Read a number wrong and you will draw exactly the comforting conclusion the number is designed to prevent.

Two reports from the HHS Office of Inspector General, released June 11, 2026, found that when Medicare Advantage enrollees appealed denials of skilled nursing care, insurers reversed themselves 95% of the time. One large plan received 42% of all such appeals and overturned 99.7% of them. Where a UnitedHealth contractor called naviHealth had issued the denial, the reversal rate was 97%.

The intuitive reading is reassuring. The appeals process works. People who were wrongly denied fought back and won. The system self-corrects.

That reading is exactly backwards, and the OIG said so in the driest possible language. A near-total overturn rate does not show the appeals process working. It shows that the original denials should almost never have been issued. If 95% of contested denials collapse the moment a human being reviews them, the denials were not close calls. They were wrong on arrival, and something other than medical judgment was producing them.

The number that makes it a business model

Here is the figure that turns a quality-control story into a profit story, and it is the one to build everything else around.

Only 18% of skilled nursing denials were appealed. For long-term care hospitals it was 36%, for inpatient rehab 31%. Which means the overwhelming majority of denied patients never challenged the decision at all.

Now hold the two numbers together. Nearly every denial that gets appealed is overturned. Most denials never get appealed. The unavoidable arithmetic is that a large number of people who were entitled to care did not get it, did not fight, and simply went without, paid out of pocket, or stayed stuck in a more expensive setting.

That is not a malfunctioning system. That is a system with a highly favorable ratio. If you deny broadly, reverse the fraction of patients with the time, energy, and health to appeal, and keep the savings on everyone else, you come out ahead. The 5% you never reverse and the 82% who never appeal are the margin. The appeals process is not a check on the denials. It is the release valve that makes the denials survivable enough to keep issuing.

These are, by definition, the people least able to fight. Post-acute care is what you need after a stroke, a serious injury, major surgery. The patient is 70, weak, frightened, often cognitively compromised, discharged from a hospital and told the nursing facility their doctor ordered is not covered. The design does not have to be malicious to be effective. It only has to know that a sick elderly person is unlikely to file paperwork.

The AI angle nobody should skip

The denials were not all coming from doctors.

naviHealth, a UnitedHealth subsidiary that uses AI to assess care needs, processed about half of all skilled nursing requests and denied 14% of them. That is higher than the plans that reviewed requests internally (11%) and higher than other contractors (9%). Then the parent insurers overturned 97% of naviHealth's denials on appeal.

Sit with that internal contradiction. UnitedHealth pays a subsidiary to deny care using an algorithm, and then, when patients push back, UnitedHealth overrules its own subsidiary 97 times out of 100. The company is, in effect, publicly certifying that its own denial engine is wrong almost every time anyone checks its work. naviHealth was the subject of a STAT investigation three years ago that found its denials often led to poor outcomes for gravely ill patients. The OIG has now supplied the number that closes the case: the insurers themselves do not trust the denials, because they reverse them the instant a human looks.

The AI here is not a neutral efficiency tool. It is a mechanism for producing denials at a volume no human review staff could match, calibrated to a business that profits when most of them stick.

Why an insurer would reverse itself at 99.7%

There is a subtler reading of the overturn rate, and KFF flagged it, and it makes the picture worse rather than better.

Why would a plan overturn a denial at 99.7%? Not necessarily because it suddenly agreed the care was needed. Because of what happens if it does not. If a plan holds firm and the enrollee keeps appealing, the case eventually reaches an independent review entity. If that outside reviewer overrules the plan, it counts against the plan's star ratings, and star ratings drive quality bonus payments.

So the rational move is to deny first, and then, for the small share of patients stubborn enough to appeal, fold before the case escapes the plan's control. Reverse it in-house and the loss is one instance of care. Let it reach the independent reviewer and lose there, and the damage hits the bonus payments that are worth billions. The 99.7% is not a conscience. It is a containment strategy.

This connects the denial machine to the other lever of Medicare Advantage profitability. Plans are paid a fixed amount per enrollee in advance. That figure does not rise if the plan spends more than expected, and does not fall if it spends less. Every denied rehab stay that stays denied is money the plan keeps. Every appeal reversed quietly in-house is a threat to the star ratings neutralized before it can cost real money. The incentives point one direction.

The delay is a cost even when the patient wins

Winning the appeal is not the same as being made whole.

The OIG found the initial denial delayed post-acute care by five to six days on average. For a stroke patient, the window in which intensive rehabilitation does the most good is measured in exactly those days. A delay is not a neutral pause. It is lost recovery, sometimes permanently.

And the delay carries a bill. Many Medicare Advantage enrollees face daily cost-sharing for hospital stays, so a patient stuck in a hospital bed waiting out a wrongful denial can be accumulating out-of-pocket costs for the privilege of being in the wrong, more expensive setting. Some appeals dragged badly: 16% of long-term care hospital appeals took ten days or more. The patient who eventually wins has already paid, in money and in recovery, for a denial that never should have happened.

The denial rates themselves are the other scandal

The overturn story can overshadow how extreme the front-end denials were.

Across the highest-cost post-acute settings, Medicare Advantage plans denied 65% of long-term care hospital requests and 54% of inpatient rehab requests. KFF's prior work put the overall Medicare Advantage denial rate across all services below 8%. So denials for the most expensive rehab care ran roughly seven to eight times the baseline. The plans were not applying uniform scrutiny. They were concentrating their refusals precisely where the care costs the most, which is what you would expect if cost, not medical necessity, were the operative variable.

The three largest insurers, UnitedHealth, Humana, and CVS Health, denied at higher rates than most peers. For-profit plans denied more than not-for-profits. Inpatient rehab overturn rates ranged wildly across plans, from 14% to 86%, which tells you the "right" answer to these requests is not some contested clinical mystery. It is being decided differently by different companies, and the companies denying hardest are reversing hardest, because they were denying things they knew would not hold up.

What the industry said, and what it leaves out

AHIP, the insurance lobby, responded that the reports fail to account for wide variation in the cost and quality of post-acute care and skilled nursing facilities, and said insurers remain committed to improving prior authorization.

There is a real point buried in there. Post-acute care quality genuinely varies, some skilled nursing facilities are poor, and prior authorization does catch low-value care. Providers do sometimes submit incomplete documentation that draws an automatic denial. All true. But none of it survives the central number. If the denials were tracking genuine quality or documentation problems, they would not collapse 95% of the time on appeal. A denial grounded in a real clinical or quality concern holds up when a reviewer examines it. A denial that evaporates the moment it is contested was never about quality. It was about seeing who would push back.

The OIG's own framing is the cleanest verdict available. Plans that inappropriately deny care are not delivering the value taxpayers pay them for. Medicare Advantage now covers more than half of all Medicare beneficiaries, roughly 35 million people heading toward 45 million.

The denial-and-reverse pattern is not a glitch in that system. On the evidence in these two reports, it is one of the ways the system makes its money.

Further reading