Every sampled Medicare Advantage stroke diagnosis failed the audit.
Summary
Medicare Advantage organizations are paid more for enrollees whose diagnoses signal costlier care -- a system that depends on the diagnoses being real. HHS's inspector general sampled 97 enrollees nationwide whose acute stroke diagnoses had raised their insurer's payment. None held up against the enrollee's own medical records. OIG's extrapolation: $462 million in potential overpayments for a single payment year, from this one diagnosis category alone -- and 91 of the 95 codes involved remain part of CMS's newest payment model.
A $494 billion program built on self-reported codes
In 2024, paid roughly 760 Medicare Advantage organizations $494 billion⧉ -- 44% of all Medicare spending that year -- using flat monthly payments adjusted for each enrollee's reported health status. doesn't verify those diagnoses at the point of payment; it relies on the MA organizations that benefit from higher-risk codes to submit them accurately in the first place.
View data as table
| History of stroke, not a new acute stroke | 68 |
|---|---|
| Records didn't support any stroke diagnosis | 22 |
| No medical records located | 4 |
| Illegible record | 1 |
| Wrong diagnosis code submitted | 1 |
| Record from an unacceptable source | 1 |
Old stroke, new payment
The single biggest reason a diagnosis failed: 68 of the 97 enrollees' records showed a history of stroke⧉ -- an old, resolved event -- with nothing supporting a new acute stroke at the time it was coded. Twenty-two more had records that didn't support any stroke diagnosis at all, and 4 organizations couldn't even locate the underlying medical record. From the sample's $187,122 in identified net overpayments -- 'net' because one wrongly-coded case actually caused an underpayment, offset against the rest -- extrapolated a national point estimate of $462 million for payment year 2021 (90% confidence interval: $442.2 million to $477.5 million). isn't recommending try to recover that money, since the sample spanned too many different organizations for individualized collection.
View data as table
| Had preventative coding techniques | 59 |
|---|---|
| Had detect-and-correct procedures | 25 |
| Began using OIG's high-risk-code toolkit | 12 |
| No attempt to address the miscoding | 1 |
Not new, and not fixed
This isn't a one-off finding. has separately audited 34 individual MA organizations for high-risk diagnosis codes, and across those, 1,146 of 1,185⧉ reviewed stroke diagnoses were unsupported -- a pattern this new, broader nationwide audit's 100% finding is consistent with. 's written response didn't concur or disagree with 's recommendation for a preventive fix; it pointed to its existing detection audits and a newly phased-in 2026 payment model. 's rebuttal: those audits catch overpayments after they happen rather than preventing them, and 91 of the 95 stroke codes this audit flagged are still part of the new model.
The takeaway
- Not a partial failure -- a complete one. All 97 sampled enrollees' diagnoses failed validation; the extrapolated national impact is about 2,469 times the sample's own dollar findings.
- Most errors weren't fraud, they were confusion. 73% of unsupported cases coded an old stroke as a new one -- a documentation problem more than a deception, per 's own case notes.
- 's fix doesn't close the gap. The codes this audit flagged remain largely intact in 's newest payment model, and didn't commit to a specific preventive step.
labels these 'potential' net overpayments because its review was limited to the specific medical records tied to the flagged physician diagnosis codes, not a comprehensive review of every record that might separately support the diagnosis. The underlying data are from service year 2020 and payment year 2021 -- fieldwork ran into January 2026, and 's own comparison to 's 2026 payment model (91 of 95 flagged codes still in use) is what establishes this audit's continued relevance despite the older claims data. Acute stroke is also only one of several high-risk diagnosis categories tracks; the 34 prior individual-organization audits referenced above covered high-risk diagnosis codes generally, of which stroke was one category. The 59/25/12/1 compliance figures are self-reported by the MA organizations involved and were not independently verified by as part of this specific audit.
Sources(1) ▾
- HHS Office of Inspector General, CMS Potentially Overpaid Medicare Advantage Organizations $462 Million Based on Certain Unsupported Acute Stroke Diagnosis Codes (2026-05-28) — Nationwide audit (Report No. A-02-23-01020) of Medicare Advantage organizations' acute stroke diagnosis code submissions, examining a stratified random sample of enrollees against their medical records, with a statistically extrapolated national overpayment estimate. Fetched directly from oig.hhs.gov and converted with pdftotext -layout; read in full. oig.hhs.gov · original document
Comments
Always open. Logged-in readers can annotate paragraphs in place.
Medicare Advantage organizations are paid more, each month, for enrollees whose diagnoses signal a need for costlier care. 's inspector general⧉ tested that system at one specific pressure point: acute stroke diagnoses submitted by a physician alone, with no matching hospital record. Auditors sampled 97 such enrollees nationwide. For all 97, the diagnosis did not hold up against the enrollee's own medical records.