Prior Authorization Denial Reasons by Payer

The first mandatory public disclosure of payer prior authorization metrics landed by March 31, 2026, and KFF's analysis of that data, published in August 2026, gives the industry something it never had before: a side-by-side look at how 14 insurers, covering roughly 71 million enrollees across Medicare Advantage, Medicaid managed care, and the ACA marketplace, actually behave when a PA request crosses their desk. The averages alone tell a story. Medicare Advantage insurers denied 12% of standard PA requests in 2025, Medicaid managed care insurers denied 14%, and ACA marketplace insurers denied 18%. But averages flatten the real news, which is the spread. In Medicare Advantage, denial rates ran from 5% at Elevance to 17% at UnitedHealth Group. In Medicaid managed care, the range stretched from 2% at one health plan. Care Health Plan to 23% at Independence Health Group. In the ACA marketplace, GuideWell sat at 3% while Centene posted 25%.

A gap that wide, within the same coverage category, covering similar patient populations and similar covered services, is not a clinical story. It's a policy story. It reflects choices about adjudication method, risk tolerance, and how aggressively a plan wants to manage utilization before it pays a claim. These published rates are aggregated across every item and service a payer covers. Payers aren't yet required to break denials out by service type, so the real exposure on, say, advanced imaging or specialty injectables is almost certainly higher than the topline number suggests. Knowing a payer's aggregate denial rate tells a practice the temperature of the room. Knowing which denial reasons that payer leans on, and why, tells the practice what to wear before walking in.

How frequently each denial reason lands across payers

Six categories account for the large majority of PA denials across commercial, Medicare Advantage, and Medicaid managed care payers, and they appear in roughly the same order everywhere, even as the raw rates shift payer to payer.

Medical necessity not established is among the most commonly cited reasons. Available data suggests it accounts for a significant but not dominant share of denials across major commercial insurers, meaning administrative and coverage-design issues, not clinical disagreement, actually make up the majority of denials industry-wide. When a plan cites medical necessity, the reviewer has concluded the submitted documentation didn't satisfy the plan's own clinical coverage policy. That's a narrower claim than it sounds like: it's not a judgment that the treatment is wrong, it's a judgment that the paperwork didn't prove it was right.

Step therapy, or fail-first, comes next, landing an estimated 20% to 25% of the time. The plan wants documented proof a patient tried and failed a cheaper, preferred option before it'll approve the one the physician actually ordered. Experimental or investigational classification follows at roughly 10% to 12%, and this one deserves attention because payer criteria can lag well behind FDA approval or the relevant clinical society's guidelines, so a request can get denied as "experimental" for a use that's already standard of care in the specialty.

Insufficient clinical information runs about 8% to 10%: the request was technically incomplete, missing chart notes, missing imaging reports, no documented failed-therapy history, or a required field left blank. Out-of-network provider denials represent a smaller share, and those require network adequacy documentation on appeal rather than more clinical evidence. Service not covered under the plan is the least common of these categories, covering treatment excluded outright or classified as non-covered.

Here's the error that costs practices the most time: treating "not medically necessary" and "insufficient clinical information" as the same problem. They often describe the identical underlying gap, a missing piece of documentation, but they call for different appeal strategies. One is a clinical argument. The other is a completeness fix. Conflating them means sending the wrong kind of appeal and burning a review cycle for nothing.

Roughly half of all denials, though, never make it to a clinical dispute at all. They trace back to front-end workflow failures. Eligibility verification errors alone account for about 22% of preventable denials, and Experian Health's 2025 State of Patient Access survey found 56% of providers cite patient information errors as a primary cause. Coding errors sit in this same bucket: an incorrect CPT, HCPCS, or ICD-10 code on the PA request generates a denial that looks clinical on its face but is purely administrative in origin. The 2026 CPT code set added 288 new codes, revised 46, and deleted 84. ICD-10-CM updates effective October 1, 2025, added 487 new diagnosis codes, deleted 28, and revised 38. Any practice that hasn't updated its charge capture and PA submission workflows against those changes is generating technically invalid requests without realizing it, and every one of those shows up in the denial log looking like a clinical rejection.

One structural change helps here. Under CMS-0057-F, effective January 1, 2026, payers covering Medicare Advantage, Medicaid managed care, and ACA marketplace plans now have to give specific written denial reasons instead of generic codes. That alone makes it far easier to map a denial to one of the six categories above and respond to the actual problem instead of guessing.

How UnitedHealth Group's denial behavior differs from other major payers

UnitedHealth Group posted the highest Medicare Advantage standard PA denial rate in the 2025 data at 17%, more than three times Elevance's 5% in the same segment. In the ACA marketplace, UnitedHealth came in at 21%, second only to Centene's 25%.

Part of this traces to AI adjudication. UnitedHealthcare runs AI systems that cross-reference PA requests against internal clinical criteria databases, including InterQual, which is owned by Optum, UnitedHealth's own subsidiary, without a physician advisor reviewing the case on initial determination. The AMA's Prior Authorization Survey found AI-adjudicated denial rates run 40% higher than human-reviewed decisions. That's a significant variance: it means a meaningful share of UnitedHealthcare's denials are coming from an algorithm applying a threshold, not a physician weighing a chart. It means a meaningful share of UnitedHealthcare's denials are coming from an algorithm applying a threshold, not a physician weighing a chart.

Speed cuts the other way for practices, too. UnitedHealthcare's peer-to-peer review window varies by case type, and the timeline can be very short, the window was reduced from 30 days to 14 days from denial date, with certain clinical PA denials requiring a P2P request within 24 hours of the coverage denial. A practice running a weekly denial review cycle, which is still common, will simply miss that window on a regular basis. Nobody looked at the denial until it was already too late to challenge it the fast way.

UnitedHealthcare dropped PA requirements for 231 procedures in December 2025 as part of the broader counterweight to that trend. UnitedHealthcare dropped PA requirements for 231 procedures in December 2025, including nuclear medicine studies, certain obstetrical ultrasounds, and echocardiogram procedures, as part of the broader industry pledge from June 2025. So the picture isn't one-directional: UnitedHealth is both the highest denier in two major segments and one of the more visible actors in walking back PA requirements on specific procedure categories.

For practices dealing with UnitedHealth, the operational read is straightforward: the denial is more likely to arrive fast, thanks to AI adjudication, and the P2P window is tighter than it used to be. That means same-day denial triage, not a weekly queue. Expect medical necessity as the leading denial reason, since AI systems apply InterQual thresholds without a human overlay on first pass.

Centene's denial patterns across Medicaid and the ACA marketplace

Centene posted the highest ACA marketplace standard PA denial rate in the 2025 data, at 25%, and a 15% rate in Medicare Advantage. Those numbers alone would suggest an insurer applying unusually tight coverage criteria on the front end. The appeal data tells a different story.

Centene's Medicare Advantage appeal overturn rate ran at 93%, the highest overturn rate in the entire MA segment. When almost every appealed denial gets reversed, the initial denial is functioning as a filter rather than a clinical judgment. It's functioning as a filter, one that assumes most practices won't bother to appeal. For practices dealing with Centene-administered plans, treating the initial denial as a speed bump rather than a verdict is the accurate read. The appeal process is where the real adjudication happens, and budgeting staff time for it isn't optional, it's the expected path.

Step therapy and medical necessity documentation are the two levers Centene pulls hardest, particularly on specialty medications across both its Medicaid and marketplace lines. Incomplete fail-first documentation shows up repeatedly as the single most correctable reason behind a Centene denial, and a large share of these are fixable with better documentation on the front end rather than requiring a full appeal.

The Medicaid angle is worth a separate note. Centene's ACA marketplace rate of 25% sits well above that same plan's floor. Care's Medicaid managed care floor is 2%, and that gap illustrates something structural: Medicaid managed care plans have wide latitude in how tightly they apply PA criteria, and Centene is at the aggressive end of that range.

How Humana, Aetna, and BCBS plans apply denial criteria differently from each other

Humana's Medicare Advantage denial rate came in at 7% in the 2025 data, below the 12% segment average, which on its face looks favorable. But a lower denial rate doesn't mean the appeal process is straightforward, as expedited timelines for urgent PA requests can compress exactly where speed matters most clinically. A lower denial rate doesn't mean a lower-stakes appeal process.

Aetna runs AI-assisted adjudication as well, alongside UnitedHealthcare and BCBS plans. Its primary denial reasons cluster around medical necessity and step therapy, especially on specialty medications and biologics. Aetna applies MCG criteria (formerly Milliman Care Guidelines) alongside its own Clinical Policy Bulletins and, for certain behavioral health and substance use cases, ASAM criteria, which means the specific criteria set in play depends heavily on the service line in question.

BCBS resists any single description because it operates as a federation of state affiliates, not one national entity. Individual BCBS state plans set their own appeal windows for non-urgent PA denials, and these can vary significantly. That means denial behavior at, say, a BCBS plan in one state tells a practice almost nothing about how a BCBS plan in a neighboring state will behave. Any operational strategy has to be built at the state-affiliate level, not the brand level.

CVS Health, through Aetna and Caremark, posted an 8% Medicare Advantage denial rate in the 2025 data. Its Medicaid managed care denial rate isn't confirmed at 17% in KFF's dataset (that 17% figure belongs to UnitedHealth Group's Medicare Advantage segment, not CVS's Medicaid book, and the two shouldn't be conflated). What is notable about CVS is its 89% overturn rate on appeal in Medicare Advantage, the second-highest in the segment. That's meaningfully lower than Centene's 93% to 95%, which suggests CVS's initial decisions hold up more often when challenged. Practically, that changes the calculus: a practice might reasonably invest more heavily in a strong initial submission to Aetna, since appeal isn't as close to a guaranteed win as it is with Centene.

Elevance posted the lowest rates across the board: 5% in Medicare Advantage, 9% in Medicaid managed care, 10% in the ACA marketplace. Consistency across segments is notable, but a low denial rate doesn't mean frictionless PA. Denial reasons at Elevance still tend to center on documentation completeness, so the volume is lower, but the underlying discipline required to avoid a denial hasn't gone away.

Independence Health Group posted the highest Medicaid managed care denial rate of any insurer in the dataset, at 23%. Practices in its service area should treat Medicaid PA requests as needing meaningfully more documentation up front than they'd budget for with most other payers.

Why AI adjudication and expanded PA lists are driving denials up across all payers in 2026

MBC's denial management analysis found PA denials jumped 31% year-over-year in 2026, and three forces are driving that number at once: AI adjudication expanding across more payers, PA requirement lists growing across more service categories, and practices failing to update their own checklists fast enough to keep pace with either.

On the AI side, UnitedHealthcare, Humana, Aetna, and BCBS plans all now run systems that return decisions in hours instead of the 3 to 5 business days that used to be standard. Speed sounds like a win for patients and practices alike, until the accuracy tradeoff appears in the data: the AMA's 2025 Prior Authorization Survey found AI-adjudicated denial rates run 40% higher than human-reviewed decisions. Faster isn't the same as better, and the gap between the two is visible directly in the denial log.

On the PA list side, every major commercial payer and MA plan expanded its prior authorization requirement list between 2024 and 2026. The categories with the sharpest growth: advanced imaging (MRI, CT, PET), specialty medications like injectables and biologics, outpatient surgical procedures, behavioral health services, durable medical equipment, and home health referrals. A practice that has ordered the same imaging study for years without needing authorization can suddenly find itself denied because the payer quietly added that CPT code to its PA list. Cardiology, orthopedics, pain management, rheumatology, dermatology, gastroenterology, and oncology are facing the sharpest expansion, which means those specialties are absorbing a disproportionate share of the increase.

The 40% AI denial gap isn't fixed. It closes when documentation is written to match the criteria an algorithm is actually scanning for, rather than the clinical narrative a physician would use to explain the same decision to a colleague. That's a mismatch problem, not a quality problem, and it's fixable at the submission stage if practices know what they're submitting into.

The AMA's 2025 physician survey puts a human cost on all of this. Physicians complete an average of 40 PA requests a week, and 32% say their requests are often or always denied. Physicians and their staff spend an average of 13 hours a week on PA work alone. Ninety-five percent say PA delays necessary care, and 79% report patients abandoning treatment because of it. Twenty-six percent report a patient suffered a serious adverse event, including hospitalization, permanent impairment, or death, attributable to PA delays or denials, and 8% specifically report disability, permanent bodily damage, or death. Those are the downstream cost of a system where speed and volume have outpaced accuracy. They're the downstream cost of a system where speed and volume have outpaced accuracy.

What the June 2025 industry pledge and CMS-0057-F changed, and where the gap remains

In June 2025, roughly 50 major insurers, including UnitedHealthcare, Aetna, Cigna, Humana, Elevance Health, and BCBS Association plans, signed a pledge to cut PA requirements, honor existing approvals for 90 days when a patient switches plans, and give clearer denial explanations and appeal instructions. The pledge also set a target: 80% of electronic PA approvals processed in real time by January 1, 2027.

AHIP reported that signatories had eliminated 11% of prior authorizations since the pledge took effect, representing 6.5 million fewer PA requests industry-wide. UnitedHealthcare's removal of PA requirements on 231 procedures in December 2025 stands as one concrete, documented action tied to that commitment.

The AMA sees a different picture from the provider side. Its position is that little has changed for clinicians actually doing the work, and the math backs that up: a 31% year-over-year increase in denials in 2026 runs directly against the pledge's stated direction. Fewer PA requests overall doesn't mean fewer denials on the requests that remain, and the data suggests the opposite is happening.

CMS-0057-F, effective January 1, 2026, is the harder regulatory floor that the voluntary pledge sits on top of. It requires Medicare Advantage, Medicaid managed care, and ACA marketplace plans to decide standard PA requests within 7 calendar days and urgent requests within 72 hours, to provide specific written denial reasons, and to build FHIR-based APIs for electronic PA processing. CMS estimates the rule will save $15 billion over ten years.

The scope limitation restricts which plans the rule covers. CMS-0057-F does not touch commercial employer-sponsored ERISA plans. Self-funded plans are regulated by the Department of Labor and operate under their own plan documents, which means a significant share of commercially insured patients simply aren't covered by any of these new federal protections. The rule also has a timing wrinkle: the faster decision deadlines are enforceable now, but full technical API implementation isn't due until January 1, 2027. The infrastructure for electronic PA is still being built even as the clock on decision speed is already running.

Two more pieces are moving through the pipeline. The CMS WISeR Model, effective January 1, 2026, introduced PA requirements for Traditional Medicare Part B for the first time in six states, covering select high-cost, high-variation services. Practices in those states that never had a Part B PA obligation before now do. And CMS-0062-P, a proposed rule published April 14, 2026, would extend electronic PA requirements to drugs under both the medical and pharmacy benefit, with a proposed target of October 1, 2027. It isn't final yet, but drug-heavy specialties should track it closely.

State legislatures aren't waiting on federal action. At least 20 states passed PA reform laws across 2024 and 2025. Gold card programs, which exempt high-performing providers from PA requirements entirely, now exist in Texas, Arkansas, Wyoming, and West Virginia. New Jersey requires 24-hour decisions on urgent requests and 72 hours on standard ones. Enforcement against Medicare Advantage plans operating in states with gold card laws remains limited, which means the protection on paper doesn't always translate into a protection in practice.

How to sequence the response when a PA denial lands

The most common mistake a practice makes is going straight to a formal written appeal the moment a denial notice arrives, skipping the peer-to-peer review entirely. That's backwards, and it costs time the practice doesn't get back.

A peer-to-peer review puts the ordering physician directly on the phone with the payer's medical director, and it resolves a real share of denials in a single conversation, particularly the ones rooted in incomplete documentation rather than a genuine clinical disagreement. A formal written appeal, by contrast, can take weeks to process and often requires the same clinical justification a P2P call would have delivered in twenty minutes. Filing the written appeal first doesn't just waste the faster option, it can burn the very P2P window some payers, UnitedHealthcare among them, close after a matter of days.

The sequence that holds up: read the denial reason first and identify which of the six categories it actually falls into, medical necessity, step therapy, experimental classification, insufficient information, out-of-network, or non-covered service. That determines whether the fix is a documentation resubmission, a P2P call, or a formal appeal, because each denial type calls for a different response, not a one-size version of "appeal everything." Then check the clock. Every payer runs a different P2P window, and missing it forecloses the fastest path back to a favorable decision, regardless of how strong the clinical case is on paper.

Getting the sequence right doesn't guarantee a reversal. But getting it wrong guarantees a slower one, and in a system where 79% of physicians report patients abandoning treatment because of PA delays, slower is not a neutral outcome. It's the outcome that costs a patient the treatment altogether.