Development and validation of an explainable machine learning model for predicting surgical intervention in pediatric intestinal obstruction
A machine learning tool predicted which children with bowel obstruction truly needed surgery with 98% accuracy, offering clearer guidance in ambiguous cases.
An explainable random forest ML model using 8 easily obtainable features predicted surgical necessity in pediatric intestinal obstruction with AUC 0.981 on external validation from two independent centers. SHAP analysis enabled clinician-interpretable decision support, and the model was translated into a clinical tool, addressing the current lack of standardized surgical criteria in this high-acuity condition.
What the study was
- Study design
- Retrospective multicenter study with internal and external validation
- Population
- Pediatric patients with intestinal obstruction (n=642 training/internal; n=137 external validation from 2 centers)
- Sample size
- 779
- Category
- Diagnostics
- Maturity
- Validated
- Journal
- European Journal of Pediatrics
Why it surfaced
Strong external validation (AUC 0.981), SHAP explainability, and clinical tool translation make this near-term implementable for pediatric emergency surgery.
A plain-language summary of published research — not medical advice. Talk to a clinician about your own care.