Medicare AI prior authorization has already changed how millions of older Americans get care. In the first six months of a new pilot programme, use of targeted medical services fell sharply in six states where traditional Medicare now requires advance approval screened with artificial intelligence, according to the first independent study of the experiment. Epidural steroid injections for back pain fell 15%, while other targeted services fell about 20% compared with trends elsewhere in the country.

The finding, published in JAMA on 26 October 2026, does not settle the argument at the heart of the programme. The researchers, from the University of Minnesota and Brown University, say their data show that care fell where approval was required. They cannot say whether the care that did not happen was wasteful, unnecessary, delayed, or badly needed.

Key Points

• Medicare AI prior authorization under the WISeR pilot cut epidural steroid injections by 15% and other targeted services by about 20% in six states.
• The six pilot states are Arizona, New Jersey, Ohio, Oklahoma, Texas and Washington, home to about 6.4 million traditional Medicare beneficiaries.
• The drop was largest straight after the 1 January 2026 launch, then faded: epidural injections fell 21% in the first quarter and 9% in the second.
• Services planned for WISeR but never implemented showed no significant decline, strengthening the case that prior authorization itself drove the fall.
• The study measured service use only, not pain relief, complications, denials or spending, so it cannot say whether patients were protected or harmed.

What Happened

The study, by Hannah Neprash of the University of Minnesota School of Public Health and Michael Barnett of Brown University School of Public Health, is titled Changes in Service Use Following the Introduction of AI-Enabled Prior Authorization in Medicare. It used Epic Cosmos, an electronic health record database covering more than 300 million patient records nationwide.

The researchers compared monthly use of WISeR-targeted services among traditional Medicare patients in the six participating states with use in all other states and Washington, D.C. The comparison covered the two years before the programme launched and the first six months after it began on 1 January 2026.

Against that baseline, epidural steroid injections, the highest-volume service in the model, declined 15% relative to trends outside the pilot. Other targeted services declined 19.5%, reported as about 20%. The researchers also found that the effect weakened over time. For epidural injections, the reduction fell from about 21% in the first quarter of 2026 to about 9% in the second quarter. For other services, it eased from about 21% to 18%.

That fading pattern matters. The authors suggest it may reflect clinicians learning how to meet the new requirements, early workflow problems being worked out, or both. Co-author Barnett, a physician, said the early results suggest doctors may be adapting rapidly to the new system, and warned that researchers must watch whether the extra work of prior authorization justifies what might be relatively little change in who ultimately gets care.

Importantly, the study was observational, not a randomised trial. It describes an association between the new approval requirement and lower service use. It does not measure how many requests were denied, how long patients waited, whether pain or function improved or worsened, or how much money Medicare saved.

Background/Context

For many years, traditional Medicare operated almost entirely without prior authorization. Commercial insurers and Medicare Advantage plans use prior authorization widely, requiring clinicians to obtain approval before certain services will be covered. Traditional Medicare, by contrast, historically almost never required it.

That changed with WISeR, short for Wasteful and Inappropriate Service Reduction, a six-year pilot run by the Centers for Medicare and Medicaid Services through its Innovation Center. It began on 1 January 2026 and is scheduled to run through 2031. It requires prior authorization or prepayment review for a set of outpatient services that the agency considers vulnerable to fraud, waste and abuse.

The targeted list includes 41 services in some descriptions of the model, and more than a dozen procedures in others, including epidural steroid injections for pain management, minimally invasive treatments for spinal stenosis in the back and osteoarthritis in the knees, nerve-stimulation therapies for conditions such as sleep apnoea and incontinence, cervical spinal fusion, electrical nerve stimulator implants, and skin substitutes for chronic wounds. Emergency services are excluded.

Contractors, mostly technology firms, conduct the reviews and are encouraged to use artificial intelligence to screen requests. The agency says licensed clinicians make every recommendation against payment, and that WISeR does not change Medicare coverage or payment policy. The pilot applies only to Original Medicare, not to Medicare Advantage, which already uses prior authorization extensively.

The political backdrop is the Trump administration's drive to cut what it calls waste, fraud and abuse in government health spending. The agency says wasteful care contributes to as much as 25% of United States health spending. Critics, including Democratic lawmakers, argue that the payment design, in which participating companies receive a share of spending tied to averted wasteful or inappropriate care, creates an incentive to deny care. In July, the Senate voted 46 to 50 against advancing a resolution to overturn the pilot, so the programme survived.

The timing also matters for patients. Medicare's annual open-enrollment period runs through 7 December 2026, when people aged 65 and older can switch between traditional Medicare and private Medicare Advantage plans, change plans, or join or change prescription-drug coverage, with changes taking effect on 1 January. The WISeR findings land while many older adults are making exactly those choices.

Medicare AI Prior Authorization: What the Numbers Do and Do Not Prove

The strongest part of the study is its control check. Services that were originally slated for WISeR but never implemented showed no significant decline in use. That pattern supports the conclusion that prior authorization itself, rather than some broader trend in back pain treatment or health care use, drove the drop in targeted services.

The weakest part, by the authors' own account, is what comes next. A fall in service use is not, on its own, evidence of savings, safety or better care. If the missing injections and procedures were low-value care that patients did not need, the pilot worked as intended. If they were care that patients needed and then went without, or received later after delays and extra paperwork, the same numbers would look very different.

Early complaints have centred on time. A report from Senator Maria Cantwell said UW Medicine reported average waits of 15 to 20 days for decisions that had taken one to three days before WISeR. The agency has said it is committed to timely decisions and would take corrective action where needed. The JAMA study does not measure waiting times, so it cannot confirm or refute those reports.

There is also a separate, earlier warning from inside the agency. A senior Centers for Medicare and Medicaid Services official acknowledged at an October conference that many parts of WISeR were not working, without specifying which parts. That admission, reported before the JAMA study, helps explain why researchers and patient advocates are pressing for outcome data, not just use data, before any decision to continue or expand the model.

For now, the honest summary is narrow but important: an AI-assisted approval requirement quickly reduced how often targeted services were used, most sharply at the start, and the effect faded as clinicians adapted. Whether that is a story of waste removed or care delayed remains unproven.

Reactions/Quotes

Lead author Hannah Neprash said traditional Medicare had operated almost entirely without prior authorization, so WISeR represented a major change in how older adults access care. She said: "Our early evidence shows that these requirements had an immediate effect, with noticeable drops in use of targeted services in participating states."

Neprash added the sentence that should frame every debate about the programme: "The key question we can't yet answer is whether the care not received was wasteful or necessary. As the Centers for Medicare and Medicaid Services considers whether to continue or expand this model, understanding its effects on patients, clinicians and Medicare spending should be a top priority."

Barnett said: "As a physician, I know how burdensome prior authorization can be. The early results suggest that doctors may be adapting rapidly to the new system. We need to watch closely to see if the additional work of prior authorization justifies what might be relatively little change in who gets care."

The agency's position, as set out on its WISeR model page and reported in coverage of the study, is that the pilot targets services with limited benefit for some patients and a history of fraud, waste and abuse, excludes emergency services, and requires licensed clinicians to make every recommendation against payment. Supporters argue that screening out low-value care protects both patients and taxpayers.

Critics, including Democratic lawmakers such as Representative Suzan DelBene, have called the programme harmful to patients, according to Medical Economics reporting cited in coverage of the study. Their argument is that paperwork, delays and a payment structure tied to averted spending can discourage needed care, even when the stated goal is to stop waste.

The wider public mood adds pressure. A Pew Research Center survey published on 29 October 2026, covering 3,488 adults surveyed in June, found that 46% of Americans said they were not open to AI playing any role in health insurance claim decisions, and only 5% said they would be comfortable with AI making those calls without a human reviewing them. Thirty-eight percent said they were open to AI handling claims if a person did a final review.

Why It Matters

WISeR matters because it tests a question that reaches far beyond six states. If AI-assisted prior authorization can reliably separate wasteful care from needed care, it could be expanded to more services or more states, as the Innovation Center has the power to consider. If it mainly adds friction, delays and denied or abandoned care, expansion would export those harms to millions more people.

The stakes are highest for older adults with chronic pain, mobility problems or slow-healing wounds. Epidural steroid injections are not a cure, but for some patients they reduce pain enough to keep moving, working, sleeping and avoiding stronger medicines. A 15% fall in use, averaged across states, hides individual stories: some patients may have avoided an unnecessary injection, while others may have waited weeks, given up, or paid out of pocket.

There is also a cost question that the study does not answer. Prior authorization itself costs money, in contractor fees, clinician time, appeals and rework. If service use falls by 15% to 20% but administrative costs and delays rise, the net saving to Medicare could be smaller than the headline drop suggests, or even negative for some services. That is why the authors call for research on spending, not just use.

Finally, the fading effect deserves attention. If clinicians simply learned to write requests in ways that pass the screen, the pilot may change paperwork more than care. If early problems were fixed and approvals became smoother, the fading effect could be a sign of a system settling down. Either way, six months is a short window in a six-year pilot, and the first numbers should not be treated as the final verdict.

What Happens Next

The researchers recommend that future work examine whether the effects last over the model's six-year run, and how WISeR affects the appropriateness of care, patient outcomes, clinician burden and Medicare spending. The agency has not yet released its own public evaluation of the pilot.

Since July, providers with high approval records in all six states have been able to qualify for exemptions from prior authorization. That change could further soften the effect over time, by removing trusted clinicians from the approval queue, and it will complicate later comparisons.

Patients in the six states can take practical steps now. Ask a doctor early whether a planned procedure falls under WISeR and whether a request has been submitted. If a request is not approved, providers can resubmit it and request a peer-to-peer review, and a denied claim can be appealed through the standard Medicare process. Keep copies of letters, note dates of calls, and seek free help through each state's State Health Insurance Assistance Program or by calling 1-800-MEDICARE.

Paperwork should never delay urgent care. New leg weakness, loss of bladder or bowel control with back pain, or a wound that becomes red, hot or draining needs prompt medical attention, and emergency services are excluded from WISeR review. The next major test will be whether the agency publishes outcome and spending data that let patients, doctors and Congress judge the pilot on health, not just on how much less care was delivered.

Frequently Asked Questions

What is Medicare AI prior authorization under WISeR?

It is a six-year pilot, launched on 1 January 2026, that requires prior authorization or prepayment review for selected outpatient services in Original Medicare in Arizona, New Jersey, Ohio, Oklahoma, Texas and Washington, with contractors encouraged to use artificial intelligence to screen requests.

How much did service use fall?

Epidural steroid injections fell 15% and other targeted services fell about 20% in the six pilot states compared with trends elsewhere, in the first six months after launch, according to the JAMA study.

Does the study prove the missing care was wasteful?

No. The study measured service use only. It did not measure pain, function, complications, denials, spending or whether patients switched to other treatments, so it cannot say whether the care not received was unnecessary or needed.

Does WISeR apply to Medicare Advantage?

No. It applies only to Original, or traditional, Medicare. Medicare Advantage plans already use prior authorization extensively.

What can patients do if a request is denied?

Providers can resubmit the request and ask for a peer-to-peer review, and denied claims can be appealed through the standard Medicare appeals process. State Health Insurance Assistance Programs and 1-800-MEDICARE offer free help.

Related coverage on US News Pulse: how AI advice in health care is already facing court scrutiny, and the FDA warning over Darwin's cat food lots linked to Salmonella. For the wider health picture, see also the United States debt story for how health and federal spending pressures are colliding.