 | ● |  |
|
|
|
| | | News and views you ought to know about:- Clinicians who come to rely on AI for decision support risk the dulling of their skills. The concern is not new. But now comes a pointed call for researchers to inquire about the particulars of the peril. And it’s just one of five avenues of investigation urged by three scholars in the Journal of the American Medical Informatics Association, aka JAMIA. Each of the five, posed as questions, “underscores the importance of balancing efficiency, accuracy and clinician expertise while mitigating bias and skill degradation,” write cognitive scientist Tad Brunyé, PhD, and colleagues. Proactively addressing the challenges spurring the questions can “maximize AI’s potential across healthcare and other high-stakes domains.”
- The other four questions: What type and format of information should AI provide? When should such information be presented? How might AI, whether explainable or black-box, affect diagnostic decisions? How can AI influence automation bias and encourage complacency?
- “Diagnostic errors remain common across medicine, with AI offering potential to reduce both perceptual and interpretive errors,” the authors write. “However, the impact of AI depends critically on how and when information is presented. Studies indicate that delayed or toggleable cues may outperform immediate ones, but attentional capture, overreliance and bias remain significant risks.”
- Abstract here. The full paper is behind a paywall, but decent coverage by UCLA Health is here.
- When clinical AI grows up, it might want to be a medical specialist in its own right. Or maybe a subspecialist in an established field like clinical informatics. Either way, think about this: The rise of imaging technologies led to the creation of radiology. The emergence of molecular assays birthed created clinical genetics. And the use of tissue microscopy shaped modern pathology. Each of these fields “had roots as a technical adjunct, then matured into a clinical discipline with distinct expertise, training pathways and standards of practice—driven by factors beyond the technology itself,” two Harvard researchers point out in a succinct opinion piece. “The question is whether artificial intelligence now requires the same professionalization.”
- Clinical AI training might complement or expand on—rather than duplicate—existing informatics training structures, David Bates, MD, and medical student Arjun Mahajan write. “Given that AI systems evolve rapidly, training would likely need to emphasize principles of critical appraisal and continuous learning rather than mastery of any one platform,” they note. Whether healthcare AI’s many and varied functions “ultimately coalesce into a distinct specialty or evolve as a track within existing training pathways remains an open question, though one that may echo how new competencies have historically found their footing in medicine,” Bates and Mahajan reiterate. “What matters now is recognizing that medical AI is not a fixed tool but an evolving practice requiring interpretation and oversight.” Hear them out.
- Publicly traded companies produce an inordinate number of medical devices that get recalled. A study backing the suspicion with hard data came out in August. It was covered in this space. This week its corresponding author is making sure the finding gets through to a critical mass of stakeholders. “The lopsided nature of these recalls should give every advocate for medical AI pause,” Tinglong Dai, PhD, tells the Johns Hopkins news operation. “Publicly traded companies, the big fish in this still-small pond, built just over half the devices but were responsible for nearly all the recalled units.”
- What’s more, recall tallies were unnervingly high among tools with “no reported clinical validation whatsoever,” Dai adds. “We just thought, ‘Wow—if AI hasn’t been tested on people, then people become the test.’”
- For many in the Western world, an AI health & nutrition coach is a nice option. For untold millions of people living in underdeveloped regions, the same tool could be a lifesaver. “AI can enable personalized nutrition guidance tailored to genetics, lifestyle and environment, far beyond one-size-fits-all interventions,” sustainability advocate Mariam Adebayo, LLB, states in an article posted by the World Economic Forum. “Digital monitoring tools can provide real-time data, allowing quicker policy responses, while mobile platforms can help expand reach, delivering education and support to even the most remote communities.”
- Malnutrition and diet-related diseases already cost the world more than $8.1 trillion annually, Adebayo notes. “Without urgent action, they risk triggering the next global health crisis. AI offers a way forward—reducing disease, cutting healthcare costs and strengthening workforce productivity. Investing in AI-powered nutrition is not just health policy; it is economic defense against a looming global crisis.” Read the piece.
- Women physicians can shape the future of healthcare AI. Male doctors can do that too, of course. But for the American Medical Association, September was Women in Medicine Month. Following up on the awareness-raising episode Oct. 30, the AMA revisited key themes with Ashley Beecy, MD, chief AI officer at Sutter Health. She fields a lot of questions from women physicians, and one seems to come up a lot: Is it too late to get involved in AI? “No, it’s not too late!” Dr. Beecy tells them. “This technology is evolving so rapidly. Just jump on the train.” Get the rest.
- Also worthwhile:
- Notable research news:
- From AIin.Healthcare’s sibling news outlets:
|
| | | |
| |
| | | - Don’t Let AI Touch Patient Data
Shadow AI is sneaking into hospitals and clinics faster than IT Teams can track. Download the free Security Checklist from Fellow to uncover where AI notetakers may be recording or storing sensitive conversations. Use it to flag compliance risks early and keep your organization on the safe side of privacy laws. Before Shadow AI spreads inside your team, download the checklist to reduce your risk here.
Nabla launches Nabla Connect: The fastest way for any EHR to deliver world-class ambient AI
Nabla Connect is a new plug-and-play module that lets EHR vendors embed Nabla’s trusted AI assistant directly into their platform, making integration fast, compliant, and effortless.
In just a few days, EHRs can: - Integrate ambient AI without heavy engineering
- Enhance clinician experience with seamless, accurate note generation
- Stay future-proof through continuous updates
Learn more about Nabla Connect
Know exactly who said what in clinical transcription 👀
Speaker attribution should not be guesswork. Speechmactics’ Medical Model pairs high-accuracy transcription with real-time diarization so orders, findings, and timings stick to the right voice. Built for multi-speaker clinical workflows and messy overlaps. Read more.
|
| | | |
|
| |
|
|
