Healthcare AI’s Evolving Dialogue
The conversation surrounding the use of artificial intelligence in healthcare is entering a more mature phase, according to Zafar Chaudry, the chief digital officer and chief AI and information officer at Seattle Children’s. In an interview last month, Chaudry noted a significant increase in discussions about the governance and regulation of healthcare AI over the past year.
“Remember, it’s not just about technology. People and process are a big problem in all of this,” Chaudry stated.
Over the past year, Chaudry has observed the healthcare industry placing greater emphasis on the ethical and safety considerations of AI, along with the need for transparency. For instance, some lawmakers have expressed concerns about payers using AI to deny claims, and organizations such as the Coalition for Health AI Coverage (CHAI) and VALID AI are working to develop standardized AI governance frameworks.
Chaudry highlighted the practical challenges of implementing AI in healthcare, explaining that AI tools rarely function as ready-to-use solutions. “I have not seen a lot of actual, tangible use cases that [AI companies] are actually solving. There’s some, but many are very much ‘If you tell us what the problem is, we’ll build the solution for you,’” he explained. “I sort of compare it to when you get your iPhone, you download a whole bunch of apps, and then you can do a whole bunch of things.”
When hospitals collaborate with AI developers to create customized tools, Chaudry believes the contracts should incorporate shared risk. He noted that this type of contract is not currently the norm, as many AI developers prefer to be compensated for their time and materials.
“The problem with the time and materials-type contract is, if I have a great idea, and you agree to build it for me, wouldn’t you love to build it for a long period of time because then my bill keeps going up? It’s very hard to cost-control that,” Chaudry clarified.
Chaudry envisions a more ideal scenario where contracts are based on achieving specific milestones, with payment contingent on the product’s performance. Currently, hospital contracts with technology vendors often do not guarantee that the product will deliver the intended results.
“If you want a blue car and they deliver you a red car, would you just accept the red car? You’d say, ‘Hey, something’s wrong here.’ But with AI in healthcare, it seems to be ‘Well, we’ll kind of give you what you want, and then we’re done,’” Chaudry concluded.
