AI was a topic that dominated discussions at this year’s ViVE conference, which was held in late February in Los Angeles.
There’s certainly a great deal of excitement surrounding the technology — but it has yet to prove itself when it comes to solving healthcare’s most pressing crises, such as the clinical burnout crisis or the sector’s massive dearth of workers. From what experts told me at the conference, health systems aren’t seeing a ton of ROI when it comes to their AI investments quite yet, as use of the technology is still in an iterative stage.
That is likely to change in the next couple years, though, according to Mount Sinai CEO Brendan Carr.
During an interview at ViVE, he described a grid-like conceptual model to organize his thinking around healthcare AI.
“I think about this 2×2 grid. In it, I think about administrative things that AI is good for and clinical things that AI is good for, crossed with whether or not these are good for the patient or good for the provider. [An AI use case] doesn’t have to be in only one of those four boxes, but this helps me to organize my thoughts,” Carr explained.
Using this organizational thinking, he has found that there seems to be a lot of relatively low-risk administrative use cases for AI that are beneficial for patients.
For example, Carr pointed out that providers are beginning to see success when using AI to nudge patients in the right direction. This includes tools that remind patients to get screenings, give them pointers on their diet or encourage them to increase their daily steps, he said.
Things get more complicated when looking at administrative AI use cases that are focused on aiding clinicians’ workflows.
Carr highlighted the EHR inbox as one of the biggest sources of stress for clinicians. It would be wonderful if doctors had tools to help them better manage this, but this enters into a territory where AI could be used to do more serious things, like dispense medical advice or read an imaging study, he noted.
Providers are beginning to use AI to reduce friction in clinicians’ EHR experience, but they certainly haven’t fully automated processes like giving medical advice or reading studies. For things like this, a human needs to be in the loop, Carr remarked.
And it’s the purely clinical side of things where AI deployment is riskiest, he declared.
“The clinical side of the house is the place where people are the most scared — because you’re going to start to either use AI to make clinical decisions, make diagnoses or recommend treatment pathways,” Carr stated. “That’s where we have to be most careful.”
All that being said, he noted that health systems “aren’t more than a couple of years away” from starting to see a sizable return on investment from their AI deployments. He predicted that radiology will be the first clinical field to start seeing marked improvements when it comes to AI-related ROI.
Photo: steved_np3, Getty Images
