Risk Adjustment

MEAT criteria (Monitor, Evaluate, Assess, Treat)

~8 min readBy Pelica Health teamPublished August 28, 2026Updated August 28, 2026Reviewed by Catherine Zhao

Short answer

MEAT stands for Monitor, Evaluate, Assess, Treat. It is a documentation habit that shows a chronic condition was actively managed at a face-to-face encounter, which is what makes the diagnosis codeable and the resulting HCC defensible. MEAT is not a CMS rule. The word appears nowhere in the FY2026 ICD-10-CM Official Guidelines or the CMS RADV Medical Record Reviewer Guidance. What binds is Section IV.J of the coding guidelines: code conditions that require or affect care at the encounter, and do not code conditions that were previously treated and no longer exist.

What MEAT is

MEAT is a four-letter check on a single line of documentation. Monitor, Evaluate, Assess, Treat. If a note shows a clinician doing at least one of those four things to a named condition at that encounter, the condition was managed and the diagnosis is supportable. If the note only names it, it was not.

The mnemonic came out of coding education and clinical documentation improvement work, not out of a regulation. It stuck because it gives clinicians and coders one sentence to agree on, and because the alternative is arguing about a whole chart.

IV.J

The ICD-10-CM guideline that actually governs: code what requires or affects care, drop what no longer exists.

Times the word “MEAT” appears in the FY2026 ICD-10-CM guidelines or the CMS RADV reviewer guidance.

~7,770

ICD-10 codes that map to a V28 HCC, down from roughly 9,700 under V24. Fewer chances to get it right.

What each letter looks like in a real note

Generic phrasing is what fails review. Each line below is tied to a named condition and to something that happened at that visit.

Letter	What it proves	Documentation that holds up
Monitor	The clinician is tracking the condition over time	“Type 2 diabetes with diabetic polyneuropathy: reviewed home glucose log, fasting readings 130 to 165 over the past month. Foot numbness unchanged since May.”
Evaluate	The clinician looked at data and drew a conclusion	“CKD stage 3b: eGFR 37 today, down from 44 in March. Reviewed BMP and urine albumin-to-creatinine ratio; decline consistent with progression rather than dehydration.”
Assess	The clinician states where the condition stands now	“COPD, moderate: stable on current inhaler regimen. No exacerbation since the March course of prednisone.”
Treat	Something was done about it at this encounter	“Chronic systolic heart failure: increased furosemide to 40 mg daily, continue sacubitril-valsartan, recheck BMP in two weeks.”

Notice what each line has in common. The condition is named with its specificity. Something at this visit is attached to it. A reviewer reading only that line can tell the condition affected care.

MEAT is not a CMS rule, and it matters that you know that

Search the FY2026 ICD-10-CM Official Guidelines for “MEAT” and you get nothing. Search the CMS Contract-Level RADV Medical Record Reviewer Guidance and you get nothing there either. CMS does not grade documentation on whether an acronym was satisfied.

What CMS does have is Section IV.J of the coding guidelines, which says: “Code all documented conditions that coexist at the time of the encounter/visit and that require or affect patient care, treatment or management. Do not code conditions that were previously treated and no longer exist.” Section IV.I adds that chronic diseases treated on an ongoing basis may be coded and reported as many times as the patient receives treatment and care for them.

Read together, those two sentences are the actual standard, and MEAT is a reliable shortcut to meeting it. The distinction matters when a vendor or a consultant tells your clinicians that CMS requires all four elements. It does not. Building a documentation policy on a rule that does not exist makes it harder to defend the policy you actually need.

Why a problem-list mention fails

A diagnosis pulled from a problem list is the single most common unsupported HCC. The reason is not that CMS bans problem lists. It is that a list entry carries no evidence the condition was in play that day.

The RADV reviewer guidance instructs auditors to evaluate a problem list “for evidence of whether the conditions are chronic or past and if they are consistent with the current encounter documentation,” and to look for support across the full record including history, medications and the final assessment. It then draws hard lines. Lists of code numbers without narratives are not acceptable. Lists of conditions written by the patient are not acceptable. Conditions from lists labeled as pertinent negatives should not be submitted. Problem lists that are not clearly dated as part of the audited face-to-face encounter get judged case by case, which is a coin flip you are choosing to take.

Two other failure modes sit alongside it and have nothing to do with clinical content. The record has to come from a face-to-face encounter with an acceptable provider type, and it has to be authenticated: a signature, legible credentials, the right member, the right date of service. Records fail on those grounds routinely, and a perfectly documented condition in an unsigned note is still a repayment.

MEAT vs TAMPER

TAMPER is the other mnemonic you will hear: Treatment, Assessment, Monitor or Medicate, Plan, Evaluate, Referral. It covers the same ground with two additions, splitting out the plan and the referral, and making prescribing explicit rather than folding it into “treat.”

Neither one is codified. Neither is referenced as validation criteria in CMS risk adjustment rulemaking. The argument between them is a training-material argument, not a compliance one. What matters is that one mnemonic is taught, used, and audited against consistently, so that a coder, a clinician and an auditor reading the same note reach the same answer.

How MEAT interacts with V28

The documentation standard did not change under V28. The consequences of missing it did.

Payment year 2026 is the first year Medicare Advantage risk scores run entirely on V28, after CMS blended roughly one third V28 in 2024 and two thirds in 2025. V28 expands to 115 payment HCCs while cutting the ICD-10-to-HCC crosswalk to about 7,770 codes from roughly 9,700 under V24, and it lowers many coefficients. Fewer diagnoses map, and the ones that still map carry more of the score.

That changes the arithmetic of a failed note. Under V24, a member with a wide spread of mapped conditions could absorb one unsupported HCC. Under V28 the spread is thinner, so the same failure takes out a larger share of the member's RAF. V28 also uses constrained groups, where related conditions share one coefficient, which means documenting a second condition in the same group adds audit surface without adding score. MEAT discipline is what stops a team from spending encounter time on codes that cannot pay and cannot survive review.

Where RADV stands right now

Two things are true at once, and they point in opposite directions.

Audit volume went up. In May 2025 CMS said it would audit all eligible Medicare Advantage contracts every payment year, roughly 550 of them, up from about 60 a year, and grow its coder pool from 40 to around 2,000. That plan is intact.

Extrapolation went away, at least for now. On September 25, 2025 the U.S. District Court for the Northern District of Texas vacated the 2023 RADV final rule in Humana Inc. v. Becerra, holding it procedurally invalid because it was not a logical outgrowth of the proposed rule. The Department of Justice appealed to the Fifth Circuit on November 21, 2025, and that appeal is pending.

Planning against the second fact would be a mistake. The exposure math is unsettled; the volume math is not. Every contract now expects a records request, and a records request is answered with notes, not with a legal position.

Common mistakes teams make with MEAT

Copy-forward status lines. “Diabetes: stable” repeated across twelve visits reads as a template, not an assessment. Reviewers notice when the language never changes while the labs do.
Attaching MEAT to the wrong condition. A medication list at the bottom of a note does not link a drug to a diagnosis. The link has to be explicit.
Coding “history of” as active. Section IV.J is direct about conditions previously treated that no longer exist. Z85 and Z86 history codes belong in the note when they influence current care; they do not risk-adjust.
Treating MEAT as the compliance standard itself. It is a proxy. The standard is the coding guidelines plus a valid, authenticated, face-to-face record.
Auditing after the year closes. A gap found in February of the following year cannot be fixed. A gap found while the encounter is still open can.
Chasing conditions inside a constrained group. Under V28 they share a coefficient, so the second one adds review risk and no revenue.

How Pelica handles MEAT documentation

Pelica's Risk Adjustment Copilot works the problem at the point of capture rather than at the point of audit. When it surfaces a suspected HCC, it shows the clinician what evidence already exists in the record and what is missing for that specific condition, and it attaches the source note, the supporting labs, the provider attestation and the coder review to the captured diagnosis as one packet.

The point is that a records request becomes a retrieval instead of a reconstruction. Pelica runs across 350,000+ members, and new customers are live in about two weeks. For how to build the program around it, read building a RADV-defensible HCC program.

Related terms

An HCC is what a documented diagnosis becomes when it maps to a payment category, and RAF is the score those categories add up to. RADV is the CMS audit that tests whether the note behind each one holds. V28 is the model that decides which diagnoses map at all. The full program build is in the RADV-defensible HCC guide.

Sources

Frequently asked questions

Common questions about MEAT criteria, what CMS actually requires, and how documentation gets tested in a RADV audit.

What does MEAT stand for in medical coding?

MEAT stands for Monitor, Evaluate, Assess, Treat. It is a shorthand used by coders and clinical documentation teams to check that a chronic condition was actively managed at the encounter rather than simply listed. Monitor is tracking signs, symptoms or a disease course. Evaluate is reviewing test results or response to treatment. Assess is stating the condition's status. Treat is the medication, procedure, referral or therapy directed at it.

Is MEAT a CMS requirement?

No. MEAT is a teaching device, not a regulation. The word does not appear anywhere in the FY2026 ICD-10-CM Official Guidelines for Coding and Reporting or in the CMS Contract-Level RADV Medical Record Reviewer Guidance. What binds is Section IV.J of the coding guidelines: code all documented conditions that coexist at the time of the encounter and that require or affect patient care, treatment or management, and do not code conditions that were previously treated and no longer exist. MEAT is a practical way to show a condition met that bar.

Why does a problem list entry fail a RADV audit?

Because a name on a list does not show the condition affected care that day. The CMS RADV Medical Record Reviewer Guidance tells reviewers to evaluate a problem list for whether the conditions are chronic or past and whether they are consistent with the current encounter documentation, and to look for support elsewhere in the record such as history, medications and the final assessment. It also states that lists of code numbers without narratives are not acceptable, that lists of conditions written by the patient are not acceptable, and that conditions from lists labeled as pertinent negatives should not be submitted.

What is the difference between MEAT and TAMPER?

They are two mnemonics for the same idea. MEAT is Monitor, Evaluate, Assess, Treat. TAMPER is Treatment, Assessment, Monitor or Medicate, Plan, Evaluate, Referral, which splits treatment into a plan and a referral and makes prescribing explicit. Neither is codified in the ICD-10-CM Official Guidelines or referenced as validation criteria in CMS risk adjustment rulemaking. Pick one and use it consistently; the audit tests the note, not the acronym.

Does MEAT change under the V28 risk adjustment model?

The documentation standard does not change. What changes is the cost of getting it wrong. Payment year 2026 is the first year Medicare Advantage risk scores run on 100% V28, after a blend of roughly one third V28 in 2024 and two thirds in 2025. V28 cuts the ICD-10-to-HCC crosswalk to about 7,770 codes from roughly 9,700 under V24 and lowers many coefficients, so fewer diagnoses pay and each surviving one carries more of the score. A thinner set of payable conditions means a single unsupported note removes a larger share of a member's RAF.

How many MEAT elements does a diagnosis need?

There is no CMS-published number, because CMS does not score documentation by acronym. In practice, one clearly documented element tied to the specific condition at that encounter is the working minimum coding teams use, and a status plus a plan is stronger. The real test is the one in the guidelines and the RADV reviewer guidance: does the note show this condition required or affected care at this visit, in a record from a face-to-face encounter with an acceptable provider type, signed with legible credentials.

Can CMS still extrapolate RADV audit findings?

Not under the 2023 rule as it stands. On September 25, 2025 the U.S. District Court for the Northern District of Texas vacated the 2023 RADV final rule in Humana Inc. v. Becerra, finding it procedurally invalid because it was not a logical outgrowth of the proposed rule. The Department of Justice filed a notice of appeal to the Fifth Circuit on November 21, 2025 and that appeal is pending. Separately, CMS's May 2025 plan to audit all eligible Medicare Advantage contracts every year, roughly 550 of them, is unaffected. The audit volume is real regardless of how the extrapolation question resolves.

The note either shows the condition was managed, or it does not.

Pelica's Risk Adjustment Copilot flags what a specific condition still needs before the encounter closes, and attaches the source note, labs, attestation and coder record to every captured HCC. Pelica runs across 350,000+ members, and new customers are live in about two weeks.

Book a Demo → See the Risk Adjustment copilot