Daily Update for September 29, 2026

Show Notes

Colorado funds rural tele-ICU with $169.6 million; coding vendors dispute insurers on AI costs; AI score predicts heart failure deaths; robot data flags surgical skill

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.

Read the post and its sources

Colorado awarded $169.6 million on Monday to 91 rural health care providers for about 250 projects, including tele-ICU, telehospitalist and remote patient monitoring services, under the federal Rural Health Transformation Program, the state said.

The awards cover 52 rural or frontier counties and three rural census tract counties and were chosen from 112 eligible applications seeking more than $250 million, according to the state's Department of Health Care Policy and Financing. The same day, the Centers for Medicare and Medicaid Services announced $51 million for chronic disease prevention through 68 rural hospital districts and authorities in Texas. Neither announcement names an artificial intelligence project.

Makers of hospital coding software disputed insurers' finding that AI-assisted billing added $942 million in costs, Healthcare Dive reported. New papers in npj Digital Medicine describe an Oxford heart failure mortality score, a University of Pennsylvania CT model and a Japanese analysis of robot motion during prostate surgery. It was otherwise a quiet day for new developments.

In this episode

• Colorado awards $169.6 million to 91 rural providers for tele-ICU, telehospitalist, remote monitoring and other projects
• Coding software makers say AI captures missed services as insurers tie $942 million in added costs to hospital AI billing
• Oxford's 11-variable AI heart failure score predicts one-year death with C-index of 0.801 in 373,389 patients, above an adapted MAGGIC score
• Penn radiology team trains CT vision-language model on more than 400,000 scan-report pairs and tests it on 20,000 held-out people
• Robot motion data separate expert from less experienced surgeons in 98 robotic prostatectomies with an AUC of 0.892
• Remote blood pressure monitoring reached 5.2% of eligible patients at a Deep South health system; a few doctors made most referrals
• Optum Health names former AI scribe company chief Wes Donohoe CEO for AI and product innovation

What is Physician in the Loop?

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 Tuesday, September 29th, 2026.

Nora: Colorado just handed out nearly $170 million to 91 rural health providers, all in one day.

James: For what, exactly, in terms of actual work?

Nora: Tele-ICU, telehospitalist coverage, remote patient monitoring, shared technology platforms, digital health tools, that kind of thing. Also today: coding vendors pushing back on insurers over AI billing costs, an Oxford heart failure score, and what robot motion data show about surgical skill.

James: That’s a lot of ground to cover in one grant round. Where’s the money actually landing, county-wise?

Nora: 52 rural or frontier counties, plus 3 rural census tract counties, according to Colorado’s Department of Health Care Policy and Financing.

James: Anything in there specifically tagged as an AI project, or is it all infrastructure?

Nora: No, neither this release nor CMS’s names one specifically.

James: Was CMS busy that same day too, or is that separate?

Nora: Same day, CMS gave $51 million for chronic disease prevention through 68 rural hospital districts and authorities in Texas. Also no AI project named.

James: So two big rural health announcements on the same day, and neither one mentions the technology we usually spend this show on.

Nora: Michelle Mills, chief executive of the Colorado Rural Health Center, called the awards “a direct investment in rural resiliency.”

James: We’ll see what that actually buys, now that the full list’s up on the state’s program page for anyone who wants to dig in.

Nora: A follow-up on the insurers’ AI billing cost story: coding software makers are now pushing back.

James: What’s their argument against that finding?

Nora: Luke Chalker, the Blue Cross Blue Shield Association’s senior vice president of product and data science, said if patients were truly sicker, you’d expect to see more treatment, not just more diagnoses. He pointed to a rise in anemia diagnoses without a matching rise in transfusions, according to Healthcare Dive.

James: And what do the vendors say in their own defense?

Nora: Dr. Travis Bias of Solventum, whose coding products Healthcare Dive says are used by more than 80% of U.S. hospitals, blames the payment system itself: “we pay for volume, not for value.”

James: So the software’s just following incentives that already existed before any of this AI showed up.

Nora: That’s his argument. Hamid Tabatabaie, chief executive of CodaMetrix, which serves more than 500 hospitals, said hospitals without these systems had simply been missing billable services all along.

James: Is anyone actually worried about where this standoff ends up?

Nora: TechCrunch reported that Abridge founder Dr. Shiv Rao described competing automated systems on both sides of a claim as a possible “horrible dystopic future.”

James: Coding AI on one side of the claim, denial AI on the other, and the patient’s bill stuck somewhere in between the two systems.

Nora: Oxford researchers built an 11-variable heart failure score called SIMPLE-HF, tested on more than 370,000 adults from UK primary care records.

James: How’d it actually perform against the older score?

Nora: For 1-year, all-cause mortality, it reached a C-index of 0.801, against 0.735 for an electronic-record version of the MAGGIC score.

James: What does that gap mean once you’re standing at the bedside?

Nora: At a set risk threshold, it caught 177 true events per 1,000 people screened, against 89 for MAGGIC.

James: Does it need an echocardiogram to run, or can any clinic use it?

Nora: No. It’s built from point-of-care variables: age, body mass index, comorbidities, medications. The authors still want external validation and prospective testing before anyone uses it.

James: Promising numbers, though with the usual asterisk about needing real-world testing first.

Nora: Last one: Kobe University researchers looked at instrument motion from the hinotori surgical robot, across 98 robot-assisted prostatectomies by 16 surgeons.

James: What exactly were they trying to find out from all that motion data?

Nora: Whether motion data alone could separate certified proctors from other surgeons. A random forest model hit an AUC of about 89%, and a slimmed-down 3-feature version still worked on surgeons it hadn’t seen before.

James: Did the proctors actually finish the operation faster, or just do it differently?

Nora: On the step they studied, bladder to urethra reconnection, yes: about 46 minutes on average, against nearly 68 minutes for everyone else.

James: That’s a real gap in time. What’s the catch with a study like this?

Nora: The authors call it exploratory: one robot platform, one surgical step, a slow 1-hertz sampling rate, and only 5 anastomotic leaks to check against outcomes.

James: So the motion data might flag surgical skill, but nobody’s using it to grade surgeons yet.

Nora: Before we go, something to watch.

James: The CHAI Fall Symposium is set for October 6th, less than two weeks out.

Nora: Not much else out on it yet.

James: We’ll see what they’re actually ready to say once the symposium starts.

Nora: Sources and corrections are at physicianintheloop.org. I'm Nora Davis.

James: I'm James Hudson. See you tomorrow.