Machine learning for medication error detection: a scoping review.

Area of Science:

• Health Informatics
• Artificial Intelligence in Medicine
• Patient Safety Research

Background:

• Medication errors pose a significant public health risk, with traditional interventions showing limited success.
• Machine learning (ML) offers advanced computational approaches to enhance medication safety.
• Existing research highlights the growing application of ML in identifying and predicting medication errors.

Purpose of the Study:

• To systematically review and categorize ML-based methods for medication error detection and prediction.
• To synthesize current advancements and identify trends in ML applications for medication safety.
• To highlight gaps and future directions in ML-driven medication error analysis.

Main Methods:

• A comprehensive literature search was conducted across PubMed, Embase, and Web of Science (2015-April 2025).
• Studies were selected based on predefined eligibility criteria following PRISMA-ScR guidelines.
• Data extraction utilized a structured framework by two independent reviewers.

Main Results:

• Twenty-two studies met the inclusion criteria, revealing two primary ML pipelines.
• Prescription error detection predominantly used structured data with tree-based models.
• Medication administration errors were addressed using unstructured multimodal data and neural networks.

Conclusions:

• ML demonstrates significant potential for medication error detection, especially in prescription workflows.
• Fragmented evidence, limited generalizability, and scarce real-world validation hinder current ML applications.
• Future advancements require high-quality datasets, transparent validation, and exploration of diverse data modalities like free text.