Medicare's AI-assisted prior authorisation pilot linked to sharp drop in targeted procedures
A JAMA analysis found that in the first six months of the WISeR pilot, traditional Medicare use of epidural steroid injections fell 15% and other targeted services 19.5% in participating states. The study cannot say whether the care that disappeared was unnecessary.
Medicare’s new pilot that uses artificial intelligence to help decide prior authorisation was followed by an immediate fall in the services it covers, according to a research letter published in JAMA on 26 October 2026.
Background
The WISeR model (Wasteful and Inappropriate Service Reduction), run by the Center for Medicare & Medicaid Innovation, began in 2026 in selected states. It requires prior approval, with AI-enabled review, for certain services in traditional Medicare and is planned to run for six years.
What the researchers found
- Hannah Neprash (University of Minnesota) and Michael Barnett used Epic Cosmos electronic health record data.
- In the first six months, encounters for epidural steroid injections fell by 15.0% and other targeted services by 19.5% relative to baseline in participating states.
- Services whose requirements had not yet started did not change, which supports a link to the pilot.
- The decline was largest at first and appeared to ease over time.
Limitations
- It is a cross-sectional, observational analysis: it shows an association, not proof of cause.
- The data cannot tell whether the procedures that did not happen were unnecessary or needed, or how patients fared; the authors say this raises questions about clinician burden and patient outcomes over the full programme.
- Epic Cosmos covers many, but not all, US health systems.
If a test or treatment you were advised to get is delayed by an authorisation request, ask your doctor about your options.
Also available in: Саха тылаРусскийEspañol中文Portuguêsالعربية