Mapping World Health Organization’s framework requirements on the HAIP Guides

WHO framework requirement	Requirement	HAIP Guides
Model Purpose & Suitability	Understanding the problem and intended use Risks and benefits Interoperability Security User testing	Identify problems across the organization Prioritize Problems Determine dimensions of the problem Determine suitability of technical approaches Evaluate internal resources for adoption Assess legal risks Audit investment decisions Design and test workflow for clinicians Adapt to enable AI implementation Identify and mitigate risks Locally validate prior to integration Determine if AI should be integrated
Algorithmic Validation	Training data External validation Performance metrics Benchmarking	Assess viability of buying or building AI Assess quality of external AI product options Define the role of AI Define performance targets Define successful use
Clinical Validation	Prospective analysis plan Evaluates the impact on the pathway Meaningful endpoint Transparent report	Identify potential downstream impacts Design and test workflow for clinicians Monitor work environment Sustain improved outcomes Determine if work environment requires adaptation
Deployment & Ongoing Monitoring	Generalisability Standards for adoption Monitors safety and effectiveness Reports adverse events Algorithmic audits	Manage changes to the work environment Prevent inappropriate use of AI Monitor AI performance Audit AI solutions and work environment Identify potential risks Determine if updating or decommissioning is necessary Determine if work environment requires adaptation Evaluate expansion to new settings Minimize disruptions from decommissioning
Economic evaluation	Economic evaluation of AI health technologies compares costs and consequences, addressing unique cost profiles and reimbursement strategies for market access.	Monitor AI performance Audit AI solutions and work environment
Communication of results	Transparently communicating clinical evaluation results is vital for safe and effective use of AI health technologies involving datasets, model descriptions, clinical studies, and post-deployment audits.	Disseminate information to end users Audit AI solutions and work environment Disseminate information about updates to end users