OIG: $462M in Potential MA Overpayments Tied to Unsupported Acute Stroke Codes

On May 28, 2026, HHS Office of Inspector General issued report A-02-23-01020, estimating $462 million in potential net overpayments to Medicare Advantage organizations based on acute stroke diagnosis codes submitted on physician data records without a corresponding hospital record in the same service year. The report covered payment year 2021 and was publicly posted June 1, 2026.

What OIG found — and the 100% sample error rate

OIG reviewed a sample of 97 MA enrollees whose acute stroke codes appeared on physician data records. Every single one was unsupported under OIG's definition.

OIG's definition is precise and matters for your desk: a diagnosis is unsupported when the acute stroke code appears on a physician data record but there is no acute stroke diagnosis on an inpatient or outpatient hospital data record during the same service year. The code is not being challenged because the physician's note is bad. It's being challenged because an acute stroke, an event requiring emergency or inpatient care, has no hospital record to confirm it happened.

OIG extrapolated from the sample to estimate the $462 million figure. That is a statistical projection, not an adjudicated recovery and not a fraud finding. OIG characterizes these as unsupported codes and recommends a preventive control; it does not allege intentional misconduct.

The I63 / I69 / Z86.73 fault line — the table you need

This is where coders get the code wrong in both directions. The three code groups each have a specific clinical and documentation meaning, and using the wrong one is what OIG's audit surfaced at scale.

The coding principle is clear: I63.x codes are for acute events that should, in practice, generate a hospital record in the same year. I63.9 (cerebral infarction, unspecified), I63.10 (cerebral infarction due to embolism of unspecified precerebral artery), and I63.50 (cerebral infarction due to unspecified occlusion or stenosis of unspecified cerebral artery) are all valid billable codes within the category, but they require the same acute event documentation. The I63 category header itself is not billable.

History-of codes like Z86.73 and sequelae codes like I69.30 (unspecified sequelae of cerebral infarction) or I69.398 (other sequelae of cerebral infarction) do not drive the same payment weight that acute stroke codes carry, which is precisely the incentive OIG's methodology targets. Coding a past stroke with no residuals as I63.x inflates the RAF and is the pattern this report describes.

The coder's rule of thumb: if the provider's note is updating or monitoring a stroke history from a prior year, I63.x is the wrong code. Z86.73 or the appropriate I69.x is what the documentation supports.

The prepayment control OIG recommended — and what CMS said

OIG's single recommendation: CMS should implement a procedure to prevent overpayments when acute stroke codes are submitted on physician data records and the enrollee has no acute stroke diagnosis on an inpatient or outpatient hospital data record for the same service year.

The logic is structural. An acute stroke without a hospital record is, clinically, almost always a documentation mismatch rather than an accurate code. Blocking payment before it goes out is more efficient than recovering it afterward.

CMS did not specify concurrence or nonconcurrence with the recommendation. CMS said it would take the report into consideration.

What that means practically: the control is not in place yet. But OIG recommending a prepayment edit is a signal that CMS is likely to look at this category in the next round of edits. Physician claims carrying I63.x without a same-year hospital record are exactly what such an edit would catch.

OIG has been here before — the 2020 precedent

This is not OIG's first time publishing on this exact pattern. A 2020 OIG report (A-07-17-01176) flagged the same documentation gap and found $14.4 million in increased payments from 580 of 582 sampled enrollees whose acute stroke codes were not supported by the medical records.

The 2026 report is larger in scope and dollar estimate but rests on the same finding. OIG has now flagged this specific gap, physician-only acute stroke codes with no same-year hospital record, across two reports spanning six years. Plans that absorbed the 2020 report and made no changes to their coding review process are in the same position today, at a much larger estimated scale.

The Mintz analysis of the 2026 report notes that OIG's focus on prevention, the prepayment control recommendation rather than a recovery demand, reflects a shift toward front-end controls in MA risk adjustment program integrity.

What this means for your queue

Every I63.x in your current queue that lives only in a physician office note is the next prepayment-edit target. That's not speculation — it's what OIG found at a 100% rate in payment year 2021 and is recommending CMS block before payment.

The documentation check is straightforward: for any I63.x code, there should be a same-year inpatient or outpatient hospital record confirming the acute stroke event. If the note is from a primary care or specialist office visit where the provider is documenting a past stroke history, that code should not be I63.x.

Check your I63.x submissions against the decision table above. For past strokes without residuals, Z86.73 is the right code. For past strokes with documented, specific residual deficits, the appropriate I69.x subcategory is the right code. Both directions are correctness issues, not just payment issues — overcoding on I63.x and undercoding documented sequelae both create problems.

For documentation review, the MEAT Criteria Guide covers what constitutes evidence a condition was addressed at the DOS, which applies to both acute and chronic conditions in your review queue. The ICD-10 Encoder lets you confirm billable status and specificity within I63.x, I69.x, and Z86.73 so you're coding to the right level. For documentation that needs to hold through a RADV review, see the Evidence Checker on what a record actually needs to show.

The broader RADV context matters here too. OIG's recommendation for prepayment controls is separate from RADV, but the documentation standard is the same: a code needs a record to back it, and for an acute event, that record is a hospital record. See the RADV quarterly audit article for what CMS's expanded audit schedule means for your plan's submission risk.