Medical Economics Expert Panel 12.2
News|Articles|November 20, 2026

Fewer patients abandon care over prior auth delays, but provider trust in AI is wearing thin

Author(s)Todd Shryock
Fact checked by: Chris Mazzolini
Listen
0:00 / 0:00

Key Takeaways

  • Frequent treatment abandonment due to authorization delays fell from 55% to 28% among clinicians, while repeated avoidable ED/hospital care reports dropped from 47% to 16%, suggesting fewer severe downstream consequences.
  • Administrative load intensified, with moderate/heavy burden rising to 51% of clinicians and 61% of administrators; missing clinical documentation became the leading cited denial driver.
SHOW MORE

A survey finds that delay-related harm fell sharply. Strong trust in AI also collapsed, and clinicians now want strict limits and human oversight.

A patient needs an MRI to rule out a torn rotator cuff. The order goes in on Monday. By Thursday, the request is still pending because the health plan wants documentation the practice didn't know it needed. By the next week, the patient has stopped calling.

That kind of quiet abandonment has long been one of the most troubling costs of prior authorization. New survey data suggest it is happening less often. The same data show that the AI tools health plans are adding to the process face a tougher audience than they did a year ago, and that the paperwork behind each request hasn't gotten any lighter.

Cohere Health, a Boston-based company that sells clinical intelligence and prior authorization technology to health plans, released its second annual National Provider Survey on Prior Authorization Nov. 16. Wakefield Research conducted the survey online Aug. 12-27, 2026, among 100 clinicians and 100 office administrators at U.S. hospitals and medical systems, and compared the results with a similar survey from April 2025.

A few details matter for how physicians read the results. Physicians and surgeons made up 32% of the clinician sample, behind nurses at 40%, and physician assistants and nurse practitioners made up most of the rest. Each wave surveyed different people, and the margin of error within each group is about plus or minus 10 percentage points.

Patient harm eased

The clearest good news is about patients. The share of clinicians who said patients abandon recommended treatment "often or all the time" because of authorization delays fell from 55% to 28%. Among office administrators, it fell from 55% to 26%.

Reports of delays causing avoidable emergency or hospital care fell even further. Clinicians who said they had seen this happen repeatedly dropped from 47% to 16%, and administrators from 56% to 10%.

"Fewer patients are abandoning care or ending up in the emergency room," Brian Covino, M.D., Cohere's chief medical officer, said in a statement.

Starting in January 2026, federal rules required Medicare Advantage, Medicaid and Children's Health Insurance Program plans to decide standard requests within seven days and urgent requests within 72 hours, give a specific reason for every denial, and publicly report prior authorization metrics. Voluntary insurer commitments made in June 2025 also began taking effect this year.

How this compares with the AMA results

Physicians may find these numbers more hopeful than their own experience. The American Medical Association's (AMA's) 2025 Prior Authorization Physician Survey, which polled 1,000 practicing physicians in December 2025, found that 79% reported patients abandoning recommended treatment because of prior authorization challenges.

The two surveys measure different things. The AMA figure counts physicians who said abandonment happens sometimes, often or always. The Cohere figure counts only "often or all the time," and most of its respondents aren't physicians. Taken together, the data suggest that frequent abandonment may be falling while occasional abandonment remains common.

The AMA data also show how much trust has eroded. Just one in three physicians believes the insurer pledge will produce meaningful change, Medical Economics reported in May.

"Physician trust in voluntary insurer pledges is deeply eroded after years of unfulfilled promises," AMA President Bobby Mukkamala, M.D., said at the time. "Anything less risks reinforcing the skepticism these pledges were meant to address."

The workload got heavier

Every respondent in the Cohere survey said prior authorization still puts a burden on their team, and the burden appears to have grown. The share of clinicians who called it moderate or heavy rose from 40% to 51%. Among administrators, it rose from 39% to 61%.

Denials are increasingly about paperwork. Missing clinical documentation is now the top reason respondents give for denials, cited by 35% of clinicians (up from 16%) and 30% of administrators (up from 14%). Failure to meet medical necessity criteria ranked second. About six in 10 respondents in each group estimated that 21% to 30% of their requests are initially denied.

Part of the problem is that practices can't easily find out what a plan wants. Nine in 10 clinicians (91%) and 87% of administrators said prior authorization requirements are at least somewhat hard to find and understand. About 85% said they are asked at least occasionally to resubmit information they have already provided.

When asked what would help most, both groups ranked clearer requirements and documentation guidelines first (65% of clinicians and 62% of administrators). Prior authorization built into the electronic health record and faster decisions followed. Among clinicians, the hardest parts of the process were differing plan requirements and waiting for decisions, each cited by 21%. Administrators most often cited tracking request status (26%).

Electronic progress also stalled. The share of clinicians submitting more than 40% of requests through electronic prior authorization platforms fell from 24% to 8%. The share who had received approvals in under an hour multiple times dropped from 34% to 8%. Respondents said the top barrier to electronic submission is that plans don't accept it consistently.

"Visibility into the PA process isn't the same as improving it," Gigi Yuen-Reed, Ph.D., Cohere's chief data and AI officer, said in the report.

Trust in AI: Wide, but shallow

The announcement says 96% of clinicians and 94% of administrators trust AI in prior authorization approvals. That is true, but the white paper shows a bigger change underneath. Strong trust fell from 46% to 12% among clinicians and from 32% to 12% among administrators. Most respondents now say they only "somewhat" trust AI.

Clinicians' support for AI playing a significant or core role in prior authorization fell from 90% to 53%. The most common clinician answer about automating prior authorization is now "only under strict restrictions and human oversight," at 43%, up from 10% a year earlier.

Providers have also gotten more specific about which jobs they want AI to do. Interest in AI suggesting alternative treatments fell from 34% to 4% among clinicians, and interest in AI anticipating upcoming prior authorization needs fell from 30% to 1%. Interest in administrative help rose. Clinicians' interest in AI that confirms the right clinical documentation rose from 42% to 57%, and in real-time status updates from 29% to 47%. Administrators' interest in pre-filling requests rose from 34% to 52%.

In short, providers want AI to handle paperwork and stay away from clinical decisions.

Covino has made a similar argument. In an August interview on AI accuracy, trust and denials, which the white paper itself cites, he told Medical Economics that "Artificial intelligence is never used to deny care. That's non-negotiable." He added: "Clinicians have to be in the loop, sitting side by side, arm in arm, with the technologists who build the systems."

Physicians aren't fully convinced. The AMA survey found that 60% of physicians worry that insurers' AI tools will increase denial rates. Madhu Reddiboina, founder and CEO of RediMinds, a health tech firm that focuses on AI-assisted contested medical review, argued in an October column on AI governance in prior authorization that with 84% of health insurers now using AI or machine learning somewhere in the review process, AI's "existence is no longer the difference-maker; governance is."

Reddiboina told Medical Economics in October that "if AI makes an assertion, the physician should be able to move back through the chain and find the underlying source." He also wrote: "Many documented failures have involved opaque criteria, poor oversight and incentives that rewarded speed over accuracy."

The survey points to one more way to build confidence. About half of clinicians (51%) and two-thirds of administrators (67%) said their confidence would rise at least moderately if specialty medical societies helped develop or endorsed the clinical review criteria plans use.

What it means for physicians

The survey has some good news for patients and a warning about the workload. Frequent abandonment and avoidable ER visits appear to be declining. Practices, though, are doing more documentation work than before, and providers' trust in AI now depends on how it is used.

A few practical points follow:

  • Track your own numbers. Vendor and association surveys measure different things. Your practice's denial reasons, turnaround times and abandoned orders will tell you more about your payers than any national figure.
  • Treat documentation as your main denial risk. If missing documentation is now the leading cause of denials, front-end checklists for your highest-volume procedures may prevent more denials than appeals can reverse.
  • Use the new transparency rules. Plans must now give specific denial reasons and publish prior authorization metrics. Compare your experience with what your largest payers report, and bring the gaps to contract talks.
  • Ask how decisions are made. When a request is denied, ask whether AI was involved and who made the final decision. Physician pressure on these questions is what's pushing vendors to talk about oversight.

The next deadline is Jan. 1, 2027, when federal rules require plans to support standardized electronic prior authorization connections. Insurers have also pledged real-time answers for at least 80% of complete electronic requests. Whether practices notice a difference will depend less on how much AI plans adopt than on whether plans make their requirements clear before a request is submitted.


Related to this article

Qasim Faizan: ©WeBill Health
Starting Jan. 1, Medicare will pay 20% less for the lower-valued service when an office visit billed with modifier 25 falls on the same day as a zero-day global procedure. One claims report priced at local 2027 rates shows independent practices their exposure, and three documentation habits protect the visits that still qualify for full payment.