Artificial Intelligence and public health seem like a promising combination on paper.
The technology steps in as an ally to cut unnecessary spending, identify fraud, and speed up processes that have historically depended on mountains of paperwork and hours of manual labor.
But what happens when that promise runs headfirst into the reality of overworked doctors, patients who need urgent treatment, and a system rolled out in a rush?
That is exactly what is happening right now in the United States with WISeR, Medicare’s new AI-powered prior authorization program.
Instead of simplifying things, the system has been generating a growing pile of confusion — with errors that look like they came from a careless algorithm, delays stretching into weeks, and patients who are literally thinking about crossing state lines just to avoid the process. 😬
Below, you will get a clear picture of what is behind this situation, how the program works, and why experts are already questioning whether the solution is creating a problem even bigger than the one it was meant to fix.
What is WISeR and how it entered the Medicare equation
WISeR — short for Wasteful and Inappropriate Service Reduction Model — is an artificial intelligence-based system designed for screening and analyzing prior authorization requests within the Medicare program, the federal health plan for Americans aged 65 and older, as well as patients with certain conditions and disabilities. The original idea was pretty reasonable: use algorithms to review requests for medical procedures before approving them, flagging suspected cases of fraud or unnecessary use of public resources. In theory, this would save the federal government money and protect patients from potentially unsafe or unnecessary treatments.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
The important detail here is that, for years, traditional Medicare deliberately avoided this very practice — prior authorization — which requires the patient or someone on the medical team to seek plan approval before moving forward with certain procedures, tests, and prescriptions. The program was announced in June 2025 and went live in mid-January 2026, launching as a pilot in six states: Oklahoma, Arizona, New Jersey, Ohio, Texas, and Washington. In total, 13 medical services are subject to this new review, chosen because the federal administration considers them more prone to fraud or misuse.
Among those procedures are epidural spinal injections, used to treat back pain, and kyphoplasty, a surgery for vertebral fractures that had already been flagged as overutilized by a Medicare advisory commission. Also on the list are so-called skin substitutes, whose spending skyrocketed nearly 700% in just two years, according to an alert from the inspector general of the Department of Health and Human Services, raising serious concerns about waste, fraud, and abuse.
To really grasp the impact, you need to keep Medicare’s scale in mind. Any dysfunction in this system, no matter how small it looks on paper, multiplies fast. And the reports coming in from hospitals, clinics, and medical offices in the pilot states suggest the transition is far from smooth — some professionals have gone as far as describing the rollout as horrible. 😕
A rollout that moved way too fast
One of the main criticisms of WISeR is simple and straightforward: everything happened too quickly. Between the announcement in June 2025 and the launch in January 2026, the gap was considered shorter than normal for the federal government’s usual pace. In practice, that meant doctors and administrative teams had to figure out how the system worked pretty much on the fly, without the adjustment period a change of this magnitude would normally require.
Even the contractors operating the technology acknowledge the tight timeline. Jeremy Friese, CEO of Humata Health, the vendor responsible for Oklahoma, admitted there was an aggressive rollout from the moment of notification to the go-live date. Tech executives serving other states said they were still adding features to their products during the spring — meaning the system was already running and serving real patients.
On the administration side, Abe Sutton, director of the center responsible for Medicare innovation, did not directly comment on the timeline but stressed that the goal of the reforms is to make prior authorization efficient, fast, and streamlined. According to him, the model aims to reduce inappropriate treatments without delaying appropriate care. Meanwhile, Mehmet Oz, who heads the Centers for Medicare and Medicaid Services, stated that the purpose of the program is not to deny care but to ensure the patient receives the treatment they actually need — not whatever some unscrupulous provider might want to administer. Worth noting that even Sutton acknowledged the percentage of providers who commit fraud or abuse is small.
Systematic errors: when the algorithm does not understand clinical context
Although vendors claim humans make the final decision on approvals, doctors and their teams believe artificial intelligence plays a central role in the process — and that some denials are the result of AI hallucinations, where the system distorts or simply fabricates information. The reports are concerning and very specific.
A doctor in Arizona, who was not authorized by his clinic to be identified, shared that he received a denial claiming his patient was not eligible for procedures in the thoracic region — the mid-back area. The problem is the patient needed an injection in the neck, a completely different region. In another case, James Webb, a musculoskeletal radiologist in Tulsa, Oklahoma, documented four times that a patient did not have numbness, yet his request was denied citing numbness as the very reason — which, in the reviewer’s interpretation, would rule out the spinal procedure.
This kind of inconsistency reveals the system’s inability to interpret a patient’s full clinical context. The algorithm works with patterns from historical data, procedure codes, and standardized diagnoses — but medical practice rarely fits neatly into such well-defined boxes. Jennifer Valle, a pre-certification supervisor at an Oklahoma radiology clinic, reported that in the case of kyphoplasties, reviewers are demanding way too much. In other situations, information the clinic sends ends up being ignored, and reviewers request imaging that is already on file. Friese, from Humata, said he had not heard of any AI hallucinations — a response that stands in stark contrast to what healthcare professionals are experiencing day to day.
Experts in artificial intelligence applied to healthcare had already been warning about this type of risk. Models trained on historical data tend to reproduce the biases and inconsistencies that already existed in that data, and when applied to decisions with a direct impact on people’s health, any margin of error becomes ethically problematic. What is being observed now is not a technical surprise — it is the practical confirmation of a risk that has already been documented and discussed in specialized publications. 😕
Delays that cost more than just time
The delays generated by the system are not just a matter of administrative inconvenience. The University of Washington medical system alone had nearly 100 patients waiting for epidural injections due to WISeR-related delays, according to an April report from Senator Maria Cantwell’s office, based on data from hospital associations. The document was blunt in pointing out that patients are now subject to delays or denials that simply did not exist before the model was introduced.
Medicare states that decisions should be returned within 72 hours and that clean claims should be paid within 15 days. In practice, the story is very different. Webb, the Tulsa radiologist, said he has been seeing reimbursement delays of six to eight weeks. Jerry Sobel, a pain management physician in the Phoenix area, described the situation as horrible from the start, with what appeared to be a total lack of organization. He reported that as of May, he still had not been paid by Medicare for nine epidural injections he had performed.
There is also a direct impact on the people seeking treatment. Bill Curry, a 65-year-old rancher from rural Oklahoma, has been making a two-and-a-half-hour drive to the state capital for years to get epidural spinal injections. In February, he was hit with the news that he would need prior approval for the procedure. The result: he had to go back more than once, logging about ten hours on the road, and was even asked to show up a third time just to fill out a form describing how he felt — something that had never been required before. Faced with all that, Curry said he might start seeking treatment in Kansas to escape the whole approval process. 🚨
In New Jersey, physical medicine and rehabilitation physician Dorota Gribbin reported an even more serious case: by the time the authorization finally came through for one of her patients who needed a back pain procedure, he had already gone to the hospital seeking far more expensive emergency care. That is exactly the kind of side effect that runs in the opposite direction of the program’s goal of saving money.
Costs that sneak back in through other doors
One of the most interesting paradoxes of WISeR is that it may actually end up increasing the government’s own costs. With more denials come more appeals, which need to be processed by Medicare’s administrative contractors — and the government pays those contractors to handle each appeal. Sutton himself acknowledged that the agency has already factored in potential changes in appeal volume because of the program and the costs associated with it.
Health policy researcher Miranda Yaver, from the University of Pittsburgh, raises a central question. In her view, prior authorization saves money for plans partly because it shifts the cost onto patients in the form of wait times and inconvenience. People end up trapped in a web of red tape, waiting on the phone and getting bounced from one department to another. The big question, according to her, is whether prior authorization truly saves resources or just pushes the cost onto patients and doctors.
It is worth remembering that the use of artificial intelligence in this context is nothing new in the private sector. According to a 2025 survey published by the National Association of Insurance Commissioners, 84% of commercial insurers already use AI tools — although they consistently claim they do not use them to deny prior authorization requests. Interestingly, while the government is testing this approach with Medicare, the same administration is trying to reduce the practice for people with private insurance. A January survey found that 69% of adults with coverage consider prior authorization a burden when seeking care.
What can be done from here
The discussion around WISeR has already reached Congress, with representatives from medical associations pushing for a review of the program. The main demand is not to abandon artificial intelligence in the process — the potential benefits are real and widely recognized — but rather to implement stronger safeguards. Among the proposals are mandatory thorough human review for any denial, maximum turnaround times that are actually enforced for appeals, and fast-track escalation channels for cases with documented clinical urgency.
From a technical standpoint, the path forward involves retraining the models with more representative and diverse data, along with a careful review of the automatic flagging criteria. AI systems applied to critical contexts like healthcare need to be calibrated with tolerance thresholds very different from those used in content recommendation systems. The cost of denying a necessary procedure is nothing like the cost of recommending the wrong video. That distinction needs to be built into the model’s architecture, not just the usage policies.
There is also an underlying concern among physicians: that the program will expand far beyond the handful of services currently flagged as fraud-prone. Mary Clarke, a family physician in Oklahoma, summed up that fear well when she said everyone knows that if the pilot works, prior authorization will be extended to basically every procedure. When asked about a potential expansion, Sutton said there are no planned changes to the list of services at the moment, but that the agency continues to evaluate whether adjustments will be needed.
The WISeR and Medicare episode serves as a valuable warning for any initiative looking to deploy artificial intelligence at scale within public health systems — and that lesson applies well beyond American borders. The technology has real potential to transform how efficiently these systems operate, but it needs to be introduced with transparency, tested at a controlled scale, and above all, kept under constant human oversight. When the pace of implementation outstrips the pace of validation, errors and delays stop being technical bugs and become real human consequences. And at that point, the problem is far bigger than any algorithm can solve on its own. 🤖
