Symposium Explores Opportunities, Challenges of AI in Health Care

As investment in and adoption of artificial intelligence barrels ahead, the health care industry is rife with both opportunity and challenge.

At the recent AI in Health Care Symposium sponsored by the Health Leadership Institute at UVA Health, University of Virginia Darden School of Business and UVA School of Data Science, leaders in AI and health care from UVA and beyond described an industry actively adopting AI applications to gain efficiencies, save money and improve patient outcomes, yet doing so in an often-fragmented ecosystem.

Leaders further described the frequently oppositional forces of needing to move quickly to harness new tools while operating in an environment in which human lives are at stake and an understandable caution pervades organizational thinking.

Dr. Girish Nadkarni, the chair of the department of AI and Human Health at Mount Sinai, described an environment in which health systems are awash in data, and increasingly able to convert the data into actionable insights using newly accessible AI models. Health centers such as Mount Sinai are actively using AI applications to uncover signs of heart failure using electrocardiograms, detect neurological injuries that may be missed by human eyes and uncover patients who fit the criteria for clinical trials, among other applications.

Despite the excitement around such innovations, Nadkarni said the industry was beset by a “paradox” he described as involving ideas that were exhaustively researched yet never made it to clinical care versus tech-driven implementations that had been insufficiently researched.

The industry needs to push forward with technology and AI innovations, Nadkarni said, but do so with appropriate regulation, governance, transparency, and an ability to quickly course-correct if deployment goes awry.

“We have to embed and codify essential principles of humanism into our health care system,” Nadkarni said. “Success in the AI age, the principles will be the same as in any era of human achievement. You need compassion, you need leadership, you need thoughtfulness, you need discipline and discipline in teamwork, and you also need luck.”

At a panel discussion with UVA leaders moderated by Darden Dean Scott Beardsley, Darden Professor Vivian Riefberg detailed examples of AI being actively used to enhance access and patient care while lowering costs — all with what she described as appropriate oversight.

Riefberg, who previously led McKinsey & Co.’s U.S. health care practice, described an AI application currently in-use at virtual urgent care practices of major health systems such as Cedars-Sinai. A patient logging on to the platform complaining of a typical urgent care need is first “seen” by an AI application that asks the most appropriate questions, considers the patient’s medical history and calculates likely diagnoses – all before a human physician joins the appointment.

A recent study in Annals of Internal Medicine described the app’s recommendations as aligned with physicians in most cases, and more often rated as being “better quality” when the recommendations of human and app differed, Riefberg noted. The study doesn’t suggest AI should replace physicians, but that tech may effectively assist in physician decision-making.

The app is designed to be “humble,” Riefberg said, “built to admit when it shouldn’t make a judgement call.”

How to divide tasks between human and AI for optimal well-being was a recurring theme of the symposium.

Dr. J. Scott Just (MBA ‘14) the CEO of UVA Physicians Group, noted the recent deployment of Microsoft Dax Copilot to 600 clinicians across the health system. The so-called ambient listening technology securely records visits and populates the clinical note – a time-consuming bane for many clinicians – immediately upon the conclusion of the appointment.

“It has improved the joy and practice of medicine,” said Just. “It has extended, and will extend, careers for clinicians. It has improved patient experience, because now the physician does not have to spend their time documenting in the computer, and the clinician is able to attend to the patient and engage with the patient in this much better and healthier way.”

The time savings from the AI assistant appears to both enhance job satisfaction while allowing for greater capacity for patient appointments, Just said.

Dr. Meg Keeley (TEP ‘16), senior associate dean for education at the UVA School of Medicine, described the impact of the rapidly changing landscape on medical school education. Given the rapid advancements in the health care landscape, health education has long had a “building the plane while flying it” feel, Keeley said. With the advent of AI, “the plane is trying to fly at the speed of sound,” she said.

There are specific, simple applications currently in use at UVA School of Medicine – including aiding physician communication skills through the use of AI patient actors and grading notes and assignments that formerly took significant human resources. The School is also working toward what Keeley called “precision medical education,” with a dashboard that helps identify trends among students, and develops individually tailored plans for remediation as needed.

Moreover, medical students now learn prompt generation as a foundational clinical skill.

“It is critical for people to understand how to generate prompts, and that you use AI as a consult tool in learning how to generate diagnosis and do clinical reasoning, but not to replace the human,” said Keeley. “Really knowing that you, the human, are responsible for the veracity of that information … we’ve got to remain diligent about keeping the human as part of the equation.”

In an industry that, like many, can be skeptical of change, Riefberg suggested physicians and health care leaders be “open-minded” about emerging tools, and consider not just what could go wrong, but also “what are all the things that can go right?”

Riefberg suggested opportunity lay in thinking beyond an institution’s walls by partnering with trustworthy partners in academia and industry.

“If you all keep reinventing the wheel by yourselves, we don’t have enough money, or time,” Riefberg said.

Multiple entities are pushing AI at UVA forward, including UVA’s LaCross Institute for Ethical Artificial Intelligence in Business, the School of Data Science and the Paul and Diane Manning Institute of Biotechnology, among other departments and schools. Phil Bourne, the founding dean of the School of Data Science and former associate director for data science at the National Institutes of Health, said further development in the area will require elevating individuals with “AI and data in their blood,” to help chart future development paths in areas such as health care.