Medicare AI prior auth test cut back injections 15%

Traditional Medicare patients in six states got 15% fewer epidural steroid injections for back pain after Medicare started requiring prior approval in January. Other services on the list fell 19.5%. That is the first hard evidence on WISeR, Medicare’s AI assisted prior authorization test, from a study published in JAMA on Oct. 26.

The authors, from the University of Minnesota and Brown University, compared monthly use of the targeted services in Arizona, New Jersey, Ohio, Oklahoma, Texas and Washington against all other states and Washington, D.C. They used Epic Cosmos, an electronic health record database with more than 300 million patient records, and looked at two years before launch and six months after.

Why it matters: for decades traditional Medicare almost never asked doctors for permission first. WISeR changed that in six states, and the first data show it changed how much care people got right away. Whether that care was waste is the open question.

What fell, and what didn’t

Epidural steroid injections are the highest volume service in the model, per the University of Minnesota. They fell 15%. The other targeted services fell 19.5%. CMS lists skin and tissue substitutes, electrical nerve stimulator implants and knee arthroscopy for knee osteoarthritis among the services in the model.

Services that were once slated for WISeR but never added showed no change. The researchers say that supports the conclusion that prior authorization, not some other trend, drove the drop.

The effect faded

The biggest drop came right after launch. For epidural steroid injections, the reduction shrank from about 21% in the first quarter of 2026 to about 9% in the second. For other services it went from 21% to 18%.

The authors offer two possible reasons: a new workflow with early kinks, and clinicians learning how to get approvals through. “The early results suggest that doctors may be adapting rapidly to the new system,” said coauthor Michael Barnett, a physician and professor at Brown.

What the study can’t tell you

The study counted services. It did not measure pain, complications, denials or costs, and it covers six months of a six year model.

“The key question we can’t yet answer is whether the care not received was wasteful or necessary,” said lead author Hannah Neprash of the University of Minnesota.

How WISeR works

Six technology companies run the reviews, one per state, according to the CMS model page. They are Cohere Health in Texas, Genzeon in New Jersey, Humata Health in Oklahoma, Innovaccer in Ohio, Virtix Health in Washington and Zyter in Arizona. They are paid a percentage of the spending tied to wasteful or inappropriate care they help avert, adjusted for performance measures that include provider experience.

CMS says WISeR does not change Medicare coverage or payment rules. It excludes emergency services, and licensed clinicians must make every recommendation not to pay. It does not apply to Medicare Advantage.

About 6.4 million people with traditional Medicare live in the six states, Medical Daily reported, citing 2024 figures.

What to watch

CMS has not published its own evaluation. The authors want the next studies to track whether the drop lasts and what it does to patient outcomes, clinician workload and Medicare spending.

For patients in the six states, the practical step is simple. Ask early whether a planned procedure needs approval under WISeR and whether the request has been sent. Denied claims can be appealed through the standard Medicare process.

The model runs six years. The first six months say the paperwork works as a brake. Nobody knows yet what it is stopping.