Background: Effectively managing trauma patients requires accurate prediction of efficient triaging and timely activation of Massive Blood Transfusion Protocols (MTP). Machine learning (ML) algorithms have emerged as highly promising tools in the domains of predicting clinical outcomes and optimizing triage decisions and intervention strategies, consistently outperforming traditional methodologies. We aimed to assess and compare ML models for the prediction of trauma triaging processes, activation of MTP, and mortality. Methods: In this 10-year retrospective study, including 17,390 patients, we analyzed the predictive capabilities of seven ML models for overall trauma patients as well as for hemodynamics data subset. Employing Python 3.9 for data preprocessing, feature scaling, and model development, we evaluated K-Nearest Neighbors (KNN), Logistic Regression (LR), Decision Tree (DT), Support Vector Machines (SVM) with RBF kernels, Random Forest (RF), Extreme Gradient Boosting (XGBoost), and Artificial Neural Network (ANN). We employed various imputation techniques and addressed data imbalance through down-sampling, up-sampling, and SMOTE. Hyperparameter tuning, coupled with 5-fold cross-validation, was performed. The evaluation included essential metrics like sensitivity, specificity, F1 score, accuracy, and AUC, ensuring robust predictive capability. Result: 3,385 patients were triaged as critical cases, 3.8% (664 patients) required MTP activation, and 7.7% (1,335 patients) had mortality. Imputation improved performance, and most models performed better with balancing techniques. The overall models demonstrated notable performance metrics for predicting mortality, triage, and MTP activation, with F1 scores of 0.79, 0.75, and 0.42, sensitivities of 0.9, 0.73, and 0.82, and AUC values of 0.99, 0.89, and 0.95, respectively. Conclusion: ML models demonstrated robust predictive capabilities for triage, MTP activation, and mortality, which may improve trauma outcomes.
none to declare