Auditors checked 97 Medicare Advantage stroke codes. Zero passed.

HHS's Office of Inspector General reviewed⧉ a nationwide sample of Medicare Advantage enrollees who had been flagged with a high-risk acute stroke diagnosis code -- a code that raises the monthly payment CMS sends a health plan on that enrollee's behalf.▸ Auditors pulled medical records for 97 of these enrollees. None of the 97 held up: 93 had records that didn't support the stroke diagnosis, and for the remaining 4, the health plan couldn't produce any supporting records at all. Extrapolated across the 240,401-enrollee nationwide pool these codes were drawn from, OIG estimates $462 million in potential net overpayments for 2021 alone.

Medicare Advantage plans aren't paid a flat rate per enrollee.▸ CMS pays a base rate adjusted by a risk score⧉ that rises with each diagnosis a health plan submits -- sicker-coded enrollees mean higher monthly payments. An acute stroke diagnosis carries real weight in that formula, but it's supposed to come from a hospital record, not just a doctor's visit note. OIG's target here was a specific pattern: enrollees whose acute stroke diagnosis appeared only on a physician's billing record, with no matching stroke diagnosis on any inpatient or outpatient hospital record from the same year -- exactly the kind of code prior audits had already flagged as high-risk for miscoding.

OIG drew a stratified random sample of 100 enrollees from the nationwide pool and narrowed it to 97 that fit the audit's scope.▸ An independent medical review contractor checked each one's physician records against hospital records for the same service year. The result, reported⧉ without qualification: every one of the 97 failed. Not most, not nearly all -- all of them. Four health plans couldn't even locate the records to defend the diagnosis they had submitted.

OIG's sampling frame -- every enrollee nationwide who fit this pattern for the 2020 service year -- came to 240,401 people across 554 Medicare Advantage organizations, tied to $477.5 million⧉ in 2021 CMS payments specifically linked to these diagnosis codes.▸ Statistically extrapolating the sample's 100% failure rate across that frame produced a point estimate of $461,958,186 in potential net overpayments, with a 90% confidence interval running from $442.2 million to $477.5 million -- the upper bound mathematically exceeded the total money in play, so OIG capped it there. That means the estimate implies nearly all of the payments tied to this one diagnosis pattern nationwide may have been improperly paid, not a small slice of them.

This is not OIG's first pass at this exact problem.▸ As of September 2025, OIG had already issued 34 separate audits⧉ of individual Medicare Advantage organizations covering high-risk diagnosis codes, including acute stroke -- and across those, 1,146 of 1,185 enrollees reviewed had unsupported stroke diagnoses. OIG made one recommendation this time: build a prepayment procedure that blocks these codes before CMS ever pays on them, since its existing audits only claw back money after the fact. CMS's written response didn't concur or disagree with that recommendation. CMS instead pointed to its existing RADV audits and a newly phased-in risk-adjustment model as reasons the problem is already being addressed, and said it would "take the report into consideration" going forward. OIG rejected that as sufficient: RADV audits only detect and claw back overpayments after they've already happened, they don't prevent them, and 91 of the 95 diagnosis codes at issue are still used in the new model -- meaning the same pattern can keep generating overpayments under the new system exactly as it did under the old one.