Artificial Intelligence in Pediatric Medication Safety: Opportunities for Detecting Dosing Errors and Adverse Drug Reactions

Abstract

Medication safety in children remains difficult because prescribing and administration depend on age, body weight, developmental physiology, indication, formulation, route, organ function, and frequently changing clinical status. Errors can occur during prescribing, transcription, dispensing, preparation, administration, reconciliation, and monitoring. Computerized physician order entry and clinical decision support can reduce some pediatric medication errors, but conventional systems may also generate inappropriate alerts, omit clinically important context, or create workflow-related problems. Artificial intelligence (AI), including machine learning and natural language processing, may extend these systems by learning from structured and unstructured clinical data. Potential applications include identification of unusual doses, prioritization of high-risk prescriptions, medication reconciliation, extraction of adverse drug events (ADEs) from clinical notes, and pharmacovigilance signal detection. Pediatric evidence specifically evaluating AI remains limited, and much of the available AI literature is based on adult or mixed-age datasets. Existing pediatric studies nevertheless demonstrate the value of electronic dosing calculators, computerized prescribing, prospective pharmacovigilance, and electronic health-record data for medication safety. The most appropriate near-term role for AI is therefore as an additional safety layer that prioritizes cases for pharmacist or clinician review rather than as an autonomous prescribing system. Future studies should emphasize pediatric and neonatal datasets, multicenter external validation, clinically meaningful outcomes, explainability, subgroup performance, human factors, cybersecurity, and prospective evaluation of whether AI reduces preventable medication harm.