The academic health system recently boosted its AI leadership by appointing a chair of AI and human health, who will work with its chief digital information officer to translate research to the clinic.
“There’s a paradox in the AI field right now,” Girish Nadkarni, MD, chair of the Windreich Department of Artificial Intelligence and Human Health at Icahn School of Medicine at Mount Sinai, told Becker’s. “Anything that’s researched never gets implemented, and anything that’s implemented rarely gets researched.”
By having the academic and clinical sides work together on AI, Mount Sinai hopes to change that. “AI and digital and information technology are so inextricably linked that you can’t untangle them,” Dr. Nadkarni said.
Mount Sinai Chief Digital Information Officer Lisa Stump told Becker’s she served on the search committee for Dr. Nadkarni’s role and aimed to find someone who “would be willing to partner with us around the sometimes onerous technical requirements that AI brings: the technical infrastructure, data security” and apply “rigor to the analysis to ensure those tools are not instituting bias or unfairness.” Dr. Nadkarni has been with the health system for over 15 years.
That “necessary rigor” will help identify which of the many AI solutions on the market — and in the research phase — can improve both health system operations and clinical decisions, Dr. Nadkarni said.
“In the case of AI, unless something is safe, effective and ethical, it’s dangerous to scale it,” he said. “Because, as a physician or provider, you can impact maybe one patient or 10 patients, but if you make a mistake in an algorithm that can impact tens of thousands of patients.”
Patients also deserve transparency around the type and accuracy of AI being used in their care, he added.
Mount Sinai has already been a leader in AI, deploying about a dozen internally developed algorithms identifying, say, malnutrition or pressure injury risks in patients, as well as commercial solutions for image reading, diabetes care management and revenue cycle.
“From the back office to the bedside is how I view the scope of our AI efforts,'” Ms. Stump said. “It’s hard to find a digital solution that doesn’t contain some component of AI anymore.”
The team is making sure “missions and incentives align,” Dr. Nadkarni said, by “building AI approaches to solve clinically relevant problems that are actual pain points for front-line providers.”
Like many health systems, ambient AI for clinical documentation has been a big focus for Mount Sinai, which has already piloted one vendor and plans to test out two more before inking an agreement over the summer.
Dr. Nadkarni said AI is essentially “buying time” in healthcare, giving providers more time back in their personal and professional lives and patients more time with their clinicians.
Ms. Stump also hopes to use AI to boost access, from streamlining appointment scheduling and call centers to matching patients with the right clinician as timely as possible.
Cost is just one factor on the “value on investment” equation for AI, Ms. Stump said. Reduction in cognitive burden for clinicians is another.
“Sometimes the harder discussion is … the thing has run its course, and we need to turn it off and take it away,” she said. “Are we getting the expected value however we’ve determined it, and if not, can we pull that solution out of use?”
In addition, Mount Sinai has invested in high-performance computing capabilities that will make future deployments of AI possible, she said.
“There’s also a culture at Mount Sinai from the leadership down, including the CEO, Lisa, clinical leadership, that just publishing isn’t enough, getting grants isn’t enough,” Dr. Nadkarni said. “Translating and improving patients’ lives at scale is critically important.”
The AI leadership structure at Mount Sinai is also unique, Ms. Stump explained. “When I speak with colleagues who are CIOs or chief digital officers across the country, there is sometimes this separateness of a chief AI role or a chief innovation role that inadvertently creates contention between the two roles, and shadow IT sprouts up under those leaders,” she said.
“Girish’s approach to this from the get-go has been that there needed to be a clean vertical, a clean governance process, so we avoid confusion among the researchers and clinicians who are eager often to bring these tools forward when they don’t know who to go to.”
