Fig. 3

Respondents identified the frequency with which their respective institutions take actions to promote health equity at every stage of the development and deployment of artificial intelligence-derived predictive models (AIDPM) in their clinical practice. The center line divides each bar into frequently performed actions (to the right of the line) and infrequently performed actions (to the left of the line). Actions to promote health equity were consistent (p-value > 0.05) across institutions except informing frontline clinicians of the health equity impacts of AIDPM, which was significantly less commonly undertaken by institutions (p-value < 0.001)