RADV audit HCC documentation MEAT

Surviving RADV Extrapolation: MEAT Documentation That Holds

Here is the short version. Under the CMS RADV Final Rule, one unsupported HCC pulled in the sample no longer costs you one recovery. It gets extrapolated across the contract. The fix is not more coders arguing after the fact. It is proving Monitor, Evaluate, Assess or Treat evidence before the chart is ever sampled.

ASP-RCM Risk Adjustment HCC Engine V28 Operator to operator

The answer, up front

A diagnosis you cannot defend with Monitor, Evaluate, Assess or Treat evidence is not a coding gap. In an extrapolated RADV audit it is a contract-level liability, because the auditor multiplies your error rate against the whole population.

Why the stakes changed

The old math forgave one chart. The new math does not.

CMS finalized contract-level extrapolation in its RADV Final Rule (CMS-4185-F2), applying extrapolated recovery beginning with payment year 2018 and declining to apply a Fee-for-Service Adjuster. That is the whole shift in one sentence: the sample now speaks for the population.

One sampled chart, three very different endings

A single HCC with no MEAT support, followed through the audit.

1 HCC sampled

→

0 MEAT evidence found

→

×N Extrapolated to cohort

Before extrapolation, that unsupported code was a single-member takeback. After the RADV Final Rule, the confirmed error rate in your sample is projected across the enrollment it represents. The chart did not get more expensive. The consequence of the same chart did.

The standard that survives review

MEAT is the four-word test an auditor runs on every code

MEAT is the documentation-integrity framework coders and auditors share for proving a condition was truly present and managed during the encounter. A code needs at least one leg to stand on. RADV-grade codes usually show more than one.

Monitor

Signs, symptoms, labs or disease status followed over time. Stable, worsening or improving all count as monitoring.

status & trend

Evaluate

Test results reviewed, response to treatment weighed, findings interpreted in the note rather than just ordered.

results reviewed

Assess / Address

The condition named in the clinical assessment with a decision attached. A problem list mention alone is the classic RADV soft spot.

named + decided

Treat

Medication, therapy, referral, procedure or an active plan. The most defensible leg because it is hard to fabricate after the visit.

plan of care

“

When I validate a sampled HCC, I am not hunting for the ICD-10 code. I am hunting for the encounter-year proof that the condition was real and managed. A diagnosis carried forward on a problem list with nothing done about it reads to me as unsupported, and unsupported is what gets extrapolated.

Risk-adjustment validation lead

Auditor archetype // payer-side RADV

Same code, two fates

What collapses under extrapolated sampling, and what holds

Documentation scenario

Under RADV sampling

The version that survives

Chronic condition on the problem list only

Carried forward, no encounter-year action. Coder cannot show MEAT. Deleted on validation.

Assessed in the note with status plus a plan, so Assess and Treat are both visible for the audited year.

Diagnosis stated, no clinical detail

A code with no monitoring, evaluation or treatment. Reads as a label, not a managed condition.

Tied to labs reviewed and medication continued, giving Monitor and Treat that an auditor can point to.

Condition in history, not this visit

Coded from prior-year context. Fails the date-of-service and current-management test.

Re-documented as active this encounter with evaluation, or correctly not coded when truly resolved.

Specificity and laterality missing

Unspecified code where the record supports more. Recodes downward, changing the HCC and the payment.

Coded to the highest documented specificity, matching the ICD-10-CM Official Guidelines the auditor uses.

“

The codes that scare me are not the wrong ones. They are the lonely ones. A perfectly valid HCC sitting in a note with no monitoring, no plan, nothing an auditor can hold. I would rather query the provider today than defend that chart during extrapolation, because by then it is not one member, it is the cohort.

Senior risk-adjustment coder

Coder archetype // provider-side CDI

Cited by name

The real 2026 rulebook behind every decision above

CMS RADV Final Rule (CMS-4185-F2)

Establishes contract-level extrapolation of RADV audit findings beginning with payment year 2018 and finalizes that CMS will not apply a Fee-for-Service Adjuster. This is why a sampled error now travels.

Extrapolation + no FFS Adjuster

MEAT documentation criteria

The Monitor, Evaluate, Assess or Treat framework that CDI teams and auditors apply to confirm a condition was present and managed during the encounter, not just carried forward on a list.

Monitor / Evaluate / Assess / Treat

ICD-10-CM Official Guidelines for Coding and Reporting, FY2026

Effective October 1, 2025 through September 30, 2026. Governs code specificity, coding only conditions that coexist and affect care, and reporting from provider documentation of the encounter.

FY2026, effective Oct 1 2025

CMS-HCC Risk Adjustment Model V28

The 2024 model phased in through 2026, reaching full weight in payment year 2026. It reshapes which conditions map to an HCC, so documentation that mapped cleanly under V24 must be re-checked against V28.

Full V28 in PY2026

Coding clinic and CDI query standards

The compliant-query discipline that lets a coder resolve a documentation gap with the provider before the record is final, rather than assuming intent after the fact.

Compliant provider query

Date-of-service and current-management rules

Risk-adjustment coding is tied to the audited service year. A condition must be documented and managed in that year to support its HCC, a point extrapolated sampling tests directly.

Encounter-year evidence

Close the gap before the audit does

Four moves that turn lonely codes into defensible ones

Score every HCC against MEAT before submission

Run each risk-adjusting code through the four-word test at the point of coding. Any code with zero legs is a query, not a submission.

Re-map your book to V28, not V24 memory

Confirm the conditions you rely on still land in an HCC under the 2024 model at full 2026 weight, and refresh documentation prompts where they moved.

Query in the encounter year, while it is still fixable

A compliant provider query closes an Assess or Treat gap when the record is open. After the sample is drawn, the same gap is an extrapolated takeback.

Pre-audit your sample the way CMS will

Pull a representative sample and validate it against the RADV Final Rule standard now. Find the unsupported codes yourself, before the error rate is the one that gets multiplied.

Our HCC engine scores MEAT before CMS ever samples it

ASP-RCM Solutions built its V28 risk-adjustment engine to flag the lonely codes, the problem-list carryovers and the V24-to-V28 drops that collapse under extrapolation, then route them to a compliant query while the record is still open. That is the difference between defending a chart and never needing to.

Pre-audit your risk-adjustment exposure

ASP-RCM Solutions // Risk Adjustment & HCC Coding Integrity