ISO 15189-aligned governance framework FOR clinical ai in medical laboratories: A proposed control matrix, evidence package, and risk-tiered maturity roadmap FOR verification, monitoring, and change control

Abstract

Background: Artificial intelligence is increasingly used in medical laboratories, yet ISO 15189:2022 does not specify artificial-intelligence-specific thresholds, lifecycle monitoring, or change control. This study developed a proposed ISO 15189-aligned governance framework that distinguishes externally supplied systems from in-house or materially modified systems. Methods: A design science approach used a bounded purposive corpus of 31 standards, regulations, professional guidance documents, reporting guidelines, and laboratory artificial-intelligence publications. Sources were coded across intended use and regulatory route, data and measurement inputs, supplier or developer evidence, local verification or validation, deployment qualification, monitoring, incident management, and change control. Evaluation was author-led and formative; no independent expert panel, prospective pilot, or accreditation-audit validation was performed. Results: Five linked artefacts were produced: a qualification and regulatory-route gate; a clause-linked control matrix separating paraphrased normative ISO requirements from author-derived operationalisation; a proposed evidence package incorporating user requirements and installation, operational, and performance qualification; a risk-tiered maturity roadmap; and a worked autoverification monitoring artefact with explicit metrics, proposed local thresholds, log fields, review cadence, and escalation actions. The framework also addresses European regulatory roles, measurement uncertainty, external quality assessment, input harmonisation, and clinical utility. Conclusions: The framework is a proposed implementation aid, not a validated accreditation instrument. It organises ISO 15189 requirements and clinical artificial-intelligence lifecycle controls while requiring local regulatory classification, risk assessment, threshold calibration, and prospective evaluation before routine use.