The Health AI Partnership (HAIP) is the first multi-stakeholder collaborative exclusively focused on empowering healthcare delivery systems to adopt AI safely, effectively, and equitably. HAIP was launched in April 2022 with initial funding from Gordon and Betty Moore Foundation. HAIP envisions to be the trusted source of actionable, contemporary guidance for healthcare professionals seeking to use AI and related technologies.

On February 15, 2023, HAIP organized the inaugural case-based workshop on a contemporary challenge that leaders across settings face: “Our health care delivery setting is considering adopting a new solution that uses AI. How do we assess the potential future impact on health inequities?” The workshop prompted divergent thinking from diverse attendees to surface the many ways that AI can worsen health inequities. The workshop was a critical step to develop actionable guidance that can be applied to a broad range of AI use cases across settings. A framework and series of case studies applying the framework will be published in PLOS Digital Health and through our website in the coming months.

Over 75 clinical, technical, operational, and regulatory leaders from 10 health systems and 7 healthcare ecosystem partner organizations in the United States and Canada attended the workshop.

The workshop featured two real-world case studies to ground the discussion: a postpartum depression (PPD) risk prediction model from NewYork-Presbyterian (NYP) and a machine learning model that segments the patient population (Know Thy Patient) from Parkland Center for Clinical Innovation (PCCI). Using the case studies, attendees discussed in depth a range of problems and approaches to mitigate the potential impact of AI adoption on health inequities.

The discussion was further enriched by reflections from a panel of experts representing diverse perspectives. Expert panel members provided targeted feedback on individual case studies, and framework developer members synthesized discussion topics across case studies to converge on actionable guidance. These external representatives were invited to participate due to their extensive lived experience and expertise working on health inequities and the safe, effective, and equitable use of AI. Participating experts included:

Overall, the workshop was a resounding success. Participants were highly satisfied with the event, felt the workshop provided a safe space to discuss difficult and sensitive topics, and expressed great interest in attending future HAIP workshops. Feedback from attendees highlighted the importance of the multidisciplinary approach, which brought together attendees from diverse backgrounds and encouraged collaboration. Combining the expertise of healthcare professionals, data scientists, technologists, bioethicists, social scientists, and community advocates, the workshop offered a unique opportunity to dissect the complex question of assessing AI products for potential impacts on health inequities.

HAIP partners recognize that AI will revolutionize healthcare delivery. However, it is important to ensure that these AI solutions are designed and implemented with guard rails so that it improves health equity. Multidisciplinary approaches, such as the case-based workshop described here, are essential for addressing these complex issues and for collaboratively developing actionable guidance that promotes health equity.

We look forward to sharing the framework and case studies in an upcoming collection of manuscripts featured in PLOS Digital Health. Meanwhile, if you would like to engage in HAIP activities or provide feedback, you can reach us at haip@duke.edu.

We want to express our deepest gratitude to the teams from NewYork-Presbyterian and PCCI that presented cases at the workshop and to the Gordon and Betty Moore Foundation for funding this work.