As tools powered by artificial intelligence (AI) become more prevalent, and calls for usage guidelines increase, the Coalition for Health AI (CHAI), whose members include leaders and experts representing health systems, startups, government and patient advocates, has released a draft framework for responsible health AI.
The Assurance Standards Guide helps ensure standards are met in the development and deployment of AI in healthcare. The companion Assurance Reporting Checklist, also a draft, provides criteria to evaluate standards across the AI lifecycle, from identifying a use case and developing a product to deployment and monitoring.
The drafts were released on June 26, 2024, and are open for public review and comment for 60 days.
AI Guidelines for Senior Living
Aging services providers can reference the new framework and checklists to guide their own organizations, as well as an example created specifically for senior living.
Last year, LeadingAge CAST Patron Cypress Living crafted 12 principles to guide its AI use. Joe Velderman, a CAST Commissioner and Vice President of Innovation at Cypress Living, took the lead in developing Responsible Use of AI in Aging Services.
Learn more about this effort in the LeadingAge article “Charting a Path to AI in Senior Living.”
CHAI’s Core Principles for Trustworthy Health AI
The CHAI standards enable organizations to assess whether AI solutions are useful and trustworthy. The standards include five core principles.
- Usefulness, Usability & Efficacy: AI solutions should be beneficial, reliable, and improve user experience. They must solve specific problems and show clear benefits for patients and healthcare providers.
- Fairness, Equity & Bias Management: AI solutions must be fair and work equally well for all demographic groups.
- Safety & Reliability: AI solutions should not harm patients or healthcare providers. This involves thorough testing and risk assessments before implementation, and continuous monitoring to detect and address any safety issues.
- Transparency, Intelligibility & Accountability: Stakeholders need clear and understandable information about AI systems and their outputs.
- Security & Privacy: AI systems must protect data confidentiality and integrity with strong security measures.
Governance and Independent Review
CHAI notes two additional priorities that help ensure trustworthy AI—AI governance within organizations and independent review of AI systems.
AI governance can ensure that the quality and ethical standards are implemented effectively and manage risks and changes and legal obligations. A bias management framework is needed to evaluate AI fairness and equity throughout its lifecycle. Training on AI ethics and risks is needed for those who choose, develop, and deploy AI tools.
Independent quality assurance is critical to help ensure safe, efficient, and trustworthy AI systems in health care. It uncovers technical flaws, biases, and unintended behaviors that developers may overlook.
Use Cases
The guide includes six diverse examples that show how organizations can apply the standards, including a Predictive EHR Risk Use Case (Pediatric Asthma Exacerbation) and Generative AI Use Case (EHR Query and Extraction).
Developing the Standards
CHAI says the guide is consensus-based and draws on the collaborative work of patient advocates, technology developers, clinicians, data scientists, civil servants, bioethicists, and others. It is written for an equally broad audience.
The principles underlying these documents align with the National Academy of Medicine’s AI Code of Conduct, the White House Blueprint for an AI Bill of Rights, several frameworks from the National Institute of Standards and Technology, and the Cybersecurity Framework from the Department of Health and Human Services Administration for Strategic Preparedness & Responses.
“This framework offers invaluable insights to ensure AI solutions are not only effective but also ethical, enhancing care quality while upholding the trust and transparency essential in aging services,” said Scott Code, vice president of LeadingAge CAST.
Please Give Feedback
CHAI encourages health care stakeholders to provide feedback, with the 60-day public comment period open until late August.
Nursing Home Staffing Mandates: What You Need to Know
