AMA says doctors shouldn't be liable for AI errors they can't detect

• FYIRegulation Oct 29, 2026
  AMA says physicians should answer only for AI judgments they can meaningfully exercise, not errors they cannot detect
  All physicians
  The American Medical Association's Center for Digital Health and AI on Oct. 29 released a framework that ties liability for harm from clinical AI to each party's "authority, access to information and ability to prevent or mitigate harm," according to the association. Developers would answer for demonstrating and maintaining safety and effectiveness, and hospitals, health systems and practices that deploy AI for how they select, implement, govern and monitor it. Physicians would be responsible for the clinical judgment they can meaningfully exercise given a tool's intended use, known limits and the information available at the point of care, and the framework states that physician oversight cannot substitute for adequate evidence or sound system design. An accompanying issue brief says physicians should not be held liable for AI errors they cannot reasonably detect, such as a triage model deprioritizing a case they never see, and that vendor contracts should not shift a developer's defects onto physicians, Becker's reported. "Patients must have meaningful avenues for recourse when clinical AI causes harm," said Dr. Willie Underwood III, the AMA's president.
  Sources: American Medical Association, AMA framework page, Becker's Hospital Review
• WatchRegulation Oct 29, 2026
  Judge blocks Medicare and Medicaid agency's stricter test for 'medically frail' patients weeks before Medicaid work requirements start

  Medicaid work requirements for expansion enrollees aged 19 to 64 begin Jan. 1; HHS and CMS had not commented on the ruling, the AP reported.

  Primary careHospitals
  Chronically ill Medicaid enrollees will not have to show that their conditions "significantly impair" their ability to work to be exempt from the program's new work requirements, after U.S. District Judge Richard Stearns in Boston ruled Oct. 29 for 25 states and the District of Columbia, the Associated Press reported. The 2025 law already exempts people who are blind or disabled, have substance use or disabling mental disorders, or have serious or complex medical conditions; the Centers for Medicare & Medicaid Services' rule had added the impairment showing. The judge upheld two other limits: states must reject proof of frailty based on a diagnosis more than a year old, and may accept only one self-attestation per continuous enrollment period. Enforcing the narrower definition "may lead to the disenrollment of otherwise eligible vulnerable individuals," Stearns wrote, according to STAT, which reported that he issued a permanent injunction and that rollout may slip past the start of the year. Expansion enrollees aged 19 to 64 must log at least 80 hours a month of work or community service, or attend school half-time, unless exempt.
  Sources: PBS NewsHour (Associated Press), STAT
• FYIEvidence Oct 29, 2026
  60% of Americans would be uncomfortable with their doctor using AI to diagnose and treat them, and only 3% would let AI diagnose alone
  All physicians
  In a Pew Research Center survey of 3,488 U.S. adults on its American Trends Panel, fielded June 22 to 28, 2026, and published Oct. 29, 30% said they would be comfortable with their doctor using AI for diagnoses and treatments and 60% uncomfortable. For diagnosis, 3% would accept AI acting on its own, 47% only with a human final review and 41% not at all; for reading medical scans the figures were 6%, 60% and 26%, and for insurance claim decisions 5%, 38% and 46%. Asked about outcomes, 37% expected AI to make patients' health better and 36% worse, and 47% said it would worsen the patient-doctor relationship. Twenty-six percent had a great deal or quite a bit of confidence that doctors would make sure AI serves patients' interests; 42% had little or no confidence in hospital leadership on the same question, Becker's reported.
  Sources: Pew Research Center, Becker's Hospital Review
• FYIEvidence Oct 29, 2026
  Faster 0/1-hour troponin pathways do not shorten emergency stays or raise discharges vs. 0/3-hour, Lancet meta-analysis of 110,933 visits finds
  Emergency medicineHospital medicine
  An individual participant data meta-analysis of five randomized trials, 110,933 presentations with suspected myocardial infarction, found 30-day death or new type 1 MI in 0.8% with a 0/1-hour high-sensitivity troponin pathway and 1.0% with a 0/3-hour pathway (OR, 0.83; 95% CI, 0.45 to 1.53), Dr. Jasper Boeddinghaus and colleagues reported Oct. 29 in The Lancet. Emergency department stays were 314 vs. 308 minutes (geometric mean ratio, 1.03; 95% CI, 0.85 to 1.25), and direct discharge 67.0% vs. 65.8% (OR, 1.31; 95% CI, 0.79 to 2.16). Heterogeneity between trials was substantial (I2 up to 99.2%), the safety estimate was imprecise, and the primary safety outcome was narrowed after the patient data were received, the authors said. They wrote that hospitals weighing a switch to 0/1 hours should consider local diagnostic and organizational factors.
  Sources: The Lancet, PubMed
• FYIDeployment Oct 29, 2026
  NYU Langone puts OpenEvidence inside its Epic record for 12.4 million outpatient visits a year, residents included
  HospitalsNew YorkFlorida
  NYU Langone Health switched on OpenEvidence's clinical question-answering tool inside its Epic record across its hospitals, ambulatory network of more than 400 sites in New York and Florida, Family Health Centers and trainees at its two medical schools, according to Unite.AI's report on the Oct. 29 announcement. Clinicians type clinical questions in plain language and get answers drawn from peer-reviewed literature that take the patient's chart context into account. The system logs 12.4 million outpatient and more than 421,000 emergency visits a year and has more than 6,000 physicians and faculty and 2,000 residents and fellows. Neither organization reported accuracy or outcome data for the deployment.
  Source: Unite.AI
• FYIDeployment Oct 29, 2026
  Stamford Health pilots FDA-cleared AI that reads routine ECGs for hidden structural heart disease
  Cardiology
  Stamford Health, a 305-bed hospital in Connecticut, began a pilot in which Pathway Labs' EchoNext software reads 12-lead ECGs for signals of structural heart disease, heart failure and pulmonary hypertension, and egnite's CardioCare platform checks each flag against the patient's record to pick out patients for echocardiography, the companies announced Oct. 29. The pilot will track echo completion and time to specialist evaluation, and Edwards Lifesciences is helping fund it. The release describes EchoNext as recently FDA cleared and the first U.S.-cleared AI tool to screen for this range of structural conditions from one ECG; it gives no clearance number or performance figures, and the pilot has reported no results.
  Source: egnite (PR Newswire)
• FYIMoney Oct 29, 2026
  "It could be $10. It could be $10,000": hospitals struggle to forecast Epic's usage-based AI bills
  Hospitals
  Part of what Epic charges for its AI tools now varies with usage, the underlying model and patient activity, and health system executives say that makes annual budgets harder to set, Becker's reported Oct. 29. Epic says more than 85% of its customers use its AI tools. "It could be $10. It could be $10,000, depending on what the usage and the price model looks like," Amy Trainor, Ochsner Health's chief information officer, said at Epic's August users meeting. Advocate Health said patients linking MyChart to ChatGPT rose more than 7,000% from July to August, generating calls against its Epic record, and Epic said in August that it is building cost-tracking and forecasting tools, according to Becker's.
  Source: Becker's Hospital Review