The Coalition for Health AI (CHAI) has released its draft framework designed to guide the responsible development and deployment of artificial intelligence (AI) in healthcare. This initiative aims to establish consensus standards and provide practical guidance to ensure AI technologies benefit all populations, including those from underserved and under-represented communities. The draft framework, consisting of a standards guide and a series of checklists, is now open for a 60-day public review and comment period.
Functional Components of the Framework
The release of this framework is a critical step towards ensuring the ethical and effective use of AI in healthcare. CHAI, launched in December 2021, previously published a Blueprint for Trustworthy AI in April 2023, developed through a consensus-based effort involving experts from academic medical centres, regional health systems, patient advocates, federal agencies, and other stakeholders. The new guide combines principles from the earlier Blueprint with federal guidance, while the checklists offer actionable steps for implementing assurance standards in daily operations.
The Assurance Standards Guide within the framework outlines industry-agreed standards for AI deployment in healthcare. It integrates principles from various authoritative sources, including the National Academy of Medicine’s AI Code of Conduct, the White House Blueprint for an AI Bill of Rights, and frameworks from the National Institute of Standards and Technology. The guide provides a common language and understanding of the AI lifecycle, exploring best practices for designing, developing, and deploying AI in healthcare workflows.
Accompanying the guide are Assurance Reporting Checklists, which assist in identifying use cases, developing healthcare AI products, and ensuring their effective, valid, and secure deployment while minimising bias. These checklists support the independent review of health AI solutions throughout their lifecycle, ensuring they remain effective and equitable.
Demonstrating Best Practices Through Use Cases
To illustrate the application of its principles, the framework presents six use cases:
- Predictive EHR Risk: Addressing paediatric asthma exacerbation.
- Imaging Diagnostic: Enhancing mammography diagnostics.
- Generative AI: Improving EHR query and extraction.
- Claims-Based Outpatient: Optimizing care management.
- Clinical Operations and Administration: Streamlining prior authorisation with medical coding.
- Genomics: Advancing precision oncology with genomic markers.
These use cases exemplify how the framework’s guidelines can be applied to various aspects of healthcare AI, demonstrating best practices and considerations for effective implementation.
Ensuring Transparency and Inclusivity
Public reporting of the results from applying the checklists will ensure transparency in the use of healthcare AI. CHAI emphasises the importance of public review and comment to refine the framework, making it effective, useful, safe, secure, fair, and equitable. This step is seen as crucial in supporting innovation while building trust in AI's capability to serve diverse populations.
The involvement of diverse and multi-sector patient voices is highlighted as essential for safeguarding against bias and unintended harmful outcomes. Ysabel Duron, founder and executive director of the Latina Cancer Institute, emphasised the necessity of including Latino patients and medical professionals in these discussions to ensure AI addresses and not exacerbates healthcare disparities.
Broader Implications and Regulatory Considerations
The framework's release coincides with increasing legislative interest in AI regulation. Lawmakers are scrutinising the use and oversight of AI in healthcare, with recent hearings addressing the U.S. Food and Drug Administration's (FDA) regulation of medical devices and biologics. Concerns have been raised about the diversity and potential regulatory capture within groups like CHAI. However, the FDA has clarified its role as a federal liaison, not an application reviewer, in its engagement with CHAI and similar organisations.
CHAI’s draft framework represents a significant step towards establishing robust, equitable, and transparent standards for AI in healthcare. It will foster an environment where AI technologies can advance healthcare while ensuring they are safe and beneficial for all populations.
Source: CHAI
Image Credit: iStock
