Are Your Medicare Advantage Charts Ready for a RADV Audit?

Key Takeaways

• CMS has expanded risk adjustment data validation (RADV) audits from roughly 60 Medicare Advantage (MA) plans per year to all 550+ eligible contracts. Audit readiness is now a permanent operational requirement, not a periodic concern.
• CMS is reviewing up to 200 records per plan per audit cycle, up from approximately 35, which means documentation gaps that once went unnoticed under random sampling now create real financial exposure.
• RADV audits assess whether diagnosis codes submitted for risk adjustment are supported by medical records. Unsupported diagnoses lead to direct overpayment recovery demands.
• AI-assisted review technology, combined with certified coding expertise, makes 100% encounter review before an audit operationally feasible at scale.

If you manage a Medicare Advantage (MA) plan, the operational reality of RADV audits shifted significantly in 2025, and it did so in a way that demands a different response than most organizations had previously built.

For years, the risk adjustment data validation (RADV) audit program was selective. CMS reviewed around 60 contracts per year. Plans that weren't selected could reasonably treat RADV as a background concern, something to review periodically, address when an audit notice arrived, and then set aside. That model no longer applies.

In May 2025, CMS announced it would expand RADV audits to all eligible MA contracts, more than 550 plans, up from roughly 60. The number of medical records reviewed per plan also increased, from approximately 35 to up to 200. CMS grew its RADV workforce from 40 coders to approximately 2,000 and announced that AI-assisted tools would support the review process, though all final coding determinations remain with certified human coders. As of early 2026, CMS has been working through Payment Year (PY) 2020 audits across eligible contracts. New audits are planned approximately every three months.

The message from CMS is not subtle. RADV audits are now routine and comprehensive, running on a cycle that covers every eligible plan, every year.

‍

What a RADV Audit Actually Evaluates

A RADV audit is how CMS confirms that the diagnosis codes MA plans submitted for risk adjustment are actually supported by enrollees' medical records. Plans receive higher payments for members with greater chronic disease burden, built into each member's risk adjustment factor (RAF) score. The RADV audit verifies that the diagnoses driving those payments are real, documented, and defensible.

During an audit, CMS selects a sample of enrollees and requests their corresponding medical records. AHIMA and AAPC-certified coders review those records against strict documentation standards, including whether conditions meet the MEAT criteria, meaning each diagnosis must show evidence that it was Monitored, Evaluated, Assessed, or Treated during a face-to-face encounter within the relevant payment year. If a diagnosis is in the system but the record doesn't support it, CMS flags it as unsupported and demands repayment.

Earlier RADV audits found error rates of 5 to 8% in payment years 2011 through 2013. With 200 records reviewed per plan, even that error rate creates meaningful repayment exposure per audit cycle, without any extrapolation applied.

The Extrapolation Question and Why It Still Matters

In September 2025, a federal district court vacated entire 2023 RADV final rule that would have allowed CMS to extrapolate findings from a sample across an entire contract population. That ruling, in Humana v. Becerra, temporarily relieved plans that had been bracing for a scenario where a 5% error rate in a 200-record sample could be projected across tens of thousands of enrollees, creating repayment demands that could reach into the tens of millions.

CMS appealed the ruling in November 2025. The court said CMS hadn't followed proper rulemaking procedures, not that extrapolation itself is impermissible. It may return, in the same or a revised form, in a future rulemaking cycle.

Critically, even without extrapolation, the current framework still generates direct financial consequences. Unsupported diagnoses identified in the sample lead to overpayment demands. The difference between a world with extrapolation and the current one is the scale of exposure, not whether exposure exists.

For MA plans, the implication is clear. RADV audit readiness has to be a standing operational posture, not a sprint activity that starts when an audit notice arrives.

What RADV Audit Readiness Actually Requires

The old random-sampling model created a specific kind of complacency. With only 35 records reviewed and a low probability of selection, many plans avoided serious scrutiny for years even with documentation gaps across their encounter data. That dynamic is over. With 200 records per audit and all plans in scope, the documentation has to hold at scale.

Readiness means a few distinct things in practice; it means all submitted diagnoses have clinical documentation support that would hold up under coder review. Provider records need to be organized and retrievable, not scattered across disconnected systems that require manual hunting when CMS submits a records request. Gaps should be identified before the auditor sees them, and the submission packages going to CMS should be complete, compliant, and defensible from the first page.

The challenge most organizations face is volume. Manual chart review, one record at a time, is not a scalable answer when a plan has tens of thousands of encounters to validate across a payment year. This is precisely where AI-assisted review technology changes what's operationally possible.

How AI-assisted Review Changes the Math on Audit Readiness

At HOM, our RADV Audit Readiness service is powered by AuditPro, our proprietary AI-assisted auditing platform. What that enables, practically, is 100% review of submitted risk adjustment encounters before a CMS audit, rather than relying on reactive record pulls or post-notice scrambles.

Our AI-assisted review, with AHIMA and AAPC-certified experts in the loop at every step, identifies documentation gaps, flags unsupported hierarchical condition categories (HCCs), and validates diagnosis accuracy against CMS guidelines. Plans see their audit exposure before CMS does, which is the only way to actually control it.

The results bear this out. Compared to manual review, our AI-assisted process cuts audit time by close to 50%. Our documentation validation accuracy is more than 90%, as measured against CMS guidelines in controlled pilot environments. When audit findings need to be addressed, our teams deliver compliant, actionable submission packages with the supporting documentation already assembled, so nothing is pulled together under deadline pressure.

We also handle provider chart retrieval and documentation management for the audit review process itself. This is a significant operational burden that most plans aren't set up to absorb, particularly with new audit cycles initiating approximately every three months. 

That operational complexity extends beyond the plan itself.

The Provider-side Exposure

RADV audit risk doesn't stop at the health plan. In value-based care arrangements, shared-savings models, and capitated contracts, audit findings can flow downstream to provider groups. A plan that identifies unsupported diagnoses in its records may recoup payments from the providers whose documentation failed to support the codes submitted.

For physician groups and health systems working within MA arrangements, this means RADV readiness is their concern too, not just the plan's. Proactive internal documentation review, provider education on MEAT criteria and HCC accuracy, and structured auditing of clinical records all protect the practice's revenue from those arrangements and its relationship with the plan that governs them.

CMS has been clear about its direction. RADV audits are running every three months, across every plan, and the documentation scrutiny is growing. The plans that are reviewing their encounters proactively before an audit notice arrives will be in a fundamentally different position than those that wait for one. The window between receiving an audit notice and the submission deadline is not long enough to fix documentation gaps that should have been caught months earlier.

Request your free audit to evaluate your current RADV readiness and identify documentation gaps before they become repayment demands.

Frequently Asked Questions

1. What is a RADV audit? 

A RADV audit is a CMS review process that confirms whether the diagnosis codes submitted by Medicare Advantage plans for risk adjustment are supported by enrollees' medical records. Unsupported diagnoses result in overpayment recovery demands from CMS.

2. Which Medicare Advantage plans are subject to RADV audits now? 

As of May 2025, CMS expanded the RADV program to cover all eligible MA contracts, more than 550 plans, up from the previous model of auditing roughly 60 plans per year. Payment Year 2020 audits began in February 2026, with new cycles initiating approximately every three months.

3. What documentation does CMS look for during a RADV audit? 

CMS verifies that each submitted diagnosis code is supported by clinical evidence in the medical record. Documentation must demonstrate that the condition was Monitored, Evaluated, Assessed, or Treated (MEAT criteria) during a face-to-face encounter in the relevant payment year. Unsupported or insufficiently documented diagnoses are flagged for repayment.

4. How can a Medicare Advantage plan reduce RADV audit risk? 

Proactive 100% encounter review before an audit notice is the most effective approach. This means validating all submitted diagnoses against documentation standards, organizing records for fast retrieval, and correcting gaps before CMS identifies them. AI-assisted review tools combined with certified coding expertise make this scalable across large chart volumes.

‍