A conceptual framework for measuring AI health equity

Area of Science:

• Health Informatics
• Artificial Intelligence
• Health Equity

Background:

• Artificial intelligence (AI) integration in global health systems offers potential benefits like improved efficiency and diagnostics.
• Uneven distribution of AI benefits, especially in LMICs, risks reinforcing existing health inequities.
• AI models trained on specific populations may exhibit poor performance in diverse settings.

Purpose of the Study:

• Propose a composite framework, the AI in Healthcare Equity Index (AIHEI), for measurable assessment of health AI equity.
• Provide a standardized scoring system to enable comparisons and incentivize improvements in AI technologies.
• Support informed decisions in AI regulation, procurement, publication, and funding.

Main Methods:

• The AIHEI assesses equity across five key domains: data representation, algorithmic fairness, transparency and explainability, governance and oversight, and community impact and benefit sharing.
• The framework is designed to generate a standardized score for quantifiable comparisons.

Main Results:

• The proposed AIHEI framework offers a structured approach to evaluate equity in health AI systems.
• The index aims to facilitate comparisons across different AI technologies and applications.
• Standardized scoring can drive improvements and inform critical decision-making processes.

Conclusions:

• Pilots are essential to validate the AIHEI's feasibility, refine its components, and ensure reliability across diverse health domains and geographic locations.
• Addressing challenges such as defining fairness contextually, ensuring data sovereignty, and implementing post-deployment monitoring is crucial.
• Quantifying health AI equity is vital for accountability, moving beyond ethical aspirations to measurable enforcement and preventing the widening of health disparities.