ICU sedation, delirium & recovery
An interpretable machine learning approach for predicting clinically important gastrointestinal bleeding in critically ill patients
Authors' conclusion
This study represents the first application of machine learning for predicting CIGIB in ICU patients, providing valuable insights into risk stratification. The model demonstrated high predictive accuracy and interpretability
Study context
- Population
- ICU patients aged ≥18 years admitted 2017–2024; excluded ICU stay <24 h or GIB within 24 h of admission171 (2.3%) with clinically important gastrointestinal bleeding (CIGIB)
- Participants
- 7,357 patients
- Exposure
- XGBoost vs Random Forest vs L1-regularized logistic regression
- Outcome
- Prediction of clinically important GI bleeding (CIGIB) during ICU stay, using data from first 24 h of ICU admission (AUROC, precision, recall, F1)
Methodological detailssingle-centre, machine learning prediction model, SHAP interpretability
- Single-centre
- Machine learning prediction model
- SHAP interpretability
Main finding
As reported in the abstract
Predictive performance for CIGIB
| Group | AUROC |
|---|---|
| XGBoost | 0.84 |
XGBoost
AUROC0.84
Observational data: associations, not effects.
Further findings
CIGIB incidence
- Patients with CIGIB
- 171 (2.3%)
Key predictors (SHAP)
- Key predictors: APACHE III score, hematocrit, APTT, creatinine, respiratory rate
- Invasive mechanical ventilation and stress ulcer prophylaxis within first 24 h not among top 20 predictors
Original publication
An interpretable machine learning approach for predicting clinically important gastrointestinal bleeding in critically ill patients. Anaesth Crit Care Pain Med. Published 9 July 2025. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →