The CMS RADV Tracker.
Audit status by payment year, live submission windows, the rule-change log, and the litigation that decides extrapolation — every entry linked to its primary source. Bookmark it; we keep it current.
Last verified: July 19, 2026 · updated on CMS’s quarterly audit cadence
Audit status by payment year
Per CMS’s published audit schedule and RADV Q&A. The order is deliberately non-sequential — CMS’s data-processing schedule dictates it.
| Payment year | Audit initiation | Status |
|---|---|---|
| PY2018 | Nov 14, 2024 · ~60 contracts, 35-enrollee samples | Findings expected from mid-2026 |
| PY2019 | Jun 12 & 25, 2025 · first all-contract cycle (~550) | Findings begin in 2026 |
| PY2020 | Mar 20, 2026 | In progress — records due Aug 28, 2026 |
| PY2021 | May 29, 2026 | In progress |
| PY2024 | August 2026 (scheduled) | Next up |
| PY2023 | November 2026 (scheduled) | Scheduled |
| PY2022 | January 2027 (scheduled) | Scheduled |
| PY2025 | April 2027 (scheduled) | Scheduled |
Live now — PY2020 key dates
Enrollee Data List in CDAT: Apr 3, 2026 · Record window: Apr 13 – Aug 28, 2026 (11:59 pm EST) · Hardship exception deadline: Sep 11, 2026 · Samples: 200 / 100 / 50 / 35 by contract-size stratum
Legacy cleanup: PY2011 audit reports issued Jan 2025; PY2012–13 payment-error results published May 13, 2025; PY2014–15 published Mar 4, 2026 — CMS RADV announcements.
The rule-change log
Every material change to the RADV program since extrapolation was codified — dated and sourced.
- Feb 1, 2023
RADV Final Rule (88 Fed. Reg. 6643): extrapolation codified from PY2018; FFS adjuster eliminated; $4.7B recoveries projected for 2023–2032. CMS fact sheet
- Nov 14, 2024
PY2018 audits initiated — ~60 contracts, 35-enrollee samples; record window through Apr 21, 2025. PY2018 Methods & Instructions
- May 21, 2025
Expansion announced: every eligible contract (~550) audited annually; coder workforce 40 → ~2,000; AI-assisted review; samples 35–200. CMS press release
- Jun 12 & 25, 2025
PY2019 audits initiated — the first all-contract cycle. PY2019 Methods & Instructions
- Sep 25, 2025
Humana v. Becerra: the 2023 Final Rule vacated in its entirety and remanded to CMS — extrapolated recoveries paused; audits continue. Georgetown litigation tracker
- Nov 21, 2025
CMS/HHS appeal filed — Fifth Circuit, No. 25-11293. Crowell alert
- Jan 27, 2026
HPMS memo: quarterly audit cadence confirmed; five-month record window restored; up to 2 records per audited HCC; AI-supported tools with human coders making all overpayment decisions. RISE summary
- Mar 4, 2026
Audit schedule published (PY2020 → PY2025, non-sequential); PY2014–15 legacy results released; RADV Q&A updated. CMS audit schedule
- Mar 20 / May 29, 2026
PY2020 and PY2021 audits initiated; PY2020 records due Aug 28, 2026. PY2020 Methods & Instructions
Litigation tracker — the case that decides extrapolation
Humana v. Becerra
- District court: N.D. Tex., No. 4:23-cv-00909-O — summary judgment for Humana, Sept 25, 2025; 2023 rule vacated as procedurally invalid (the FFS-adjuster rationale was not a “logical outgrowth” of the proposed rule). Order
- Appeal: Fifth Circuit, No. 25-11293 — filed Nov 21, 2025; government’s opening brief Mar 21, 2026; no argument date as of our last verification. Docket tracker
- What it means today: CMS collects sampled-enrollee overpayments only, designs audits to support extrapolation, and reserves the right to extrapolate later “if legally permissible.” Plans planning around a sample-only ceiling are betting on the appeal.
What OIG has found, by plan
HHS-OIG audits high-risk diagnosis codes at named Medicare Advantage plans and publishes the results. Across the series, roughly 70% of audited high-risk codes were unsupported by the medical record. OIG high-risk toolkit
| OIG audit | Years | Sampled | Unsupported | Finding |
|---|---|---|---|---|
| Humana Health Plan (H2649) | 2017–18 | 240 | 84% | $6.8M refund |
| Humana of Louisiana (H1951) | 2017–18 | 240 | 91% | ≥$10.5M est. |
| UCare Minnesota (H2459) | 2018–19 | 294 | 86% | $4.7M refund |
| Blue Care Network of Michigan (H5883) | 2017–18 | 210 | 91% | $3.4M refund |
| Triple-S Advantage (H5774) | 2016–17 | 281 | 73% | $297K refund |
| Industry-wide acute-stroke audit | PY2021 | 97 | 100% | ~$462M est.* |
Each row links to the published HHS-OIG report, and the figures are OIG’s findings and recommendations as stated in each one. *Estimated industry-wide overpayment, where OIG recommended preventive edits. In their published responses, the audited plans generally disputed OIG’s findings, methodology, or recommendations, and each report includes the plan’s response.
The patterns behind the findings
- Acute stroke or myocardial infarction coded from an office claim with no matching inpatient claim. OIG found over 90% of these unsupported.
- Cancers coded as active with no treatment or management in the service year, where a history-of code would apply and does not risk-adjust.
- Sepsis without an inpatient claim, and embolism without anticoagulant management.
- Major depressive disorder without documented severity or treatment.
- Conditions supported only by a problem list, with no evidence in the encounter itself.
How extrapolation is calculated
Under CMS’s published method, the average validated risk-score error across the sample is multiplied by the payments for every enrollee in the sampling frame, and applied only when the lower bound of the 90% confidence interval stays above zero. The sample prices the whole contract, which is why the arithmetic matters more than the sample size suggests.
Worked example — CMS methodology, illustrative numbers
- 5–8%
payment error CMS found in its completed PY2011–13 audits
- $4.7B
CMS's projected RADV recoveries, 2023–2032
- $76–84B
MedPAC-estimated annual MA overpayments, 2025–26
Where this stands legally: the 2023 rule authorizing extrapolation was vacated in September 2025 and is on appeal at the Fifth Circuit. For now CMS collects sampled-enrollee overpayments and designs audits “to support the collection of extrapolated recoveries,” reserving the right to extrapolate later if it prevails. CMS PY2020 Methods, §9.2
Quick reference
From the PY2020 Audit Methods & Instructions and the Medical Record Reviewer Guidance.
Record submission rules
- Up to 2 medical records per audited HCC; one valid record validates
- A record = one face-to-face encounter (one DOS outpatient; one stay inpatient)
- Any acceptable provider and any DOS in the collection year — even encounters never submitted for payment
- No amending records, and no asking providers to amend, once the audit begins
- CMS attestation form is the only signature cure — outpatient records, single DOS, cannot validate diagnoses
- Intake feedback posts in CDAT; invalid submissions replaceable while the window is open
- Hardship exceptions available per contract, enrollee, or HCC — with their own deadline
What CMS coders check
- Correct beneficiary on every page
- Acceptable provider type, source, and physician specialty for a face-to-face encounter
- Dates of service inside the data-collection year
- Valid signature with credentials — stamps banned; e-signatures need authentication language; audit-time signatures invalid
- Coding per ICD-10-CM Official Guidelines and AHA Coding Clinic at the encounter date
- Not acceptable: superbills, standalone problem lists, claim forms, lab results alone, unattended home-health forms
Common questions
The questions compliance and risk-adjustment teams ask most often about how the program works.
What is a RADV audit?
Risk Adjustment Data Validation (RADV) is CMS's audit program for verifying that the diagnoses Medicare Advantage plans submitted for payment are supported by medical records. CMS samples 35 to 200 enrollees per contract, requests records for each audited HCC, and certified coders re-validate every diagnosis against the record. Since 2025, CMS audits every eligible MA contract every payment year, which is roughly 550 contracts rather than the 60 or so it audited before.
Is CMS extrapolating audit findings right now?
Not currently. The February 2023 rule that authorized extrapolation was vacated by a federal court in September 2025 (Humana v. Becerra), and CMS's appeal is pending at the Fifth Circuit. In the meantime CMS collects overpayments for the sampled enrollees only, but it designs audits to support extrapolation and has reserved the right to extrapolate later if it prevails, so it is worth preparing as though extrapolation returns.
My contract was selected. What happens now?
CMS notifies you through HPMS, then posts your Enrollee Data List in CDAT with the sampled enrollees and audited HCCs. You have a five-month window to retrieve and submit medical records, with up to two records per audited HCC and one valid record enough to support it. For the PY2020 cycle now underway, records are due August 28, 2026. After certified-coder review CMS issues findings, and you have 60 days to request reconsideration.
What records does CMS accept as evidence?
Documentation of a single face-to-face encounter, or qualifying real-time audio-video telehealth, from an acceptable provider type and physician specialty, with dates of service inside the data-collection year and a valid, credentialed signature. Superbills, standalone problem lists, claim forms and lab results on their own do not qualify. A missing or illegible signature on an outpatient record can be cured only with the CMS-generated attestation form.
Can documentation be fixed after selection?
In most cases no. Plans may not amend medical records or ask providers to amend them once an audit begins, and signatures obtained in response to the record request are invalid. The only cure CMS accepts is its own attestation form, for missing or illegible signatures on outpatient records, and that form cannot validate a diagnosis. Gaps found before selection can still be fixed at the source.
The primary-source library
Don’t take our word for any of it — here is everything we cite, in one place.
Hub changelog
July 19, 2026 — Hub launched. Schedule current through CMS’s Mar 4, 2026 publication; PY2020/PY2021 windows per their Audit Methods & Instructions; litigation status through the government’s Mar 21, 2026 opening brief.
Maintained by Martlet AI as an informational reference for risk-adjustment and compliance teams. Not legal advice — verify against the linked primary sources before acting.
Selected, or expecting to be?
Bring one contract. We'll run a mock RADV on it — sampled per CMS's methodology, with evidence packets and an exposure estimate — inside your environment.
- 95%
closed automatically
- 99%
precision on automated codes
- 100,000+
lives on the platform