AI Is Denying Medicare Care in 6 States. Here's What That Actually Means.

He found out a bot denied his pain treatment on the third try. Not a doctor. A bot.

When the Algorithm Says No: The AI That's Deciding Who Gets Treated in America

Keith Magnuson is 83 years old. He used to be a rock climber. Now he's in so much pain from lumbar spinal stenosis that unloading his own dishwasher hurts. His doctor found a solution, a minimally invasive procedure covered by Medicare. It should have been simple. It was denied three times.

On the third denial, Magnuson found out something that should stop every one of us cold. It wasn't a person who made that decision. It was AI. "I was outraged", he said. "I was like, wait a minute, it's not even another person at the other end? It's AI? It's a bot?"

I want to sit with that sentence for a moment, because I think it captures exactly what's gone wrong here, and exactly why I need to write about it.

A Pilot Program Nobody Voted For

Six states, Washington, Arizona, New Jersey, Ohio, Oklahoma, and Texas, have been involuntarily enrolled in a six-year federal pilot called the WISeR Model, short for Wasteful and Inappropriate Service Reduction. The goal, according to the Department of Health and Human Services, is to use AI to catch Medicare spending on treatments patients supposedly don't need, citing an estimate that $5.8 billion was spent in 2022 on "services with minimal benefit".

Here's what that looks like on the ground; doctors now have to upload their clinical reasoning to a portal, and an AI system run by a third-party tech company decides whether Medicare will cover the treatment. Fifteen procedures are covered under the program, from epidural steroid injections to knee arthroscopy. Treatments that used to take a single day to authorize are now taking weeks. A Washington state medical association director put it plainly; "This was sold to the physician community as, we're going to turn this around in 72 hours, it's super fast, and instead we've had people waiting four weeks or beyond".

This wasn't voted on by Congress. It didn't go through formal rulemaking. It was classified as "voluntary", but only voluntary for the tech companies providing the AI. Doctors who opt out don't get Medicare funding at all. As one Stanford health policy professor explained it, the administration found a legal side door and walked straight through it.

The Incentive That Should Actually Terrify You

Here's the part of this story that I think deserves to be said plainly and without hedging; the tech companies running these AI systems are paid, in part, based on how much money they save Medicare by denying treatments. Read that again. A profit motive is directly attached to the number of denials an algorithm produces.

"You're incentivizing an entity to deny care", says Jeb Shepard of the Washington State Medical Association. A scrub technician in Richland, Washington, who submits these authorization requests every day, put it even more bluntly; "Every single appeal that we have submitted has been overturned". If every appeal against an AI's decision gets overturned once a human actually looks at it, that's not a rounding error. That's a system that got the answer wrong by default, and left elderly patients in ten-out-of-ten pain waiting weeks to find out.

I'm going to say what I think needs to be said here, because I don't think softening this serves anyone; we are allowing a piece of software with a financial incentive to say no, to decide who receives medical treatment and who waits, who suffers, and eventually, in the most severe cases, who lives long enough for that treatment to still matter. That is not hyperbole. When a treatment for a fragile, elderly population is delayed by weeks over and over, deterioration isn't a hypothetical outcome. It's the expected one. A health economist at Brown University called the whole system "a complete black box". Nobody, not the doctors, not the researchers, not the patients, can fully explain how these algorithms actually decide.

Other Countries Are Proving This Doesn't Have to Be the Model

Here's what makes this especially hard to accept. Other countries with a genuine, structural commitment to universal healthcare are using AI in ways that actually work, because they've kept a human being at the center of the decision instead of removing them from it. In Sweden, a study found that using AI in breast cancer screening reduced missed diagnoses by 12 percent, because the AI was catching things at the initial screening stage, alongside a radiologist, not instead of one. In the UK, the NHS has signed a data-organization deal to help hospital systems work more efficiently. In Australia, machine-reading technology is helping radiologists interpret mammograms, helping them, not replacing their judgment or denying coverage before they even see a patient.

The difference isn't the technology. The AI itself is roughly the same category of tool in every one of these examples. The difference is what each system chose to protect. Countries with a real, longstanding infrastructure of universal healthcare used AI to catch things humans might miss. The United States, a country that has never fully committed to healthcare as a right in the first place, is using AI as a cost-cutting gatekeeper with a profit incentive baked directly into its decisions. That is not a coincidence. That is a reflection of priorities, and I think it's a genuine travesty that the country with arguably the least stable healthcare safety net in the developed world is the one experimenting with letting an algorithm decide who gets care and who doesn't.

Why This Isn't a Slippery Slope Argument, It's Already Happening

I try not to reach for the most extreme framing available, but I don't think this needs any exaggeration. When an algorithm with a cost-saving incentive is standing between a patient and a covered, doctor-recommended treatment, and that algorithm is denying care that gets overturned on appeal essentially every time, we are not looking at a hypothetical about a dystopian AI future. We are looking at a live system, operating in six states right now, where a bot's initial answer can mean weeks of unnecessary suffering for someone in their 80s with a deteriorating condition. Extend that same structure to more serious conditions, more states, more time, and the outcome isn't abstract anymore. It's a system quietly deciding who gets timely treatment and who doesn't, which for a fragile population is functionally a decision about who lives comfortably and who doesn't.

The CMS administrator overseeing this, Dr. Mehmet Oz, said the program "helps bring Medicare into the 21st century". I'd ask him to talk to Magnuson, who ended up offering to pay $600 out of pocket just to get treated on time, and is still waiting to find out if he even has to.

What I Actually Want You to Take From This

I don't think the answer is rejecting AI in healthcare all together. The Swedish breast cancer data alone tells you this technology can save lives when it's built to support a doctor's judgment instead of replacing it. The problem here was never the algorithm. It's the decision to remove a human being from the most consequential moment in someone's care, and to attach a profit motive to the number of times that algorithm says no. That is a policy choice, not a technological inevitability, and policy choices can be reversed.

If you or someone you love is on Medicare in Washington, Arizona, New Jersey, Ohio, Oklahoma, or Texas, know that you have the right to request a peer-to-peer review and to appeal a denial. Push for it. The data so far suggests the appeals are winning.

➡️ Related: [AI Can Now Map Which Parts of Your Brain Are Aging Faster] 

➡️ Related: [Using AI to Shape the Midterm Elections] 

➡️ Join the free Tech 4 Grown-Ups community: tech4grownups.com/community

Michael Routhier is the founder of Tech 4 Grown-Ups, providing honest, unfiltered digital literacy for adults 55+, and host of The Virtuous Machine, exploring the ethics and human cost of AI. Read by tech-curious readers in 50+ countries. Explore more at tech4grownups.com.