Machine Learning-Based Mortality Prediction for Acute Gastrointestinal Bleeding Patients Admitted to Intensive Care
Zhou Liu1, Liang Zhang2, Gui-Jun Jiang1
1Department of Critical Care Medicine, Renmin Hospital of Wuhan University, Wuhan, 430060, China.
Current Medical Science
|February 27, 2025
Summary
Machine learning models accurately predict mortality in acute gastrointestinal bleeding (AGIB) patients. The Gradient Boosting Decision Tree (GBDT) model, especially when combined with the APACHE-II score, offers superior prognostic performance for ICU patients.
Area of Science:
- Critical Care Medicine
- Medical Informatics
- Machine Learning in Healthcare
Background:
- Acute gastrointestinal bleeding (AGIB) is a significant cause of mortality in intensive care units (ICUs).
- Accurate prognostic models are crucial for timely intervention and improved patient outcomes.
- Existing scoring systems like APACHE-II may have limitations in predicting mortality for AGIB patients.
Purpose of the Study:
- To develop and evaluate machine learning (ML) models for predicting in-hospital mortality in AGIB patients admitted to the ICU.
- To compare the performance of ML models against the established APACHE-II scoring system.
- To investigate the potential of an ensemble model combining ML with APACHE-II for enhanced prediction.
Main Methods:
- A cohort of 961 AGIB patients admitted to the ICU was retrospectively analyzed.
- Patients were divided into training (n=768) and validation (n=193) cohorts.
- XGBoost, Random Forest (RF), and Gradient Boosting Decision Tree (GBDT) models were developed using clinical data from the first 24 hours of ICU admission.
- Model performance was assessed using the area under the receiver operating characteristic curve (AUC).
Main Results:
- The overall mortality rate among AGIB patients was 9.78%.
- The GBDT model achieved the highest AUC (0.95) among the individual ML models, outperforming XGBoost (0.89) and RF (0.90).
- The APACHE-II score showed a lower AUC (0.74).
- An ensemble model integrating the GBDT algorithm with the APACHE-II score yielded a superior AUC of 0.98, with high sensitivity and specificity.
Conclusions:
- The GBDT model demonstrates significant potential as a reliable tool for predicting mortality in AGIB patients.
- Integrating the APACHE-II score with the GBDT algorithm further improves predictive accuracy, offering valuable insights for clinical prognostication.
- These ML-driven approaches can enhance the management of critically ill AGIB patients.
