Healthcare AI newswatch: Great expectations for GenAI, disaster-ready AI, Dr. Oz on payer AI, more

Buzzworthy developments of the past few days.

• Investors are betting on generative AI to solve costly provider workflow inefficiencies. So are provider executives. The American Hospital Association has taken note, citing a recent analysis by market intelligencer CB Insights. The scrutiny shows healthcare leaders increasingly looking to AI for help with staffing shortages, rising costs and administrative burdens. CB notes that clinical documentation startups comprised four of healthcare’s five biggest GenAI deals last year. The CB Insights report names vendors. The AHA summary offers insights from the company’s report. Among these: Specialized AI models—“powered by advanced clinical reasoning and domain-specific knowledge”—have demonstrated high accuracy in healthcare workflows. And drug development is poised to accelerate “as generative AI discovery platforms secure major deals.” AHA also cites a recent Gradient Flow survey that showed well more than half of respondents—65%—actively considering or implementing generative AI products. 
   
• Translating healthcare AI from English into any one of the thousands of tongues spoken in Africa is no easy thing. And that’s only one complication vexing efforts to ease severe shortages of human clinicians in the Cradle of Humankind. But, say some hearty souls at Stanford Medicine, many Africans have little choice. If you were given a choice between an unfamiliar AI bot and a trusted human doctor, of course you’d pick the human. Or, at the very least, you’d want to know just how good the AI has shown itself to be. “But if you don’t have that alternative or if your alternative is waiting nine months, what counts as good enough is different,” explains Stanford blogger Rachel Tompa. “In many settings, especially if people are suffering, we may not have time to wait for the perfect AI model.” The picture of AI’s developing role in third-world population health comes even clearer here. 
   
• If necessity is the mother of invention, modern AI has some measure of paternal rights. The technology is increasingly used to anticipate hospitals’ supply-chain needs after disasters strike—or even before. Modern Healthcare looks at the development, spotlighting GHX (whose model predicted IV shortages in the immediate wake of Hurricane Helene in North Carolina), Sg2 (which forecasts patient volumes to match supplies with demand) and Premier (which was spurred to innovate when COVID-19 caused shortages of personal protective equipment). GHX chief product officer Archie Mayani tells the outlet the idea is to use AI to “not only predict the impact of such an event, [but to] aid [provider orgs] in a very proactive way” such that disruptions to clinical operations are minimized. Story here (behind paywall). 
   
• Dr. Mehmet Oz is all for using AI to thwart payers from enlisting AI to help them deny care. His preferred approach is, evidently, to fight fire with fire. Testifying on his nomination for CMS secretary before the Senate Finance Committee last week, Oz said: “We should be using AI within the [CMS] agency to identify that [bad behavior] early enough so that we can prevent it.” He also said the U.S. has “a generational opportunity to fix our healthcare system and help people stay healthy for longer.” 
   
• When Kaiser Permanente models responsible AI, people emulate. Why wouldn’t they? The sprawling integrated healthcare system serves more than 12 million people in 40 hospitals, more than 600 medical sites and upwards of 25,000 physicians working and learning across nine states and D.C. Kaiser has come up with a 7-principle checklist to make sure they’re doing AI right. The stripped-down version goes like this. 1.) Start with privacy. 2.) Continually assess for reliability. 3.) Focus on outcomes. 4.) Strive to deploy tools transparently. 5.) Promote equity. 6.) Design tools for customers—not only patients and families but also healthcare workers who use the tools. 7.) Build trust. Kaiser hopes others will learn from its example: “To realize AI’s full potential, we and all healthcare organizations must use the technology responsibly.” Read the whole thing.
   
• Tell me again why insurance companies are taking heat for using AI to review claims. Simple. For every $10 billion in revenues, health insurers that use the technology for this purpose could save themselves $150 million to $300 million in administrative costs and $380 million to $970 million in medical costs. In these ways, AI “could help generate between $260 million to $1,240 million in additional revenue” for such payers. The numbers are from McKinsey via reporting by Newsweek (which incorrectly identifies McKinsey as a law firm). 
   
• AI is being trained on large datasets of sounds to differentiate between noise and speech for those with hearing impairments. Hearing aid makers have been working with AI for some time, but now they’re getting really good at it. One product on the market “sorts out relevant conversation using directional microphones and a deep neural network trained on 13.5 million spoken sentences and other background sounds,” Healthcare Brew reports. Another manufacturer says its AI “helps reduce the effort needed to focus on listening, which has been linked to cognitive decline.” Get the rest. 
   
• The greatest threat to AI ‘uptake’ in healthcare is the off switch. As used here, uptake is a Britishism for what many Americans would call adoption. A white paper promoted by the University of York explains: “If frontline clinicians see the technology as burdensome, unfit for purpose or are wary about how it will impact upon their decision-making, their patients and their licenses, then they are unlikely to want to use it.” Read the rest. 
   
• Recent research in the news: 
   
  • Stanford: AI can help doctors give intravenous nutrition to preemies, study finds
     
  • University of Texas: Who’s to blame when AI makes a medical error?
     
  • Icahn School of Medicine: Mount Sinai launches center for AI in children’s health to advance AI-driven outcomes in children
     
  • Hiroshima University: Japanese researchers argue ChatGPT is ready to teach medical ethics 
     
  • Guangzhou Medical University: DeepSeek: the ‘Watson’ to doctors—from assistance to collaboration
     
• Notable FDA approval activity:
   
  • Coreline Soft wins FDA nod for upgraded AVIEW 2.0 imaging AI 
     
  • Diagnos seeks Saudi FDA Approval for AI-based medical device
     
• Funding news of note:
   
  • Clinical AI leader Navina secures $55M Series C funding
     
  • Inspiren banks $35M to scale AI-powered senior living technology 
     
  • Silna Health announces $27M in funding to tackle America's prior authorization crisis
     
  • Radiology artificial intelligence firm Brainomix raises $18M, expands in US
     
  • Olio Labs Raises $4.5M Funding
     
  • Arlo raises $4M to expand small business health plans
     
  • University of Tennessee research foundation invests $150K in professor’s AI healthcare claims processing startup
     
• From AIin.Healthcare’s news partners:
   
  • Radiology Business: AI accurately routes interventional radiology procedure requests at pennies apiece
     
  • Cardiovascular Business: Opportunistic screening: AI highlights key heart findings in mammography images
     
  • Health Imaging: AI ‘not safe’ to be implemented as a solo reader for breast cancer screening exams
     
  • Radiology Business: Siemens Healthineers inks $560M imaging and AI deal with Canadian government
     
  • Cardiovascular Business: FDA sees promise in new AI model for heart attack detection