CMS’s Wasteful and Inappropriate Service Reduction (WISeR) model is nearing the end of its first year testing AI-enabled prior authorization in Traditional Medicare across six states.

A new JAMA study offers an early look at what happened: use of targeted services fell, but we don’t yet know whether patients avoided unnecessary care or missed care they needed.⁠⁠

In this article, I’ll recap how WISeR works, break down the findings, and share my take on what they mean for patients, physicians, and the healthcare system.

CMS launched WISeR in January 2026 as a six-year pilot in New Jersey, Ohio, Oklahoma, Texas, Arizona, and Washington. It expands prior authorization in Traditional Medicare for select services considered vulnerable to unnecessary use, fraud, and abuse:

CMS contracts with tech companies to review requests using AI-enabled tools and existing Medicare coverage criteria. A licensed clinician must review any recommended denial. Emergency services, inpatient-only services, and services that would pose a substantial risk if delayed are excluded.

The compensation structure is a little sketchy because participating companies receive a share of the savings from avoiding noncovered services, with payment adjusted for performance measures such as timeliness and appeal outcomes. As I wrote before launch, that creates a financial incentive to restrict utilization, and makes patient outcomes and access essential to judging whether the model works.

In JAMA, researchers used Epic Cosmos records to compare changes in targeted service use in the six WISeR states with the rest of the country. They examined two years before launch and the first six months afterward, focusing on Traditional Medicare patients.

Compared with changes in nonparticipating states, WISeR was associated with:

The reductions were largest early on and weakened over time, particularly for steroid injections. The study couldn’t determine whether the reduced use reflected unnecessary care avoided or necessary care delayed or forgone. It also couldn’t distinguish denials from fewer orders or referrals, or isolate AI’s contribution.

Prior authorization is a necessary evil. If we deliver unnecessary care (which does happen), someone still has to pay for it—so there’s incentive to curb such unnecessary care. But when healthcare organizations report billions in profits while patients and practices spend hours fighting for approval, I think we should scrutinize who benefits from that friction.

Financial performance alone tells us very little about whether patients are getting appropriate care, which is my concern with WISeR. This JAMA study gives us evidence that utilization fell after implementation. But we still don’t know which patients went without a procedure, whether they needed it, or what happened to them afterward.

For a patient who avoids an unnecessary injection, that could be a good outcome. For someone who gives up on an appropriate procedure because their physician’s office can’t get it approved, the same reduction in utilization could mean months of untreated symptoms. We need data on pain, function, complications, alternative treatments, and subsequent healthcare use before we can judge the clinical value of this model.

I’d also want to know how much work it took to get care approved. As I wrote recently, access can depend partly on whether a patient and their practice have the time and resources to challenge a denial. An eventual approval doesn’t erase the delay or the staff hours spent obtaining it.

And I suspect this will increasingly come down to AI fighting AI. Vendors will use AI to review requests. Practices will use AI to assemble documentation and appeal unfavorable decisions. These tools may save us time, which I welcome, but both sides will keep investing in ways to respond to the other. I covered this arms race last year, and WISeR gives it another place to develop. This particular study couldn’t tell us what AI contributed.

I’ll also be watching whether CMS expands the model to more services. Medicare Advantage already shows how broadly prior authorization can reach. If lower utilization becomes the measure of success, I worry it could support expansion before we understand the consequences for patients.

In summary, I’m open to evidence that WISeR reduces unnecessary care. Before expanding it, I’d want evidence that patients receive the care they need, outcomes are preserved or improved, and any savings justify the administrative work we’ve added.