Diagnostic agent keeps 49.4% of cases at 98.9% accuracy; House hears Medicare pay bills; 22 states fund remote monitoring as CMS moves to bar vendor staffing; Abridge adds pre-bill review
Diagnostic agent keeps 49.4% of cases at 98.9% accuracy; House hears Medicare pay bills; 22 states fund remote monitoring as CMS moves to bar vendor staffing; Abridge adds pre-bill review
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.
A diagnostic agent that scores its own consistency kept 49.4% of cases at 98.9% accuracy and routed the rest to a physician, a European group reported in Nature Medicine, dated Sept. 15. The group, led by Jakob Nikolas Kather, built an agent that takes a history, examines the patient, orders labs and imaging, and commits to a diagnosis, all on open-weight models running inside the hospital firewall. The agent reached about 90% on a seven-disease emergency task, close to a cloud GPT-5.2 baseline, the authors said. When the researchers ran the agent five times on each case, its agreement with itself predicted whether it was right better than any other measure they tested; the 49.4% of cases the agent kept were those that cleared a strict threshold on that score, and the other half went to a human. The paper appeared one week after the Advanced Research Projects Agency for Health funded heart-failure agents built on exception-based oversight, the design in which the machine keeps the cases it is sure of and hands the rest to a physician.
Chris Klomp, the deputy secretary nominee, appeared before the Senate Finance Committee on Tuesday and the Senate Health, Education, Labor and Pensions Committee on Wednesday morning, and coverage focused on vaccines and measles, not WISeR or AI. The House Energy and Commerce Health Subcommittee held a legislative hearing Tuesday on the Patients First Act and three other physician-payment bills. Sacramento has said nothing about AB 1979 with 14 days left on the clock. The Food and Drug Administration's only press release announced an expedited investigational new drug (IND) pilot. No major funding round or deal crossed the wire.
In other developments, states are putting rural transformation money into remote patient monitoring while the Centers for Medicare & Medicaid Services (CMS) proposes to bar the vendor-staffed version of the service; Abridge is selling a pre-bill review tool built on its ambient notes to coding departments; and 24 revenue cycle leaders told Becker's Hospital Review that their chief concern is payer automation that denies claims faster than they can answer, not autonomous coding.
In this episode
How technological innovation is changing medicine for patients and the people who practice on the front lines.
Nora: This is Physician in the Loop, the daily briefing on AI in medicine. I'm Nora Davis.
James: And I'm James Hudson. It's Wednesday, September 16th, 2026.
Nora: So a diagnostic agent just kept half its own caseload and was right 98.9% of the time on those cases.
James: Half the cases, at that accuracy, with a machine deciding which half. That's a strange sentence to say out loud.
Nora: It is. And the other half went straight to a physician, no argument. I'm leading with that diagnostic agent story, then the remote monitoring fight between the states and CMS.
James: I've got the House hearing on physician pay, then Abridge moving into the billing office.
Nora: So this comes from Nature Medicine, dated September 15. Jakob Nikolas Kather's group built an agent that takes a history, examines the patient, orders labs and imaging, and commits to a diagnosis, all running on open-weight models inside the hospital firewall.
James: Inside the firewall meaning the data never leaves the building.
Nora: That's the design. On a seven-disease emergency task it hit about 90%, close to a cloud GPT-5.2 baseline. But the real trick was running each case five times and checking whether the agent agreed with itself.
James: And that self-agreement predicted accuracy better than anything else they tried?
Nora: Better than any other measure they tested. At a strict threshold, it kept 49.4% of cases at 98.9% accuracy. Blinded physician review of 181 cases agreed with the automated grading 92.3% of the time.
James: What about the cases it didn't keep?
Nora: Those went to a person. And the authors were upfront about limits: this was retrospective simulation, text only, one institution's data, weaker in older patients, and about five times the compute.
James: So it's cautious by construction. Keep what you're sure of, hand off the rest.
Nora: Right, and the paper landed a week after the Advanced Research Projects Agency for Health funded heart-failure agents built on that same idea, keeping the sure cases and routing the rest to a physician.
James: So two separate efforts landing on the same design within a week of each other.
Nora: That's what the post notes, yes.
James: Alright, my turn. The House Energy and Commerce Health Subcommittee held a legislative hearing Tuesday on 17 bills, including the Patients First Act.
Nora: Which does what?
James: Sponsored by three physician members of the House, John Joyce, Kim Schrier and Greg Murphy. It would tie the fee schedule to the Medicare Economic Index minus one point, minus half a point for alternative payment model participants, and phase out MIPS over five years.
Nora: And the budget-neutrality trigger?
James: Raised from $20 million to $57.64 million in 2028. There's a companion bill, the Provider Reimbursement Stability Act, that only touches that budget-neutrality piece.
Nora: Where does that leave physician pay right now?
James: The AMA says Medicare physician payment has fallen about 33% against practice-cost inflation since 2001. And the conversion factor is set to drop another 1.19% on January 1.
Nora: Was there a vote?
James: No. A legislative hearing isn't a markup, and none is scheduled.
Nora: So testimony now, decisions later.
James: That's the pattern so far.
Nora: Back to me. Fierce Healthcare reports 22 states have put roughly $240 million of rural transformation money into remote patient monitoring, and 35 states have committed more than $2.4 billion to broader remote-care technology, all out of the $50 billion Rural Health Transformation Program.
James: States are betting on this while CMS is doing what, exactly?
Nora: Proposing to bar the vendor-staffed version. The 2027 fee schedule would limit remote physiologic and therapeutic monitoring to clinical staff employed by the billing practice itself.
James: That sounds like it hits small practices hardest.
Nora: The American Telemedicine Association says exactly that, that it would make the service unworkable for them. CMS got nearly 40,000 comments on it.
James: And the codes at stake?
Nora: 99454 and 99457, the ones practices bill through vendors. Medicare RPM payments rose 31% to $536 million in 2024, reaching about 1 million beneficiaries.
James: So states are funding growth in a service CMS might restructure.
Nora: Right as the money's flowing in. The final rule is expected around November 1.
James: Worth watching how those two collide. My last one: Abridge is moving into revenue cycle.
Nora: With what?
James: A pre-bill review tool, Fierce Healthcare reports, that checks coded DRGs against what its ambient scribe actually recorded, before the claim goes out. It doesn't change the note or the codes, it just surfaces evidence for coding and documentation integrity teams.
Nora: Who's using it?
James: Reid Health in Indiana is the named early customer. Abridge says it works with 300 large health systems and handles more than 100 million conversations a year, though those are the company's own figures.
Nora: Shiv Rao had a line about this, didn't he?
James: He did. Ryan's read on this is that Rao's framing captures it: health systems “aren't reimbursed for the care they deliver. They're reimbursed for the care they document that they delivered.”
Nora: So the same recording that writes the note now also protects the bill.
James: One tool, two jobs.
Nora: Before we go, two things to watch.
James: First, AB 1979. Sacramento has said nothing about it, and there are 14 days left on the clock.
Nora: Fourteen days is not a lot of runway for something that's already been quiet this long.
James: No word yet either way.
Nora: Second, Chris Klomp, the deputy secretary nominee, appeared before the Senate Finance Committee Tuesday and Senate HELP Wednesday morning.
James: What came up?
Nora: Coverage focused on vaccines and measles, not WISeR or AI oversight.
James: So the AI questions physicians might want answered didn't get asked this round.
Nora: Not according to the coverage so far.
James: Then that's the gap to watch, whether those questions come up at his next appearance.
Nora: Sources and corrections are at physicianintheloop.org. I'm Nora Davis.
James: I'm James Hudson. See you tomorrow.