Prior authorization denials are overturned four times in five on appeal — and nearly nine in ten are never appealed

What is happening?

Medicare Advantage insurers made 52.8 million prior authorization determinations in 2024, against 49.8 million requests in 2023. Insurers denied 4.1 million of them in full or in part — 7.7 percent, a larger share than the 6.4 percent denied in 2023.

Only 11.5 percent of those denials were appealed. Of the appeals that were filed, 80.7 percent were partially or fully overturned.

Those two figures are published as shares rather than counts. Applied to the 4.1 million denials, they give roughly 470,000 appeals filed during 2024, of which roughly 380,000 fell. Roughly 3.6 million denials were never tested at all.

Physicians describe what the delay does. In a survey of 1,000 practicing physicians fielded in December 2025, 95 percent said prior authorization delays access to necessary care, 79 percent said patients abandon a recommended course of treatment because of it, and 26 percent said it has led to a serious adverse event for a patient in their care.

Whose problem is this?

The party that judges whether care is necessary is the same party that would pay for it, and the party that waits is neither of them.

Where does this problem end?

The boundary matters because this is not a coverage limit. A denial here refuses care that the plan otherwise covers, on the ground that it has not been shown to be necessary.

What is the state now, and what should it be?

Now

What is being aimed at. The federal interoperability and prior authorization rule sets deadlines for speed. From January 1, 2026, impacted payers must decide expedited requests within 72 hours and standard requests within seven calendar days, and must state a specific reason for every denial. The voluntary industry pledge adds format targets for January 1, 2027 — standardized electronic submission and expanded real-time responses.

Needs a new measurementthe target state for accuracy is missing. No source opened in this round publishes a target for what share of requests should be denied, or for what share of denials should survive review. Every announced target governs speed and format, not whether the decision is right.

How big is it?

The figure below covers Medicare Advantage only. Commercial and Medicaid managed care apply the same process and are not counted here, so it is a floor on the phenomenon rather than a measure of it.

The affected population is 16.8 million to 33 million people, as of 2024 to 2025.

The upper anchor is every Medicare Advantage enrollee. There were 33 million of them in 2024, and enrollment carries the requirement whether or not a request happens to be filed in a given year.

The lower anchor applies a survey rate to that enrollment. A national health policy tracking poll found in July 2025 that 51 percent of insured adults had needed prior authorization from their insurer in the past two years. Applying that share to 33 million enrollees gives 16.83 million.

Two mismatches sit inside the lower anchor and they pull in opposite directions. The 51 percent was measured across all insured adults rather than Medicare Advantage enrollees, and Medicare Advantage is the line of business physicians rate heaviest for prior authorization, at 69 percent calling the burden high or extremely high, so the transfer probably understates. Against that, the poll asks about two years while the enrollment figure is a single year, which pushes the derived number up.

A sharper count is not available. 4.1 million requests were denied in 2024, but one enrollee can be denied more than once, so the number of distinct people denied is at or below 4.1 million and cannot be narrowed further from what is published.

Under what conditions does it arise?

1. The reviewer is the payer. The health plan decides whether a service is medically necessary and also keeps the premium it does not spend on that service. No third party sits between the judgment and the money. 2. Challenging a denial costs the challenger, not the denier. 59 percent of physicians say they do not appeal because they do not expect the appeal to succeed, 52 percent cite insufficient staff time, and 49 percent say patient care cannot wait for the plan to decide again. 3. Volume outruns individual scrutiny. 52.8 million determinations landed in one line of business in a single year. Only 24 percent of physicians agree that denials based on medical necessity are reviewed by a licensed and qualified clinician, and among those who take part in peer-to-peer review, only 16 percent say the reviewer for the plan often or always holds the appropriate qualifications. 4. The harm does not appear in the reported numbers. What is published is counts and rates. Abandoned treatment, the extra office visit, and the emergency room visit that follows a delay are recorded nowhere in the determination data.

The decision to deny is made by the party that keeps the money it does not spend.

What has been tried?

Two of these five are more than seven years apart and cover much of the same ground. The 2018 consensus statement already promised selective application, transparency, and continuity of care, and those same promises reappear in the 2025 pledge.

What was found?

Why is it still unsolved?

Incentive inversion — the party that decides whether care is necessary is the party that keeps the money when the answer is no.

Denials are overturned four times in five when a physician appeals them. That figure is not evidence that the appeal process works. Among the denials that are appealed, the initial decision does not survive review four times out of five, and appealed denials are only about one in nine of the total, so the remaining eight are never tested by anyone. What this dossier treats as the failure is the design — the party that reviews a request also pays for the care it approves — and not the conduct or the intent of any particular insurer.

A denial is cheap for the plan that issues it and expensive for everyone else. Filing an appeal costs a practice staff time it does not have, and most physicians do not expect to win, so the appeal rate holds near one in nine. That is the mechanism in one sentence: the process is profitable in proportion to how rarely it is contested, and it is contested rarely because contesting it is costly for the side that would have to do it.

The reforms that have landed act on speed and on format rather than on that arithmetic. The federal rule sets a clock and demands a reason. The industry pledge standardizes submission and shortens the list of services needing approval. Neither changes who benefits from a denial, and neither makes the cost of a wrong denial land on the party that issued it. The 2018 consensus statement made much the same promises and, on the evidence of the December 2025 survey, changed little.

What observation would mean it is solved?

Candidates — (a) the denial rate falls from 7.7 percent, (b) the overturn rate on appeal falls from 80.7 percent, (c) the appeal rate rises from 11.5 percent.

None of the three works alone. (a) falls when a plan drops rarely denied services from the approval list while keeping the expensive ones, which moves the ratio without changing anything for the patient who is denied. (b) falls when fewer marginal denials get appealed, and it reaches zero when nobody appeals at all. (c) rises when denials rise, so a higher appeal rate can equally mean the process is getting worse and being fought harder.

The measure that would actually settle it is in none of the three. It is the share of denials that turn out to be correct when tested by a review the plan does not control, applied whether or not the patient thinks to ask for it. The federal rule does require payers to publish their own prior authorization metrics, on a start date the available summaries disagree about, and that is a step toward visibility. It still leaves the plan as the source of its own scorecard.

What is it connected to?

Fills with researchplausible links to medical debt, physician burnout and workforce exit, hospital administrative cost, and the Medicare Advantage payment model itself. Relation type and evidence grade were not established in this round.

What these sources do not say

• How many people, as against how many requests. Everything published counts determinations and denials. 4.1 million denials could be 4.1 million people or a far smaller number denied repeatedly, and nothing opened here resolves it.
• What was denied. The totals are not broken out by service type, so there is no telling whether the reversals concentrate in post-acute care, imaging, drugs, or something else. That matters because a denial of a rehabilitation stay and a denial of a repeat scan are not the same event.
• What became of the 3.6 million denials nobody appealed. No source follows those patients to find out whether the care was eventually delivered, paid for privately, or never given.
• Whether the 11 percent reduction in preapprovals removed anything that mattered. That number counts requirements withdrawn, not care restored, and nothing says whether those requirements were rarely denied in the first place.
• What commercial insurance looks like. The 52.8 million figure exists because Medicare Advantage plans must report to CMS. Employer and individual market plans publish nothing comparable, so the largest insured population in the country is simply uncounted here.
• Whether automated review is driving denials. 60 percent of physicians say they are concerned that augmented intelligence increases or will increase denial rates, but that is a perception measure and no source opened in this round quantified automated denial volume.

See the evidence

The physician survey report was read directly as a PDF and its figures are taken from that document rather than from coverage of it. Its headline numbers were independently restated by a hospital association news service on 2026-05-14, matching on 95 percent, 26 percent, 92 percent, 32 percent and 88 percent. The determination analysis and the tracking poll were both read directly on the site of the research organization that published them.

The federal rule text was read directly, but only in part. The rule as printed in the Federal Register opened on govinfo.gov and confirms four things: that this is CMS-0057-F, that impacted payers must give a specific reason when they deny a prior authorization request, that impacted payers must publicly report metrics about their prior authorization processes, and that the compliance date is January 1, 2026 for Medicare Advantage organizations and state Medicaid and CHIP fee-for-service programs. Three other government routes failed on the same day — cms.gov returned HTTP 403, and federalregister.gov and ecfr.gov both redirected to a block page. The numeric decision deadlines were not in the portion of the rule text that could be retrieved, so the figures of 72 hours expedited and seven calendar days standard rest on three compliance-vendor summaries rather than on the rule itself. Those three agree on those deadlines, on the requirement to give a specific reason for each denial, and on the January 1, 2027 date for the data interfaces. They disagree on two points, both recorded above: two place the first public metrics report at 2026-03-31 covering calendar 2025 while the third places public reporting in 2027, and two count four required interfaces while the third counts five. A ten-year savings estimate appeared in only one of the three and is therefore not reported in this dossier. One further caution about the rule text: the passage that could be retrieved from govinfo.gov quotes decision timeframes that were already in force under 42 CFR 422.568 and 422.572, which are not the deadlines reported in this dossier and are not used here.

One arithmetic caution. The 52.8 million determination count and the 33 million enrollment figure are not divided against each other anywhere above, because the published determination analysis computes requests per enrollee on a narrower base and reports 1.7 for 2024. The appeal counts in the opening block are derived from the published shares and are marked as derived.

People affected

Estimated range 16,830,000–33,000,000 As of 2024~2025

Sensitivity The range is not a confidence interval. It is the distance between two definitions of affected: everyone enrolled in a plan that applies prior authorization, and the smaller group that actually went through the process. The two known mismatches in the lower bound pull in opposite directions and neither magnitude is known. The 51 percent rate was measured across all insured adults, while Medicare Advantage enrollees are older, use more care, and belong to the line of business physicians rate heaviest for prior authorization, with 69 percent calling the burden high or extremely high in the December 2025 physician survey, so the transfer probably understates. Against that, the poll asks about the past two years while the enrollment anchor is a single year, which pushes the derived figure up. A sharper number is not available from published data: 4.1 million requests were denied in full or in part in 2024, but one enrollee can be denied more than once, so the count of distinct people denied is at or below 4.1 million and cannot be narrowed further. Commercial and Medicaid managed care enrollees face the same process and are not inside this range at all, because no comparable public count exists for them.

Regional breakdown The published determination analysis reports prior authorization volumes and denial rates nationally and by insurer, not by state. Medicare Advantage penetration varies widely across states, but the source gives no state-level determination or denial counts, so a state breakdown cannot be measured. Splitting the national count by state population would be proportional allocation with no basis in the data.