CMS is launching an artificial intelligence challenge in partnership with the American Academy of Family Physicians to refine the agency's predictive modeling practices, the head of the CMS innovation arm announced at HIMSS on Wednesday.
The goal of the AI initiative is to spur private interest in the tech, improve the government's grasp on healthcare data, eliminate unnecessary care and cut costs, Adam Boehler, who also advises HHS Secretary Alex Azar on value-based care, said.
The challenge should be officially unveiled over the next few weeks. Boehler's teaser follows President Donald Trump's executive order earlier this week directing his agencies to prioritize and set aside funds for AI programs.
Boehler's dual roles as innovation chief and HHS deputy secretary put him in a unique position for nudging the healthcare industry, whether it's introducing new payment models, pushing Congress to enact legislation or oiling AI adoption in the private health tech sector and the federal government alike.
Though details on the upcoming competition are scarce, CMS would provide de-identified data to participating groups who would then devise AI solutions for delivering better, more intelligent care. There will be a "significant" cash prize to the winner, Boehler said, to incentivize groups to help redefine the quality of CMS predictive algorithms.
XPRIZE, a nonprofit organization that designs and manages competitions to encourage technological development, will be working with CMS and AAFP to organize the AI challenge.
The competition could be officially rolled out in the next few weeks, Boehler told media, though he couldn't give a more detailed timeline. CMS expects significant physician participation given the reward and the long-reaching implications of the tech.
"Our job, really, is not to redefine AI for everybody, it's really to popularize it, provide data and then create an incentive so that the private market" can take the reins, Boehler said. "This is an important enough issue where we can at least get people to 60, 70, 80% and then let the private market take it to 99%."
Some use cases Boehler mentioned for more streamlined analytics included helping predict onset of life-threatening but avoidable conditions such as septic shock, or the risk of expensive medical emergencies.
CMS has already trimmed some of the fat when it comes to predictive metrics, data points that serve as performance indicators on the success of a program, but there's still a ways to go, Boehler said. "[Administrator Seema Verma has] gone from 600 some odd metrics down to 400," Boehler told reporters, "which is a huge improvement. But it's still 400 some odd metrics."
Boehler pointed to a Google Brain/University of California, San Francisco study on leveraging derived metrics and AI to predict quality, which found results more predictive than any of CMS' best quality metrics. That attracted the government's attention and got it searching for a way to get feedback from the private industry about redefining quality, he said.
Some providers are wary that artificial technology capabilities will heap more burden onto their practices instead of removing it, but Boehler stressed the potential tech won't be hard to understand for the average health worker. "We want to make sure it's not a black box because we want clinicians to be able to learn from it," he said.
"AI and technology, when used well, does not dehumanize. It's quite the opposite," Boehler said to applause at the panel closing Wednesday's HIMSS festivities. "It lets you focus on the portion of healthcare that is human."