Exclusive Interview with the Researcher: Anna Zink on AI and Heart Disease Diagnoses

Exclusive Interview

In 2018, Medicare established coverage and reimbursement for its first AI-enabled service: computed tomography (CT) fractional flow reserve (FFR-CT). A new Health Affairs publication by Anna Zink, PhD, Hannah Neprash, PhD and Michael Chernew, PhD, explores the adoption of FFR-CT and its impact on the diagnosis of coronary artery disease through cardiac imaging. Learn more in this exclusive interview with Anna Zink.

Q: What impact did the AI diagnostic tool have on utilization, costs, and health outcomes? 

A: Clinicians who adopted the AI diagnostic tool for diagnosing coronary artery disease (CAD) changed the basket of tests they used, moving away from invasive testing and toward AI-assisted testing and its required underlying imaging. Ultimately, this increased spending per newly diagnosed patient because the savings from fewer invasive tests were offset by the cost of the AI diagnostic tool and its associated imaging. We did see some evidence that adoption of this tool was associated with improved health outcomes in the year following use, and we hope to do more research to understand the long-term outcomes of AI adoption. We also saw some early evidence that the AI diagnostic tool might increase physician productivity, in the form of higher visit volume post-adoption.

Q: Why are these findings important for the current policy discussion around AI in health care?

A: The Centers for Medicare & Medicaid Services (CMS) is thinking carefully about how to price these new AI diagnostic tools (as they decide whether to reimburse for their use) and getting that pricing right is challenging! In our paper, we saw the mechanisms working as expected (i.e., a decrease in expensive, invasive testing), but because the price of the AI diagnostic tool was set so high (in addition to the imaging required for the test), no net savings were realized. There may be savings down the line in the form of avoided treatments and adverse health events, but CMS is unlikely to have that evidence available when setting initial pricing.