Taking a page from the private insurance industry playbook, the Trump administration will launch a program next year to see how much artificial intelligence algorithms can save the federal government by refusing to care for Medicare patients.
Designed to eliminate wasted “low-value” services, a pilot programme amounts to a federal expansion of an unpopular process called pre-authorization, where a patient or someone on the health care team must seek insurance approval before proceeding with a specific procedure, testing, and prescription. It will impact Medicare patients and doctors and hospitals who care about them in Arizona, Ohio, Oklahoma, New Jersey, Texas and Washington, running from January 1st until 2031.
The move has raised eyebrows among politicians and policy experts. The traditional version of Medicare, which covers some adults over the age of 65 and those with disabilities, has largely avoided prior permission. Still, it is widely used by private insurance companies, especially in the Medicare Advantage market.
And the timing was amazing. The pilot was announced in late June a few days after the Trump administration announced voluntary efforts to revamp and reduce advance approvals by private health insurance companies.
“It erodes public trust in the health care system,” Oz told the media. “It's something that cannot be tolerated by this administration.”
However, some critics, such as Ohio State physician and policy researcher Dr. Vinay Rathi, have accused the Trump administration of sending mixed messages.
On the one hand, he said the federal government wants to borrow cost-saving measures used in private insurance. “On the other hand, I slam it against the wrist.”
Rep. Susan Delvene, a Washington Democrat, said the executives were “talking from both sides of their mouths.” “That's very concerning.”
Patients, doctors and other lawmakers are also critical of what they consider to be delay or abuse tactics, which can slow or block access to care, and cause irreparable harm and death.
“The insurance company puts it in their mantra to do their final things to deny taking the patient's money and giving it to those who provide care,” said Rep. Greg Murphy, a North Carolina Republican and urologist. “It continues in the boardrooms of all insurance companies.”
Insurance companies have long argued that advance approval reduces fraud and unnecessary spending and prevents potential harm. The public's discomfort over refusing to insurance ruled the news in December. This was the shooting death of the CEO of United Healthcare, which led to many anointing him as his murderer and folk hero.
And the public hates practices a lot. Nearly three-quarters of respondents considered it a “major” issue in a July poll published by KFF, a health information nonprofit organization that includes KFF Health News.
In fact, Oz said at a press conference in June that “street violence” had urged the Trump administration to take on the issue of advance approval reforms in the private insurance industry.
Still, the administration is expanding its use of advance approval in Medicare. CMS spokesman Alexx Pons said both initiatives “fulfil the same goal of protecting patients and Medicare dollars.”
Unanswered Questions
A sensible pilot program, short for “wasteful and inappropriate service reduction,” tests the use of AI algorithms when making pre-certification decisions for some Medicare services, including skin and tissue replacements, electrical nerve stimulation implants, and knee arthroscopy.
The federal government says such procedures are particularly vulnerable to “fraud, waste, and abuse” and could be curtailed by prior permission.
You can add other steps to the list. However, a federal release shows that hospitalization only, emergency, emergency, or services that “substantial delays will pose a significant risk to patients” are not eligible for evaluation of the AI model.
Although the use of AI in health insurance is not new, Medicare is slow to adopt private sector tools. Medicare has historically been limited to previous permits. But experts who studied the plan believe federal pilots can change it.
Pons told KFF Health News that Medicare requests will not be denied before being reviewed by a “qualified human clinician,” and that vendors are “banned from refusal arrangements related to refusal rates.” The government says vendors will be rewarded for their savings, but Pons says multiple safeguards will “remove incentives to refuse medically appropriate care.”
“A shared savings arrangement means that a shared savings arrangement means that a vendor will benefit financially if less care is provided,” says Jennifer Bracken, senior director of government affairs at the Washington State Hospital Association, a structure that allows businesses to create strong incentives to reject medically necessary care.
And doctors and policy experts say it's just one concern.
Lati said the plan was “not fully embodied” and relied on “disorganized and subjective” measures. The model said it would ultimately rely on contractors to assess their outcomes. This is the choice that the outcome could potentially be suspected.
“I don't know if they know how they understand whether this is helping patients or hurting them,” he said.
Pons said the use of AI in Medicare pilots is “subject to strict surveillance to ensure transparency, accountability and integrity between Medicare rules and patient protection.”
“CMS remains committed to ensuring that automated tools support clinically sound decisions,” he said.
Experts agree that AI can in theory promote what was an cumbersome process marked by delays and rejections that could damage the health of a patient. Health insurers claim that AI eliminates human error and bias and saves money in the health system. These companies also argue that humans, not computers, are ultimately considering coverage decisions.
However, some scholars suspect that it happens on a daily basis.
“We're a part of the world,” said Amy Killerea, assistant professor at Georgetown University's Center for Health Insurance Reform.
A 2023 report published by Propublica found that for two months, doctors who reviewed payment requests only spent an average of 1.2 seconds in each case.
Cigna spokesman Justine Sessions told KFF Health News that the company has not used AI to deny care or claims. The Propublica study “referred to a simple software-driven process that helps accelerate clinician payments for common, relatively low-cost testing and treatments. “It was not used for advance approval.”
Still, class action lawsuits filed against major health insurance companies claim that the AI model has failed to undermine the recommendations of flawed physicians, failing to consider the unique needs of patients, forcing some people to put a financial burden of care.
Meanwhile, a doctor survey published by the American Medical Association in February found that 61% believe that AI is “increasing refusal to prior approval, exacerbating avoidable patient harm, and escalating unnecessary waste for the present future.”
Chris Bond, a spokesman for the insurance company's industry group Ahip, told KFF Health News that the organization has “no” to implement commitments made to the government. These include reducing the scope of prior permission and ensuring that communication with patients regarding denials and appeals is easier to understand.
“This is a pilot.”
The Medicare Pilot Program highlights ongoing concerns about advance approvals and raises new approvals.
Private health insurance companies were unclear about how AI is used and how much they use prior permissions, but policy researchers believe that these algorithms are often automatically programmed to deny high-cost care.
“The more expensive it is, the more likely it is to be denied,” said Jennifer Oliva, professor at Maurer Law School of School at Indiana University Bloomington, which focuses on AI regulation and health insurance.
Oliva explained in a recent paper in the Indiana Law Journal that health insurance companies are “motivated to rely on algorithms” when patients are expected to die within a few years. As time passes and the patient or her provider is forced to sue denial, the chances of a patient dying during the process increase. The longer the appeal, the less likely the health insurance company will pay the claim, Oliva said.
“The first thing is making it very, very difficult for people to get high-cost services,” she said.
With health insurance companies poised to grow in AI use, insurance companies' algorithms represent “regulatory blind spots” and require more scrutiny, said Carmel Shacher, a faculty member at Harvard Law School's Center for Health Law and Policy Innovation.
The clever pilot is an “interesting step” to help Medicare dollars use AI to buy high-quality healthcare, she said. However, the lack of details makes it difficult to determine whether it works.
Politicians are addressing some of the same questions.
“How is this being tested in the first place? How do you make sure it's working and not deny care or create a higher rate of rejection of care?” asked Delbene. Delbene signed an August letter to Oz with other Democrats demanding answers about the AI program. But it's not just the Democrats who are worried.
Murphy, co-chairing the House GOP Doctors Caucus, acknowledged that many doctors are concerned that wise pilots can challenge medical practice if AI algorithms reject the care they recommend.
Meanwhile, both parties supported the measures proposed by Florida Democrat Rep. Lois Frankel, who blocked pilot funding in the Department of Health and Human Services' 2026 budget.
Healthcare AI will remain here, Murphy said, but it is still unclear whether smart pilots will save Medicare money or contribute to the issues posed by prior permission.
“This is a pilot and I'm open to see what happens with this,” Murphy said.
