Medicare's AI prior authorization pilot cuts service use, JAMA study finds
Medicare's AI prior authorization pilot cuts service use, JAMA study finds
The voices on this episode are generated, and so is the script, from the day's post at Physician in the Loop. Every fact in the post links to its source.
Use of services covered by Medicare's AI-assisted prior authorization pilot fell about 15% to 20% in its first six months in the six participating states, according to a JAMA research letter published Oct. 26.
The authors, Hannah Neprash of the University of Minnesota and Michael Barnett of Brown University, said the data cannot show whether the care patients did not receive was wasteful or necessary. Services that were scheduled for the Wasteful and Inappropriate Service Reduction (WISeR) model but never put under review showed no decline, and the reduction in epidural steroid injections narrowed from about 21% to about 9% between the first and second quarters, the University of Minnesota said.
The same day, the Department of Health and Human Services finalized its HTI-5 rule, which removes the requirement that certified health record software carry standard information on how predictive AI tools were built and tested.
In this episode
How technological innovation is changing medicine for patients and the people who practice on the front lines.
Voice 1: This is Physician in the Loop, the daily briefing on AI in medicine. It's Tuesday, October 27th, 2026.
Voice 1: Use of services in Medicare's AI-assisted prior authorization pilot fell between about 15% and 20% in its first six months, according to a new JAMA research letter.
Voice 2: Which services, and compared to what?
Voice 1: Hannah Neprash at the University of Minnesota and Michael Barnett at Brown looked at the Wasteful and Inappropriate Service Reduction model, called WISeR, running in Arizona, New Jersey, Ohio, Oklahoma, Texas and Washington. Epidural steroid injections, the highest-volume service in the pilot, fell about 15% relative to trends in non-pilot states. The other targeted services fell about 20%.
Voice 2: Does the data say whether the care people didn't get was actually unnecessary?
Voice 1: No, and the authors say plainly it can't. Services that were scheduled for WISeR but never actually put under review showed no decline at all, which at least suggests the drop tracks the review itself.
Voice 2: Did the effect hold steady over the six months?
Voice 1: Not for epidural injections. The reduction narrowed from about 21% in the first quarter to about 9% in the second. Barnett said that suggests doctors may be adapting rapidly to the new system.
Voice 2: Adapting could mean learning what gets denied, or just learning how to ask differently.
Voice 1: The letter doesn't separate those two possibilities. It just notes the gap closed fast.
Voice 2: The same day, a different agency was rewriting what software companies have to disclose about their own AI.
Voice 1: HHS finalized its HTI-5 rule on Thursday. It removes 14 of the 60 certification criteria for electronic health record software and revises seven more. The original December proposal would have removed 34, so this is a narrower cut than first floated.
Voice 2: What actually goes away for clinicians using these systems?
Voice 1: The rule strikes the requirement, adopted back in HTI-1, that certified software support and let users view and modify source attributes, the standard information on how a predictive decision support tool was trained and tested.
Voice 2: Why pull that out now?
Voice 1: ONC cited limited demonstrated benefit and the cost of the transparency requirements, and said it may revisit some of this in future rulemaking. HHS projected more than $1.5 billion in savings for developers.
Voice 2: And information blocking, the part everyone was watching?
Voice 1: Left out of the final rule. Those proposals remain under consideration, according to HHS.
Voice 2: Different continent, much higher stakes. The Ebola outbreak in Congo keeps widening.
Voice 1: The World Health Organization said Congo had 9,704 confirmed cases of Bundibugyo virus disease as of October 20th, including 4,687 deaths. Kenya has nine cases, eight of them in a cluster that includes four health and care workers, and Nigeria confirmed an imported case in Abuja on October 20th.
Voice 2: Is there a vaccine for this strain yet?
Voice 1: No vaccine or specific treatment is approved. A phase 2/3 vaccine trial did enroll its first participants on October 23rd, so that work is just starting.
Voice 2: What's happening at the borders while that trial gets underway?
Voice 1: Uganda ordered anyone returning from Congo into 21-day quarantine at their own expense and barred transit for anyone who's been there. The patient in Nigeria was flown to Switzerland on October 26th. WHO itself advises against travel or trade restrictions.
Voice 2: So the agency's own guidance and some countries' actions aren't lining up.
Voice 1: That's the gap the post lays out, yes.
Voice 2: On a smaller scale, measles in the U.S. just hit a milestone no one wanted.
Voice 1: The CDC counted 4,588 confirmed cases in 2026 as of October 22nd, more than double the 2,289 cases for all of 2025. Forty-three new outbreaks this year, 96% of cases tied to an outbreak, and three deaths.
Voice 2: Any sense of where the non-outbreak cases came from?
Voice 1: Eighteen were among international visitors. The agency didn't break down the rest.
Voice 2: A number that keeps climbing, with no sign in this report of it slowing.
Voice 1: Before we go, something to watch.
Voice 2: A CMS deputy administrator, Kimberly Brandt, said earlier this month that there are, in her words, a lot of things about WISeR that are not working.
Voice 1: Is anyone writing that into law?
Voice 2: Alabama already has. Its SB 63 requires a licensed physician or other health care professional to make any medical necessity denial of prior authorization. Pennsylvania has similar legislation, HB 1925, now sitting in the Senate.
Voice 1: One state has it on the books, one state is still waiting.
Voice 2: The other date worth marking is November 1st, when the 2027 Physician Fee Schedule and OPPS final rules are expected.
Voice 1: What's riding on those?
Voice 2: The remote monitoring employed-staff rule, the conversion factor, the G2211 modifier, payment for software as a medical service, and a request for information on AI. That's the statutory target date, not a guarantee.
Voice 1: A rule covering how AI tools actually get paid for is a bigger lever than any single study.
Voice 2: Especially next to a prior authorization pilot the agency's own deputy administrator is already second-guessing.
Voice 1: Sources and corrections are at physicianintheloop.org, where you can sign up for the free email briefing. This briefing was read by AI voices.