Integrating Clinical Modeling and Machine Learning for Risk Assessment of Paracetamol and Other Nonsteroidal Anti-Inflammatory Drug Hypersensitivity in Children - PubMed
5 hours ago
- #Pediatrics
- #Machine Learning
- #Drug Hypersensitivity
- Study focuses on risk assessment of paracetamol and NSAID hypersensitivity in children.
- Aim to develop a clinically interpretable risk stratification tool (nomogram and simplified score).
- Retrospective cohort study (2014-2025) involving 507 index reactions from 487 children.
- 17.7% of cases confirmed hypersensitivity.
- Key predictors: age ≥82.5 months, coexisting asthma/allergic rhinitis, latency ≤60 minutes, angioedema, respiratory symptoms, hypotension/syncope.
- Nomogram and simplified score showed strong discrimination (ROC-AUC=0.877).
- Ensemble ML models (Gradient Boosting, Random Forest, AdaBoost) improved sensitivity and performance after class balancing.
- Tool provides practical pre-DPT risk stratification; ML enhances sensitivity to reduce false negatives.
- Multicenter external validation and prospective studies needed before clinical implementation.