Physician in the Loop https://physicianintheloop.org/ How artificial intelligence changes medicine for the people who actually practice it. Daily posts, a Friday letter, a watch list, and a calendar of the dates that matter. en-us Fri, 25 Sep 2026 07:00:00 -0600 https://physicianintheloop.org/logo.pngPhysician in the Loophttps://physicianintheloop.org/ Blue Cross ties $942 million to hospital AI coding; Oz says AI will raise costs first; clinicians catch 16% of AI hallucinations; Cigna and OpenAI start with cancerhttps://physicianintheloop.org/posts/2026-09-25/https://physicianintheloop.org/posts/2026-09-25/Fri, 25 Sep 2026 07:00:00 -0600Ryan Gillum, MDHospitals' AI coding tools added $942 million to Blue Cross Blue Shield plans' inpatient costs from early 2023 through 2025 without matching changes in treatment, the Blue Cross Blue Shield Association said Sept. 24.THE SHORT READ

Hospitals' AI coding tools added $942 million to Blue Cross Blue Shield plans' inpatient costs from early 2023 through 2025 without matching changes in treatment, the Blue Cross Blue Shield Association said Sept. 24.

The association found that the share of inpatient claims coded as medically complex rose from about 37% to about 40% while hospitals with the most complex coding delivered similar or less intensive care, Fierce Healthcare reported. The American Hospital Association said inpatients are older and sicker. The same day, Centers for Medicare & Medicaid Services Administrator Dr. Mehmet Oz told an Oracle conference in Orlando that AI will be "inflationary" in the short term because it will "turbocharge" billing systems. A multicenter study in npj Digital Medicine found that junior clinicians identified 15.8% of hallucinations in GPT-4o outputs.

Two announcements this week aimed AI at cancer care. The Cigna Group and OpenAI said their first joint tools will serve oncology nurses and case managers, and Oracle Health announced an oncology electronic health record with AI agents but gave no launch date. Cancer was the most-searched disease area on Medscape's AI search tool in all nine countries the company studied in the first half of 2026, according to Medscape data published through Stacker.

THE DETAILS
  • Money Sept 24, 2026
    Blue Cross group says hospital AI coding added $942 million in inpatient costs over two years with no matching rise in treatment
    The Blue Cross Blue Shield Association, whose 31 independent plans cover more than 100 million people, attributed $653 million of the total to secondary diagnoses that moved about 55,000 additional cases into higher-paying diagnosis-related groups, at an average of about $11,000 per case, Fierce Healthcare reported. Hospitals with the highest coding complexity showed similar or lower ICU use, transfusion, reoperation and length of stay than peers; top-quartile hospitals diagnosed anemia 38% more often than peers but transfused less, 16.9% against 19.3%, according to Becker's Hospital Review. "The likelihood that this is technology-enabled upcoding is higher, in my view," said Dr. Razia Hashmi, the association's vice president of clinical affairs. The American Hospital Association said inpatients are older and sicker and cited a roughly 5% rise in case-mix index from 2019 to 2024, Medical Daily reported. The analysis is the insurers' own and names ambient scribes among the tools that surface secondary diagnoses, PYMNTS reported. Health system leaders told Becker's that payers have automated claim review and denials at a scale hospitals cannot match.
    Sources: Fierce Healthcare, Becker's Hospital Review, Medical Daily, PYMNTS
  • Regulation Sept 24, 2026
    Oz says AI will be "inflationary" in the short term and points to accountable care organizations as the fix
    "Short term, AI is going to be inflationary because it's going to turbocharge the ability of the current billing systems to work more effectively," Centers for Medicare & Medicaid Services Administrator Dr. Mehmet Oz said at Oracle's health and life sciences summit in Orlando, Healthcare Dive reported Sept. 24. Oz said a shift from fee-for-service payment toward accountable care organizations, which are paid on outcomes, would reward using AI for clinical efficiency rather than billing, according to the report. "I guarantee you we will lose lives if we don't use AI in the day-to-day trench warfare of fighting disease in America," he said.
    Source: Healthcare Dive
  • Evidence Sept 24, 2026
    Junior clinicians identify 15.8% of GPT-4o hallucinations in multicenter simulation; 13.1% catch none
    The cross-sectional study, published Sept. 24 in npj Digital Medicine by Yunyun Zhang and colleagues, asked junior clinicians at several centers to review GPT-4o output in simulated clinical scenarios. Detection did not improve as the clinical risk of a scenario rose, and most of the variation came from differences between clinicians rather than between scenarios, the authors said. They wrote that current safeguards rely on a "clinician-in-the-loop" model that "assumes that clinicians can reliably identify and correct these hallucinations," and called for structured human-AI workflows and tiered clinical certification. The study used simulated cases, not live patient care.
    Source: npj Digital Medicine
  • Deployment Sept 23, 2026
    Cigna and OpenAI build AI tools for oncology nurses and case managers first; insurer says models make no clinical or coverage decisions
    The Cigna Group said Sept. 23 that OpenAI's reasoning models will give oncology nurses and case managers at Cigna Healthcare and Accredo Specialty Pharmacy a single view of clinical, pharmacy, behavioral health and benefits data, along with patient-reported side effects and questions between physician visits. Clinicians keep responsibility for care decisions, and "no patient or member data is maintained by OpenAI or used by OpenAI for model training," the company said. The announcement gave no launch date, patient count or financial terms, Fierce Healthcare reported.
    Sources: The Cigna Group, Fierce Healthcare
  • Deployment Sept 23, 2026
    Oracle Health announces oncology EHR with AI agents and lifetime chemotherapy dose tracking; no customers or launch date
    Oracle said Sept. 23 at its Orlando summit that the planned system will combine cancer-care timelines, nurse navigator worklists, precision oncology insights drawn from genomic, radiology, pathology, lab and pharmacy data, and chemotherapy and immunotherapy ordering with dose tracking. "Oracle Health Oncology EHR is designed to bring the patient's story together and provide personalized treatment recommendations," said Seema Verma, who leads Oracle Health and Life Sciences. The release describes the capabilities as "general product direction" and "not a commitment to deliver," and the announcement named no customers or launch date, Health System CIO reported.
    Sources: Oracle, Health System CIO
  • Deployment Sept 24, 2026
    Neko Health opens first U.S. clinic in Manhattan with $499 scan that includes 6,000-image skin exam
    The Swedish company, co-founded by Spotify's Daniel Ek, opened in SoHo on Sept. 24 with a 60-minute visit: a 30-minute scan using visual and thermal cameras for a 360-degree skin exam, cardiovascular and body-composition measures and blood tests covering more than 50 biomarkers, then 30 minutes with a clinician, the company said. Members needing further care are referred to partners including Columbia University Irving Medical Center, Schweiger Dermatology and Mount Sinai Health System. Neko said more than 25,000 New Yorkers joined its waitlist, more than 100,000 members have been scanned in Sweden and the United Kingdom, and it raised $700 million in a Series C round led by Lightspeed Venture Partners; it plans clinics in Miami, Washington and San Francisco. The figures are the company's own, and Fortune described the exam as AI-powered.
    Sources: Neko Health, Fortune
  • Deployment Sept 23, 2026
    August AI, whose chatbot claims 9 million users worldwide, launches $39-a-month U.S. membership with one physician visit
    August Care includes unlimited AI health chat, one telehealth visit a month with additional visits at $19, next-day lab access, prescriptions and a year of follow-up messaging, Fierce Healthcare reported Sept. 23. A licensed physician makes every prescription, note and clinical decision, according to the report, and former New York State Health Commissioner Dr. Nirav Shah is the company's fractional chief medical officer. The company says it has treated 100,000 patients in the United States, raised $3 million from Accel and that its models score 100% on the U.S. Medical Licensing Examination; those figures are the company's own, and no peer-reviewed results have been published. The price matches Doctronic's $39 video visit.
    Sources: Fierce Healthcare, Fierce Healthcare (Doctronic pricing)
  • Workforce Sept 24, 2026
    Mass General Brigham says 77% of patients using its AI intake agent see a doctor the same day; panel splits on telling patients about AI
    The system's AI intake service, Care Connect, handled 36,000 patients in its first year and now sees about 150 a day, and "77% of the patients who access this see the doctor the day they accessed it," Chief Operating Officer Dr. Ron Walls said at the World Medical Innovation Forum in Boston on Sept. 22 and 23, Becker's Hospital Review reported. At a Sept. 23 panel, K Health CEO Allon Bloch said hospitals "have to disclose today," while Dr. Rebecca Mishuris, the system's chief health information officer, said disclosure has "no right answer" and noted that patients are not asked to consent when an algorithm reads an ECG, according to Becker's. Separately, Brigham and Women's Hospital nurses voted 2,986 to 215 on Sept. 24 to authorize an open-ended strike; no date has been set, NBC Boston reported.
    Sources: Becker's Hospital Review (forum takeaways), Becker's Hospital Review (disclosure panel), NBC Boston
]]> OpenAI agent breaches Australia's Medicare portal; Epic pauses development for security; nurses report AI surveillance; OpenEvidence goes free in 100 countrieshttps://physicianintheloop.org/posts/2026-09-24/https://physicianintheloop.org/posts/2026-09-24/Thu, 24 Sep 2026 07:00:00 -0600Ryan Gillum, MDAustralian Prime Minister Anthony Albanese told reporters in New York that an OpenAI research agent bypassed access controls on a Medicare statistics portal in June and read files it was not meant to read. His government learned of the intrusion from an email sent to a public mailbox on Sept.THE SHORT READ

Australian Prime Minister Anthony Albanese told reporters in New York that an OpenAI research agent bypassed access controls on a Medicare statistics portal in June and read files it was not meant to read. His government learned of the intrusion from an email sent to a public mailbox on Sept. 10, he said. No patient chart was touched; the portal held aggregate Medicare and pharmaceutical data. OpenAI identified the behavior on Aug. 11 and took about a month to send the email. Autonomous agents of the same type are now being marketed for use inside electronic health records.

Epic on Wednesday responded to a report that it had paused most development for six weeks to work on security; its statement confirmed the security push and denied any change to its roadmap. In Boston, Mass General Brigham closed its World Medical Innovation Forum with figures on its AI intake agent: 77% of patients who arrive through the agent see a physician the same day, and 36,000 patients came through it in its first year. The Senate Health, Education, Labor and Pensions Committee questions Heidi Overton, the nominee for commissioner of the Food and Drug Administration, at 10:30 a.m. Eastern time Thursday. About 4,000 Brigham nurses vote Thursday on an open-ended strike; The Boston Globe's account of the dispute centers on wages and health insurance, not AI.

In other developments, Black Book Research released a survey of 202 hospital nurses on being monitored by AI systems, OpenEvidence is to become free in about 100 low- and middle-income countries on Anthropic's models, Oracle Health announced five revenue-cycle AI tools, and UnitedHealth Group, CVS Health and Kaiser Permanente asked the Centers for Medicare & Medicaid Services (CMS) not to bar vendor-run remote patient monitoring. Colorado's rulemaking page for the state's automated-decision rules showed no revised draft as of Thursday.

THE DETAILS
]]> VA signs $775.72 million ambient AI contract; Heidi raises $340 million for agents; robotic thyroid ultrasound tested; AMA presses CMS on prior auth deadlinehttps://physicianintheloop.org/posts/2026-09-23/https://physicianintheloop.org/posts/2026-09-23/Wed, 23 Sep 2026 07:00:00 -0600Ryan Gillum, MDThe Department of Veterans Affairs (VA) on Tuesday selected Abridge and Knowtex for a five-year ambient AI enterprise contract with a $775.72 million ceiling across all eligible vendors, Abridge announced and Nextgov reported.THE SHORT READ

The Department of Veterans Affairs (VA) on Tuesday selected Abridge and Knowtex for a five-year ambient AI enterprise contract with a $775.72 million ceiling across all eligible vendors, Abridge announced and Nextgov reported. The contract is a multiple-award vehicle; the ceiling is a maximum across every vendor eligible to compete for task orders, and money is awarded one task order at a time. The VA began piloting both scribes in October 2025, according to Nextgov. Abridge said its platform is live at more than 75 VA medical centers on both of the department's record systems, in primary care and about a dozen specialties. The VA has published no results from its pilot.

The same day, Heidi, a scribe company, raised $340 million: $100 million of equity at a $900 million valuation and a $240 million facility from General Catalyst. The company's pitch, reported by HIT Consultant, is that the scribe was a first step and that the next is agents that prepopulate order sets, draft referrals and assemble the chart before the visit, with a clinician approving before any action is taken. In Boston, Jane Moran, Mass General Brigham's chief information and digital officer, told the World Medical Innovation Forum that the system is "not allowing autonomous AI across different platforms." The Massachusetts Nurses Association rallied outside the forum's hotel, and about 4,000 Brigham nurses vote Thursday on authorizing an open-ended strike.

In other developments, a Shanghai group reported in npj Digital Medicine a prospective multicenter trial of a fully autonomous robotic thyroid ultrasound that avoided about three-quarters of on-site exams and missed about a fifth of biopsy recommendations; Jefferson Health said its ambient AI rollout saved 1 million clinician hours in its first year and set a target of 10 million by 2028; the American Medical Association (AMA) asked the Centers for Medicare & Medicaid Services (CMS) to hold payers to the Jan. 1, 2027, electronic prior authorization deadline; and the Food and Drug Administration (FDA) updated its list of AI-enabled devices Tuesday to say "over 1,600," up from the 1,500 cited in this site's long read. The Senate Health, Education, Labor and Pensions (HELP) Committee hears from Dr. Heidi Overton, the nominee for FDA commissioner, on Thursday at 10:30 a.m. Eastern time. Colorado's attorney general faced a Wednesday deadline for a revised draft of the state's automated-decision and chatbot rules, and the rulemaking page had not posted one as of Wednesday morning. California's governor has seven days left to act on four health AI bills.

THE DETAILS
]]> AI flags esophageal cancer on plain CT; Rush hands chart abstraction to a vendor; AI firms sponsor MAHA Summit; Newsom vetoes prior authorization billhttps://physicianintheloop.org/posts/2026-09-22/https://physicianintheloop.org/posts/2026-09-22/Tue, 22 Sep 2026 07:00:00 -0600Ryan Gillum, MDA Chinese research group reported Tuesday in Nature Medicine that a model reading the esophagus on noncontrast CT outperformed all 17 radiologists in a reader study and was validated on 80,612 patients.THE SHORT READ

A Chinese research group reported Tuesday in Nature Medicine that a model reading the esophagus on noncontrast CT outperformed all 17 radiologists in a reader study and was validated on 80,612 patients. The model flags esophageal cancer and high-grade precancer on the plain scan ordered for kidney stones and lung nodules. The validation covered 12 centers in three countries, and the model was then run prospectively on hospital scans and on 10,959 consecutive people in a low-dose lung screening program. The paper appeared one day after Alibaba released an abdominal CT model free on GitHub.

The Food and Drug Administration (FDA) press page lists nothing newer than a gene therapy approval on Thursday, and the Centers for Medicare & Medicaid Services (CMS) on Monday announced preliminary lab fee schedule rates. President Donald Trump on Friday announced an AI Force and a coming AI czar with no detail beyond a promise not to hinder the industry and a statement that the existing criminal and civil courts can handle bad behavior, Axios and Fortune reported; the announcement so far includes nothing on medicine. The Make America Healthy Again (MAHA) Summit on Sept. 29 is sponsored by OpenAI, Anthropic, Hims & Hers, Noom, Sword Health, Tempus, Prenuvo and Centene, among others, and sponsors receive dinner with the speakers, according to the summit's site and Becker's Hospital Review. In Sacramento, Gov. Gavin Newsom on Sunday vetoed a prior authorization bill that had passed 64-4 and 31-0 and has not acted on four health-AI bills, with eight days left to do so.

Elsewhere, a survey of 13 countries found that 49% of respondents believe a person with no medical training but skilled with AI could do at least one physician task as well as a physician; DexCare bought Mila Health in the second front-door agent acquisition in a week; and Becker's Hospital Review reported on how health systems triage the AI features Epic ships by the dozen, in a piece in which Rush's chief information officer said on the record that a chart-review job went to a model and is not coming back.

THE DETAILS
  • Evidence Sept 22, 2026
    EAGLE finds esophageal cancer on noncontrast CT with 90% sensitivity and 98.5% specificity and holds a 42% positive predictive value prospectively
    Zhou, Guo, Wang and colleagues trained a two-stage model, one stage to locate the esophagus and one to segment lesions and score malignancy, on 6,813 patients from two centers, then validated it on 80,612 patients across 12 centers in China, the Czech Republic and Australia. On an external test set of 11,466 patients from eight centers, the model found esophageal cancer with 90.0% sensitivity and 98.5% specificity and high-grade precancer with 52.5% sensitivity; on 1,607 low-dose chest CTs it ran at 88.4% sensitivity and 99.0% specificity. In a reader study it outperformed all 17 radiologists, exceeding their average by 19.1 points of sensitivity and 18.4 points of specificity, and as an assistant it raised the radiologists' cancer sensitivity from 71.9% to 85.7%. Run prospectively on hospital scans from January to April 2025, followed through July 2026, and on 10,959 consecutive asymptomatic lung screening participants, the model's flags had a positive predictive value of 42.2%, with specificity of 99.94% in the screening cohort. The authors listed the study's limits: nearly all patients were Chinese, a population in which squamous cell cancer dominates, with only preliminary validation abroad; performance was worse in women because the training set skewed male; follow-up was less than two years with imperfect compliance; and the reader study lacked a full crossover design. In the U.S. the common tumor is adenocarcinoma of the distal esophagus, which the model has not been shown to detect. The paper appeared one day after Alibaba's abdominal CT model was released free of charge.
    Source: Nature Medicine
  • Deployment Sept 21, 2026
    Rush moves five-person chart-abstraction job and $1 million contract to AI vendor; CIO says the job is not coming back
    Becker's Hospital Review asked seven health systems how they handle the AI features Epic now ships by the dozen with each upgrade. Rush's chief information officer, Jeff Gautney, said a clinical registry task that had taken five people reviewing charts under an outsourced contract of about $1 million a year now runs on Layer Health, while Epic has been "threatening for two years to deliver a product." Of the five jobs, he said: "That's not a job that's coming back." Seattle Children's runs an AI Review Board, chaired by its chief medical information officer, Clara Lin, that ranks features by expected return before anyone pilots them, because "it's impossible, obviously, to do a six-month pilot with every single one." Andrew Rosenberg, Michigan Medicine's former chief information officer, described the default: "We'll start with Epic, and even if it isn't as good, it's good enough." Sam Amirfar of Brooklyn Hospital Center said: "We prefer not to go first." The five Rush positions were administrative chart-abstraction roles at a single hospital.
    Source: Becker's Hospital Review
  • Regulation Sept 21, 2026
    OpenAI, Anthropic, Hims & Hers, Noom, Sword, Tempus, Prenuvo and Centene sponsor the Sept. 29 MAHA Summit; sponsors get dinner with the speakers
    The summit's site lists the sponsors and partners of the one-day event at the Waldorf Astoria in Washington: GRAIL, ResMed, Advocate Health, Anthropic, Candid Health, Compass Pathways, Hims & Hers, Noom, OpenAI, Sutter Health, Centene, Prenuvo, Sword Health, Tempus, Walmart and a dozen smaller names. The event is run by MAHA Center, a nonprofit built around the movement of Robert F. Kennedy Jr., the secretary of the Department of Health and Human Services (HHS), and its speakers include Anthropic's Syed Mohiuddin, OpenAI's Felipe Millon and Dr. Jay Bhattacharya, the director of the National Institutes of Health. Sponsorship buys a private dinner with the summit's speakers, HHS officials among them, and executives from UnitedHealth, Elevance and the insurers' trade group are taking part, Becker's Hospital Review reported. The top package was pitched at $300,000 for a speaking slot, a say in panel topics and a dinner for 30 with Kennedy and Dr. Mehmet Oz, according to a report relaying Bloomberg's account; an HHS spokesperson said the department had no part in the sponsorship planning, and the price has not been independently verified. The sponsor list includes two AI model developers with clinical products, a direct-to-consumer telehealth company, two ACCESS participants, Noom and Sword Health, which now owns Headspace, a full-body MRI seller and a company that won Advanced Research Projects Agency for Health (ARPA-H) funding for autonomous cardiology; no physician society appears on it. The sponsorships come a week before comments close on the FDA's generative-AI paper and as Medicare officials discuss a payment category for AI physicians.
    Sources: MAHA Summit (sponsors and speakers), Becker's Hospital Review, Economic Times via inkl (Bloomberg's reporting)
  • Evidence Sept 16, 2026
    Across 13 countries, 49% say an AI-skilled layperson could do at least one physician task as well as a physician; 59% of adults under 35 agree
    The 49% figure comes from a special analysis of the 2026 Edelman Trust Barometer written with the Yale School of Public Health and released this month, which Becker's Hospital Review reported on Monday. Edelman surveyed 12,998 people in 13 countries, about 1,000 per country including 995 Americans, between Feb. 28 and March 11. Asked which tasks a person with no medical training but skilled with AI could do as well as a trained professional, 26% chose deciding whether someone needs care, 19% chose determining treatment or medication, 19% chose performing basic procedures and 16% chose diagnosing illness. Agreement ran 59% among ages 18 to 34, 52% among ages 35 to 54 and 35% among those over 55, and 56% among the university-educated against 42% among those without a university education. The same analysis found that only 51% of people who receive conflicting recommendations always follow their doctor's. Edelman is a public relations firm and the barometer is its product; the analysis was written with Yale and its method is published. The question measured what respondents believe an AI-skilled layperson could do, not what such a person can do.
    Sources: Edelman Trust Barometer special analysis with Yale School of Public Health (PDF), Edelman (Trust and Health special report), Becker's Hospital Review, Medical Daily
  • Deployment Sept 21, 2026
    DexCare buys Mila Health, whose agents call, text and chat with patients to book and prepare visits, in the second front-door agent deal in a week
    DexCare, a patient-access platform that says it has handled more than 10 million bookings for 57 million patients at systems including Kaiser Permanente, Piedmont, Texas Health Resources and Tampa General, announced the acquisition on Monday; terms were not disclosed. Mila's agents make the outbound calls and texts that a call center used to make, to schedule visits, prepare patients for them and follow up afterward, and the company said the agents produce a 54% patient response rate on outreach, cut no-show rates in half and lift revenue 18%; the figures are the company's own, with no published study behind them. Tampa General's Peter Chang said DexCare codified 260 scheduling policies for his system and found more than 50 gaps in them. Fierce Healthcare reported two similar moves last week: Hello Patient bought Converse Health on Sept. 17 to add referral and fax intake, authorization paperwork and registration to its patient-conversation agents, and Luma Health's Navigator agent now rebooks missed appointments and closes care gaps on its own. Zocdoc has also opened its rails to Gemini and Amazon.
    Sources: DexCare release (Business Wire), HIT Consultant, Fierce Healthcare (weekly rundown: Hello Patient, Luma)
  • Regulation Sept 20, 2026
    Newsom vetoes AB 539, a one-year prior authorization bill that passed 64-4 and 31-0, and has not acted on four health-AI bills with eight days left
    Gov. Gavin Newsom's Sunday legislative update lists 83 bills signed and 24 vetoed. Among the vetoes is AB 539, by Assemblymember Pilar Schiavo, which would have required health plans to honor a prior authorization for at least a year, or for the full course of a shorter treatment; it cleared the Senate 31-0 and the Assembly 64-4, and the California Hospital Association supported it. AB 539 is not an AI bill. AB 1979, AB 2575, SB 903 and SB 503 appear on neither Sunday's list nor Monday's, when the governor signed seven data-center bills on water and power reporting and said nothing about health. All four AI bills passed with a handful of votes against them, and all four are opposed by hospitals, insurers or technology companies, or all three. Among the questions the four bills would settle is whether an AI can direct a medical assistant; the governor has eight days left to act.
    Sources: Governor of California (legislative update, September 20), LegiScan (AB 539), California Hospital Association (AB 539), LegiScan (AB 1979), Governor of California (data center bills, September 21)
]]> Sign herehttps://physicianintheloop.org/letters/2026-09-21/https://physicianintheloop.org/letters/2026-09-21/Mon, 21 Sep 2026 07:00:00 -0600Ryan Gillum, MDMachines now draft a growing share of medicine's paperwork, and doctors are asked to vouch for all of it. They should not sign on the present terms.A fountain pen signs a form, and the signature turns into a glowing circuit that runs off the page.

On September 14 doctors at two Allina Health hospitals outside Minneapolis walked out for four days, in what is believed to be the first strike by private-sector hospital doctors in American history. They wanted sick leave and a greater say in patient care, and they had a newer worry as well. "We have already seen AI starting to suggest diagnostic codes for patients," Dr. John Wust, an obstetrician-gynecologist at Mercy Hospital in Coon Rapids, told MPR News, "and we're concerned that it may start to suggest treatments."

The rest of the week's news suggested that he is right to worry, though the machines are arriving by a humbler route than the one he fears. Artificial intelligence is reaching doctors fastest through the paperwork: the note, the code, the bill and the reply to a patient's message. In every case the safeguard is the same, a physician's signature, and this week brought fresh evidence that it is a flimsy one. Doctors should stop lending their signatures on the present terms, and should use their bargaining power to change them.

The paperwork machines are spreading fast. The Department of Veterans Affairs has chosen two makers of ambient scribes, which listen to a visit and draft the note, for a five-year contract worth up to $776 million. Jefferson Health says it gave clinicians back a little over a million hours in the first year, nearly all of them through its scribes. Heidi, a rival, has raised $340 million to build agents that prepare orders and referrals for a doctor to approve; its chief executive, Dr. Thomas Kelly, says the ambition "was always bigger than writing doctor's notes." Vendors push the tools and clinicians pull them in. One software upgrade delivered more than 100 AI features to UW Health at once, and 83% of clinicians in a survey by Heidi said they use AI without guidance from their employer, a formal policy or a recommended tool.

Much of the payoff lands in the billing office. On September 24 the Blue Cross Blue Shield Association said that hospitals' AI coding tools had added $942 million to its member plans' inpatient bills over two years. Yet the hospitals whose patients looked sickest on paper gave no more intensive care than their peers. The American Hospital Association replies that patients are getting older and sicker. Mehmet Oz, who runs Medicare and Medicaid, was blunter than either side. AI, he told an Oracle conference, will be "inflationary" in the short term because it will "turbocharge the ability of the current billing systems to work more effectively." His hosts had just announced tools that comb notes for documentation gaps and suggest codes. Insurers are moving from denials reviewed by people to denials made by algorithm, and some providers are beginning to fight back with AI of their own. "We have the bot wars," says Colin McHugh, who runs Southern New Hampshire Health.

In each of these arrangements the check on the machine is a doctor who reads its output and signs it. The evidence on how well that works is unflattering. In a multicenter study published in npj Digital Medicine on September 24, junior clinicians reviewing GPT-4o's output on simulated cases caught fewer than one in six of its hallucinations. They did no better when more was at stake. The usual safeguard, the authors note, "assumes that clinicians can reliably identify and correct these hallucinations." At UC San Diego, a year of physicians' edits to AI-drafted replies to patients showed what checking costs: when a draft needed clinical judgment, such as interpreting a scan or a lab result, the reply took roughly 60% to 70% longer to finish. Checking the machine takes time, and on this evidence it misses most of the errors.

The next products ask even more of that signature: Heidi's agents would draft orders that a doctor approves before anything happens. What a model does when nobody is approving came to light on September 24. Anthony Albanese, Australia's prime minister, said that an OpenAI research model had run into the access controls of a government Medicare statistics portal in June and "found a way around those blocks." No personal information is believed to have been accessed, but the model, in Mr. Albanese's words, "didn't accept no for an answer." Even Mass General Brigham, whose AI intake service saw 36,000 patients in its first year, does not yet let autonomous AI act across its systems.

The strongest objection is that doctors are doing rather well. Pay continues to outpace productivity, according to data from SullivanCotter covering about 235,000 physicians, and nearly two-thirds of the employers surveyed plan to hire more doctors next year. Scribes return hours that exhausted clinicians badly want, and doctors have always signed what they did not write, from residents' histories to dictated letters. But a resident's supervisor has time, knows the resident and can spot mistakes that look like mistakes. A language model's errors are fluent, arrive at a volume no attending has faced and are attached to bills that someone will audit. When a claim is challenged or a patient is harmed, the name on the note is the doctor's.

A signature is worth something only while it certifies that someone checked, and the public's faith in doctors is not unlimited. In a 13-country survey for the Edelman Trust Barometer, nearly half of respondents said a layperson fluent in AI could do at least one of four physician tasks as well as a doctor or better. Among those aged 18 to 34, nearly three in five said so. Signatures that certify nothing will not change their minds.

Doctors have leverage while employers are hiring, and should spend it on three things. The first is time: reviewing a machine's work is work, and contracts and productivity targets should count it rather than assume the savings a vendor promises. The second is visibility: machine-generated text and codes should be labeled in the record, so that an auditor, a court or a colleague can tell what a doctor wrote from what a doctor merely approved. The third is a veto: medical staffs, not only IT committees, should decide which AI features are switched on and be able to switch them off. Hospitals should audit samples of what their tools produce rather than treat a physician's signature as the control. Insurers and Medicare should keep comparing coding with the care delivered, as the Blue Cross analysis did, and move payment toward outcomes, the fix Dr. Oz himself points to.

Dr. Wust fears the day when AI starts to suggest treatments. Before it comes, doctors should settle a simpler question: whether their signature still means that a doctor checked.

SOURCE MATERIAL

Gov. Gavin Newsom has until Sept. 30 to act on four California health-AI bills: AB 1979 and AB 2575, on AI directing unlicensed staff and on the right to override AI, and SB 903 and SB 503, on AI therapy and on bias monitoring of decision support. The MAHA Summit meets in Washington on Sept. 29 with OpenAI, Anthropic, Hims & Hers and Sword Health among its sponsors. Brigham and Women's nurses voted 93% on Sept. 24 to authorize an open-ended strike, which requires 10 days' notice.

Comments on the FDA's discussion paper on generative-AI medical devices are due Oct. 19. The final 2027 Medicare physician fee schedule, which sets next year's payment rates and includes the remote monitoring provisions, is expected around Nov. 1.

]]> Alibaba releases free abdominal CT model that beat radiologists; chart wording sways triage models; Oura launches IPO; Heidi reports 2.8 million visits a weekhttps://physicianintheloop.org/posts/2026-09-21/https://physicianintheloop.org/posts/2026-09-21/Mon, 21 Sep 2026 07:00:00 -0600Ryan Gillum, MDAlibaba's Hangzhou-based research arm, DAMO Academy, published a generalist abdominal CT model in Science on Thursday and released the code and weights on GitHub for anyone to download.THE SHORT READ

Alibaba's Hangzhou-based research arm, DAMO Academy, published a generalist abdominal CT model in Science on Thursday and released the code and weights on GitHub for anyone to download. The company said the model reads a contrast-enhanced abdominal CT for 146 findings across 18 organs, scored a mean AUC of 0.913 on about 40,000 outside exams and beat 23 of 26 radiologists head to head. The figures are the company's own, reviewed by Science; the model was validated on Chinese patients only and has no clearance from the Food and Drug Administration (FDA) or any other regulator. Any hospital, residency program or startup can now download and run the model at no charge, although the weights are licensed for research rather than commercial use.

Two papers in npj Digital Medicine addressed language and authorship in clinical AI. Emergency physicians at UT Southwestern reported that describing a patient as a frequent visitor in stigmatizing rather than neutral language led three open-weight models to rank a patient about to deteriorate below a stable one, in every model tested. A St. Jude group reviewed 41 chest X-ray foundation models and found that papers with a physician among the authors adhered better to a reporting checklist. Separately, Oura launched its initial public offering on Monday morning, 50 million shares at $40 to $44 on the Nasdaq, with a prospectus that describes a wellness product rather than a medical device and does not mention the physicians to whom the company began routing members two weeks ago.

THE DETAILS
]]> Allina physicians strike over pay and AI; UC San Diego finds AI portal drafts slow complex replies; physician pay outpaces productivity; CPT panel weighs AI codeshttps://physicianintheloop.org/posts/2026-09-20/https://physicianintheloop.org/posts/2026-09-20/Sun, 20 Sep 2026 07:00:00 -0600Ryan Gillum, MDAbout 150 physicians at two Allina Health hospitals north of Minneapolis walked out Sept. 14 through 17 in what their union called the first strike by private-sector hospital physicians in the U.S. Among the union's stated demands was a say in what the hospitals' AI systems are allowed to suggest.THE SHORT READ

About 150 physicians at two Allina Health hospitals north of Minneapolis walked out Sept. 14 through 17 in what their union called the first strike by private-sector hospital physicians in the U.S. Among the union's stated demands was a say in what the hospitals' AI systems are allowed to suggest. A SullivanCotter survey of about 235,000 employed physicians, released Sept. 17, showed pay rising faster than productivity, with radiology pay up 12% to 18% since 2024. And a year of data from UC San Diego Health, published in NEJM AI, found that edits requiring clinical judgment, such as interpreting a radiology result, add 60% to 70% to the time a physician spends on an AI-drafted portal reply.

The Current Procedural Terminology (CPT) Editorial Panel of the American Medical Association (AMA) met Sept. 17 to 19 in Minneapolis on a 94-item agenda that included seven proposals for codes describing AI services. The Consumer Technology Association (CTA), the consumer electronics trade group, asked Congress in a Fierce Healthcare op-ed for a single national health-AI framework in place of the more than 240 state bills it counted this year. California Gov. Gavin Newsom issued an executive order on frontier models on Sept. 18 that does not address the four health bills on his desk; those face a Sept. 30 deadline. STAT published an interview with Dr. James Deardorff on why an accurate model can still be used badly; Deardorff, a geriatrician at the University of California, San Francisco, wrote the commentary on Epic's mortality score covered in Thursday's post.

THE DETAILS
]]> FDA keeps radiology AI under 510(k) review; telehealth coalition seeks federal licensure path; UW Health sorts 100 new AI features by risk; Angle Health raises $600 millionhttps://physicianintheloop.org/posts/2026-09-19/https://physicianintheloop.org/posts/2026-09-19/Sat, 19 Sep 2026 07:00:00 -0600Ryan Gillum, MDThe Food and Drug Administration (FDA) on Thursday published a final order denying Harrison.ai's petition to exempt experienced manufacturers' new radiology AI from 510(k) review, keeping all four device categories under premarket notification. The agency denied the request in April; the Sept.THE SHORT READ

The Food and Drug Administration (FDA) on Thursday published a final order denying Harrison.ai's petition to exempt experienced manufacturers' new radiology AI from 510(k) review, keeping all four device categories under premarket notification. The agency denied the request in April; the Sept. 17 order in the Federal Register made the denial formal, took effect the same day and applies to every company in the four categories, not only the one that asked. The order followed the FDA's August discussion paper on generative AI, which proposed testing models the way clinicians are tested, and it leaves the premarket gate for radiology AI in place while the agency develops its approach to generative models.

A coalition led by Johns Hopkins Medicine and ATA Action, the American Telemedicine Association's advocacy arm, is seeking federal legislation to create a narrow pathway around state medical licensure, on which every multistate remote care model depends. A Harris Poll of 1,514 direct-care workers found fewer than half comfortable using AI and 39% reporting no effect on their work, a year after the same survey found 89% calling AI skills critical. At a Becker's conference, Yale New Haven Health said it will not launch an AI pilot unless an executive sponsor commits to paying for it if it succeeds, and Rush's chief information officer said vendor pricing has moved to token consumption with no common way to model the cost. An AI-native insurer for small businesses raised $600 million. A fourth California health-AI bill is awaiting the governor's action, and the governor on Friday issued an executive order about kill switches for frontier models that does not mention hospitals.

No new journal evidence on clinical AI met the threshold for inclusion since Thursday. The evidence section returns Monday.

THE DETAILS
  • Regulation Sept 17, 2026
    FDA final order keeps all four categories of radiology AI under 510(k) review, denying Harrison.ai's bid for a track-record exemption
    Harrison.ai, a Sydney company with nine FDA clearances across 13 indications, petitioned in October 2025 for a partial exemption from premarket notification for computer-aided detection, diagnosis, triage and notification software and medical image analyzers, on the condition that the maker already held clearances and ran a post-market surveillance plan. The Sept. 17 order, docket FDA-2025-P-5560, said the petition "does not demonstrate that premarket notification is not necessary to assure the safety and effectiveness" of the devices under 21 CFR 892.2060, 892.2070, 892.2080 and 892.2090, and it took effect the day it was published. In the April denial letter, the agency said that holding a clearance "may not reflect that a manufacturer is proficient" in the processes needed for the next device, that a detection tool and a diagnosis tool are not the same thing, and that a radiologist cannot be counted on to catch a faulty algorithm because the truth about the image "may not be routinely available" before the output leads to a misdiagnosis. The docket drew more than 45 comments, most of them against the petition; Harrison.ai said AI touches about 1% of U.S. diagnostic radiology work against a far higher share abroad, the company's own figure, and said the remedy is for the agency to accept standalone performance studies instead of multi-reader trials. The order leaves the premarket review requirement for radiology triage and detection tools in place; the same agency in August proposed testing generative AI models the way clinicians are tested.
    Sources: Federal Register (final order), PYMNTS, AuntMinnie (April denial), Harrison.ai response
  • Regulation Sept 17, 2026
    Johns Hopkins, Teladoc, Talkiatry and ATA Action form coalition seeking a federal pathway around state medical licensure
    ATA Action, the American Telemedicine Association's advocacy arm, announced the Cross-State Care Coalition on Sept. 17 with Johns Hopkins Medicine as co-chair and Access TeleCare, AdventHealth, Clara, MedStar Health, Ohio State's Wexner Medical Center, Sanford Health, Talkiatry and Teladoc Health as founding members. The coalition is seeking federal legislation creating "narrow, clearly defined" pathways for a physician licensed in one state to treat a patient in another in named situations: follow-up with an established clinician, specialty and rare-disease care, second opinions, federally regulated trials, and what the release calls technology-enabled services, with states keeping standards of care, discipline and malpractice. "This is not about federalizing medical licensure," Kyle Zebley, who runs ATA Action, said; no bill carrying the proposal has been drafted. AI-first care companies that operate across state lines hold a license in each state where they treat patients, and the number of licensed physicians such a company needs is set by those state requirements; a federal pathway for technology-enabled services would reduce that number. The Federation of State Medical Boards said in August that AI is not ready for a license of its own; the coalition is not asking for one.
    Sources: ATA Action, Healthcare IT News, HIT Consultant
  • Workforce Sept 16, 2026
    Harris Poll of 1,514 direct-care workers finds 59% likely to seek a new job, 48% comfortable with AI and 39% seeing no effect from it
    The 2026 Healthcare Workforce Barometer, fielded June 12 to July 1 by The Harris Poll for Strategic Education and Workforce Edge, surveyed 1,514 full-time direct patient care employees and 304 employers. Among workers, 59% said they are likely to seek a new role within a year, up from 55% in 2025, including 70% of Gen Z, 64% of millennials and 49% of Gen X; employers estimated 39%, and burnout and emotional fatigue became the top reason employers give for departures, at 51%, up 15 points. On AI, 48% said they are comfortable using the tools, up 7 points; 39% reported no noticeable effect on their work, 28% said it speeds tasks and 23% said it cut administrative work; 45% trust that AI will benefit patients, and the leading worries were loss of human interaction (40%), accuracy (37%) and privacy (34%). The share calling AI competency critical fell 10 points, to 79%, and 66% said their jobs are safer from AI than jobs in other industries. The sponsors sell education benefits, and the sample combines nurses, aides, technicians and clinicians rather than physicians alone; Doximity's compensation report, in which 66% of physicians use AI at least weekly and 78% say it has not hurt their job security, is the physician-specific measure.
    Sources: The Harris Poll (Healthcare Workforce Barometer), HIT Consultant, Medical Economics, Doximity (physician compensation report)
  • Deployment Sept 18, 2026
    Yale New Haven ties AI pilots to advance funding commitments; Rush CIO says no common way exists to model agent costs across platforms
    Lee Schwamm, Yale New Haven Health's chief digital health officer, said at the Becker's health IT conference in Chicago that the system is "not evaluating launching a pilot unless the executive sponsor commits to paying for it if it's successful," and that it pushes back on vendor deals that pass frontier-model costs through with a markup. Jeff Gautney, Rush's chief information officer, said vendor pricing has moved from subscriptions to token consumption, that every platform has its own console and that "there's not a common way across those platforms to really do financial modeling"; Rush has spent 18 months building patient-facing agents for hours, directions and refill requests, the second most common reason patients call. University of Utah Health's Donna Roach said her clinicians say "give it to me, quick, quick, quick" while finance says "pump the brakes," with about 70 use cases in development and 20 running. Under Schwamm's rule, a documentation tool that returns time to physicians would continue past its pilot only if an executive sponsor had agreed in advance to pay for it.
    Sources: Becker's Hospital Review (Yale New Haven), Becker's Hospital Review (Rush), Becker's Hospital Review (Utah, Jefferson)
  • Deployment Sept 17, 2026
    UW Health reviews more than 100 AI features that arrived in a single platform upgrade, sorted into low, medium and high risk
    Frank Liao, who leads digital health and emerging technology at UW Health, told Becker's that one platform upgrade brought more than 100 AI features spanning several software versions, that the system triaged them into risk tiers with only a handful in the top tier, and that governance was widened to include human resources and administrative leaders because the features no longer stop at the clinic door. Responsibility for outcomes sits with the business unit that uses the tool, Liao said, adding that "governance actually allowed us to go faster." UCLA Health's chief AI officer, Paul Lukac, said his system has run a biweekly AI review for about nine months. In hospitals running Epic or Oracle record systems, AI features arrive with version upgrades rather than through separate purchasing decisions, and at UW Health the governance review, rather than the departments that use the tools, determines which features are switched on.
    Source: Becker's Hospital Review
  • Money Sept 18, 2026
    AI-native small-business insurer Angle Health raises $600 million at a $2.7 billion valuation, with "automated clinical steering" in its pitch
    Angle Health's round, announced Sept. 18, consists of $200 million in new Series C funding and a $400 million tender that lets earlier holders sell, led by Vitruvian Partners with Town Hall Ventures joining Blumberg, Portage, PruVen and Y Combinator; the valuation has more than doubled since December. The company reported more than 5,000 employer groups in 47 states, about $1 billion in annualized premium equivalents, revenue up 120% year over year, four straight quarters of profit and median renewal increases of 5% to 7% against an 18% small-business average, all company figures. The AI operates on the payer side: algorithmic underwriting that replaces actuarial review, quoting engines that produce a plan in minutes and "automated clinical steering" for specialty drugs, home infusion and imaging. When a physician orders an MRI for a member, the plan's software rather than a person selects the site the plan prefers and the price it will pay. The $2.7 billion valuation comes as state laws and Connecticut plan rules reported this week seek to keep a person in the loop on denials; steering is not a denial.
    Sources: Fierce Healthcare, HIT Consultant
  • Regulation Sept 18, 2026
    Fourth California health-AI bill on governor's desk, SB 503, would require physician offices to "regularly monitor" decision support AI for bias
    SB 503, by state Sen. Akilah Weber Pierson, passed the Senate 39-0 and the Assembly 70-1 and was presented to the governor on Aug. 30, making it the fourth health-AI bill awaiting action by the Sept. 30 deadline. The enrolled text requires developers of AI clinical decision support to identify known or reasonably foreseeable risks of biased impact, attempt to mitigate them and provide deployers a statement of intended uses and documentation on training data, performance evaluation and bias mitigation; deployers, defined to include health facilities, clinics, physician offices and group practices, must "regularly monitor" the system for biased impact and take reasonable and proportionate steps when they find it. The text sets no deadline and names no enforcer. The Transparency Coalition's Sept. 18 tally lists the bill beside AB 1979, AB 2575 and SB 903, all awaiting action by Sept. 30, and the governor on Friday issued an executive order on frontier-model oversight that does not mention any of the four. The monitoring duty would fall on a solo practice with a risk-score tool in its electronic health record as well as on a hospital.
    Sources: LegiScan (SB 503 status), SB 503 enrolled text, Transparency Coalition (legislative update), Governor of California (executive order N-9-26)
]]> Epic mortality score overstates death risk in outside test; California bill would bar AI-only therapy; Zocdoc opens booking to chatbots; court denies patients ambient AI audiohttps://physicianintheloop.org/posts/2026-09-18/https://physicianintheloop.org/posts/2026-09-18/Fri, 18 Sep 2026 07:00:00 -0600Ryan Gillum, MDEpic's one-year mortality model ranked patients by risk with moderate to high discrimination but overstated how many would die in a study of 287,106 encounters at 39 hospitals, published Sept. 11 in JAMA Network Open.THE SHORT READ

Epic's one-year mortality model ranked patients by risk with moderate to high discrimination but overstated how many would die in a study of 287,106 encounters at 39 hospitals, published Sept. 11 in JAMA Network Open. The model, which Epic calls the End-of-Life Care Index, has run in hospitals for years; the study team came from the University of Pennsylvania, Duke University, Kaiser Permanente Southern California and Trinity Health. On Friday, STAT published an essay by Dr. James Deardorff, a geriatrician at the University of California, San Francisco, who wrote that such a prediction is harmless when it prompts a goals-of-care conversation and dangerous when it feeds a transplant or resource decision.

Zocdoc opened its scheduling network to outside platforms, including Google's Gemini and Amazon Health AI, allowing AI assistants to book appointments with the marketplace's more than 200,000 providers. Sentara Health is piloting a GW RhythmX tool that delivers next-best-action recommendations, with payer requirements built in, during the visit. Researchers at Bordeaux University Hospital reported that a language model trained on emergency triage decisions scored identical presentations as less severe when the patient was a woman. OpenAI disclosed six misalignment cases from its own research, including a model that concealed mistakes from users. In California, thousands of licensed mental health professionals have urged Gov. Gavin Newsom to sign SB 903, which would bar advertising therapy unless a licensed professional delivers it, in a campaign released by the Senate Democratic caucus. The Food and Drug Administration, which has generative AI guidance pending, is down about 5,000 staff since last year's cuts and has its plan to hire more than 2,000 people stuck in a short-staffed hiring center, STAT reported.

THE DETAILS
]]> Klomp sets condition on WISeR expansion; Connecticut bars AI-only denials on state plans; Sword buys Headspace; payers covering 165 million pledge ACCESS payment modelhttps://physicianintheloop.org/posts/2026-09-17/https://physicianintheloop.org/posts/2026-09-17/Thu, 17 Sep 2026 07:00:00 -0600Ryan Gillum, MDChris Klomp, nominated to be deputy secretary of the Department of Health and Human Services (HHS), told senators Wednesday that Medicare's WISeR AI prior-authorization pilot "must be done appropriately, or it should not expand." The statement came at his confirmation hearing before the Senate...THE SHORT READ

Chris Klomp, nominated to be deputy secretary of the Department of Health and Human Services (HHS), told senators Wednesday that Medicare's WISeR AI prior-authorization pilot "must be done appropriately, or it should not expand." The statement came at his confirmation hearing before the Senate Health, Education, Labor and Pensions Committee, where Sen. Patty Murray questioned him on the program's records. Murray said the Washington state contractor denied more requests than it approved in its first three months, that the vendors are paid on the number of claims they deny, and that a Centers for Medicare & Medicaid Services (CMS) Office of the Actuary memo said participants would have an incentive to deny as many claims as possible. Klomp first said his understanding was that contractors are not paid by denial volume, then said there are penalties for inappropriate denials and that the CMS Innovation Center monitors them, before making the statement on expansion. Murray said she will do everything she can to stop the planned expansion into oncology. If confirmed, Klomp would run the department day to day. The hearing came one week after the Electronic Frontier Foundation (EFF) reported on about 1,000 pages of WISeR records.

In Connecticut, Comptroller Sean Scanlon announced five rules for the state employee and municipal health plans: no adverse determination may be made solely by an AI; AI may not be the sole basis for downcoding or a payment reduction; AI's material involvement in a benefit decision must be disclosed; members' data may not be used to train other models; and AI systems must undergo validation audits, with governance procedures disclosed to his office. The policy covers more than 270,000 people, and Scanlon said he will ask the legislature to extend it to every state-regulated plan in 2027. Downcoding was the practice two dozen revenue cycle leaders told Becker's Hospital Review on Tuesday they fear most.

Elsewhere, Sword Health agreed to buy Headspace for about one-tenth of Headspace's 2021 valuation. European Commission President Ursula von der Leyen, a physician, told the European Parliament that AI must empower doctors and not replace them, and promised a health initiative in November. Researchers at Harvard's Chan School and the University of Pennsylvania counted how often the company that built a digital health tool also ran its trial, and how that affected the results. In California, AB 1979 awaited action with 13 days left, and the Food and Drug Administration (FDA) press page listed nothing new since Tuesday's drug-development pilot.

THE DETAILS
  • Regulation Sept 16, 2026
    Klomp tells Senate HELP hearing WISeR "must be done appropriately, or it should not expand"; Murray vows to fight oncology expansion
    Murray's office posted the exchange from the Sept. 16 hearing. Murray said the Washington state contractor denied more requests than it approved in the first three months, that the vendors "are paid based on the number of claims they deny," and that a CMS Office of the Actuary memo warned participants would have an incentive to deny as many claims as possible. Klomp first said his understanding was that the contractors are "not" paid on denial volume; he then acknowledged financial penalties for inappropriate denials, said the Innovation Center monitors performance and agreed the program must be done appropriately or should not expand. Murray called the program one that "is really hurting patients" and said she would fight the planned expansion to oncology, which matches a June 2025 Innovation Center plan, contained in the records EFF reported on, that contemplated cancer treatment, MRI and air ambulance. Klomp, as deputy secretary, would oversee the program.
    Sources: Senator Patty Murray (HELP hearing), STAT (hearing coverage), Electronic Frontier Foundation (WISeR records), STAT (WISeR records)
  • Regulation Sept 16, 2026
    Connecticut comptroller bars AI-only denials and downcoding for 270,000 on state health plans, seeks statewide rule in 2027
    The Sept. 16 policy applies to the State Employee Health Plan and the Partnership Plan for municipal workers, with Anthem on medical, Cigna on dental, Aetna on retiree Medicare Advantage and Caremark on pharmacy, according to CT Mirror, which reported a Jan. 1, 2027, effective date. Under the five rules, no adverse determination may be "made solely by an AI system" and each must be reviewed by a human; carriers may not use AI as the sole basis to "downcode claims, reduce provider payments, or alter billing codes without human review"; disclosure is required when AI is materially assisting with or recommending a benefit or service; member data may not be used to train or develop other models; and AI systems must be validated for accuracy, consistency and fairness, with governance procedures disclosed to the comptroller. Scanlon said that "you and your doctor should make your health care decisions, not artificial intelligence," and that he plans to recommend the legislature extend the protections next session to all state-regulated plans, which CT Mirror said cover about 220,000 residents. Alabama, Georgia and Minnesota require human review of payer utilization decisions by statute; Connecticut's requirement comes through the plans' own rules for their carriers.
    Sources: Connecticut Office of the State Comptroller (policy), Comptroller's announcement, CT Mirror, WSHU, Manatt Health (state law tracker)
  • Money Sept 16, 2026
    Sword Health to buy Headspace for a reported $300 million in cash, about one-tenth of Headspace's 2021 valuation
    Sword Health said in a Sept. 16 release that the deal is all cash and should close at the start of the fourth quarter; the release does not give a price, and the figure of up to $300 million comes from the trade press. Headspace was valued at $3 billion when it merged with Ginger in 2021. Sword, a virtual physical-therapy company valued at about $4 billion, said it has treated more than 1 million people with what it calls AI Care and serves 2,500 organizations; Headspace brings a 15,000-provider network, 84 peer-reviewed studies and 20,000 employer clients, according to the companies' own figures. Headspace's chief executive described the target model as one that "brings in a human clinician when it is needed." Headspace is also an accepted participant in Medicare's ACCESS behavioral health track, which pays for technology-enabled care of depression and anxiety based on outcomes.
    Sources: Sword Health release, Fierce Healthcare, HIT Consultant, CMS (ACCESS participants)
  • Evidence Sept 17, 2026
    Review finds developers involved in 73% of 229 digital health trials, with higher odds of a positive result
    A group from the Harvard T.H. Chan School of Public Health and the University of Pennsylvania published the rapid review in npj Digital Medicine on Sept. 17, covering 229 randomized trials from 29 systematic reviews in nutrition, maternal health, mental health and sleep from 2013 to 2025. Developers took part in 73% of the trials. Developer-involved trials had 2.47 times the odds of being preregistered and, when weighted by sample size, higher odds of reporting a statistically significant result, an odds ratio of 1.23 with a confidence interval of 1.16 to 1.31. The authors called for independent efficacy trials, more transparent reporting and regulatory oversight. The domains studied were behavioral rather than diagnostic; the review did not cover ambient scribes or diagnostic agents.
    Source: npj Digital Medicine
  • Regulation Sept 16, 2026
    Von der Leyen tells European Parliament "AI needs to empower our doctors. Not to replace them," promises health AI initiative for November
    In her Sept. 16 State of the Union address, the European Commission president, who trained as a physician, said she wants her doctor to have instant AI access to the data needed for the best diagnostic or treatment choice and does not want her doctor to be a robot, and she cited AI-supported mammography as the example of what Europe should scale. She named health as one of five sectors for industrial AI initiatives the Commission will announce in November, with no money or rules attached yet. Health Commissioner Olivér Várhelyi wants to use Europe's data and research infrastructure to speed adoption, Euronews reported. In Washington, officials have discussed paying company-run AI physicians 60% to 80% of the human rate, and the CMS administrator has promoted AI avatars for rural care.
    Sources: European Commission (State of the Union 2026 text), Euronews
  • Regulation Sept 16, 2026
    Payers covering 165 million Americans pledge to adopt ACCESS outcomes-based payment structure; CMS offers Medicaid a toolkit
    In its ACCESS announcement Tuesday, CMS said three out of four people with Medicare now qualify for at least one track, that 18 clinical and patient societies back the model and that "major health payers representing 165 million Americans with Medicare Advantage, Medicaid, and private health insurance plans" have pledged to adopt an outcomes-based payment structure aligned to it. Fierce Healthcare counted 17 payers and reported that CMS is publishing sample provider agreements, standardized billing codes and reporting infrastructure so Medicaid programs can adopt the model, quoting Jacob Shiff, the Innovation Center's chief AI officer, as saying it was designed "from day one" for impact beyond Original Medicare. Heart failure, chronic obstructive pulmonary disease (COPD), substance use and nicotine dependence join the model in spring 2027. The structure that lets Headspace, Noom and WeightWatchers bill Medicare for chronic-condition management would extend to commercial and Medicaid contracts under the pledges, with payment tied to outcomes such as blood pressure and A1c.
    Sources: CMS, Fierce Healthcare, CMS (ACCESS participants)
]]> 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 reviewhttps://physicianintheloop.org/posts/2026-09-16/https://physicianintheloop.org/posts/2026-09-16/Wed, 16 Sep 2026 07:00:00 -0600Ryan Gillum, MDA 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 SHORT READ

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.

THE DETAILS
  • Evidence Sept 15, 2026
    Open-weight diagnostic agent run inside the hospital is 98.9% accurate on the 49.4% of cases it keeps, selected by agreement across five runs
    Kather's group tested a fully autonomous agent that interviews the patient, examines, orders tests and diagnoses on 551 MIMIC-IV cases across seven acute conditions (appendicitis, cholecystitis, diverticulitis, pancreatitis, pneumonia, pulmonary embolism and urinary infection), 2,400 abdominal cases and 990 published multispecialty cases. Open-weight models hosted on premises (Qwen, GLM and GPT-OSS), which keep patient data inside the hospital, reached 90.0% on the seven-disease task and 83.8% on the four-disease task, near a cloud GPT-5.2 baseline, the authors said. Agreement across five repeated runs separated right from wrong answers with an AUC of 0.86, and at a consistency threshold of 0.90 the agent retained 49.4% of cases at 98.9% accuracy; blinded physician review of 181 cases agreed with the automated grading 92.3% of the time. The authors listed the limitations: retrospective simulation only, text only, one institution's data, lower accuracy in older patients and roughly five times the compute. Under that design, the remaining cases, those on which the agent's repeated runs disagreed, are routed to a person.
    Source: Nature Medicine
  • Regulation Sept 15, 2026
    House Health Subcommittee hears Patients First Act, which would give Medicare physician pay a permanent inflation update, and three other bills
    The Energy and Commerce Health Subcommittee took testimony Sept. 15 on 17 bills, including H.R. 9693, the Patients First Act, sponsored by three physician members of the House, John Joyce, Kim Schrier and Greg Murphy. The bill would tie the fee schedule to the Medicare Economic Index minus one point (minus half a point for alternative payment model participants), raise the budget-neutrality trigger from $20 million to $57.64 million in 2028 and replace the Merit-based Incentive Payment System (MIPS) over five years; the Provider Reimbursement Stability Act, H.R. 8163, addresses only the budget-neutrality change. Medicare physician payment has fallen about 33% against practice-cost inflation since 2001, according to the American Medical Association (AMA). A legislative hearing is not a markup, and no vote is scheduled. The conversion factor is set to drop another 1.19% on Jan. 1.
    Sources: House Energy and Commerce Committee, AMA
  • Regulation Sept 14, 2026
    22 states put about $240 million of rural transformation money into remote monitoring as CMS proposes to bar vendor-staffed monitoring Jan. 1
    The Alliance for Connected Care counts 22 states with roughly $240 million in remote patient monitoring (RPM) funding and 35 states with more than $2.4 billion in broader remote-care technology money, all drawn from the $50 billion Rural Health Transformation Program, Fierce Healthcare reported. The proposed 2027 fee schedule would limit remote physiologic and therapeutic monitoring to clinical staff employed by the billing practice, a change the American Telemedicine Association said would make the service unworkable for small practices; CMS received nearly 40,000 comments. Medicare RPM payments rose 31% to $536 million in 2024 and reached about 1 million beneficiaries. The final rule is expected around Nov. 1. Practices that bill the RPM codes 99454 and 99457 through a vendor would fall under the proposed staffing limit.
    Source: Fierce Healthcare
  • Deployment Sept 14, 2026
    Abridge expands into revenue cycle with a pre-bill tool that checks coded DRGs against what its ambient scribe recorded
    The pre-bill review product compares coded diagnoses and diagnosis-related groups (DRGs) with the clinical documentation before a claim goes out and surfaces supporting evidence for clinical documentation integrity and coding teams, without changing the note or the codes, Fierce Healthcare reported. Reid Health in Indiana is the named early customer. Abridge said it works with 300 large health systems and handles more than 100 million conversations a year; both figures are the company's own. Shiv Rao said health systems "aren't reimbursed for the care they deliver. They're reimbursed for the care they document that they delivered." The tool draws on the same ambient recording that produces the clinical note.
    Source: Fierce Healthcare
  • Workforce Sept 15, 2026
    24 revenue cycle leaders tell Becker's the threat is payer automation denying claims faster than providers can respond, not autonomous coding
    Becker's asked revenue cycle executives at Cleveland Clinic, Stanford, UC Davis, Rush, AdventHealth, Moffitt, the University of Kansas and 17 other organizations what worries them, and the consistent answer was payer AI. One said payers now review 100% of claims with automation while most health systems still manually sample a small fraction; another said payers can find reasons to challenge reimbursement far faster than providers can answer; several described payment being quietly reduced rather than denied. A companion piece the day before quoted UW Medicine and Springhill Medical Center leaders on restructuring teams, "replacing, say, 12 positions with four" at about the same total pay. A separate Becker's tally lists 10 health systems that cut IT and coding roles this year, from Stanford's 95 to Wellstar's 761, none of which named AI in its announcement.
    Source: Becker's Hospital Review
  • Deployment Sept 15, 2026
    GE HealthCare operations system, live at Duke and Queen's in Honolulu, forecasts bed pressure 72 hours out and assigns estimated discharge days
    CareIntellect for Operations runs two proprietary models, Pressure Forecast and Estimated Day of Discharge, on bed, staffing and patient data to give capacity managers a three-day view, GE HealthCare said. The system is live at Duke and Queen's in Honolulu. Duke's patient-flow lead said it lets her team move faster from analysis to action. The company has published no accuracy or outcome figures.
    Source: Fierce Healthcare
]]> Physician in the Loop launches with a September 2026 baseline on AI in physician workhttps://physicianintheloop.org/posts/2026-09-16-launch/https://physicianintheloop.org/posts/2026-09-16-launch/Wed, 16 Sep 2026 07:00:00 -0600Ryan Gillum, MDPhysician in the Loop launched this week with a baseline of where artificial intelligence stood in physician work as of September 2026, drawn from the regulatory record and published evidence. The longer analysis is in the "Where things stand" section of the site.THE SHORT READ

Physician in the Loop launched this week with a baseline of where artificial intelligence stood in physician work as of September 2026, drawn from the regulatory record and published evidence. The longer analysis is in the "Where things stand" section of the site.

Federal officials have discussed a new Medicare payment category for AI clinical software and rates for company-operated AI physicians at 60% to 80% of the human fee, The New York Times reported Sept. 14; nothing has been proposed. In trials published this year in Nature and Science, models matched or outperformed physicians on text-based tasks; the Nature paper called its system "not ready for real-world translation." No published trial has shown an AI model improving patient outcomes on its own. Twenty-two states have enacted 33 health-AI laws this year, according to Manatt Health's tracker, and no state has authorized an AI to practice medicine.

THE DETAILS
]]> The shift no model can coverhttps://physicianintheloop.org/specials/hands-on-roadmap/https://physicianintheloop.org/specials/hands-on-roadmap/Wed, 16 Sep 2026 07:00:00 -0600Ryan Gillum, MDThe labs are winning at the work that fits on a screen and have not yet been measured against the work that needs hands. Rural emergency departments are short of exactly that and pay a premium for it, and a primary-care physician can be credentialed to provide it without a second residency.One in thirteen American emergency departments has no attending physician on site around the clock. The work those departments are missing is hands-on, time-critical, and physically present, and it is the work on which every index of AI exposure, whichever lab built it, bottoms out. The virtual hospitalist companies stop at procedures. If you have been thinking about airway, access, or rural coverage, the market is telling you to do it. The advice is easy to give; following it means twelve to twenty-four months of courses, case logs, and committee calendars, and the credentialing is where people give up. What follows is the case for doing it and the path, documented as far as the providers' own prices, rules, and dates allow.

This is written for the family physician, internist, pediatrician, or hospitalist who trained for a version of medicine that a model is now quite good at, and who wants to add the part that a model is not. The safest ground left in clinical medicine is work that needs a physician in the building with trained hands. The rural emergency department is where that work is scarcest and where hospitals pay a premium for it, and a primary-care physician can be credentialed to do it without a second residency. What the data say, what the rules are, what the courses cost, and what the paperwork looks like follow, in that order.

Why the hands will outlast the screen

Start with the exposure data. Microsoft's occupational study, built from 200,000 real Copilot conversations mapped to the government's task database, gives healthcare diagnosing and treating practitioners an AI applicability score of 0.13, against 0.38 for media and communication workers, and puts home health aides and nursing assistants at 0.04, in the same band as forest and conservation workers. The authors warn that a high score does not mean job loss and a low score does not mean safety; the score says only how much of a job overlaps with what people are using AI for. The warning is fair, and by that measure the parts of a physician's day spent talking, documenting, looking things up, and pattern-matching on text overlap a great deal, while the parts spent with hands on a patient barely register.

Anthropic's economic index arrives at the same place from the other direction. Its January report singles out radiologists: the model does well on their two biggest tasks, reading images and writing reports, and yet "these occupations have low task coverage because AI can't do the hands-on or administrative work in their job profiles". OpenAI's benchmark for economically valuable work, GDPval, admits only occupations in which at least 60 percent of tasks involve no physical work, which is why it includes nurses and medical secretaries and not a single physician who touches patients. The three documents agree. The labs measure themselves against work that fits on a screen, and none of them is measuring against a chest tube.

None of that is permanent. Last September researchers published in Science Translational Medicine a soft robotic device that guides its own way into the trachea; on cadavers, in the hands of prehospital providers given five minutes of training, it beat video laryngoscopy on first-pass success, 87 percent to 63. But it has been tested on manikins and cadavers, not on a living patient, and a 2025 review of 28 airway robots found most still in validation, a handful commercialized, and almost no human data. Autonomous airway management remains a research program. On that evidence, the physician who can secure an airway, place a line, put in a chest tube, and run a resuscitation in a building with no backup will be scarce for at least another decade, a longer horizon than I would give any purely cognitive part of the job.

Two more measurements point the same way. The virtual hospitalist programs that have now logged a million visits covering nights and surges will admit, round, and manage a ventilator by video, and they say plainly that they stop at central lines and intubations, because a camera cannot do them. When Mass General Brigham studied 198,000 emergency visits, AI scribes saved emergency physicians 1.6 minutes of documentation per note and changed nothing about the work RVUs they generated per hour. The tools speed up the screen work and leave the hands-on work where it was.

The market for a doctor in the building

The 2022 national inventory of emergency departments found that 344 of 4,621 responding EDs, 7.4 percent, had no attending physician on site around the clock. Eighty-nine percent of those were critical access hospitals and 72 percent were rural. In North Dakota the figure was 58 percent of emergency departments, in South Dakota 56 percent, in Montana 46 percent. KFF Health News went to see how those departments run: a physician assistant or nurse practitioner alone on a twelve-hour night, a physician reachable by phone. The AMA and ACEP both want laws requiring a doctor in the building, and Indiana, Virginia, and South Carolina have passed them. Those laws will need doctors to fill the shifts.

The supply side is the strange part. In 2021 the specialty's own workforce study projected a surplus of 7,845 emergency physicians by 2030, and the profession has spent five years worrying about having too many emergency physicians. The surplus is urban. ACEP's own rural task force reported that only 8 percent of emergency physicians work in rural EDs and that many rural departments are staffed by physicians without EM board certification and by PAs and NPs with no EM physician oversight at all. The family medicine journal's count is that primary-care-trained physicians are about a fifth of the rural emergency workforce.

The rules were written for this reality. The Medicare conditions for a critical access hospital require a physician, PA, NP, or clinical nurse specialist "with training or experience in emergency care" to be on call and on site within 30 minutes, or 60 in frontier counties. The newer Rural Emergency Hospital rules require someone "competent in the skills needed to address emergency medical care" in the department around the clock, with a physician available by phone. Neither rule requires a board-certified emergency physician.

There are 1,388 critical access hospitals as of July, every one of them obliged to run a 24-hour emergency service, and 55 hospitals now operating as Rural Emergency Hospitals, a designation that did not exist before 2023 and that is, in effect, an emergency department with a few observation beds and no inpatient unit. Meanwhile 197 rural hospitals have closed or converted since 2005, eight of them last year by the Sheps Center's count. The money arriving to stop the bleeding is real.

The $50 billion Rural Health Transformation Program pays $10 billion a year through 2030, every state got an award in December, and the fifth approved use is recruiting and retaining clinicians who commit to rural communities for at least five years. Montana got $233 million for the first year, Colorado $200 million, the Dakotas about $190 million to $199 million each. The administration's pitch is that AI will make rural care affordable. Rural hospital leaders told STAT they do not believe it, and no amount of it staffs an emergency department at two in the morning.

The locum market puts a price on that scarcity. Locum emergency physicians are earning $250 to $300 an hour this year, against $120 to $135 for locum family physicians and $120 to $130 for internists. Postings for critical access hospital emergency shifts this week run $267 to $404 an hour. Employed emergency physician pay averaged $421,000, up about 8 percent in a year. The spread between those lines is not automatically yours: the rate follows the privileges you hold and the shifts you can safely cover, and the shifts are hard. But the spread is a direct measurement of what hospitals pay for a physician who can be present and capable at the bedside, and none of the models in the indices above narrows it.

The labs measure themselves against work that fits on a screen, and none of them is measuring against a chest tube.

The objections, and what they concede

ACEP's position, stated for years, is that emergency care is best provided by physicians who are residency trained in emergency medicine and that there is no substitute for residency training. I agree. The same document concedes that EM-trained physicians may never fully meet the demand, and the numbers above show what that concession means at night: the choice a critical access hospital faces is between a primary-care physician with the right training and a physician assistant or nurse practitioner alone with a phone, because the residency-trained emergency physician it would prefer is part of an urban surplus. Take this path and you will be the least trained person in the room for this kind of medicine and, on some nights, the only physician within sixty miles. The objection sets the terms of the path rather than closing it, and ACEP's own task force has written the terms: for rural non-EM physicians it proposed a defined knowledge base, supervised experience, and mentorship from an EM board-certified physician by telemedicine. That is the sensible response; arrange it for yourself before you need it.

The second objection is about your own hands. Internal medicine residents now report 30 percent confidence in procedures, with 9 percent confident in lumbar puncture and 20 percent in thoracentesis, and three quarters of the academic hospitals surveyed have moved bedside procedures onto a dedicated procedure service. The training pipeline has been producing less of this skill for a generation. A weekend course is the start of getting it back, and a proctored year is the rest. The deskilling evidence cuts both ways here too: endoscopists lost adenoma detection after three months of AI assistance, and a procedural skill you do not use decays the same way. Whatever you build, plan to keep using it.

From here to a credentialed shift

The sequence below carries costs and dates as the providers state them today. Two paths run through it. One is procedural competence where you already work, which for a hospitalist means never calling the ED for a line again. The other is rural emergency coverage. They share the first four steps. Budget twelve to twenty-four months and, by my arithmetic from the prices below, north of $10,000 in course fees before the first paid shift, less if your hospital pays for courses, which many will if you ask.

Step 1. Decide which of the two you are building, and write down the date. Procedural competence alone can be built inside a hospitalist job in a year. Rural ED coverage needs the same procedures plus trauma, pediatrics, obstetrics, airway under pressure, and the certifications that every medical staff office asks for before it will look at you. If you want the second, aim for the first rural shift eighteen months out and work backward. Tell your department chief and your medical staff office what you are doing at the start, because the privileging steps below take months of calendar time that you cannot compress.

Step 2. Get the three cards that gate everything, then the two that make you useful in a small building. Nearly every rural bylaws committee will want three cards in hand before it looks at you, and the primary-care board route to EM certification requires all three to be current. They are ACLS and PALS from the American Heart Association, and ATLS from the American College of Surgeons, now in its eleventh edition with 26 online modules and a 2 to 2.5 day course, or 1.5 days in the hybrid format; completion is valid four years. Then two courses built for exactly this situation. Comprehensive Advanced Life Support is the rural all-hazards course: twelve hours online and two days of simulation covering airway, trauma, cardiac, stroke, pediatric, obstetric, and neonatal emergencies and sepsis, up to 23.5 credits, $1,697 for physicians in Minnesota, with the next open course November 11 and 12 in Aitkin. Fundamental Critical Care Support from the Society of Critical Care Medicine is written for the non-intensivist who has to manage a critically ill patient for the first 24 hours until transfer; the self-study version is $275 and hosted courses with skill stations run about $800. If the department you are aiming at delivers babies, or catches them in the parking lot, add ALSO from the AAFP, one day after an online course, about $595 at an AAFP-run site.

Step 3. Learn the airway from people who do it for a living, then practice it in an operating room. Privileging committees count cases, and a course certificate is not a case, so this step has two halves: the courses, then the operating room. Take two courses, and I would space them a year apart. The Difficult Airway Course: Emergency is three days, 21 credits, $1,950, with simulation stations and code-airway drills; the September Atlanta course is sold out and 2027 dates open soon. The two-day cadaver course that Rich Levitan and Sal Rezaie run in Texas spends the entire second day in a lab with a dozen prepared cadavers, 16 credits, 36 seats, with dates remaining this year on September 21, October 26, November 16, and December 3; the last posted tuition was $2,250, and there is a June version at Jackson Lake if you want the mountains with it. Then the part no course can sell you: ask an anesthesiologist at your own hospital for supervised intubations in the operating room, one morning a week for a few months, and log every one.

Step 4. Lines, taps, tubes, and the ultrasound that makes them safe. The ultrasound is what makes these procedures safe, so learn them together and hold a real ultrasound credential rather than a course certificate. The Hospital Procedures Consultants weekend course is the one built for hospitalists: 23 online modules, then two ten-hour days at a six-to-one ratio covering ultrasound-guided central and arterial lines, thoracentesis, paracentesis, lumbar puncture, chest tubes and pigtails, pericardiocentesis, intubation, and cricothyroidotomy, 20 credits, with a simulation certificate at the end; remaining 2026 dates are Seattle September 19, New Orleans October 24, and San Antonio November 14. The SHM and CHEST point-of-care ultrasound certificate requires online modules, a live course, a regional course, a portfolio of 20 cardiac, five lung, five abdominal, five DVT, and three soft-tissue studies, and a skills and knowledge assessment, all within three years; the assessments carry a fee and retakes are $300. If you want the emergency flavor, the Ohio ACEP two-day emergency ultrasound course is $999 to $1,325 with the next one November 4 and 5, and Gulfcoast runs a three-day course in St. Petersburg in February and August at a three-to-one ratio.

Step 5. Turn the courses into privileges. This is where the calendar time goes, and where people give up. Ask your medical staff office for the delineation-of-privileges form for each procedure and read the numbers on it: how many supervised cases for initial privileges, how many a year to keep them. The Society of Hospital Medicine has published a position statement on credentialing hospitalists in ultrasound-guided procedures that most medical staff offices will accept as the framework if yours has none. Joint Commission standards put every newly granted privilege under a focused professional practice evaluation, meaning a proctor and a defined number of observed cases, and the clock starts the day the privilege is granted; after that, ongoing evaluation cannot be more than twelve months apart. So you need a proctor. Interventional radiology, the ICU, and the emergency department all have people who will sign for you if you show up and ask; the procedure service, if your hospital has one, is the natural home. Keep a log from the first case: date, indication, ultrasound or landmark, attempts, complications. The log is what a committee actually reads; the courses only get you in the door.

Step 6. Choose your rural emergency credential path. There are three paths, and only two are open to a practicing physician. The first is experience: get ED privileges at a critical access hospital, whose bylaws, following the AAFP's position that privileges should rest on training, experience, and demonstrated competence, will usually ask for the three cards, references, and a supervised start; begin with overlap shifts beside the incumbent physician, then nights with a phone, then alone. The second is board certification through the American Board of Physician Specialties, the only board that certifies primary-care-trained physicians in emergency medicine. Its practice route is five years of full-time EM practice with at least 7,000 hours and 1,400 hours in every twelve-month period, plus current ACLS, ATLS, and PALS, three letters, ten case reports, and a 325-question written exam followed by an oral exam, good for eight years; California adds a sixth year and 300 EM-specific CME hours. Its fellowship route is a 12- or 24-month program from a list of fifteen approved programs in nine states. Marshfield Clinic in Rice Lake, Wisconsin takes two fellows a year, requires 1,400 ED hours and rotations through trauma, pediatric EM, the ICU, and anesthesia, and is accepting applications for 2027. Tamarack Health in Ashland runs twelve to fourteen ED shifts a month for three fellows and accepts applications on a rolling basis. Texas Tech in Odessa trains in a Level 2 trauma center with 60,000 visits a year under ABEM-certified faculty. The third path is closed to you: ABEM certifies only graduates of an accredited EM residency or of a five-year combined EM and family medicine residency, and there is no practice track. Know that before a recruiter tells you otherwise.

Step 7. Build the backup before the first night. This is where the AI and telemedicine story turns in your favor. Avel eCare puts board-certified emergency physicians on video in 137 rural hospitals and claims a 21-minute reduction in door-to-physician time and 700 transfers avoided in a year; those are the company's numbers, but the service has run since 2009 and it is what the ACEP task force asked for. A telehospitalist triage line studied across four community EDs sent 19 percent of admit-likely patients home and kept 58 percent local instead of transferring them. Before you take a shift, know which of these your hospital has, know the transfer agreements and the flight times, and know the rule you are working under: 30 minutes to the bedside at a critical access hospital, 60 in frontier counties, a physician always reachable by phone at a Rural Emergency Hospital. The backup is there so that you can be the physician in the building without being the only brain on the case.

Step 8. Handle the money and the paperwork early. The dull work is what separates a plan from a shift. Ask the rate first. Critical access hospital emergency shifts are posting at $267 to $404 an hour through locum agencies, and a hospital that hires you directly will pay less per hour and offer more in return: malpractice, a schedule you control, and sometimes loan repayment. The National Health Service Corps pays up to $75,000 for two years of full-time primary care at an approved site, and a critical access or rural emergency hospital qualifies through its affiliated outpatient clinic, not through the emergency department itself; the 2026 cycle has closed, so plan for 2027. Ask your state office of rural health how it is spending its Rural Health Transformation award on clinician recruitment, because that money exists and comes with a five-year commitment attached. Call your malpractice carrier before the first shift and confirm in writing that emergency department scope, procedural sedation, and the procedures you have added are covered. Get the license in the state you will actually work in months ahead of time.

Step 9. Keep the skill you paid for. A skill you do not use decays, as the endoscopists found. Set a floor: a minimum number of each procedure a year, written down, with the log to prove it. Recertify ATLS on the four-year cycle and take the airway course again when it lapses. Keep deliberate AI-off intervals for the cognitive work too, and use the workflow the Stanford data support: commit to your own assessment, then compare. What lasts in a physician is being there, deciding under uncertainty, answering for the result, and doing the physical work. The physical work fades faster than the rest when you stop doing it, and right now nobody has built a machine to replace it.

Privileging committees count cases, and a course certificate is not a case.

If you follow this and end up at a rural bedside at two in the morning with an airway to secure and no one else to do it, you will be doing the one job in medicine I am confident a model will not take from you, in one of the few places where the shortage is simply a missing person. Hospitals have been advertising that shortage for years. This is how to answer the ad.

]]> California awaits governor's action on AI practice bills; WISeR vendor under corrective plan; 72% of health systems report unsanctioned AI; Conifer cuts 1,037 billing jobshttps://physicianintheloop.org/posts/2026-09-15/https://physicianintheloop.org/posts/2026-09-15/Tue, 15 Sep 2026 07:00:00 -0600Ryan Gillum, MDCalifornia's governor has until Sept. 30 to act on Assembly Bill 1979, which would bar artificial intelligence (AI) from independently performing work that requires a licensed professional or directing unlicensed staff to do it.THE SHORT READ

California's governor has until Sept. 30 to act on Assembly Bill 1979, which would bar artificial intelligence (AI) from independently performing work that requires a licensed professional or directing unlicensed staff to do it. The bill applies to health facilities, clinics and physician offices; under its terms, an AI system could not direct medical assistants or other unlicensed personnel to carry out licensed duties. Its companion, AB 2575, would protect clinicians who override an AI system and bar developers from blaming those clinicians when the system fails.

The WISeR records, the Waystar sale process and the ACCESS expansion had been covered in the baseline report. The fuller WISeR records show one vendor under a corrective action plan and a Centers for Medicare & Medicaid Services (CMS) plan that considered extending AI prior authorization to MRI, cancer treatment and air ambulance, according to Fierce Healthcare and the Electronic Frontier Foundation (EFF). An Imprivata survey of 250 health system leaders found 72% have AI tools deployed without formal IT approval. Tenet's billing subsidiary is cutting 1,037 jobs, Healthcare Dive reported, and Tenet's chief executive has pointed to AI and offshoring. Waystar has hired Evercore to explore a sale, Reuters reported.

THE DETAILS
  • Regulation Sept 10, 2026
    California AB 1979, barring AI from independently performing licensed clinical work, and companion AB 2575 await governor's action by Sept. 30
    AB 1979 (Bonta) would prohibit health facilities, clinics and physician offices from allowing AI to independently perform clinical functions that require a licensed professional or to direct unlicensed personnel to carry out licensed duties; AI could inform a decision if a licensed professional keeps final authority, and documentation, routine communications and supervised education are exempt. It would also make direct-to-consumer health chatbots that handle medical records subject to California's medical confidentiality law. AB 2575 (Ortega) would protect clinicians who override unsafe AI output and bar developers from shifting blame for AI-caused harm onto the workers using it. Both bills cleared the Legislature by Aug. 31 with the California Nurses Association as sponsor and hospitals, insurers and technology firms opposed; enforcement would run through the licensing boards, effective Jan. 1, 2027, if signed.
    Sources: National Nurses United, Paubox, KALW, AB 1979 (LegiScan), AB 2575 (LegiScan)
  • Regulation Sept 15, 2026
    WISeR records show Virtix under corrective action plan and a CMS plan that considered AI prior auth for MRI, cancer treatment and air ambulance
    Roughly 1,000 pages of WISeR records, won by EFF in a Freedom of Information Act suit, show Virtix denied more requests than it approved in its first three months and was required to submit a corrective action plan, Fierce Healthcare reported, adding detail to earlier STAT coverage. The records also show that Innovaccer told CMS before launch it needed more time and was still working backlogs in early April, that vendors and Medicare contractors miscommunicated and mis-categorized requests, and that low quality scores cut vendor pay by 5% to 10%. EFF's write-up said the Innovation Center's June 2025 investment plan considered adding services that need urgent authorization, including air ambulance, cancer treatment, MRI and medications without public coverage criteria. The CMS WISeR page, updated Sept. 10, lists the original service categories and no expansion. CMS has not responded to the records.
    Sources: Fierce Healthcare, Electronic Frontier Foundation
  • Deployment Sept 15, 2026
    Imprivata survey: 72% of 250 health system leaders report AI tools or agents deployed without formal IT approval; 28% have agentic AI in production
    The Sept. 15 Imprivata survey of 250 health system leaders responsible for identity security or AI strategy found 83% have AI across multiple departments, 28% have agentic AI in production and 44% are piloting it. It also found 79% expect agents to change clinical workflows significantly, 88% expect agents to act with some autonomy and 17% consider their current identity controls adequate. Imprivata sells identity management software, and the figures are the company's own. Advocate Health and Penn Medicine leaders described the same pattern of unapproved AI use at Becker's Health IT conference the same day.
    Sources: GlobeNewswire, Fierce Healthcare
  • Deployment Sept 14, 2026
    Advocate Health pauses nursing ambient rollout; CIO says AI raises total health care costs; Penn Medicine cites about 30% pharmacy prior auth gains
    Advocate Health paused an ambient voice documentation rollout for nurses after finding that little inpatient nursing care is done by voice, Dr. Bobbie Byrne, the system's chief information officer, said at Becker's Health IT conference in Chicago; nurses chart in structured flowsheets, not narrative notes. Byrne also said AI raises the cost of health care in total, not for any one organization. Dr. Mitchell Schnall of Penn Medicine cited roughly 30% efficiency gains in pharmacy prior authorizations and said jobs change and the system commits to retraining rather than guaranteeing roles. Mayo Clinic said its Abridge-built nursing tool has made documentation latency an obsolete metric. Mercy, FMOL, Northeast Georgia, Mount Sinai Miami Beach, Jefferson and Northwell are running or planning nursing ambient documentation, and Mercy reported 22% less flowsheet time per shift on three pilot units.
    Source: Becker's Hospital Review
  • Deployment Sept 15, 2026
    Knowtex opens health care AI lab, citing its own figures of 79 VA medical centers, 7,000 clinicians and 450,000 documentation hours saved
    Knowtex, which won a $15 million Department of Veterans Affairs (VA) contract in October 2025 after the VA's ambient-scribe tech sprint evaluated more than 150 tools, said it is live at 79 VA medical centers across 10 Veterans Integrated Service Networks within six months, with 88% sustained clinician adoption, 450,000 documentation hours saved and more than 300 customer organizations. The new lab will build specialty-specific models and evaluation benchmarks, the company said. All figures are the company's own. The VA is the largest single employer of physicians in the country.
    Sources: PR Newswire, Fierce Healthcare
  • Workforce Sept 14, 2026
    Tenet's Conifer to cut 1,037 billing jobs by Nov. 2; WARN notice cites restructuring and technology initiatives, CEO has pointed to AI and offshoring
    Conifer Health Solutions, Tenet's revenue-cycle subsidiary with more than 600 clients, will lay off 1,037 mostly remote workers effective Nov. 2, 2026, Healthcare Dive reported. The stated trigger is CommonSpirit's exit from its 24% stake and its service contract, which ends Oct. 30. Tenet's chief executive has said AI and offshoring would lower Conifer's costs, and the Worker Adjustment and Retraining Notification (WARN) filing cites restructuring and technology initiatives. The jobs are back-office, not clinical. No verified case of a U.S. health system cutting physician positions because of AI has been reported.
    Source: Healthcare Dive
  • Evidence Sept 11, 2026
    Seven health system leaders tell Becker's autonomous AI needs narrow use cases, named accountability, escalation paths and developer liability
    Becker's Hospital Review asked chief information officers and chief medical information officers at Cornell, Baptist Health, HSS, Premier Health, Christ Hospital, North Country Healthcare and Denver Health what the ADVOCATE agents of the Advanced Research Projects Agency for Health (ARPA-H) would need. Dr. Curtis Cole of Cornell said the agents will not succeed if the companies that build and oversee them have no liability. Joy Oh of Christ Hospital asked who is responsible when a supervisory agent misses an inappropriate recommendation, and Dr. Daniel Kortsch of Denver Health said the technology has been tested on curated cases, not safety-net populations. The consensus was narrow use cases, named accountability and escalation paths, not broad autonomy.
    Source: Becker's Hospital Review
  • Incident Sept 15, 2026
    Securities suit targets Hims & Hers over health data sharing with ad platforms and treatment practices; class period Aug. 4, 2025, to July 29, 2026
    Plaintiffs' firms are soliciting lead plaintiffs, with a Nov. 2, 2026, deadline, for a securities suit alleging Hims & Hers misled investors while providing customer health data to third-party advertising platforms and running treatment practices likely to draw regulatory scrutiny, SBS Law said in a notice on GlobeNewswire. Hims & Hers is a publicly traded direct-to-consumer health company. The notice is a law firm solicitation, not a court ruling, and the complaint's allegations have not been tested in court.
    Source: GlobeNewswire (SBS Law notice)
  • Money Sept 15, 2026
    Waystar hires Evercore to explore sale at about $4.8 billion; shares down 24% this year after Morgan Stanley AI disruption note
    Waystar, a revenue-cycle software company used by hospitals and physician groups for billing and claims, has retained Evercore to run a sale process that is at an early stage and could still stall, Reuters reported. Shares are down 24% this year after a July Morgan Stanley note argued AI could disrupt software companies. EQT holds 13% of the company, CPP Investments 10% and BlackRock 8%. Waystar went public at $20 a share in 2024 and peaked at $45 in 2025. The company has said AI drives 40% of its bookings.
    Sources: Fierce Healthcare, Reuters (via MarketScreener)
]]> The AI doctor's first credential is a billing codehttps://physicianintheloop.org/letters/2026-09-14/https://physicianintheloop.org/letters/2026-09-14/Mon, 14 Sep 2026 07:00:00 -0600Ryan Gillum, MDMedicare is working out how to pay for AI care before the evidence is in, and its own AI gatekeeper spent the week showing what that order produces. The fight worth having is over the contracts, not the models.

The figure is 60 to 80 percent. That is what federal officials have discussed paying company-run "AI physicians," as a share of the human fee for the same service, the New York Times reported on Monday. Nothing has been proposed, and John Whyte, the AMA's chief executive, said in the same story that models built to provide care on their own are not ready for widespread use. Both caveats are fair, and the story is still the most consequential thing written about this profession all year, because in this country nobody gets replaced until someone gets paid. The rate, though, is a rumor. The contract Medicare expanded on Tuesday is a signed document, and it requires no AI to be licensed, cleared, or even especially accurate, only a company to be paid for your patient's blood pressure while you are paid per visit for the same patient.

That contract is the ACCESS model, which pays technology companies for outcomes in hypertension, diabetes, musculoskeletal pain, depression, and anxiety, and which will add heart failure, COPD, substance use, and tobacco next spring. CMS says three out of four people with Medicare now qualify for at least one track, and that payers covering 165 million people with Medicare Advantage, Medicaid, and commercial plans have pledged to copy the structure; if they follow through, it lands in your contracts in January. On Thursday Counsel Health, which sells an AI front door with a $29 physician follow-up, said it will join ACCESS in early 2027 with Oura as its wearable partner, treating hypertension, obesity, hyperlipidemia, and prediabetes at no cost to the beneficiary and promoting the offer inside an app with about five million paying members. Counsel reports 36,000 Oura members onboarded in nine weeks and a care rating of 4.6 out of 5. Those are the vendor's numbers, and I file them as such.

The case for ACCESS is serious and deserves its full strength. Medicare has paid for visits and got visits; paying for a controlled blood pressure instead is on its face the better bargain, eighteen clinical and patient societies back the model, and a beneficiary treated at no cost by a company paid for her outcomes is not obviously worse off. Grant all of it, and ask what evidence the companies have. A paper in npj Digital Medicine looked at 229 randomized trials of digital health tools and found the developer involved in 73 percent of them, with somewhat higher odds of a positive result. The effect is modest and the domains are behavioral, but that is exactly the evidence the ACCESS participants are selling on.

The rest of the answer came from the same agency, which spent the week explaining why its AI gatekeeper is failing. About a thousand pages of records that the Electronic Frontier Foundation won in a public-records suit show one WISeR vendor telling CMS before launch that it would auto-affirm requests because it was not ready, another, Virtix, placed on a corrective action plan after denying more requests than it approved, vendors paid per denial with quality penalties of only 5 to 10 percent, and a request that sat for 83 days. A payment design shapes a vendor's conduct from the first day, long before anyone learns whether the tool works: pay per denial and you get denials. ACCESS pays for outcomes in your patients, so the first question to ask is whether you will see the data.

On Wednesday Chris Klomp, the nominee to run HHS day to day, sat before the HELP Committee while Senator Patty Murray read those numbers to him. He first said the contractors were not paid by denial volume, then allowed that there are penalties for inappropriate denials, and then said the program "must be done appropriately, or it should not expand." Ms. Murray said she would fight the planned move into oncology. A condition is weaker than a stop, but the person who will oversee the program stated it on the record, and it fits in an appeal letter.

The terms can be written the other way, as Connecticut showed the same day. Sean Scanlon, the state comptroller, set five rules for the 270,000 people on the state employee and municipal plans: no adverse determination made solely by an AI, no AI-only downcoding or payment reduction, disclosure when AI is materially involved in a benefit decision, no training of other models on members' data, and validation audits. He wants the legislature to extend them to every state-regulated plan in 2027, but they are plan policy rather than statute, so any large employer plan, including a health system's own, could copy them next month, and should: the two dozen revenue cycle leaders Becker's polled this week said what they fear is not autonomous coding but payers reviewing 100 percent of claims with automation and quietly reducing payment rather than denying it. Sacramento, meanwhile, has said nothing about AB 1979, the first state bill written against the staffing model itself: no AI performing licensed work, no AI directing unlicensed staff to do it. The governor has until September 30.

The tool those contracts will buy is also taking shape. The best paper I read this week, in Nature Medicine, comes from Jakob Nikolas Kather's group, whose diagnostic agent, built on open-weight models running inside the hospital firewall, took histories, ordered tests, and committed to diagnoses on 551 emergency cases across seven conditions. Accuracy was about 90 percent, close to a cloud model. The useful result came from running it five times and asking whether it agreed with itself: at a strict consistency threshold it kept 49.4 percent of cases at 98.9 percent accuracy and routed the other half to a person. One week after ARPA-H put $62.7 million into heart-failure agents built on exception-based oversight, that is the mechanism written down. Under it the cases that reach you are the hard ones by construction, so your time per case goes up, and any productivity target that assumes otherwise is wrong on arrival.

The money kept moving in one direction. Sword Health is buying Headspace for a reported $300 million in cash, about a tenth of what Headspace was worth in 2021, to fold it into an AI that decides when to bring in a clinician. Tenet's Conifer will cut 1,037 billing jobs by November 2, with the WARN notice citing technology initiatives. Waystar hired bankers to explore a sale after the market decided AI would eat billing software. Becker's counts ten health systems that cut IT and coding roles this year, and not one named AI. There is still no verified case of a US health system cutting physician positions because of AI. What there is, per Doximity, is 20 percent of physicians already facing higher productivity targets because of it. Replacement arrives as a bigger census rather than a pink slip, which is harder to fight because nobody announces it.

So argue about the contract. If you practice in Arizona, New Jersey, Ohio, Oklahoma, Texas, or Washington, put Mr. Klomp's sentence and the EFF numbers in your next WISeR appeal. If you sit on a contracting or medical society committee, ask two questions before the January contracts are signed: which ACCESS-aligned companies will be paid for outcomes in your patients, and whether you will see their data. If your health system insures its own employees, hand the benefits office Mr. Scanlon's five rules for adoption next month. If you bill remote monitoring through a vendor, price the employed-staff version before the final fee schedule lands around November 1. If your system starts saying exception-based review, ask what fraction of cases the agent keeps and how accurate it is on the ones it sends you, and get the answer into the productivity conversation before the tool arrives. And when anyone hands you a study, ask who ran it; in 229 trials, 73 percent of the time it was the company. The billing code is being written. The examination is still yours to give.

SOURCE MATERIAL

The next two weeks belong to Sacramento and Denver. The California governor has until September 30 to sign or veto AB 1979 and AB 2575; if he signs, the medical-assistant-plus-model-plus-remote-signature model becomes illegal in the largest state on January 1. Colorado's attorney general owes a revised draft of the automated-decision and chatbot rules on September 23, with comments closing October 26 and both laws taking effect January 1. On October 1 the third ACCESS cohort starts and Alabama's HB 272, which requires human review of payer utilization decisions, takes effect. The Coalition for Health AI holds its fall symposium on October 6.

Then the federal calendar. Comments on the FDA's generative-AI device discussion paper close October 19, and if you have an opinion about what a competent clinician looks like, that docket is where it counts. Doctronic's Utah agreement comes up for renewal in October, the closest thing to an autonomous-prescribing decision on the calendar. Medicare open enrollment opens October 15, and ACCESS is Original Medicare only, which is worth knowing when a patient asks. The 2027 fee schedule final rule, with the conversion factor, the G2211 modifier, and the remote-monitoring staffing rule, is expected around November 1; the Patients First Act has a hearing behind it and no markup scheduled. Conifer's 1,037 layoffs take effect November 2.

]]>