On June 26, 2024, the Coalition for Health AI (“CHAI”) released draft v0.3 of their Assurance Standards Guide (“Guide”), which is meant to serve as a “playbook for the development and deployment of AI in healthcare.” It seeks to do so by providing “actionable guidance on ethics and quality assurance.”
CHAI describes the Guide as the “next crucial phase” of its healthcare AI framework, and says it builds on its “Blueprint for Trustworthy AI in Healthcare,” released on April 4, 2023.
As a companion document, CHAI has also drafted the Assurance Reporting Checklist (ARC), which “elaborates the considerations found in [the] Guide at a finer level of detail, providing evaluation criteria for best practices across the AI lifecycle.”
Other AI Standards include the National Academy of Medicine’s (NAM’s) AI Code of Conduct, the National Institute of Standards and Technology (NIST) Artificial Intelligence Risk Management Framework, and the White House Blueprint for an AI Bill of Rights. Internationally, there is also the World Health Organization’s (“WHOs”) Ethics and governance of artificial intelligence for health: Guidance on large multi-modal models.
PURPOSE OF GUIDE
The Guide is said to have been written with multiple stakeholders in mind from Data Science Developers, such as Product Managers, to End Users such as Patients and Health Care Providers. It seeks to build a “shared understanding” among them on “important considerations when selecting, developing and using AI solutions intended for patient care and related health system processes.”
The Guide also seeks to address issues surrounding data privacy, biased or inaccurate results, the non-transparency of AI models, model drift, and workflow misalignment.
THE 6-STAGE AI LIFECYCLE & CORE PRINCIPLES
Focusing on a 6-stage AI Lifecycle, the Guide discusses how to implement core principles related to healthcare AI at each stage. The 6-stages identified are as follows:
- Define Problem & Plan;
- Design the AI System;
- Engineer the AI Solution;
- Assess;
- Pilot; and,
- Deploy and Monitor.
The core principles identified by the Guide consist of the following:
- Usefulness, Usability, and Efficacy;
- Fairness and Equity;
- Safety and Reliability;
- Transparency, Intelligibility, and Accountability; and,
- Security and Privacy.
ABOUT CHAI
On its website, CHAI describes itself as a “community of academic health systems, organizations, and expert practitioners in AI and data science, launched … to identify priority areas where standards, best practices, and norms need to be developed and guidance needs to be developed to frame for directions in research technology and policy.”
Members include John Hopkins University, Google, the Mayo Clinic, and Microsoft, among others.
According to the Version Table in the Guide draft v0.4 will incorporate feedback from public comments and be released on a date to be determined before a “Final Public Release” is published.
Categories: Uncategorized
Leave a Reply