Healthcare AI today: AI’s odds of curing cancer, why it shouldn’t even try, agentic AI’s legal exposure, more

News and views you ought to know about:

• If not for off-limits data, AI might have a decent shot at curing cancer. Not all forms of it, of course. After all, cancer isn’t just one disease. It’s many. And they’re all unique. But the technology could go far toward beating the living daylights out of some of them. One major hurdle is the locked status of patient data troves that could train cancer-busting models. Getting over that bump “will take more than just smart algorithms; it will require strong, enforceable guarantees around privacy and security,” a subject matter expert reminds Pymnts. In an Aug. 14 post, the outlet gives a fresh ask to a recurrent question—Can AI cure cancer?—and reports back with insightful answers from knowledgeable sources.  
   
  • “Most cancer ‘cures’ are heavily individualized,” one expert points out. “[Effective treatments] are dependent on patient genetics, immune responses and countless environmental variables no AI model fully comprehends today.”
     
  • “Decisions we make on the treatment of an oncologic patient are made with a multidisciplinary team and individualizing every patient,” notes a surgical oncologist. “Sometimes medicine is not as mathematical as it might seem.”
     
  • And: “The goal [has to] shift from finding a universal cure to designing a unique cure for every single patient.” 
     
  • Read the whole thing.
     
• If AI can’t cure cancer, maybe it should just stand down. That wouldn’t be an unreasonable suggestion in light of new research confirming a sneaking suspicion: Physicians who relied for a time on AI ended up “de-skilled” at cancer diagnostics when the AI was taken away. As put by the European researchers with scientific (and polite) understatement: “Continuous exposure to AI might reduce the adenoma detection rate of standard non-AI assisted colonoscopy, suggesting a negative effect on endoscopist behavior.” The study was published Aug. 12 in The Lancet Gastroenterology & Hepatology. Its corresponding author tells TechSpot the findings are “concerning”—especially “given the adoption of AI in medicine is rapidly spreading.” 
   
• More than 34 million patients are served by more than 17,000 community health centers. Two groups want to bring the blessings of safe and responsible AI to that sector of U.S. healthcare. They’re the Coalition for Health AI, aka CHAI, and the National Association of Community Health Centers (NACHC). The orgs announced a new collaboration Aug. 14. “Partnering with NACHC means we can help move the needle on progressing AI adoption with our nation’s largest primary care system,” says Brian Anderson, MD, president and CEO of CHAI. 
   
  • NACHC head Kyu Rhee sounds a similarly sanguine note. “As AI continues to transform healthcare, we [want] to ensure all community health centers delivering care to at least 1 in 10 people across our nation, including 1 in 5 in rural America—and the more than 326,000 professionals who work within their walls—are at the table,” Rhee says in the announcement. “Our collaboration with CHAI aims to leverage responsible AI playbooks that are transparent, ethical and positively impact the populations they’re meant to serve.”
     
• When it comes to deploying ambient AI scribes in healthcare, picking the right vendor partner is critical. That’s one of the lessons learned at Cleveland Clinic, which is offering its experience as a case study in how to adopt ambient AI at scale. Another lesson: Be patient with end-users. “We had quite a few skeptics who doubted the technology would work,” recalls associate CMIO Eric Boose, MD, who served as primary physician champion during the institution’s yearlong pilot. “And now they’re like, wow, it actually does.” Find out which product they settled on and what else they’ve learned. 
   
• Agentic AI is advancing in healthcare. Lawyers are licking their chops. “Traditionally, product liability case law has looked to a standard of care as the level of care that a reasonably prudent person would exercise in similar circumstances,” notes Meghan O’Connor, JD, an attorney with the Quarles & Brady firm. “In the context of [agentic] AI, would the standard of care need to be higher because the AI software should be held to a higher standard than a reasonably prudent person? And how can we distinguish between the hardware, software and human healthcare provider when [agentic] AI is integrated into care delivery models?” Time will tell when it comes to product liability implications, O’Connor adds, “but we’ll have a front-row seat as this new body of caselaw develops.” Pass the popcorn.
   
• Point-of-care ultrasound machines and smart ECG monitors are among the categories of medical devices increasingly using edge AI. Which is to say they’re able to give preliminary diagnoses before a human even takes a look at the sonograms and readouts. The technology is called edge AI because it does its thing right where the data is generated, often at a considerable distance from a hospital’s centralized information systems. “The question is no longer whether AI belongs in healthcare,” remarks one edge AI aficionado. “It’s how close to the patient we can safely and effectively deploy it.”
   
• Few actions draw heat from compliance auditors like upcoding. That’s when someone tries to bill for a costlier procedure than the one that was, in fact, performed. As AI does more of the work for human coders, there’s a growing risk it will inadvertently upcode, running afoul of those dreaded auditors. “AI is great at speed but not at nuance,” explains regulatory watchdog Timothy Powell, CPA, in ICD-10 Monitor. “When applied to documentation or coding, natural language processing and machine learning tools can misinterpret a physician’s notes or assign codes based on incomplete context. In many cases, AI may infer services or diagnoses that were implied but not explicitly documented, thereby inflating claim values.” Read it all. 
   
• One of the world’s most powerful tech execs has a bone to pick with AI in healthcare. Or with AI that should be in healthcare. Lisa Su, billionaire CEO of the semiconductor giant AMD, says it’s a “travesty” that medicine is so rich in specialists but so lacking in generalists who “can pull it all together.” By “it” she means patient data. She’s emotional about the subject after seeing her mother languish in intensive care with a dire prognosis only to survive another two years. In tech, she tells Wired, “we take complex systems, put them together and make them work.” By comparison, medicine is “often only looking at one aspect of health. It’s my firm belief that, if we can use [AI] technology to help pull all of that expertise together, we’ll be able to treat people better. … In my ‘next life,’ I hope to be someone who can help bridge the divide and use [healthcare AI] technology for what it’s actually capable of.” Q&A here. 
   
• Also worth your while:
   
  • $5.3 billion Abridge is eyeing acquisitions. Here’s what the healthcare AI startup is looking for. (Business Insider)
     
  • The real cost of plugging GPT into healthcare (ITPro Today) 
     
  • The latest ChatGPT is supposed to be ‘PhD level’ smart. It can’t even label a map (CNN)
     
• From AIin.Healthcare’s sibling news outlets:
   
  • Cardiovascular Business: AI now embedded in nearly every cardiovascular IT system
     
  • Health Imaging: AI ultrasound software accurately predicts expectant mothers’ delivery timelines
     
  • Cardiovascular Business: Heartflow raises $364M with IPO, rings Nasdaq opening bell
     
  • Health Imaging: Could ChatGPT be up to the task of monitoring AI drift?