Two of the largest Medicare Advantage fraud settlements ever announced landed in the first quarter of 2026. One was $556 million in January, the largest risk-adjustment recovery the Justice Department has put on record. The other, $117.7 million in March, came from a single national insurer. Both turned on the same allegation: chart review programs that added diagnoses but never removed the ones the medical record didn’t support. For taxpayers, who underwrite Medicare Advantage to the tune of more than half a trillion dollars a year, that pattern is the story. RADV audits, the federal government’s main tool for clawing back overpayments, just shifted from periodic to permanent.
What RADV is, and why it suddenly matters
Medicare Advantage now covers more than half of Medicare’s 67 million enrollees. Plans get paid more for sicker members, based on the diagnoses they submit. Risk Adjustment Data Validation, or RADV, is how CMS checks whether those diagnoses actually show up in the patient’s chart. If they don’t, the plan owes the money back.
For a decade, RADV ran slowly. That’s over. On January 27, 2026, CMS confirmed Payment Year 2020 audits began the following month, with new audits launching about every three months from then on. Sample sizes range from 35 to 200 enrollees per contract. Plans get five months to submit records. AI assists, but final calls stay with human certified coders. The schedule runs through Payment Year 2024.
The OIG audits that made the headlines
In early 2026, the HHS Office of Inspector General published a series of compliance audits on individual Medicare Advantage contracts. The pattern across them was hard to miss.
| Contract audit | Sample finding | Estimated overpayment |
| Southern-state MA contract | 247 of 271 enrollee-years with unsupported codes | ~$7 million projected |
| Midwestern MA contract | 252 of 300 enrollee-years with unsupported codes | ~$4.4 million projected |
| Mid-Atlantic MA contract | 232 of 286 enrollee-years with unsupported codes | ~$830,000 identified |
Error rates in those samples ran from 81 to 91 percent. The biggest drivers were high-paying conditions: lung cancer, pressure ulcers, sepsis, embolisms, colon cancer, acute strokes, and acute heart attacks. The most common failure was simple. History-of conditions, like a stroke years earlier, were coded as active in the audit year, and the chart never caught up.
The shift from coding to compliance
On February 3, 2026, OIG issued the first Medicare Advantage compliance program guidance the agency had published in more than 25 years. It flagged chart reviews, in-home health risk assessments, and prompts inside electronic medical records to increase risk scores as suspect activity. Not banned. Suspect. The guidance reinforced that plans must certify the accuracy of their risk adjustment data, and that submitting or failing to delete inaccurate diagnoses can trigger False Claims Act exposure.
Read alongside the year’s two largest settlements, the message is consistent. A retrospective program that only adds codes, never deletes, and can’t tie diagnoses to real patient encounters is now treated as a compliance failure on its face.
Why this matters beyond the insurance industry
Medicare Advantage spending tops $500 billion a year. Recent estimates from federal advisors put the cost of MA above what equivalent traditional Medicare coverage would cost taxpayers, with a meaningful share of that gap attributed to coding intensity. Every dollar an insurer collects for a diagnosis the medical record doesn’t support is a dollar moved from public funds to private margins. That’s why the Justice Department, OIG, and CMS have all aligned on the same operating principle this year. Every submitted diagnosis must trace to a real encounter. Retrospective programs must work in both directions, adds and deletes. The medical record is the only source of truth.
Patients also have skin in this. Members coded as morbidly obese, diabetic, or post-stroke when they’re none of those things end up with care plans, prior authorizations, and insurance records that follow them around. Inaccurate coding isn’t a victimless paperwork issue.
What comes next
The plans that survive the next four audit years will be the ones that already treat risk adjustment as a clinical and compliance discipline, not a revenue function. That means encounter-grounded documentation, two-way retrospective review that removes unsupported codes as well as adds missed ones, and AI that produces an audit trail every credentialed coder and CMS reviewer can follow.
For everyone else, the question on the table is no longer how much can be captured. It’s whether what’s already been submitted can be defended. After two record settlements and OIG error rates north of 80 percent, that’s a question the public, taxpayers, and policymakers should expect Medicare Advantage to answer.
