Related Experiment Video
Updated: Feb 17, 2026
07:42
A Data-Driven Approach to Quantifying Immune States in Sepsis
Published on: February 7, 2025
553
Personalized sepsis mortality prediction: An interpretable machine learning nomogram.
Lulu Weng1, Haidong Li2, Yonglai Lv1
1Department of Critical Care Medicine, First Affiliated Hospital of Huzhou University, the First People's Hospital of Huzhou, Zhejiang, China.
Clinics (Sao Paulo, Brazil)
|February 15, 2026
Summary
This study developed a machine learning nomogram to predict sepsis mortality in ICU patients. The interpretable tool aids clinical decisions for better patient outcomes.
Area of Science:
- Critical Care Medicine
- Machine Learning in Healthcare
- Prognostic Modeling
Background:
- Sepsis is a leading cause of intensive care unit (ICU) mortality.
- Accurate prognostic tools are crucial for effective sepsis management.
- Existing tools may lack interpretability for clinical decision-making.
Purpose of the Study:
- To develop and validate an interpretable machine learning-based nomogram.
- To predict in-hospital mortality in sepsis patients.
- To guide clinical decision-making for critically ill patients.
Main Methods:
- Retrospective cohort study of 407 adult sepsis patients.
- Random division into training (n=284) and validation (n=123) cohorts.
- LASSO-based multivariate logistic regression for nomogram construction.
Main Results:
- Identified seven predictors: age, abdominal infection, vasopressor use, WBC, BNP, APACHE II score, mechanical ventilation.
- Nomogram demonstrated strong discrimination (AUC 0.900 training, 0.796 validation).
- Model performance and clinical utility confirmed by calibration, decision curve, and SHAP analysis.
Conclusions:
- A novel, interpretable machine learning nomogram for sepsis mortality prediction was developed.
- This tool offers practical risk assessment for early intervention.
- Potential to improve outcomes in critically ill sepsis patients through enhanced clinical understanding.
