Responsible implementation of artificial intelligence across the medication-use process: an evidence-informed framework for medication safety.

چکیده اصلی

BACKGROUND: Artificial intelligence (AI) is increasingly used across medication-use workflows, but discrimination alone does not establish safe or useful implementation. OBJECTIVE: To propose an evidence-informed framework for responsible implementation of AI across the medication-use process. METHODS: A targeted critical evidence scan through 28 July 2026, updated on 17 September 2026, charted 17 clinical studies. Domains were derived through inductive coding, comparison with clinical-AI and pharmacy guidance, and structured study-level appraisal. RESULTS: The Medication-Use AI Safety and Implementation Framework contains seven linked domains: clinically meaningful target and label; transportable validation; calibration and threshold justification; clinically actionable information and explanation; workflow integration and human oversight; prospective clinical evaluation; and lifecycle governance and resilience. Human factors and equity are cross-cutting. Medication-specific criteria emphasize severity-weighted false negatives, alert burden, pharmacy capacity, formulary and workflow dependence, safety nets for low-risk suppression, and contingency planning. CONCLUSIONS: The framework is a proposed implementation structure, not a validated minimum standard or readiness score. It is intended to help medication-use professionals judge whether AI systems have sufficient clinical, operational, and governance evidence to progress responsibly toward prospective use. Stakeholder consensus, reliability testing, and prospective validation are required.

متن کامل اصلی