Healthcare AI today: Healthcare AI ROI, Microsoft mainlines healthcare AI, Google gets slapped back from within, more

News and views you ought to know about:

• What good is AI in healthcare if it doesn’t improve financial margins? It’s no surprise when vendors ask themselves that question. But note well: It’s increasingly getting posed in the C-suites of provider organizations, including the not-for-profit variety, as well as health insurance companies. So report Bain & Company and KLAS, which collaborated to survey 228 U.S. healthcare executives. Among their key findings: 
   
  • Some 70% of providers and about 80% of payers have an AI strategy in place or in development. That’s a notable leap from last year, when the slice was around 60%. The authors of the survey analysis say the rise reflects “a shift from broad exploration to focused execution. While uncertainty persists, leaders agree that inaction represents the riskiest path.”
     
  • Many provider leaders say it’s still too early to quantify precise financial returns, but “the qualitative read is positive,” the authors remark. Less than 5% of respondents feel the technology is not meeting expectations for categories in which AI solutions have been introduced, and “most are excited or very excited to scale up AI in those areas.”
     
  • ROI outcomes have generally been positive on the payer side too, although “there are more instances of payers saying AI use cases have not met expectations, likely owing to the scale of internal IT and other operating resources available to assess the impact of AI,” Bain and Klas state. “Still, payers appear excited to scale up generative AI use cases—notably for various forms of member/care navigation, in which they see an opportunity to engage members at a scale and level of personalization previously unimaginable.”
     
  • Meanwhile provider-payer pilots in automated prior authorization are demonstrating material impact, the authors report. Here they offer some illustrative if anecdotal evidence. “In one Epic-Humana pilot, 75% of coverage checks were processed automatically, and registration time fell by about 90 seconds per patient,” they write. “These are win-win outcomes, reducing administrative costs for both sides of the system.”
     
  • More numbers and analysis here. 
     
• Microsoft has had just about enough of OpenAI. Either that or it feels it’s outgrown its dependence on ChatGPT. Maybe it’s both those things. Regardless, this week Microsoft—which has invested well north of $10B in OpenAI—announced it’s paying a licensing fee to Harvard Medical School in exchange for access to consumer health content held by Harvard Health Publishing. The move signals Microsoft’s intentions to dive deeper into healthcare AI on its own terms, and it comes with the Big Tech player’s stated aim to update its Copilot AI assistant for better service to healthcare AI users. That’s better read as more reliably accurate, especially for clinicians looking for help with clinical decision-making. 
   
  • The Wall Street Journal broke the (paywalled) story Oct. 8. The business newspaper of record spoke with Microsoft’s VP of health AI, Dominic King, who said his team wants Copilot to rival expert human physicians at earning users’ confidence. King also noted Microsoft has been developing its own AI models while also integrating Anthropic’s Claude, presumably in preparation for a phaseout of OpenAI. 
     
  • The training data Microsoft wants to glean from Harvard will be limited, at least initially, to some specific diseases and wellness topics. Microsoft surely wants to build on the momentum it evidenced in healthcare this summer, when it said one of its AI models had handily outperformed doctors at diagnosing diseases—and for a smaller bill, at that. 
     
  • “No single physician can span the full complexity” of the New England Journal of Medicine case series on which Microsoft tested its AI Diagnostic Orchestrator, the company said. “AI, on the other hand, doesn’t face this tradeoff. It can blend both breadth and depth of expertise, demonstrating clinical reasoning capabilities that, across many aspects of clinical reasoning, exceed those of any individual physician.” 
     
• Fellow Big Tech traveler Google is learning the hard way not to force employees into using healthcare AI. The staffers cried foul when Google announced everyone on the payroll had to let an AI tool access their personal data. The cost of opting out would have been disqualification from benefits in the coming year. After Business Insider reported on the policy and its discontents, Google reversed course. The tech behemoth said, in essence, Oops. Never mind. 
   
  • The outlet quoted a few employee messages posted to an internal communications site. “This is a very dark pattern,” one worker wrote. “I cannot meaningfully consent to my data being shared with this company, and I do not want to consent in this manner.” 
     
  • Another: “Consent for an optional feature like ‘benefits usage optimization’ is not meaningful if it’s coupled to a must-have feature like Google’s HEALTH PLANS! The word you’re thinking of is coercive.” 
     
  • The tool such employees can now reject is from the New York-based healthcare AI startup Nayya. It encourages users to share information about their health and lifestyle, then makes coverage and benefits recommendations. Of course, many Google employees will be OK with that. “For those who still choose to opt in,” BI reports, “data such as pay, gender and social security number will be shared with the tool.” 
     
• It’s easy to forget how far patient knowledge has come in our lifetime. The internet continued the incremental evolution of the democratization of information. AI has put the process into hyperdrive. And yet these digital aids will never make healthcare professionals optional, much less obsolete. Spencer Dorn, MD, MPH, MHA, makes the case at Forbes. “[P]eople don’t just want information—they want solutions,” he writes. “And for that, clinicians remain essential.” 
   
  • But only barely, in a growing number of use cases. “By loosening our grip on knowledge as the core marker of our worth,” Dorn suggests, “we [clinicians] can refine the skills that now matter more—asking better questions, curating information, exercising sound judgment—while embracing our evolving roles as guides, interpreters and trusted partners.” More: 
     
  • “Healthcare is, at its core, built on trust—a blend of reliability, responsibility, security and accountability. Today that trust remains local: People place far more confidence in their doctors and friends than in governments or corporations. But how do we preserve trust in a world where convincing answers—right or wrong—come instantly? And how do we safeguard the value of human expertise while widening access to knowledge? The future of medicine hinges on how we answer those questions.”
     
  • Read the whole thing.
     
• Also worth your while:
   
  • Sen. Bill Cassidy has an idea on how to regulate artificial intelligence: Let AI do it. (Politico) 
     
  • Family offices still bet on AI and healthcare even as deals slow down (CNBC)
     
  • Meet the Southlake teen already making his mark in the medical world (Fort Worth Star-Telegram)
     
• Noteworthy research news:  
   
  • UT-Arlington: New AI tools help scientists track how diseases start
     
• From AIin.Healthcare’s sibling outlets:
   
  • Cardiovascular Business: FDA clears new AI model for MAC detection in routine imaging results
     
  • Radiology Business: Deep learning automates head CT reformatting, bolstering workflow efficiencies
     
  • Cardiovascular Business: ‘Every second counts’: New Viz.ai platform brings the entire heart attack team together