In this HealthLeaders podcast, Jim Blum, CMIO at University of Iowa Health Care and a Mastermind participant, talks about giving clinicians the tools to improve their workflows and patient care without hitting them over the head with governance.
Is it possible to do too much with AI governance?
James Blum, MD, thinks so. The CMIO at University of Iowa Health Care and a participant in the HealthLeaders Mastermind program on AI in clinical care, says the technology needs to be treated with the proper safeguards, but that doesn't mean separating it from all other innovative tools and processes.
"AI and healthcare probably shouldn't exist in the vacuum," he said during a recent HealthLeaders podcast. "And we shouldn't be acquiring AI for the sake of AI. We should probably be looking to solve problems that people have, and if that involves AI, great. If it doesn't, that's probably in many ways better because it takes out display of governance and potentially a lot of additional expense."
UI Health Care has launched two AI tools for clinical care: An ambient transcription platform developed by Nabla, which roughly 1,100 of the health system's 3,000 doctors are now using, and a chart mining platform from Evidently that collects all relevant data on a patient from multiple sources to give clinicians a concise view of the patient.
"I can see these very sick patients [with] long, complex medical histories and, really, I'm able to know as much about that patient as the intern that was up all night trying to comb through their entire medical history," he points out.
Blum says both tools were carefully reviewed by UI Health Care through a normal process for reviewing new vendors. With AI, that includes bringing in clinicians and IT personnel who understand the nuances of the technology.
"It is with a group of individuals that are qualified to review the AI right and really understand the performance characteristics and what can be expected of the technology in addition to our typical acquisition process," he said.
"And that's where the AI committee can get engaged and say, ‘OK. Let's look at the performance characteristics and training set and those types of things and give a thumbs up or thumbs down' as to [whether] they think the science behind the AI algorithms is good. And if it's not, then convey that to the groups that are doing the acquisition and say we really don't endorse this because we think it's not a product that's really going to deliver what they're purporting it will deliver to you."
Blum says AI is such an intuitive technology that it doesn't need the intensive resources and training for adoption that health systems typically set aside.
"With all types of departmental meetings and going to individual clinics and a lot of hand-holding … we would we would get nowhere near this level of adoption this quickly if we did it the way we typically roll out," he said.
Instead, with the ambient dictation tool, the health system held one training session and made that available on video. This takes the pressure off of the IT department and gives clinicians more of a responsibility to understand how AI will work for them.
"If it's not working for you, go back and do things [the old way}," he said. "So you don't need to have this really comprehensive, elaborate rollout and you can go ahead and basically turn the technology on and let the use of it grow organically."
"It's a much more targeted intervention and I think results in a much greater utilization," Blum added, noting the health system can identify who isn't using the new technology and help them if needed. "We're not going to go ahead and force you to use the technology you don't want to use, and we're also not going to spoon-feed you. But if you want to be better, you want to be more efficient, it takes a little bit of self-motivation. How many people went to a class on how to use their smartphone before they started using it?"
Blum says he's excited to see how generative and predictive AI evolve and are worked into the clinical space. And while UI Health Care is always looking for new partnerships, the health system will balance doesn't want to buy or build a new tool that will be rendered obsolete by a new capability from its EHR partner in a few years.
More important, he said, are the cost and data storage concerns that come with AI.
"All this stuff, all these large language models and all this processing and the tokenization just requires a ton of computational power," he said. "And this introduces an entire new realm of expense for health systems, without there being necessarily a great financial upside on these things."
Blum says health systems will have a hard time defining an ROI that will please everyone. AI might be great at reducing physician stress, but is that enough for CFOs and CEOs who want the health system to meet increasing patient demands?
"This leads to millions of dollars a year that you're spending on these types of technologies and you may not see a financial upside for that," he says. "And in this era of 1% margins, this gets to be a real challenge."
To listen to the HealthLeaders podcast, click here.
The HealthLeaders Mastermind program is an exclusive series of calls and events with healthcare executives. This Mastermind series features ideas, solutions, and insights on excelling in your AI programs.
To inquire about participating in an upcoming Mastermind series or attending a HealtLeaders Exchange event, email us at exchange@healthleadersmedia.com.
Eric Wicklund is the associate content manager and senior editor for Innovation at HealthLeaders.
KEY TAKEAWAYS
UI Health Care is using AI tools that transcribe the doctor-patient encounter and help physicians assemble and view patient data.
Jim Blum, the health system's CMIO, says AI is governed carefully, and always with a human in the loop, but executives aren't overwhelming clinicians with guidance or instruction on how to use the technology.
The real challenge, he says, will be finding an ROI in tools that satisfy both clinicians and those in control of the purse-strings.