Prior Auth Denial Rates Hit Record Highs in 2026 — Here's What Changed

One in five prior authorization requests is now being denied.

That's not a rounding error. AMA survey data has tracked denial rates climbing steadily for years, and by 2025-2026 the industry crossed a threshold that was unthinkable a decade ago: more than 20% of PA requests rejected on first submission, up from roughly 15% just three years prior.

For a rheumatologist trying to get a patient on a biologic for severe RA, that means one in five requests triggers a denial letter, an appeal, a delay — and sometimes a patient who just gives up.

Why Denials Are Spiking Right Now

The increase isn't random. Three converging forces pushed denial rates to where they are today.

Payer automation scaled faster than anyone expected. Insurers have deployed AI-powered review systems that flag documentation gaps in seconds. What used to require a human reviewer now gets rejected algorithmically before a clinician ever sees it. The speed is the problem — these systems are optimized for throughput, not clinical nuance.

Biologics and specialty drugs dominate the PA queue. The CMS program data shows biologic therapies now account for a disproportionate share of PA volume relative to their prescription volume. These are complex drugs with complex coverage criteria — and payers know providers often struggle to document them completely on the first try.

Step therapy requirements have gotten stricter. "Fail first" policies — where a patient must try and fail cheaper drugs before accessing the one their doctor actually wants to prescribe — have expanded. Documenting that a patient already tried and failed methotrexate, or that they have a contraindication that makes it inappropriate, requires specific clinical detail that isn't always in the default PA request.

The Biologics Problem Is Particularly Acute

Specialty drugs get denied at higher rates than almost any other category. That's partly by design.

Payers have detailed LCD/NCD criteria for biologics — disease activity thresholds, lab value ranges, prior treatment requirements. For a drug like adalimumab or ustekinumab, a complete PA request might need to include DAS28 scores, CRP levels, documentation of DMARD failure, and comorbidity considerations. Miss one element and the automated system flags it as incomplete.

Here's the part that should make you angry: research published in Health Affairs found that the majority of initial PA denials are overturned on appeal. The treatment wasn't wrong. The documentation was incomplete. The patient still went weeks or months without their medication while the paperwork caught up.

That's not a clinical problem. That's an administrative one — and administrative problems are fixable.

The Documentation Quality Gap

Ask any rheumatologist or dermatologist what happens to their PA requests and they'll describe the same pattern: initial denial, rushed appeal, eventual approval. The clinical case was solid from day one. The documentation just didn't survive contact with the payer's intake system.

A few specific gaps show up constantly:

Missing or vague prior treatment history. "Patient tried methotrexate" doesn't cut it. Payers want doses, durations, and documented failure reasons. An incomplete record triggers a denial even when the underlying history is real.

Lab values without context. A CRP of 28 mg/L means something different in a patient with RA that's been undertreated for two years versus one who just had surgery. Payers increasingly require that clinical data be framed within the patient's disease course, not just reported as isolated numbers.

Outdated payer-specific criteria. Each insurer publishes their own clinical criteria for each biologic — and those criteria get updated regularly. What satisfied Aetna's criteria in 2024 might fall short in 2026. Keeping up with those changes manually is unrealistic for most practices.

Generic letter templates. Many practices still use the same boilerplate PA letter for every request, swapping in patient names and drug names. Payers have seen these templates thousands of times. An automated review system isn't fooled by formatting — it's looking for specific data fields.

What Actually Moves the Needle

There's no single fix, but the practices that hold denial rates well below 20% tend to do a few things differently.

They treat the first submission as the only submission. Every element the payer's criteria requires gets addressed upfront, not after the denial letter arrives. This sounds obvious, but it requires knowing exactly what each payer requires for each specific drug — which changes constantly.

They document disease severity with precision. Validated scoring tools (DAS28, PASI, CDAI) give payers the clinical anchors they're looking for. "Severe disease activity" backed by a score is fundamentally different documentation than "patient has significant symptoms."

They keep prior treatment records current and detailed. Not just drug names, but doses, duration, response, and reason for discontinuation. If a patient couldn't tolerate a drug due to side effects, that's documented — not just implied by the switch to something else.

They use the payer's language, not their own. Every payer's clinical policy document describes what they're looking for, down to specific phrases. A well-constructed PA request mirrors that language back.

This is genuinely a lot of work per request. A biologic PA for a Medicare Advantage patient can easily require 45-60 minutes of documentation time done properly. At high-volume practices, that load falls on staff who are already stretched thin.

Where AI Tools Fit In

This is where tools like purpose-built AI platforms have started making a real difference — not by replacing clinical judgment, but by making sure that judgment actually shows up in the documentation.

The core problem is translation: a clinician knows exactly why a patient needs a biologic. Getting that knowledge into the specific format a payer's system requires is a documentation task, not a medical one. AI can handle the translation layer — pulling the relevant clinical data, mapping it to payer-specific criteria, and generating structured documentation that addresses every required element.

The outcome isn't magic. It's just that the first submission actually contains what the payer is looking for, which means fewer denials and fewer appeals. For specialty practices dealing with dozens of biologics PAs per week, that's the difference between a sustainable workflow and one that's burning out staff.

The Bigger Picture

Denial rates won't fix themselves. Payer automation will only get more sophisticated, and biologic PA criteria will keep expanding as new drugs come to market.

The practices that fare best won't be the ones who get better at writing appeals. They'll be the ones who stop generating denials in the first place — by treating documentation quality as a clinical workflow, not an afterthought.

The data is clear on one thing: most of these denials were preventable. That's actually an optimistic finding, if you're willing to act on it.

Sources:
American Medical Association. (2025). 2025 AMA Prior Authorization Physician Survey. ama-assn.org
Centers for Medicare & Medicaid Services. (2025). CMS Program Statistics — Medicare Part D. cms.gov
Dusetzina, S.B., et al. (2023). Prior Authorization Denials and Appeals for Specialty Drugs. Health Affairs. healthaffairs.org
American Journal of Managed Care. (2024). Step Therapy and Access to Biologic Therapies in Inflammatory Diseases. ajmc.com