Related Experiment Video
Updated: Aug 22, 2025
07:42
A Data-Driven Approach to Quantifying Immune States in Sepsis
Published on: February 7, 2025
300
Machine-learning models for prediction of sepsis patients mortality
1Xiangya Hospital, Department of Critical Care Medicine & National Clinical Research Center for Geriatric Disorders, Central South University, Hainan General Hospital, Department of Emergency, Hainan Medical University, Haikou, Hainan, China.
Medicina Intensiva
|November 7, 2022
Summary
Machine learning models accurately predict sepsis patient mortality using electronic health records. The light gradient boosting machine (LGBM) model achieved high accuracy, offering potential for improved clinical decision tools.
Area of Science:
- Critical Care Medicine
- Health Informatics
- Machine Learning
Background:
- Sepsis is a life-threatening organ dysfunction caused by infection.
- Predicting sepsis patient mortality is crucial for timely intervention and improved outcomes.
- Large-scale data analysis can enhance predictive model development.
Purpose of the Study:
- To develop and evaluate machine learning models for predicting sepsis patient mortality.
- To identify the most effective machine learning algorithm for sepsis mortality prediction.
Main Methods:
- Utilized two large databases: MIMIC-IV (training) and eICU (testing).
- Included 21,680 sepsis-3 patients from US ICUs (2012-2019).
- Built seven models (SVM, Decision Tree, Random Forest, Gradients Boosting, MLP, XGBoost, LightGBM) using patient data.
Main Results:
- LightGBM, Gradient Boosting Machine (GBM), and XGBoost showed the highest AUC values in the test set.
- The LightGBM model demonstrated superior performance with an AUC of 0.99 (train) and 0.96 (test) after parameter tuning.
- The developed models accurately predicted patient mortality.
Conclusions:
- Machine learning models, particularly LightGBM, can accurately predict sepsis patient mortality from electronic health records.
- These predictive models can be integrated into clinical decision support systems.
- Enhanced prediction can lead to better patient prognosis and prevention of adverse events.
