CMS Overhauls RADV Audit Program for Medicare Advantage Plans
In May, the Centers for Medicare and Medicaid Services (CMS) announced a major overhaul of the Risk Adjustment Data Validation (RADV) audit program, expanding the size, scope and timeline. CMS conducts annual RADV audits to ensure the accuracy of diagnosis codes submitted by MA plans. Diagnosis codes directly impact patient risk scores, which determines CMS’ risk-adjusted payment amounts to MA plans per enrollee.
CMS plans to audit all Medicare Advantage plans – more than 500 plus, which is a sharp increase from the approximately 60 plans previously reviewed each year. In addition, the number of records reviewed per plan has increased from about 35 to 200, substantially raising documentation demands. To support this expansion, CMS is growing its RADV workforce from 40 to 2,000 coders by September 2025 and plans to use artificial intelligence and machine learning to identify unsupported diagnoses.
CMS uses extrapolation to estimate overpayments based on a sample of medical records and applies those findings across the full plan population. Further pressure comes from CMS’ goal to complete audits for Payment Years 2018 through 2024 by early 2026, compressing the timeline for plans and providers to respond.
Provider Impacts
The RADV program expansion may pose challenges for long-term and post-acute care providers, especially those serving large MA populations or participating in risk-based contracts. Some providers are already being impacted, as many MA plans have begun requesting records from facilities in preparation for or to respond to the audits. This may increase documentation pressure on providers, who must respond to a higher volume of record requests with tighter turnaround times. In addition, MA plan contracts may include claw back clauses, allowing plans to recover payments from providers if CMS identifies overpayments. AHCA has met with CMS and highlighted the system-wide impact these audits are having. We will keep you updated if changes develop.