Leaders within healthcare organizations are confronting huge financial and workforce challenges and are eager to harness AI to improve clinical outcomes, operational efficiency, and clinician well-being. But AI is no silver bullet. Healthcare leaders must mitigate the risk of potential harms that arise from the use of AI-based solutions. while coping with rapidly changing regulations and governance issues. There is an urgent need for community-informed best practices to guide the procurement, development and adaptation, clinical integration, and lifecycle management of AI solutions.

To address this need and to enable healthcare professionals to use AI in a safe, effective, and equitable way through community-informed up-to-date standards, our team at the Duke Institute for Health Innovation launched the Health AI Partnership (HAIP) in 2022. HAIP was launched by leaders from DLA Piper, Duke Health, Mayo Clinic Center for Digital Health, and the University of California at Berkeley. The Gordon and Betty Moore Foundation provided an initial seed grant and, today is fortifying its investment to support the launch of phase II, which further advances the responsible use of AI in healthcare.

Phase I Impact:

HAIP was launched in April 2022 with ten healthcare organizations and four ecosystem partners across the United States. These organizations facilitated interviews of key leadership team members, participated in design research, and shared learnings through case-based workshops. The major contributions and accomplishments so far are:

Launch of the Best Practice Guides for responsible AI in healthcare: A Comprehensive Resource for Healthcare Professionals

In May 2023, the HAIP community released an extensive repository of best practice guides across eight key decision points throughout the use of AI solutions. The content and best practice topics were surfaced through community-centered design research in partnership with IDEO.org and extensive interviews with AI practitioners and healthcare experts. The process by which we developed the topic guides is published in the proceedings of the Fairness, Accountability, and Transparency Conference and also covered in a STAT News Special Report.

Alignment of best practices with Contemporary Standards: Providing a Transparent Mechanism to Comply with Emerging Standards and National Recommendations.

In August 2023, we released mappings of the components of HAIP best practices to the NIST risk management framework, Blueprint for an AI Bill of Rights, Executive Order 13960, and the WHO’s framework for generating evidence for AI-based medical devices. These cross-reference maps enable users of HAIP guides to ensure compliance with principles contained in the various national guidance documents and standards.

The HEAAL Framework: A Pragmatic Framework to Assess the Impact of AI Solutions on Health Disparities.

In October 2023, we released the Health Equity Across the AI Lifecycle (HEAAL) framework, currently available as a pre-print. Our interviews with stakeholders across the country revealed that healthcare leaders struggle to assess the potential impact of AI on health disparities. There is an impending need to describe the steps that healthcare organizations need to take to assess the health equity impacts of AI as several regulatory bodies are organizing to hold healthcare organizations accountable for the propagation of health inequities. The HEAAL framework was thoughtfully developed through a multi-case-based workshop and with extensive engagement of clinical, operational, technical, and regulatory leaders across the US healthcare ecosystem. This framework will be updated after the ongoing peer review process.

Launching Phase II: Scale and Fortify the Community of Practice

We are thrilled to announce the launch of the next phase of the HAIP community of practice. This comes at the heels of successfully establishing HAIP and bringing together an enthusiastic community of practitioners who learn, develop, and share best practices for the responsible use of AI in healthcare. The two-year phase II of HAIP is enabled by follow-on funding from the Gordon and Betty Moore Foundation.

The major components furthering the mission of HAIP for our next phase includes:

1: Continuous AI Best Practices Development

We (HAIP) are formalizing a robust and repeatable process for sourcing, updating, and crowdsourcing content related to AI best practices. HAIP is recruiting topic guide authors from diverse backgrounds to curate and develop pragmatic case studies from the community of participating organizations. Through this approach, HAIP aims to engage with a diverse set of community members and establish a system for regular content updates to ensure the resources remain actionable, relevant, and timely.

2: Scale Participation and Adoption

To increase its impact, HAIP will expand the partnership. A select group of sites will be invited to join the corps, where best practices are identified, refined, and disseminated. A new track for participation will be established to help organizations adopt and implement the HAIP best practices. Through surveys and interviews, HAIP will gather feedback, identify challenges, and enhance the feasibility of implementing AI best practices across sites.

3: Align with Regulatory Agencies, Professional Societies, Trade Associations and Industry

AI is creating new opportunities and challenges for the healthcare sector, and many stakeholders are struggling to most effectively adapt to this new technology. HAIP will deepen relationships with the AI industry, regulatory agencies, professional societies, and trade associations and will seek to align our resources and programs with the priorities of our ecosystem. HAIP will serve as a primary bridge for healthcare organizations.

4:  Sustaining Impact

HAIP will create a sustainability plan to ensure that healthcare professionals can have continuous access to up-to-date and actionable guidance to use AI safely, effectively, and equitably. AI technology is poised to have a profound and lasting impact on healthcare which will be facilitated by partnerships and extensive stakeholder engagement. Through collective experience and expertise, we are developing a rapid pathway to sustainability.

Looking Beyond

We (HAIP) envision a future where the use of AI in healthcare is deliberate, thoughtful, ethical, and responsible. HAIP aims to be the primary source of robust content and a community for healthcare professionals seeking to advance health together through the safe, equitable, and responsible use of AI. The promise of AI to advance health equity, improve patient care, and improve the workplace requires a diverse and engaged community of practice. We invite you to join our community of practice as we begin the next phase of our journey.