The Insurance Denial That Lost on Appeal Most of the Time and Won Anyway

Most prior authorization denials are wrong. Insurers know it. Patients don't act on it. That gap is the entire business model.

In August 2026, the health policy research group KFF published the first public analysis of prior authorization denial data, made available after a new CMS rule required insurers to disclose it. The numbers are stark. Across Medicare Advantage, Medicaid managed care, and ACA marketplace plans, insurers denied at least one in eight standard prior authorization requests last year. Denial rates varied enormously by insurer, from 2% to 25% for functionally similar coverage.

Here's the number that reframes everything else: when patients actually appeal a denial, somewhere between 43% and 82% get overturned, depending on the analysis and plan type. The treatment was medically appropriate the whole time. The insurer's own appeals process says so, in a majority of cases, once someone bothers to ask twice.

Almost nobody asks twice. Fewer than 11% of denied patients ever file an appeal.

Insurers pledged last year to ease prior authorization requirements and report having cut authorization volume by 11% as of this spring. That number describes fewer requests being flagged for review in the first place. It says nothing about what happens to the requests that are still denied, and the appeal data suggests the answer is: mostly nothing, because most people never appeal, regardless of their odds.

This isn't one insurer's process problem. KFF's analysis covered the largest carriers across three major coverage categories and found the same structural pattern in all of them: meaningful denial rates, wide variation between insurers offering comparable coverage, and a documented gap between how often a denial is wrong and how often anyone challenges it.

The regulatory response has been substantial precisely because the pattern is systemic. A CMS rule finalized in January 2024 now requires faster decisions, standard requests must be answered in seven days instead of fourteen, urgent ones within 72 hours, specific written reason codes tied to clinical criteria for every denial, and public reporting of approval, denial, and appeal outcomes. Multiple states have passed their own prior authorization reforms in 2026. None of this activity happens in response to an isolated bad actor. It happens when regulators conclude an entire market is running the same play.

The mechanism here isn't really about clinical judgment. It's about which outcome requires effort and which one doesn't.

Default effect: When one outcome occurs automatically unless a person takes deliberate action to change it, that outcome becomes far more likely to persist, independent of whether it's the outcome the person would actually choose if the choice were made with no friction attached. The default isn't neutral. It's a decision that's already been made on the person's behalf.

A denial is the default. Approval, once denied, requires the patient to notice the denial, understand that it's often wrong, gather documentation, and formally contest it, usually while managing an actual medical condition at the same time. Nothing about that process is automatic. Every step depends on the patient doing something extra. Meanwhile, doing nothing costs the insurer nothing and, in the aggregate, saves the insurer money, since a meaningful share of people who would have won on appeal simply never file one.

That gap between the denial and the appeal isn't accidental friction. It's structural, and behavioral economics has a specific term for friction that makes a beneficial action harder to take, whether or not anyone designed it that way on purpose.

Sludge: Friction embedded in a process that makes a beneficial action more difficult, effortful, or confusing to complete than it needs to be, discouraging people from taking an action that would otherwise be in their interest, even when the option is technically available to them.

An appeals process that exists on paper but requires forms, deadlines, and clinical documentation most patients don't know how to assemble is sludge, regardless of whether any single insurer built it deliberately to suppress appeals or simply never invested in making it easier. The effect on outcomes is identical either way. A process that is technically available but practically unusable produces the same result as a process that doesn't exist: most people don't use it, and the default holds.

There's a specific accountability mechanism that could show whether this burden falls unevenly on people least equipped to fight it, and it was turned off. A 2025 CMS rule would have required Medicare Advantage insurers to publish health equity analyses breaking down denial and approval rates by population, including dually eligible enrollees, low-income beneficiaries, and people with disabilities. The current administration announced in June 2025 that it would not enforce that requirement.

The practical effect is that the populations most likely to lack the time, health literacy, or resources to navigate an appeals process, exactly the population sludge burdens hardest, are also the population for which there is no longer a public mechanism to even measure the disparity. It's not that the equity gap has been shown not to exist. It's that the specific tool built to detect it has been shelved, which means the system is now structurally unable to tell you whether the friction it created is landing evenly or not.

That's a trust problem layered on top of a design problem. Patients can't verify whether a denial was reasonable. Now, in a meaningful segment of the market, nobody outside the insurer can verify whether the burden of that uncertainty is distributed fairly either.

The CMS rule addresses speed and disclosure. It does relatively little, on its own, to address the sludge sitting between a wrong denial and the appeal that would fix it.

• Make appeal the default response to a denial, not a separate action the patient has to initiate. A number of state reforms already require automatic escalation or independent review for denials above a certain threshold. Removing the requirement that the patient personally trigger the correction is the single highest-leverage fix available, because it attacks the actual mechanism, not just the visibility of it.

• Report overturn rates alongside denial rates, prominently, not buried in a separate dataset. A denial rate without its corresponding overturn rate tells a patient nothing about how likely the decision is to be wrong. Pairing the two numbers publicly, plan by plan, gives patients and regulators the one comparison that actually reveals how reliable a given insurer's first decision is.

• Restore, or independently replicate, the equity reporting requirement that was suspended. If a federal agency won't enforce it, state regulators, patient advocacy organizations, or researchers with access to the underlying data can build the same analysis. The absence of an official requirement doesn't have to mean the absence of the data.

• Simplify the appeal process itself, not just the decision timeline. Faster denials that remain equally hard to contest don't reduce sludge. They just deliver the same friction on a shorter clock. A meaningful fix targets the number of steps and the documentation burden in the appeals process directly, not only how quickly the initial denial arrives.

• Automatic appeal escalation has a real cost that someone has to absorb. Independent review at scale requires staffing and infrastructure insurers, and in public programs, taxpayers, would need to fund. The 11% reduction in prior authorization volume that insurers are touting is partly a response to cost pressure. A fix that increases review volume runs directly against that same pressure.

• Publishing overturn rates assumes patients will actually look, and most currently don't even file the appeal that's already available to them. A more visible statistic doesn't automatically translate into different patient behavior if the underlying sludge in the appeals process itself remains unchanged. Transparency addresses trust. It doesn't, by itself, address friction.

• Equity data collected outside an official federal requirement carries less enforcement weight. A researcher-built or state-built version of the suspended analysis can document the disparity, but without the regulatory teeth of the original CMS rule, documentation doesn't automatically translate into a policy consequence for the insurers it identifies.

• Faster decision timelines can create their own denial pressure. A seven-day standard review, down from fourteen, gives reviewers less time to gather complete clinical information before deciding, which could plausibly increase erroneous denials even as it satisfies the letter of the reform. Speed and accuracy aren't guaranteed to move in the same direction.

If any process in your organization relies on the person affected to notice something went wrong and personally initiate the fix, genuine question: have you ever measured how often that default outcome would have been overturned, if only someone had asked twice?