CMS opened PY2024 RADV audits built to extrapolate. Records are due February 5, 2027.

On August 28, 2026, CMS published its Payment Year 2024 Medicare Advantage contract-specific RADV Audit Methods and Instructions, the document CMS gives to each contract selected for a PY2024 audit. This is what the August PY2020 deadline piece told coders to watch for. It sets the PY2024 deadlines and says CMS designed these audits to calculate extrapolated overpayments.

The dates that start the PY2024 clock

The document is dated August 28, but the operational clock runs this month. A plan can't see its sample until the enrollee data list posts in CDAT on September 11, and the submission window doesn't open until September 18.

That looks like a long runway. Record retrieval across multiple provider systems is the slow part, so February is really the deadline to find out whether a 2023 encounter actually supports the audited HCC.

CMS designed these audits to extrapolate

In the background section, CMS says it "designed its PY 2024 MA contract-specific RADV audits to calculate extrapolated overpayments," and that it "has not yet decided whether it will collect extrapolated overpayment amounts or only overpayments associated with the sampled enrollees."

Here is how the extrapolation works when CMS applies it. CMS takes the average change in risk score across the sampled enrollees and multiplies it by the sum of county rates for the whole sampling frame. That gives the estimated frame-wide payment error.

The lower bound of the 90 percent confidence interval is the trigger, not the multiplier. CMS applies the extrapolated amount only when that lower bound is greater than zero. If it is at or below zero, CMS doesn't extrapolate and collects only the overpayments tied to the sampled enrollees. Extrapolation also drops out when the sampling frame has fewer than 30 enrollees and CMS is auditing all their HCCs.

For this cycle, CMS says the total it will collect is the sum of the sampled-enrollee payment errors, and it "reserves the right to collect extrapolated overpayments at a later date ... if legally permissible." That last phrase matters, because the rule that authorized extrapolation is itself in litigation.

The litigation behind the extrapolation question

The 2023 rule that authorized statistical sampling and extrapolation in these audits was vacated by a federal district court in 2025, in a decision that is now under appeal (Humana Inc. v. Becerra). CMS is calculating the extrapolated amount for PY2024 while that appeal is pending, which is why it separated calculating the number from deciding whether to collect it.

Who is in the PY2024 sample

CMS didn't audit everyone. The sampling frame is enrollees continuously enrolled in the audited contract from January 2023 through January 2024, with Part B for all of 2023 and at least one 2023 diagnosis that mapped to a PY2024 HCC. CMS excludes ESRD, hospice, C-SNP, I-SNP, and FIDE-SNP enrollees, and anyone already in an OIG audit or settlement.

From that frame, CMS kept only enrollees "ranked in the top quartile" by its improper-payment prediction models, meaning the enrollees predicted to have the greatest reduction in risk score from an audit.

Sample sizes are 35, 50, 100, or 200 enrollees per contract, set by stratum: the ten contracts with the largest sampling frames draw 200, and the rest fall into thirds at 100, 50, and 35. The count of contracts selected is published separately in CMS's list of audited contracts, which this piece doesn't restate.

What a valid PY2024 record has to show

A valid record is a legibly signed record of a face-to-face visit with an appropriately credentialed provider inside the 2023 data collection period. For each sampled enrollee, a plan submits at least one record per audited HCC and no more than twice the number of audited HCCs in total, in any combination. Three audited HCCs means at most six records, not an open chart dump.

The abstraction is done by people. Each valid record can go through up to three independent reviews by certified coders before an HCC is called discrepant.

Two PY2024 details change what counts. Telehealth satisfies the face-to-face requirement only for dates of service through May 11, 2023, when the COVID-19 public health emergency expired. And because PY2024 ran under both the 2020 (V24) and 2024 (V28) CMS-HCC models, audited HCCs carry a model-version label like V24HCC19 on the coversheet.

How this fits the RADV coverage so far

This builds on two earlier pieces. The January 2026 HPMS memo coverage explains the shift to auditing all RADV-eligible contracts on a quarterly cadence, and the blog RADV schedule guide walks the payment-year sequence. The PY2024 methods document is what fills in the specifics: the deadlines above and the extrapolation design. If a record looks thin, run it through the MEAT criteria guide before it goes in.