AI adoption for real-world healthcare settings can’t happen in a vacuum. Behind every successful implementation is a work culture open to innovation, a leadership team supportive of continuous improvement and a PPT framework primed to be built upon.
PPT, of course, is people, process and technology—the building blocks of both optimized operations and change management. All of this comes through, even if between the lines, with many a healthcare technology leader who has lived these concepts through to meaningful AI adoption.
Case in point: Vincent Liu, MD, MS. The critical care physician, senior research scientist and clinical informaticist for the Permanente Medical Group of Kaiser Permanente recently took questions from Todd Unger, chief experience officer for the American Medical Association. The topic of their discussion: What it takes to bring an AI initiative to life. Here are five of Dr. Liu’s pointers, lightly edited for clarity.
1. Involve physicians in decision-making around AI tools.
“Our physicians are on the front lines [with this technology], and they are leading as well,” Liu tells Unger. “When you have that kind of structure, you find tremendous opportunities to understand the enhancements that [AI] can bring to the everyday work that we have and to our objectives for achieving the Quadruple Aim.” More:
Whether it’s patient outcomes, population management, healthcare costs or provider wellness—all of those [quadruple aim objectives] are more important than they’ve ever been.
2. Speaking of the Quadruple Aim, it’s a good guide for deciding where to start—and prioritizing the AI projects you want to pursue down the line.
Liu uses ambient AI as an example because his group recently published on their deployment of the technology for all Kaiser physicians in Northern California. Those doctors are now using ambient AI to capture and file conversations with patients, he explains.
Ambient AI represents a key step toward reducing the burden of manual documentation, and it frees up doctors to do what they love to do: interact with patients.
3. Let the A in AI stand for augmented. This will help keep people at the center of your AI picture.
Liu says there are three core competencies for AI understood this way—clinical integration, technology enablement and data science. “It’s all about, what is the method?” he says. “How do we rigorously test and evaluate performance metrics of those methods?” More:
How do we evaluate this to know that the AI technology—integrated with our clinical workforce and our data scientists—will ultimately produce something that’s valuable for patient health or provider wellness?
4. To train your teams on AI appropriate use, extend existing educational efforts.
“We’ve been doing quality and performance improvement for a long time,” Liu says. “And we’ve had a long history of technology implementation.” Those experiences have proven instrumental, he suggests, in coordinating AI training across complex, multidepartmental teams.
What distinguishes AI from what’s come before are uncertainties about how the new methods will work and, more importantly, what their fail points might be. Where can we apply AI with confidence? And where should we be especially cautious with it?
5. Sure, some of the excitement over large language models is hype. Don’t let that keep you from embracing this ‘next big thing’ in AI.
“How do we identify the best use cases, the ones that maximally, again, help us to reach the Quadruple Aim?” Liu says.
Anything that can reduce physicians’ avoidable burden—especially administrative or clerical tasks—has got to be top of mind for every health system and every medical group.
View the AMA video interview.
