Machine Learning Predicts Mortality and Respiratory Failure in Patients Admitted With Rib Fractures.

Area of Science:

• Trauma surgery
• Medical informatics
• Machine learning in healthcare

Background:

• Rib fractures are common injuries with high morbidity and mortality.
• Existing clinical risk models for rib fracture patients have poor external validation.
• Predicting adverse outcomes like mortality and respiratory failure remains challenging.

Purpose of the Study:

• To develop and validate machine learning models for predicting in-hospital mortality and respiratory failure in trauma patients with rib fractures.
• To utilize admission variables for early risk stratification.
• To assess the performance of machine learning algorithms in this patient population.

Main Methods:

• Utilized the National Trauma Data Bank (2021-2022) for adult trauma patients with rib fractures.
• Developed predictive models using Light Gradient Boosting Machine and Extreme Gradient Boosting algorithms.
• Evaluated model performance using cross-validation and metrics including AUC-ROC and AUC-PR; employed SHAP for interpretability.

Main Results:

• The study included 260,771 patients; 3.4% experienced mortality and 10.1% required mechanical ventilation.
• Light Gradient Boosting Machine models achieved high predictive performance (AUC-ROC 0.90 for mortality, 0.87 for respiratory failure).
• Key predictors identified include age, hemorrhage severity, pulmonary dysfunction, and neurologic status.

Conclusions:

• Machine learning models demonstrate high accuracy in predicting mortality and respiratory failure in trauma patients with rib fractures.
• These data-driven models offer a sophisticated approach to risk stratification.
• Further research should enhance model interpretability for clinical integration.