Breakthrough in Sepsis-Associated Liver Injury Prediction through Machine Learning

Strategic Insights -

Researchers at the Third Xiangya Hospital of Central South University have made a significant breakthrough in predicting sepsis-associated liver injury (SALI) using machine learning techniques. The study, published in the Journal of Medical Internet Research, aimed to develop an explainable machine learning model that could predict the occurrence of liver injury in patients with sepsis and provide decision support for early intervention and personalized treatment strategies.

The researchers trained nine machine learning models on a dataset of 8834 patients with sepsis from the Medical Information Mart for Intensive Care IV (MIMIC-IV) database and evaluated their performance using various metrics such as accuracy, balanced accuracy, and area under the receiver operating characteristic curve (ROC-AUC). The stacking ensemble model, which combined the predictions of three base learners (LightGBM, XGBoost, and RF), demonstrated the best performance, achieving ROC-AUCs of 0.995, 0.838, and 0.721 in the training, internal validation, and external validation sets, respectively.

Key Takeaways:

  • The stacking ensemble model developed in this study yields accurate and robust predictions of SALI in patients with sepsis, demonstrating potential clinical utility for early intervention and personalized treatment strategies.
  • The key predictors of SALI identified across models were total bilirubin, lactate, prothrombin time, and mechanical ventilation status.
  • The models were trained and validated using data from two large databases: MIMIC-IV and eICU-Collaborative Research Database (eICU-CRD).
  • The study adhered to the TRIPOD+AI guidelines for transparent reporting of prediction models.
  • The research was conducted by a team of researchers from the Third Xiangya Hospital of Central South University, led by Jia Zhai.

Statistics:

  • The study retrospectively analyzed data from 8834 patients with sepsis in the MIMIC-IV database.
  • The stacking ensemble model achieved ROC-AUCs of 0.995, 0.838, and 0.721 in the training, internal validation, and external validation sets, respectively.
  • The base learners (LightGBM, XGBoost, and RF) achieved ROC-AUCs of 0.9977, 0.9311, and 0.9847, respectively, in the training set.
  • The most important predictors of SALI identified by the models were total bilirubin (mean SHAP value: 0.457), lactate (mean SHAP value: 0.398), prothrombin time (mean SHAP value: 0.346), and mechanical ventilation status (mean SHAP value: 0.323).

Sources:

  • Supervised Machine Learning Models for Predicting Sepsis-Associated Liver Injury in Patients With Sepsis: Development and Validation Study Based on a Multicenter Cohort Study. Journal of Medical Internet Research, 2025;27.
  • Jia Zhai, Jingchao Lei, Yao Zhang, Jing Qi, Chuanzheng Sun. Third Xiangya Hospital of Central South University. People's Republic of China.
  • Jmir Publications, Inc. 130 Queens Quay East, Unit 1100, Toronto, On M5A 0P6, Canada.