What does CMS check during a Medicare Advantage RADV audit?

CMS uses Medicare Advantage Risk Adjustment Data Validation to confirm that diagnoses submitted by MA organizations for risk adjustment are supported in enrollee medical records. If the medical record does not support a submitted diagnosis, CMS may collect overpayments. For coders, this means each HCC needs current encounter support, provider documentation, and enough detail to support the ICD-10-CM code submitted.

Can HCC coders code from the problem list alone?

A diagnosis on the problem list is not enough by itself for defensible HCC capture. Coders should look for current provider documentation that shows the condition was addressed during the encounter. That support may include monitoring, evaluation, assessment, or treatment. If a diagnosis appears only as a copied-forward problem list item with no current discussion, it is weak support for risk adjustment coding.

What should coders recheck before submitting an HCC?

Coders should confirm the diagnosis is tied to a valid encounter, addressed by the provider, specific enough for the documentation, and supported in the current record. Common areas to slow down include CKD staging, heart failure type and acuity, diabetes complications, obesity with BMI support, vascular specificity, resolved conditions, and medication clues that are not linked to a documented diagnosis.

RADV Audit Schedule 2026: What HCC Coders Should

CMS has posted RADV audit timing for several payment years. Here is what HCC coders should recheck before unsupported diagnoses become expensive.

By HCC Buddy Coding Team , Certified Professional Coder (CPC)
Reviewed: May 26, 2026

Quick Answer

CMS has posted a Medicare Advantage RADV audit schedule, and HCC teams should treat it as a reminder to tighten documentation review now.

The current schedule lists audit initiation months for multiple payment years, including payment year 2020 in March 2026, payment year 2021 in May 2026, payment year 2024 in August 2026, payment year 2023 in November 2026, payment year 2022 in January 2027, and payment year 2025 in April 2027. CMS notes that these dates can change.

For coders, the practical takeaway is simple. Every submitted HCC needs support in the medical record. If the record does not support the diagnosis, the code can become an overpayment issue.

Why RADV Is Back in the Conversation

RADV is not new. It just gets louder when the calendar starts filling up.

CMS describes Medicare Advantage Risk Adjustment Data Validation as its main program for checking whether diagnoses submitted by MA organizations for risk adjustment are supported in enrollee medical records. CMS also says unsupported diagnoses can lead to overpayment collection.

That is the piece coders need to keep close. RADV is not about whether a diagnosis sounds clinically possible. It is about whether the submitted diagnosis is supported by the record CMS is reviewing.

That means the everyday chart review habits matter. The small misses matter. The copied-forward condition nobody addressed matters. The unspecified code that looked close but did not match the documentation matters.

What CMS Has Published

CMS has published a RADV audit schedule listing intended audit initiation months by payment year. CMS also published payment year 2020 RADV audit methods and instructions in March 2026.

CMS says RADV audits happen after the final risk adjustment data submission deadline for the MA contract year. In plain coder language, the review comes later, but the support has to be there in the original record.

That is why clean coding habits now matter even when the audit letter comes later.

The Coder Checklist Before a Diagnosis Leaves Your Desk

Before an HCC gets submitted, run the diagnosis through these checks.

1. Is There a Valid Encounter?

The diagnosis should be tied to a valid encounter, not floating in a chart review note without a service connection.

Check the date of service, provider type, encounter context, signature status, and whether the diagnosis appears in the encounter being coded.

Do not let a diagnosis drift away from the visit that supports it.

2. Did the Provider Address It?

A problem list entry is not enough by itself.

Look for the provider doing something with the condition. That can be monitoring, evaluating, assessing, or treating. Coders usually shorthand this as MEAT.

Examples that help include A1c reviewed for diabetes, eGFR reviewed for CKD, medication continued for CHF, symptoms assessed for COPD, or a specialist note reviewed for an active condition.

Examples that should make you pause include a diagnosis appearing only in past medical history, a copied-forward condition with no current discussion, a medication list without provider linkage, or a diagnosis in the assessment with no current support.

3. Is the Code Specific Enough?

RADV is not only about whether a condition exists. It is also about whether the submitted code matches the documentation.

Common places coders should slow down include CKD without stage, heart failure without type or acuity when the chart supports more detail, diabetes without linked complications, obesity with only a BMI code, and vascular disease with unspecified wording.

If the chart supports a more specific code, code to the documentation. If it does not, do not invent specificity from labs, medications, or old notes.

4. Is the Support Current?

Risk adjustment is calendar-year work, but RADV reviews still come back to the record.

Current support matters. A diagnosis from last year, a copied problem list, or an old specialist assessment may not be enough if the current provider did not address the condition.

Ask one plain question.

Can I point to the sentence, lab review, treatment plan, or assessment that supports this diagnosis for this encounter?

If the answer is no, the code is not ready.

The Mistakes That Create RADV Trouble

Coding From the Problem List Alone

This is the classic trap. The diagnosis may be real. The patient may have had it for years. The problem list may be accurate.

Still, if the provider did not address it during the encounter, it is weak support for risk adjustment.

Treating Medication Clues Like Provider Documentation

Medication lists are useful. They can help a coder spot a missed diagnosis or prepare a query.

But a medication list does not replace provider documentation. Insulin may suggest diabetes. Entresto may suggest heart failure. Inhalers may suggest COPD. The provider still needs to document and address the condition.

Ignoring Specificity Because the HCC Looks the Same

Sometimes two codes map to the same HCC. That does not mean specificity stops mattering.

If the documentation supports a specific condition and the submitted code is vague, the chart can still look sloppy under review. Specific coding is part of defensible coding.

Carrying Forward Resolved Conditions

History of cancer is not active cancer. Old myocardial infarction is not acute MI. Prior stroke does not automatically support an active CVA code.

Resolved and historical conditions need the right code path.

What Team Leads Should Do This Week

If you manage HCC work, do a small audit before this becomes a big one.

Pull 10 charts with high-risk HCCs and ask whether every HCC is tied to the correct encounter, whether each diagnosis has provider support, whether the note shows MEAT, whether specificity elements are present when required, whether old problem-list conditions are being captured without current support, and whether coder queries are being used when documentation is close but not enough.

This does not need to be dramatic. A tight 10-chart review can show whether your team has a documentation habit problem, a specificity problem, or a workflow problem.

Where HCC Buddy Fits

HCC Buddy helps coders check ICD-10-CM codes, HCC mapping, RAF impact, and related coding details faster.

It does not replace coder judgment. It does not decide whether the provider documentation supports the diagnosis. That part still belongs to the coder.

Use HCC Buddy to speed up the lookup work, then use the medical record to answer the audit question.

Is the diagnosis supported here, today, by this provider, in this encounter?

That is the question RADV eventually comes back to.