For instance, researchers at Stanford Medicine already are demonstrating how algorithms could help doctors differentiate new subtypes of disease, allowing them to precisely treat individual patients. AI could help health care teams provide better end-of-life counseling or help smooth out wrinkles in scheduling, anticipating who might miss an appointment and sending additional reminders. Algorithms could help physicians proactively monitor patient record and flag patients who might have undiagnosed genetic disorders.
“It’s all about improving patient care,” Shah said. “I want patients to say, ‘This was the most proactive care I’ve ever had,’ or ‘Scheduling was a breeze,’ or ‘My surgery ran late, but thank goodness my wife was informed that she should come an hour later.’ These things aren’t necessarily going to be the basis for big flashy papers, but that’s okay.”
Implementation of artificial intelligence across the health delivery system undoubtedly will be a boon to patients and providers alike, Shah said. The trick, he added, is to integrate AI in a way that does not disrupt an already stretched health care ecosystem.
“In addition, we need to be cognizant of equity and fairness when considering adopting AI-guided decision making and be open to the possibility that there will be situations in which we should not be using AI,” Shah said. “Building and integrating an algorithm into any workflow will have ripple effects that are beyond just what the algorithm does. So we’re thinking about AI integration as an overall delivery science. It’s not just the algorithms we need to consider; the algorithms and clinicians must work as partners.”
The aim is not to have every decision made by an algorithm but to have every decision supported by one, he added.
A (hypothetical) case in point
Shah further illustrates the potential of artificial intelligence through a somewhat futuristic, hypothetical patient scenario. Vera, a 60-year-old woman with a history of high blood pressure and asthma, comes to the hospital. She arrives with shortness of breath. Her physician must diagnose and treat her condition and consider how to monitor her future health risks, such as heart failure, as well as evaluate her risk for chronic diseases, such as heart attack, stroke and kidney failure. That’s a lot of data to collect.
But what if Vera could don a wearable device that monitors her heart rhythm, respiration, blood glucose levels and blood pressure? This continuous flow of data would provide a real-time view of Vera’s health, and with AI-powered algorithms, her care-team could precisely and accurately monitor her health. (Perhaps an extended, irregular heart rhythm would trigger an alert for her doctor, and a scheduling system would automatically contact her with times for an available appointment.)
While this case is hypothetical, and Vera fictional, the situation she finds herself in is common for both patients and providers.
“What if we could do that — and more — for every single patient? Broadly speaking, part of this is looking at how AI can support personalization of health care at a massive scale,” Shah said. “The point is to bring AI into clinical use safely, ethically and cost effectively, writ large.”