Will Artificial Intelligence Decrease Physicians' Workload?
When investors, providers, and vendors gathered at KLAS’ Digital Health Investment Symposium, we got to discuss one of HIT’s most infamous topics: artificial intelligence (AI). Rumors about far-out possibilities of AI in healthcare have been flying for decades and have only increased in numbers over the past few years.
But providers and investors don’t want sparkling dreams of what could eventually be. They want a view of how AI could realistically help them in the near future. That’s why we were excited to have two expert panelists—Lonny Northrup, Senior Health Informaticist at Intermountain Healthcare, and Jason Wiesner, Medical Director at Sutter Health—tell us what we can really expect from AI.
Will AI Take Physicians’ Jobs?
Toward the end of the AI panel I was facilitating, I opened up the floor to questions from the audience. One eager attendee asked the following:
“It always made sense to me that any action based on healthcare data would require human oversight for liability purposes. But is there a single activity that can be done now or will be able to be done in the future without a human signing off on it? It could be in administration, the financial side, consorting, or anything else. It could be the most menial task ever. But is there anything in healthcare that AI can do completely by itself and without human oversight?”
“Absolutely not,” one panelist said, “if the task involves delivering clinical care.” As Northrup and Wiesner went on to explain, things like glucose readings, MRIs, and blood analysis can be handled well by technology alone until the patient steps back into the picture. “Diagnostics and medical devices are great examples of things that are automated but insufficient.”
While possibly discouraging to some, this outlook may at least give hope to any radiologists or cardiologists worried about AI making their work obsolete. Our panelists don’t anticipate that happening anytime soon. Why? Because of regulations, for starters. “If you’re expecting 100% accuracy from an AI machine in diagnosing breast cancer, that technology will require a lot of FDA approval and oversight. The legal ramifications would be huge. Setting the bar a little lower and using the machine as an aid to physicians requires little oversight and may ultimately be more helpful.”
One panelist listed teleradiology—something technically capable of standing on its own for about 25 years—as evidence that the regulatory environment will always step in before letting AI take over completely. “I’m not necessarily trying to protect my turf,” he clarified. “I do think that anything AI can give to diagnosticians to help interpret our data is a value add.”
Can AI Increase Physician Productivity?
One provider introduced themselves as a primary care doctor and brought up the pervasive issue of physician burnout and AI’s potential to help.
“It seems to me,” this physician said, “that there is a lot of opportunity in AI to start making providers more efficient and more effective. I have vast amounts of data in my system because someone made me put it there. But I feel that I have become a forensics expert. I spend all of my time going over the data. Computer systems ought to be able to pull out the data I need and be adjustable by the providers. There’s opportunity there. Do you see development happening in this area over the next 10–15 years?”
“It’s already starting,” our panelists said. “There are things happening in AI that will hopefully help with capturing data that as part of a clinical visit—things like noise-to-text technology. I think those innovations will simplify interactions between the doctor and patient and improve what we have with current EMRs.”
Northrup and Wiesner also discussed one project involving clinical dictation for colon screenings. The results showed that providers could get 85% of the data that was needed for reporting from the health record. Providers just needed to get the remaining 15% from the actual encounter.
“However, because of the way most EMRs are set up today, providers have to manually put in that 85%,” mourned one panelist. “We need to create solutions that are smarter and faster at taking advantage of what we already know, augmenting it, and capturing the data we need.”
Conclusion
Some HIT enthusiasts might feel let down by the realism from Northrup and Wiesner. I, however, found it refreshing to know how AI might actually help providers in not too many tomorrows. I’m also eager to see how investors, providers, and vendors will work together to create solutions that will ease physicians’ burdens without taking the human touch away from areas that need it. AI is still giving us plenty to be excited about.
Interested in KLAS DHIS19? Let us know about it!