From principles to practice: an Implementation Decision-Tool for Ethical AI in clinical governance

Healthcare institutions assessing clinical artificial intelligence (AI) must turn ethical principles, evidence standards and implementation guidance into decisions about specific systems in specific contexts. This paper proposes the Implementation Decision-Tool for Ethical AI (IDEA), an author-developed normative framework for clinical governance committees. IDEA organises assessment across 10 domains and 3 item-level priority tiers: Tier 1 blockers, Tier 2 concerns that can be addressed through mitigation and Tier 3 opportunities for institutional excellence. A mandatory synthesis then connects these assessments to a documented institutional recommendation. Where unresolved Tier 1 uncertainty rules out clinical implementation, IDEA provides a separate route for governed non-clinical evaluation. It also sets minimum procedures for authority, disagreement, conflicts of interest and decision recording. A worked hypothetical application to published evidence on a real sepsis prediction algorithm illustrates how IDEA is intended to operate under evidential uncertainty. IDEA complements broader institutional governance and implementation processes by giving committees an explicitly ethical structure for judgement and authorisation.