Coalition for Health AI (CHAI™) Updates Progress and Plans to Issue Guidelines for the Responsible Use of AI in Healthcare

The Coalition for Health AI (CHAI™) is making strides in the responsible integration of artificial intelligence (AI) into healthcare, focusing on the critical issues of bias and health equity. This initiative, launched in spring 2023, brings together a diverse group of stakeholders, including academic health systems, technology companies, and expert practitioners in AI and data science, all aiming to develop guidelines that ensure equitable outcomes in healthcare delivery.

The potential of AI to enhance clinical research and improve healthcare services is significant. However, there is a growing concern that without a structured framework addressing algorithmic bias, these AI tools could inadvertently perpetuate existing disparities in healthcare. The CHAI™ coalition was formed to tackle these challenges head-on, identifying priority areas where standards and best practices are urgently needed. Notable members of the coalition include Change Healthcare, Duke AI Health, Google, Johns Hopkins University, Mayo Clinic, Microsoft, MITRE, Stanford Medicine, UC Berkeley, and UC San Francisco.

Foundational Themes and Initial Workshops

The coalition's inaugural workshop focused on foundational themes of Bias, Equity, and Fairness, setting the stage for a series of discussions aimed at formulating guidelines for AI's responsible use in healthcare. This initial meeting was supported by funding from the Gordon and Betty Moore Foundation and conducted in collaboration with the Partnership on AI.

Over the course of two days, participants engaged in presentations, group discussions, and breakout sessions that culminated in a 15-page topic paper. This document, which is now available for public input, addresses crucial topics such as Health Equity by Design, Bias and Fairness Processes and Metrics, and strategies to mitigate the impact of bias on marginalized groups.

Michael Pencina, cofounder of the coalition and director of Duke AI Health, expressed satisfaction with the engagement observed during the meeting. He emphasized the importance of including diverse perspectives in the development of ethical frameworks for AI in healthcare, highlighting that while AI has the potential to improve care for all, it is essential to ensure that all stakeholders—including physicians, scientists, researchers, programmers, manufacturers, and patients—are actively involved in the process.

Ongoing Efforts and Future Directions

In addition to the initial workshop, the coalition has conducted virtual workshops addressing various themes such as Testability, Usability, and Safety; Transparency; and Reliability and Monitoring. Papers associated with these workshops have also been published for public feedback, indicating a commitment to transparency and community involvement in the guideline development process.

John Halamka, M.D., president of Mayo Clinic Platform and cofounder of the coalition, noted that while AI can significantly enhance patient care, it also poses risks of exacerbating inequities. He stressed that ethical guidelines for AI applications must be integral to the development process rather than an afterthought. The coalition aims to ensure that patient-centered and stakeholder-informed guidelines lead to equitable outcomes across all populations.

The coalition has also welcomed the Office of the National Coordinator for Health Information Technology as a third federal observer, joining the U.S. Food and Drug Administration and the National Institutes of Health. Dr. Brian Anderson, cofounder of the coalition and chief digital health physician at MITRE, remarked on the national interest in ensuring that Health AI serves the entire population, highlighting the coalition's role in establishing a national framework that promotes transparency and trustworthiness.

The coalition's efforts are geared towards creating an AI framework that emphasizes intentionality, ensuring that AI applications are safe, reliable, and equitable. This framework will guide the development of tools and guidelines applicable throughout the patient care journey, from chatbots to patient records, with a commitment to preventing algorithmic bias from adversely affecting any population.

Eric Horvitz, M.D., Ph.D., Microsoft's Chief Scientist and CHAI™ cofounder, underscored the dual responsibility of harnessing AI's potential while ensuring that systems meet safety and reliability expectations. He noted that the opportunities presented by AI are substantial, but the stakes are equally high, necessitating a focus on equity, transparency, and trust.

As the coalition prepares to convene again in mid-October, it aims to finalize its framework and share recommendations by the end of the year. This ongoing collaboration represents a significant step towards establishing ethical standards in AI for healthcare, aligning with the broader goals of the White House and the U.S. Department of Health and Human Services to eliminate bias in health-focused algorithms and promote equitable healthcare outcomes.

Why it matters

The work of the Coalition for Health AI is crucial for creators and technologists involved in the development of AI systems for healthcare. By addressing algorithmic bias and prioritizing health equity, the coalition is setting a precedent for responsible AI deployment in a sector where the stakes are particularly high. As AI continues to evolve, the guidelines and frameworks developed by CHAI™ will serve as essential resources for ensuring that technological advancements benefit all populations equitably. This initiative not only fosters innovation but also reinforces the ethical responsibility of creators and technologists to prioritize fairness and inclusivity in their work.