Google AI passes supervised clinic test in Lancet; Vanderbilt tallies chatbot psychosis; CMS proposes remote pathology review; UT System widens Abridge
Listen to this post
read by AI voices ยท all episodesTHE SHORT READ
A Google conversational AI held supervised pre-visit conversations with about 100 patients at a Boston hospital without a single safety stop, according to a prospective study The Lancet published Oct. 8.
Physicians supervised every conversation in the single-arm study, which Alphabet funded, the authors reported. A day earlier, Vanderbilt University Medical Center researchers reported in JAMA Psychiatry that 28 patients in its records had psychosis worsened by chatbot use, most of them in a first episode. The Centers for Medicare & Medicaid Services and the Centers for Disease Control and Prevention proposed letting pathologists review digital slides and images from outside the laboratory, and the University of Texas System opened Abridge's ambient documentation software to all of its health institutions.
Everlywell began selling a mammogram risk model authorized by the Food and Drug Administration directly to consumers, Suki added an UpToDate-based question tool to its scribe app, and Healthleap raised $38 million for hospital screening software, the companies said.
THE DETAILS
- FYIEvidence Oct 8, 2026Adults booked for single-complaint urgent primary care visits at Beth Israel Deaconess Medical Center in Boston talked with Google's Articulate Medical Intelligence Explorer, or AMIE, up to five days before the appointment while physician safety supervisors watched every conversation; of 114 patients enrolled from April to November 2025, 98 completed both the chat and the visit, according to the single-arm study published Oct. 8 in The Lancet. No conversation met the prespecified criteria for a safety stop, though supervisors noted one hallucination and added clinical information in five chats, the authors reported. Clinical evaluators rated the conversations favorably in 87% to 100% of cases across 17 criteria, and primary care clinicians who read the transcript beforehand found it helpful for visit preparation in 33 of 44 cases. In its summary of the study's preprint in March, Google said AMIE's differential included the final diagnosis in 90% of cases and ranked it first in 56%, and that evaluators judged physicians' management plans more practical and cost-effective. AMIE had no access to the medical record and could not examine patients, the company said; the single-arm study was funded by Alphabet and had no comparison group.
- FYIEvidence Oct 7, 2026Researchers at Vanderbilt University Medical Center searched behavioral health notes written from Dec. 1, 2022, to April 15, 2026, for 215,712 patients and classified 28 as having psychosis worsened by conversational AI, about 0.013% of those receiving mental health care, according to the cohort study published Oct. 7 in JAMA Psychiatry. Seventeen of the 28 were in a first psychotic episode, significantly more than among patients with neutral chatbot use or with AI-themed delusions and no documented chatbot use, and their median age was 27.5. ChatGPT was the product named in 15 cases, and 24 of the interactions were documented after OpenAI released GPT-4o in May 2024, which the authors said coincides with documented sycophancy in that model, though the design cannot establish causation. In 18 cases the chatbot amplified existing symptoms by validating distorted ideas, the authors reported. They listed ascertainment bias, a single site and an unvalidated rating system as limitations and wrote that routine assessment of AI use during psychiatric encounters "appears warranted."
- ActRegulation Oct 8, 2026
Comments on the proposed CLIA rule are due Dec. 7, 2026, according to the Federal Register notice.
The proposed rule, published Oct. 8 in the Federal Register, would amend the Clinical Laboratory Improvement Amendments, the federal quality standards for clinical laboratories, so that pathologists and other qualified staff can review digital data, results, live imaging and digital images while "not physically on the premises" of the testing laboratory, without a separate certificate for the remote site. The change would codify enforcement discretion CMS announced in 2023; the laboratory director would remain responsible for the testing, connections would have to be secure and labs would have to record who reviewed remotely, on which device and with which software version. Virtual access would not be permitted in cytology, the agencies wrote. The rule would also allow gynecologic cytology proficiency testing on digital images of glass slides, citing FDA's 2024 authorization of the first digital cytology screening system, and would require labs to keep digital images with cytology and histopathology slides.Sources: Federal Register, Rural Health Information Hub - FYIDeployment Oct 8, 2026The agreement, announced Oct. 8, sets a systemwide framework under which each UT health institution, eight of them named in the announcement, including UT Southwestern Medical Center, UT MD Anderson Cancer Center and UTHealth Houston, decides how to implement the ambient documentation software, according to Abridge. A pilot that began in 2025 covered five locations, more than 50 specialties and outpatient, inpatient and emergency settings; over the past 12 months the number of clinicians recording encounters rose 73% and monthly notes rose 80%, the company said. The system's hospitals and clinics provided nearly 11.7 million outpatient visits and more than 2 million hospital days in the past year, Becker's reported. The deal was coordinated through UT REAL Health AI, a state-funded initiative created in 2025, and could later extend to care gaps, payer workflows and trial matching.Sources: Abridge, Becker's Hospital Review
- FYIDeployment Oct 1, 2026Women with a screening mammogram from the past 12 months can order Clairity Breast through Everlywell, where a clinician in the company's affiliated network reviews and orders the test and another releases the results with next steps, according to the Oct. 1 release; the tool estimates future risk and does not detect cancer. FDA granted the model De Novo authorization in May 2025, and it was previously offered only at Beth Israel Deaconess Medical Center and an imaging center in Colorado, AuntMinnie reported. The companies said the model was validated on more than 77,000 mammograms at five U.S. sites for authorization and more than 120,000 at 10 facilities since, and that 2026 NCCN guidelines list AI mammogram risk assessment as an option from age 35. The release gives no price. Clairity chose the consumer route because waiting for physicians to prescribe the tool and insurers to cover it would take too long, STAT reported Oct. 8.
- FYIDeployment Oct 8, 2026The feature, called Ask, lets clinicians pose clinical questions inside the Suki app and returns answers grounded in UpToDate content, with sourcing, dosing considerations and links, Fierce Healthcare reported Oct. 8. The companies described it as the first product to use Wolters Kluwer's UpToDate Expert AI programming interface. Clinicians selected patient-specific insights 83% of the time across questions on diagnoses, dosages and safety, the companies said; no independent evaluation has been published. Abhi Pathak, Suki's chief product officer, said standard guidelines are "inherently limited" because clinicians must tailor them to each patient.Source: Fierce Healthcare
- FYIMoney Oct 8, 2026Sequoia Capital, First Round Capital and Hummingbird Ventures backed the combined seed and Series A rounds, Fierce Healthcare reported Oct. 8. The software reads notes, labs, vitals, orders and problem lists and sends risk scores into the electronic health record, starting with disease-related malnutrition; Houston Methodist announced a systemwide launch in June. The company said one Penn Medicine hospital saw about $23.8 million a year from added reimbursement and shorter stays, and Sequoia Capital partner Alfred Lin said an unnamed hospital identified and treated 39% more malnutrition cases; those figures come from the company and an investor. Chief Executive Josiah Meyer said Healthleap plans to reach hundreds of hospitals in the coming year and to add conditions such as aspiration pneumonia and pressure injuries. Correction (Oct. 13): This item has been corrected. An earlier version said the company reported that an unnamed system identified and treated 39% more malnutrition cases; Sequoia Capital partner Alfred Lin said in a statement that one hospital saw 39% more malnutrition cases identified and treated.Sources: Fierce Healthcare, Healthleap
- FYIEvidence Oct 9, 2026Researchers from Johns Hopkins, the Uganda Cancer Institute and Makerere University screened six databases from 2000 through 2025 and found 18 studies of AI deployed in radiation oncology in low- and middle-income countries, according to the systematic review published Oct. 9 in npj Digital Medicine. Nine stopped at single-task applications, six reached clinician-supervised workflow integration, one coordinated work across stages of care and none met their highest level of integration, the authors reported. Validation was mostly retrospective and no study reported continuous performance monitoring. The authors attributed the early-stage pattern to limits in validation, workflow maturity, infrastructure and governance rather than a lack of innovation.Source: npj Digital Medicine