Prediction models for the mortality risk in traumatic hemorrhage based on machine learning
10.3760/cma.j.cn114656-20250314-00193
- VernacularTitle:基于机器学习构建创伤出血死亡风险预测模型研究
- Author:
Yiquan WANG
1
;
Sijia TIAN
;
Shengmei NIU
;
Zhipei HUANG
;
Fei QIN
;
Jinjun ZHANG
Author Information
1. 北京急救中心，北京 100031
- Keywords:
Trauma;
Hemorrhage;
Machine learning;
Risk of death
- From:
Chinese Journal of Emergency Medicine
2025;34(11):1574-1578
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To evaluate the predictive performance of machine learning methods for predicting the risk of death in traumatic hemorrhage, and address the low prediction accuracy of traditional trauma scores, provide a reference for developing a more robust prediction method for severe trauma patients.Methods:Clinical data of severe trauma patients from the National Trauma Medical Center between April 1, 2023, and March 31, 2024 were collected. ElasticNet, Recursive Feature Elimination, and Mutual Information-based feature selection methods were used to screen variables and compared with traditional hypothesis testing methods. Built the prediction models for mortality risk in traumatic hemorrhage using Logistic Regression, ElasticNet, and Support Vector Machine (SVM) and compared the predictive performance.Results:The study included 5,601 trauma patients, the results of the variable screening and importance ranking were consistent with three feature selection methods. The classification accuracy and AUC values for the three models were as follows: Overall accuracy was 83.2%, survival accuracy was 84.0%, death accuracy was 76.3%, and an AUC was 0.86 in logistic regression; Overall accuracy was 78.9%, survival accuracy was 78.5%, death accuracy was 81.7%, and an AUC was 0.88 in ElasticNet; Overall accuracy was 84.7%, survival accuracy was 86.1%, death accuracy was 72.4%, and an AUC was 0.86 in SVM. The prediction performance of three models is quite little.Conclusion:Machine learning methods can effectively improve the prediction of death risk for traumatic hemorrhage,and has wide applications.
