Healthcare AI today: 95% fail rate, epic at Epic, AI to the rescue of short staffs, more

News and views you ought to know about:

• Most organizations trying out generative AI—a full 95%—are failing to receive a measurable return on the investment. Researchers at MIT came up with the figure after interviewing 150 executives, surveying 350 information workers and analyzing 300 public AI deployments. Drilling into the findings, Fortune reports that many leaders blame such Gen AI disappointments on regulation or model performance—but the MIT results suggest a bigger problem is flawed enterprise integration. “Generic tools like ChatGPT excel for individuals because of their flexibility, but they stall in enterprise use since they don’t learn from or adapt to workflows,” Aditya Challapally of Stanford Online, the lead author of the MIT report, tells Fortune. The magazine also notes MIT’s documenting of a “misalignment” in allocations of resources for Gen AI: “More than half of generative AI budgets are devoted to sales and marketing tools, yet the MIT researchers found the biggest ROI in back-office automation.” 
   
  • Staying on the findings for a second story, an intrepid Fortune reporter writes of a “shadow AI” economy that’s booming at the edges of the conventional one. The MIT study, the magazine explains, finds that workers at more than 90% of companies are using personal chatbot accounts for daily work tasks—“often without approval from IT.” In fact, only 40% of companies have a paid LLM subscription. “Many shadow users describe interacting with LLMs multiple times a day, every workday,” Fortune reports. Such adoption “often far outpaces these workers’ sanctioned AI initiatives, which remain stuck in pilot stages.” 
     
  • Notably, the Fortune article was drafted by AI and edited by a human.
     
  • The MIT report comes from the institution’s Project Nanda (for Networked AI Agents in Decentralized Architecture), which is working to build infrastructure for the “Internet of AI Agents.” Read the research report here. 
     
• It was an announcement so (ahem) epic, it could only be made to 10,000-plus healthcare executives jammed into a gleaming underground arena. “We are combining the intelligence and curiosity of the human being with the investigative capabilities of gen AI,” came the voice over the crystal-clear sound system. The speaker was Judy Faulkner, the 82-years-young founder and CEO of Epic Systems. The event was the EHR giant’s annual users’ group meeting at the company’s sprawling Wisconsin campus—a setting so grand it’s a tourist destination in its own right. Faulker and co-leaders described a number of AI features from a slate of 200 or so that it’s working on in an effort to please patients, providers and payers. 
   
  • For many in attendance Aug. 19, one highlight was the unveiling of an AI charting tool Epic is building with Microsoft. Later, in prepared remarks, Microsoft VP Joe Petro said the collaboration with Epic is part of a vision to “help transform the healthcare experience through innovation that delivers meaningful outcomes.” The Microsoft-Epic alliance, he added, will enable the melding of Microsoft’s Dragon ambient AI technology with Epic’s new AI Charting offering. 
     
  • Faulkner, dressed in something of a Buzz Lightyear getup for the sci-fi themed event, said Epic’s MyChart Central component will let patients access their medical records with a single set of login credentials even if they’ve been seen by multiple providers unaffiliated with one another. “This is a huge time of growth for technology and medicine,” Faulkner said. “Things” are being done, she added, “that couldn’t be done before.” 
     
  • Get additional event coverage from CNBC and/or Fierce.
     
• Healthcare pay rates shot up 20% over the last four years—yet only 52% of healthcare employees feel they are compensated fairly. That’s the lowest among all economic sectors. Further, many hospitals continue to need more nurses even though healthcare added 2.6 million jobs while rebounding from the Covid crisis. The notable numbers are from the Teachers Insurance and Annuity Association of America. The group’s research arm, the TIAA Institute, arrived at its findings by interviewing healthcare leaders and other experts. They also analyzed relevant research and, in a report, wove the results into a “strategic framework to help health systems meet current and future challenges.” 
   
  • U.S. healthcare is “at a crossroads, grappling with the demands of caring for a rapidly aging population,” the report’s authors remark. “Staffing shortages are creating stress and burnout, productivity remains stagnant, and operating costs are growing—all of which are unsustainable in the long run.” 
     
  • The path forward, the analysts believe, “requires cultivating new sources of talent, enhancing the employee value proposition—and using AI and other advanced technologies to streamline workflows and increase efficiency.”
     
  • Mull the principles and particulars of that strategic framework after downloading the full report. 
     
• Margaret Lozovatsky, MD, was shocked. Why was AI warning her of an 85% mortality risk in a mostly healthy 6-year-old patient? The child was hospitalized, sure, but only with the usually non-life-threatening RSV virus. From her clinical perspective, she knew the AI’s assessment was inaccurate. When she and some colleagues dove into the case, they found that particular AI tool had never been tested on pediatric patients. “Imagine the potential risk that could have happened if I used that value to change the way I cared for this patient,” says Lozovatsky, who happens to be the AMA’s chief medical information officer and vice president of digital health innovations. “This is why you need clinicians in the conversation.” The AMA posted the item Aug. 21 to get the word out about its toolkit designed to steer provider orgs through appropriate intake and evaluation processes for new AI tools. The helps include a working group with physicians and other health professionals. Learn more here. 
   
• Robert Uzzo, MD, MBA, is concerned. Specifically, he’s unsettled by AI’s unchecked incursions into urology and nephrology. The EVP for cancer services at Temple University Health System fleshes out some important flash points in a piece published Aug. 21 in Renal + Urology News. Uzzo says the aspect of clinical AI he finds most worrisome is the use of AI-driven decision support in treatment recommendations. “We’re now seeing systems that recommend surgical vs conservative management based on population-level outcomes,” he writes. “But these systems don’t—and can’t—grasp patient goals, cultural values or subtleties in comorbidity profiles.” Machines “don’t sit with the patient,” he adds. “We do.” 
   
• From AIin.Healthcare’s sibling outlets:
   
  • Cardiovascular Business: FDA clears CCTA AI coronary plaque assessment from Artrya 
     
  • Radiology Business: ‘Hybrid’ AI reading strategy cuts radiologists’ mammography workload by nearly 40%
     
  • Cardiovascular Business: Cardiovascular device approvals continue to lead the FDA's Breakthrough Device Program