptoms, disease progression or regression tracked in the note over time.

Labs · vitals · trend

Evaluate

Test results, response to treatment, medication effectiveness reviewed by the provider.

Results · exam · review

Assess

Clinical judgment: the condition named in the assessment, counseling or discussion documented.

Impression · A&P

Treat

Medications, therapies, referrals or procedures ordered to manage the condition.

Rx · order · referral

One MEAT element, tied to a diagnosis by a credentialed provider on a valid face-to-face encounter, is what turns a coded HCC into a defensible one. The workflow below builds that check into the pipeline.

The unique angle

One chart, through the engine, past the MEAT gate

Follow a single encounter as the V28 engine moves it from intake to a RADV-ready packet. The gate is the point of the diagram: no HCC leaves the pipeline until its evidence would survive a medical-record review.

STEP 01

Chart intake

Encounter note and problem list land in the queue with date of service and rendering provider.

STEP 02

PHI de-ID boundary

Protected data is stripped at a fail-closed boundary before any model call. No PHI leaves the machine.

STEP 03

Extract & map V28

Conditions are lifted from the note and mapped to ICD-10-CM, then to CMS-HCC V28 categories.

STEP 04

MEAT evidence gate

Each candidate HCC is held until the note proves Monitor, Evaluate, Assess or Treat.

GATE // MEAT-standard evidence check — run per diagnosis, before submission

MonitorIs a trend, symptom or status change tied to this condition?

EvaluateAre results or treatment response reviewed against it?

AssessIs it named in the provider's assessment or plan?

TreatIs a med, order, therapy or referral attached?

Evidence found → submitThe HCC advances with the supporting sentence flagged and the provider signature and DOS confirmed. This is what a RADV validator opens the chart to find.

Evidence thin → hold + queryThe code is parked, not submitted. A documentation query goes to the provider. Nothing unsupported reaches the risk score, so nothing unsupported can be extrapolated against.

STEP 05 · OUTPUT

RADV-ready packet

Submitted HCCs ship with their MEAT citation, provider, and date of service linked. When the audit letter arrives, the "one best medical record" is already assembled, not hunted for.

Survives vs. gets recovered

What a RADV reviewer accepts, line by line

The difference between a code that holds and a code that gets clawed back is rarely the diagnosis itself. It is the evidence attached to it. Same condition, two outcomes.

Coded condition (archetype)	Fails RADV review	Survives RADV review
Chronic condition on the problem list	Carried forward, not addressed in the note on the DOS.	Assessed & treated at the visit with a MEAT element documented.
Diagnosis in a scanned attachment	No signature or credential tying it to a valid encounter.	Provider-signed face-to-face note with date of service.
Condition coded from history	"History of" language for an active, monitored condition.	Status documented as current with monitoring or evaluation.
HCC-driving specificity	Unsupported specificity the note does not substantiate.	Coded to the note, mapped to the correct V28 category.
Discrepant record	Conflicting notes with no single defensible source.	One best record pre-selected and packaged for the pull.

The guidelines this workflow is built on

Named, current, and load-bearing

RADV METHODOLOGY

CMS RADV Final Rule (CMS-4185-F2)

Establishes the Risk Adjustment Data Validation audit method, sampling, and the "one best medical record" review that MA-submitted diagnoses are validated against.

EXTRAPOLATION POLICY

Extrapolation from PY2018, no FFS Adjuster

The same Final Rule extrapolates audit recovery findings across the contract beginning with payment year 2018 and finalizes audits without a Fee-For-Service Adjuster.

RISK MODEL

2024 CMS-HCC Model (V28)

The revised risk model, phased in through the CY2024–CY2026 Rate Announcements and fully weighted in payment year 2026, which the engine maps every diagnosis to.

CODING RULES

ICD-10-CM Official Guidelines & MEAT criteria

The FY2026 ICD-10-CM Official Guidelines for Coding and Reporting govern code selection; MEAT is the industry documentation standard for demonstrating a condition was actively addressed.

ASP-RCM Solutions · HCC Risk Adjustment Engine V28

We built the MEAT gate into the workflow, so audit readiness is the default, not the fire drill

Our V28 risk-adjustment engine de-identifies before any model call, maps to CMS-HCC V28, and holds every candidate diagnosis at the MEAT evidence gate. Codes that clear it ship with their supporting citation attached. Codes that do not become a provider query instead of an extrapolation exposure. When the RADV letter comes, the defensible record is already assembled.

Book a walkthrough on your own charts →

ASP-RCM Solutions — Senior Partner, Frisco · Risk Adjustment & RADV Readiness