Health AI Partnership Guides

This week, the Health AI Partnership released a collection of actionable guides for healthcare professionals looking to use AI. The guides span the entire AI product lifecycle, from identifying a problem within a healthcare delivery setting to updating or decommissioning an AI product. The content is contemporary and grounded in practice. We surfaced learnings from ten healthcare settings across the United States and incorporate the insights from over 90 healthcare experts who participated in interviews and over 75 industry leaders who participated in a recent workshop.

Health AI Partnership (HAIP) was launched in April 2022 to empower healthcare professionals across diverse delivery settings to safely, effectively, and equitably engage and use AI-based solutions. Our mission is more important than ever before given the exponential rise in interest and evolution of these technologies. AI is advancing at a dizzying speed and regulatory actions proposed or enacted by the Food and Drug Administration, Office of the National Coordinator, Centers for Medicare and Medicaid Services, Federal Trade Commission, and state attorney generals are also rapidly evolving. Healthcare leaders are simply unable to keep pace with changes in technology, the surrounding ecosystem, and the regulatory guardrails to most safely, effectively, and equitably use AI-based solutions.

For the first time, leaders and practitioners within healthcare delivery settings have banded together to surface and disseminate up-to-date best practices related to AI used in healthcare and to cultivate a community of practice. The HAIP leadership team consists of clinicians, engineers, lawyers, and social scientists from DLA Piper, Duke Health, Mayo Clinic, and UC Berkeley. To catalyze the diffusion of best practices most effectively across the community, Health AI Partnership embraces four values.

First, Health AI Partnership is committed to advancing equity. We prioritize solutions and implementation contexts that advance health equity. All guides targeting healthcare delivery leaders are freely available and the partnership features organizations that focus on safety net contexts, including OCHIN and Parkland Center for Clinical Innovation (PCCI). The first workshop focused on developing a framework to assess potential impact of AI solutions on health inequities. And the content explicitly covers both AI solutions developed in-house and those procured from external vendors.

Second, Health AI Partnership is committed to improving patient care. Our focus is to enhance responsible adoption of AI-based solutions to advance patient care where applicable, and not to promote any type of AI technology. AI is always framed as a component of a solution that often includes other components, such as clinical workflows, organizational change management, education campaigns, and feedback loops to clinical end-users and patients. Lastly, current content focuses on AI-based solutions that assist with diagnosis or treatment decisions and that prioritize patients for scarce clinical resources.

Third, Health AI Partnership is committed to improving the clinical work environment. The content is written by people familiar with the use of AI-based solutions in clinical contexts for individuals who manage clinical work environments. Significant attention is paid to ensure that front-line talent are involved throughout the lifecycle and that use of AI alleviates burden for front-line clinicians. AI is not framed as a substitute for human labor or to replace clinical expertise but rather to improve the safety and efficiency of care. In fact, much of the content helps healthcare delivery leaders prevent any damage to the work environment as a result of integrating AI into practice.

Fourth, Health AI Partnership is committed to facilitating participation. At present, content is grounded in the lived experience of leaders and practitioners across ten healthcare delivery settings (Duke Health, Hackensack Meridian Health, Jefferson Health, Kaiser Permanente, Mayo Clinic, Michigan Medicine, New York-Presbyterian, OCHIN, PCCI, and UCSF), three ecosystem partners (American Medical Association, DLA Piper, UC Berkeley), and one federal observer (Office of the National Coordinator of Health Information Technology). At regular cadence, the content will be refreshed with information gathered through new rounds of interviews and workshops. Case studies are being solicited from collaborating organizations and healthcare delivery leaders are invited to add and refine new guides to keep the resources evergreen.

The Health AI Partnership represents a new opportunity and standard for healthcare delivery organizations to advance the safe, effective, and equitable use of AI-based solutions in healthcare. No longer are health systems forced into siloed conversations with vendors who promote poorly validated technologies. Neither are healthcare delivery organizations forced to develop redundant solutions to shared problems.

We are inviting healthcare delivery professionals to engage with the HAIP, put to use the guides and best practices, and provide feedback. An initial public request for comments will run through June 16, 2023. Please let us know if you would like for your organization to join Health AI Partnership, contribute a case study, or augment the collection of guides with additional topics. We look forward to working together to cultivate a learning community for responsible, equitable and effective use of health AI solutions.