The Audit That Doesn't Stop
CMS has scaled its Medicare Advantage audit infrastructure from 40 coders to roughly 2,000, moved to universal plan coverage, and locked in a quarterly cadence. The question is no longer whether your plan gets audited — it's whether your documentation, coding governance, and data traceability can hold up under continuous scrutiny.
Payer Compliance · Program Integrity · RADVFor most of the past decade, RADV audits operated like a regulatory lottery. CMS audited a sample of plans. Most plans calculated their odds and managed accordingly. Some invested in documentation. Others hoped the selection process wouldn't land on them.
That logic no longer applies.
CMS has fundamentally changed the structure of its Risk Adjustment Data Validation program. It has scaled its certified coder workforce from approximately 40 to roughly 2,000. It has expanded audit coverage to encompass all 550-plus eligible Medicare Advantage contracts. It has established a cadence of new audit initiations approximately every three months.
Payment Year 2020 audits began in February 2026 and are already underway. AI-enabled tools are now supporting coder efficiency, though final overpayment determinations remain with human coders.
It is a redesign.
The operational logic of what CMS has built
Plans could previously respond to RADV with short-term documentation intensification, then relax between cycles. That model is structurally over.
The five-month medical record submission window CMS restored provides more operational runway. But that concession does not change the fundamental dynamic. Plans that approach RADV as a compliance event they prepare for when selected are already behind.
The audit machine is now running continuously.
The OIG data is consistent, and it is uncomfortable.
These are not outliers. OIG has documented for years that approximately 70 percent of high-risk diagnosis codes lack adequate supporting documentation in medical records. The question RADV industrialization raises is not whether improper payments exist — they clearly do, at scale, and across many plans.
The question is what systematic documentation failure means when it is reviewed quarterly, at full coverage, by 2,000 coders using AI-assisted tools.
Where the exposure
actually lives
CMS's own estimate of improper Medicare Advantage payments — primarily unsupported diagnoses. Roughly $30 billion annually across the full program.
If a plan cannot demonstrate that its FDRs are producing clinical documentation that meets CMS standards, the audit exposure is not just financial. It is a governance and oversight failure.
OIG has documented that approximately 70% of high-risk diagnosis codes audited lack adequate supporting documentation in medical records.
What AI-Assisted Auditing Changes
CMS's use of AI tools to support coder efficiency deserves careful interpretation. The agency has been clear that AI assists coders but does not replace them — all overpayment determinations are made by certified humans. That distinction matters legally and procedurally.
But AI-assisted auditing changes the efficiency profile of the audit program substantially. Coders can review more records in less time. Pattern recognition tools can flag high-risk diagnoses or unusual coding patterns for prioritized review. Documentation completeness can be assessed at volume before any human coder sees a record.
The audit can now reach everything
The volume of records CMS can review per audit cycle is larger than any previous program. Documentation that might have been technically reviewable but practically unlikely to be examined is now more likely to surface.
Thin documentation no longer passes unnoticed
Plans that have historically relied on documentation that is technically present but clinically thin are more exposed. Pattern-based review will find systematic gaps faster than sampling-based audit ever could.
Human coders decide. AI finds the targets.
Final overpayment determinations remain with certified coders — but AI narrows the field to the highest-risk records first. That changes what gets reviewed, and what gets found.
What the quarterly cadence means in practice
Under the previous selective model, a plan that completed an audit cycle could expect a meaningful interval before the next one. Staff, vendors, and systems could be recalibrated based on findings before the next exposure window.
That interval is gone. The administrative burden is real. The staffing implications for compliance, clinical documentation, and audit response are significant.
Structural change requires a structural response.
Plans that treat documentation improvement as a project — something done before an expected audit and then deprioritized — will cycle through corrective action plans indefinitely.
The more durable response is to embed documentation standards, coding governance, and FDR accountability into the operating model as permanent functions, not audit-triggered responses.
That means establishing ongoing coding audits at the physician and vendor level, not just the plan level. It means building data infrastructure that can trace a diagnosis submission from a clinical encounter to a CMS submission with full documentation trail. It means treating FDR documentation standards as a compliance obligation with financial consequences, not a contractual formality.
And it means having a clear, current understanding of where the plan's documentation gaps are — because CMS now has the capacity to find them regardless.