This study shows how a locally fine-tuned language model layered on an auditable electronic trigger can narrow adverse drug reaction review volume, while exposing residual miss risk that pharmacist oversight must absorb.
📅 Journal publication: Dec 2026(online Oct 20, 2026)·Last updated: Nov 2, 2026
adverse drug reactionpharmacovigilanceelectronic triggerlarge language modelQwen3LoRA fine-tuningmedication safetyhospital-prepared herbal formulationQingyanhou Mixtureretrospective cohortdiagnostic accuracyEHR surveillanceMedication Safety
Background Information
LLM-assisted ADR surveillance boosts precision while cutting review workload.
Improving adverse drug reaction (ADR) detection can enhance medication safety by reducing false positives and unnecessary reviews. The trigger-first method achieved 100% sensitivity and 98.25% specificity; LoRA fine-tuning further improved specificity to 94.96%. Although promising, these tools require pharmacist oversight and cannot replace human review entirely. Limited by small sample size. Trigger-first and LoRA fine-tuning can streamline ADR surveillance but require pharmacist oversight.
📊 Deep Analysis Available
Subscribers get the full breakdown: study design, outcomes, pharmacist implications, clinical pearls, GRADE evidence grade, and a downloadable PDF report. Cancel anytime.
Free account, no credit card. Includes 1 full deep analysis every month.
Authors:Liao C, Lai X, Chen W, He J, Yu X
Citation:Liao C, Lai X, Chen W, et al. Trigger-first large language model-assisted adverse drug reaction surveillance: A pharmacist-supervised retrospective cohort study. Exploratory research in clinical and social pharmacy. 2026;24:100866. doi:10.1016/j.rcsop.2026.100866