AINEWS
A federal pilot in six Medicare Advantage states is using an algorithm to pre approve prescriptions, scans, and surgeries — and doctors want a written clinical reason for every denial it issues.
The doctor's note goes in. The insurer has to approve it before the patient gets the prescription, the scan, or the surgery. This pre-approval step, called prior authorization, runs heaviest in Medicare Advantage plans, where about 55% of Medicare-eligible seniors and disabled people are enrolled and insurers issue millions of full or partial denials every year.
That system already has a problem the government has documented. A 2026 HHS Office of Inspector General audit found that Medicare Advantage insurers overturned nearly every prior-authorization denial for skilled nursing facility admissions once patients appealed. The denials were not medically correct; they were procedurally easy to reverse. The OIG finding is specific to SNF admissions, not the whole MA denial pile, but it is the freshest federal signal on how often the first answer is wrong.
Now the Trump administration is testing whether an algorithm can do that first step faster. The CMS Innovation Center's WISeR (Wasteful and Inappropriate Service Reduction) Model is a six-state pilot, run through the federal Centers for Medicare & Medicaid Services' Innovation Center, that uses AI to review requests for selected services before a human reviewer sees them. The model's stated goal is to catch wasteful or inappropriate care. The risk physicians name is that the pilot runs the same first-pass denials the auditors just showed get reversed on appeal, without a person explaining the call.
Health-policy analyst Camm Epstein drew the line plainly: "AI should be used to make appropriate care easier to approve, not necessary care easier to deny." That is the patient-side version of the American Medical Association's 2025 survey, in which 61% of physicians said they worry AI will worsen denials of necessary care. The 61% number is physician perception, not a measured denial rate, but it lines up with the OIG pattern: doctors see denials that look defensible at first and fall apart on appeal.
The AMA's 2026 reform agenda does not ask CMS to scrap the WISeR pilot. It asks for two concrete accountability hooks. Insurers must give the treating physician and the patient a written clinical reason for any denial, in language a doctor can rebut. The AI itself has to be auditable: which model, trained on which data, against which standard. The point is not transparency theater. It is a way to make the algorithm answerable when the OIG-style pattern shows up at scale.
A legal analysis from Reed Smith describes WISeR as built on the existing prior-authorization infrastructure, with the AI layer sitting on top of the same clinical criteria and review pipelines insurers already use. The novelty is volume and speed, not the underlying logic. An independent health-policy newsletter, Healthcare Uncovered, argues that putting an algorithm in front of a reviewer in a system where the first answer is often wrong is a structural choice, not a technical one. The model will inherit the same incentives the human reviewers had.
For a reader, the practical test is short. When a denial arrives, ask whether the clinical reason was given in writing, in terms a doctor can answer, and whether it was reversed on appeal. Under the AMA's reform agenda, both questions have to be answerable on every letter, for every plan. The WISeR pilot does not settle whether AI can fix prior authorization. It settles whether anyone has to explain the call.