BRIDGING ARTIFICIAL INTELLIGENCE ETHICS AND ACCREDITATION: ASSESSING ARTIFICIAL INTELLIGENCE OVERSIGHT IN HEALTHCARE ACCREDITATION FRAMEWORKS

Abstract

Artificial intelligence (AI) is rapidly transforming healthcare delivery by improving diagnostics, clinical decision-making, and system efficiency. However, its integration introduces critical ethical challenges, including safety, transparency, accountability, and equity. Although global regulatory initiatives such as the WHO ethical framework and the EU AI Act provide high-level guidance, there is limited empirical evidence on how these principles are operationalized within quality and accreditation standards (QAS). This study aims to evaluate the ethical oversight capacity of two accreditation frameworks in auditing the safe use of AI in healthcare and to identify areas requiring further development. A directed qualitative content analysis was conducted using the WHO’s six ethical principles for AI in health as the analytical framework. Two major accreditation systems—TÜSKA SAS and JCI-AS (8th Edition, 2024) —were purposively selected. A coding scheme consisting of 25 subthemes was developed and applied to standards and measurable elements (MEs). Each ME was classified as auditable, partially auditable, or not auditable in relation to AI-specific ethical oversight. Coding was performed independently by two researchers, and inter-coder reliability was assessed using Cohen’s kappa coefficient, demonstrating strong agreement (κ = 0.88). Discrepancies were resolved through consensus and expert consultation. A total of 73 MEs from TÜSKA SAS and 121 MEs from JCI-AS were mapped to at least one ethical subtheme. JCI-AS showed higher auditability (11/25 subthemes) compared to TÜSKA SAS (7/25). Both frameworks demonstrated strengths in human oversight, informed consent, privacy, and accountability. However, significant gaps were identified in AI-specific domains, including algorithmic transparency, bias mitigation, equity governance, environmental sustainability, and crisis preparedness. Some subthemes, such as public health optimization and crisis mitigation, were entirely absent. Ultimately the selected accreditation frameworks provide a solid foundation for AI governance but require further development through explicit, AI-specific audit criteria. Strengthening these systems with measurable standards for transparency, fairness, and accountability is essential. Leveraging their continuous improvement structure offers a strategic opportunity to transform accreditation into a proactive governance instrument, ensuring the safe, equitable, and trustworthy integration of AI into healthcare systems.