An AI dispatch from HIMSS25
I'm here at the Healthcare Information and Management Systems Society conference in Las Vegas. So far I've eaten a subpar casino restaurant won ton soup, seen the "Postcard from Earth" show at Sphere, and talked to a lot of people about AI and interoperability in health care (more on the latter in tomorrow's Health Tech newsletter).
I've been trying to get a pulse on how people across health care are thinking about AI. What's different between last year's HIMSS and this year's meeting? Last year, executives told me, people were waffling on AI and were worried about the technology's propensity for hallucinations. This year, it is full speed ahead.
"A year ago, I think people were still in the mindset of like, 'AI: Yes?' versus 'AI: No?' And now, to me, the vibe shift is, 'AI: How much, how fast and how big?'" said Aaron Neinstein, chief medical officer at Notable, which sells a platform for health care AI automation and agents. "I'm not hearing a, 'Should we, will we, can we, where do we start?'...What people are thinking now is, 'This is happening. I need to figure out what it's going to mean and how to start getting ahead of it.'"
In the case of tools for answering medical questions, clinicians are starting to use LLM technology on their own, and health systems are now trying to control the rogue behavior by getting approved tools in place, said Rhett Alden, chief technology officer at Elsevier Health, which sells ClinicalKey — an AI-powered, UpToDate-like clinical information tool. When clinicians use non-HIPAA-secured tools or tools that haven't been approved by the institution, health systems get worried about liability and unauthorized access to protected health information. "I think there's this anxiety from the C-suite, probably, to get ahead of that and get some governance in place so they can actually manage it," Alden told me.
What do health systems want out of AI? I asked several companies and organizations this question, and the overwhelming answer was operational efficiency.
The Coalition for Health AI surveyed its members on this and got a couple hundred respondents, said Brenton Hill, head of operations and general counsel at the non-profit industry group. Health systems firstly are demanding generative AI, followed by administrative simplification and then payer functions, like prior authorization and payment integrity.
Executives at Epic Systems concurred. Coming out of Covid and staff shortages, clinical efficiency and effectiveness are top of mind, said Garrett Adams, vice president of research and development at Epic. Reducing operational friction for both the health care organization and for patients is also a priority.
"We're thinking about how to get patients better access. We're thinking, 'How do we help clinicians be more efficient?' We're looking at helping save health care organizations money. And then outside of that, we're looking at how can we use the data and the technology to really advance medicine," said Jackie Gerhart, vice president of clinical informatics at Epic.
Aneesh Chopra at Arcadia and Mark Sendak of Duke University emphasized to me how financially squeezed health systems are and will be in the near future. There are the cuts to NIH grant indirect funding — though the 15% cap is now halted by court order, many expect that the government's share of money that pays for research facilities and staff to shrink, hurting academic medical centers and med schools. Hospitals are also staring down the barrel of potential Medicaid cuts and potential further reforms to Medicare and other federally funded programs. And when hospitals are squeezed, what's one of the biggest spending areas to cut? Labor.
Executives of companies selling AI products assured me that their products would simply augment existing workforces, not replace them. "There is an acute, persistent shortage of medical coders in the industry, and the typical health system has a substantial backlog of billing that is queued up because the coding workforce is inadequate," said Robin Lloyd, chief operating officer at Codametrix, a company that automates medical coding. He argued that his company's tools could help fill that gap.
Still, Lloyd knows how rapid adoption of new health technology can play out. He remembers that from his time at Nuance, the medical transcription/dictation company now owned by Microsoft. When Nuance's technology took off, "the amount of work for medical transcriptionists very quickly evaporated. And that created social disruption and disruption to ways of life for medical transcriptionists that I have real ambivalence about in retrospect," he said. "I think there [was] a lot more we could have done in partnership with the health systems themselves and with the workforce, it's workers being obviated."
He hesitated to say the same thing about AI and his current company. "Coding is different," he insisted.
From the STAT archives:
- If you missed it, last week I reported that dozens of federal health officials due to present at HIMSS dropped out of the conference. Though I did see one VA official onstage at the interoperability pre-conference forum, HHS officials are noticeably absent. I ran into a group of conference attendees who were gathering confusedly outside a room where Medicare officials were supposed to be speaking. This is what the room looked like instead:
- For a retrospective look down memory lane, check out Casey Ross' recap of HIMSS in 2021, which seems quaint now: There were whispers of what would eventually become CHAI, and industry leaders calling for machine learning watchdogs.
No comments