On August 28, 2026, CMS notified Medicare Advantage organizations selected for Payment Year 2024 Risk Adjustment Data Validation (RADV) audits. It’s the fourth audit-year notification CMS has issued in twelve months — PY2020 in March, PY2021 in May, and now PY2024 — and it lands while the legal fight over how CMS is allowed to calculate what MA plans owe is still unresolved. For coding and compliance teams, the PY2024 round isn’t just another audit cycle. It’s a live test of whether the diagnosis codes submitted for payment two years ago can survive a records review under rules that are themselves being litigated.
What CMS actually announced
Each RADV audit cycle follows a similar mechanic: CMS pulls a sample of enrollees per audited contract — industry guidance puts the range at roughly 35 to 200 enrollees depending on contract size — and requests the medical records supporting the HCC diagnosis codes tied to those enrollees’ risk scores. Plans generally get a submission window of around five months to produce records, and can submit up to two records per condition, though only one valid record is required to support a code.
Why the timeline matters
A five-month window sounds generous until you consider what it actually requires: pulling charts that may be scattered across multiple provider groups and EHRs, verifying that each HCC code has a note that meets CMS’s documentation standard, and doing it for every sampled enrollee simultaneously. Coding teams that can’t quickly trace a diagnosis code back to its supporting encounter are the ones who run out of runway.
The extrapolation question hasn’t gone away
The reason PY2024 audits are being watched so closely is what happens after the records come back. In September 2025, a federal court in Humana Inc. v. Becerra vacated CMS’s 2023 RADV final rule, ruling that CMS could not extrapolate audit findings across an entire contract’s enrollee population without having properly justified that approach during rulemaking. CMS appealed that decision that December. The appeal is still pending, which means the audits CMS is running right now — including PY2024 — are proceeding under a cloud: MA plans don’t yet have a final answer on whether a handful of coding errors found in a sample can be scaled up into a contract-wide repayment demand, or whether recoveries will be capped closer to the errors actually found.
For coding teams, that uncertainty is not a reason to relax. Whichever way the appeal lands, the underlying documentation standard doesn’t change — a diagnosis code still needs a chart note that supports it. What extrapolation determines is how expensive it is when that standard isn’t met.
What “not met” costs, in one recent example
Monogram Health, a Tennessee-based provider, agreed on August 24, 2026 to pay $2.4 million to settle a False Claims Act suit alleging it submitted clinically unsupported diagnosis codes between 2021 and 2023, according to the Department of Justice. The codes at issue spanned four HCC categories: protein-calorie malnutrition, substance use disorder, coagulation defects, and angina pectoris. DOJ’s allegation was straightforward — the codes inflated beneficiary risk scores and, with them, the payments CMS made to the Medicare Advantage organizations involved. A whistleblower, a physician, received roughly $380,000 of the settlement.
The Monogram Health case wasn’t a RADV audit — it was a whistleblower-driven FCA action — but it illustrates the exact failure mode RADV audits are designed to catch: HCC codes that look fine in a claims file until someone asks to see the chart behind them.
What this means for coding and compliance teams right now
With PY2024 records due on a rolling basis over the coming months, a few practical priorities stand out:
- Re-verify that every high-value HCC code — particularly chronic and comorbidity categories like malnutrition, substance use disorder, and cardiovascular conditions — has a note meeting MEAT (monitor, evaluate, assess/address, treat) documentation criteria, not just a problem-list entry.
- Build a fast path from diagnosis code to source encounter. If it takes coders longer than a few minutes to locate the chart behind a code, that’s a records-request bottleneck waiting to happen.
- Treat query and CDI workflows as audit preparation, not a separate function — the ACDIS/AHIMA 2026 query guidelines released September 1 apply directly to the kind of ambiguous documentation RADV reviewers flag.
- Track which contracts and payment years have open audits, so records requests don’t arrive as a surprise to whoever owns the underlying charts.
- Don’t assume the extrapolation appeal will resolve favorably before PY2024 findings are final — document as if full recovery exposure applies.
Where automation actually helps
The common thread across PY2024’s records timeline, the extrapolation fight, and cases like Monogram Health is traceability: can a plan show, quickly and for every sampled enrollee, exactly which encounter note supports which diagnosis code. That’s a harder problem at scale than it sounds, especially across contracts with years of accumulated coding history and multiple documentation sources. It’s also precisely where agentic AI coding platforms are being put to work — not to make coding decisions unsupervised, but to flag codes lacking adequate supporting documentation before a payer or auditor does, and to keep an audit-ready trail from every code back to its source note.
That’s the kind of defensibility Medikode’s automated medical coding platform is built around — surfacing documentation gaps at the point of coding, so RADV season doesn’t become a scramble.