Interpretable machine learning for postoperative sepsis prediction in ICU patients following intracranial hematoma evacuation.
5 affiliations
- 1The Second School of Clinical Medicine, Nanjing Medical University, Nanjing, Jiangsu, China.
- 2Department of Neurosurgery, Affiliated Hospital of Integrated Traditional Chinese and Western Medicine, Nanjing University of Chinese Medicine, Nanjing, Jiangsu, China.
- 3Department of Neurosurgery, Jiangsu Province Academy of Traditional Chinese Medicine, Nanjing, Jiangsu, China.
- 4School of Life Sciences, Xiamen University, Xiamen, Fujian, China.
- 5Department of Critical Care Medicine, Jiangsu Province Academy of Traditional Chinese Medicine, Nanjing, Jiangsu, China.
Abstract
Background
Sepsis is a prevalent and life-threatening complication in ICU patients after intracranial hematoma evacuation. This study aimed to develop and externally validate an interpretable machine learning model for predicting sepsis risk in this high-risk neurocritical care population.Methods
A multicenter retrospective study was designed. The MIMIC-IV was based on for model development and internal testing; MIMIC-III, eICU and an independent clinical cohort from Jiangsu Provincial Hospital of Integrated Traditional Chinese and Western Medicine for external validation. Sepsis during ICU stay was defined according to the operationalized Sepsis-3 criteria. Early ICU variables were selected using Boruta and LASSO, and six machine learning algorithms were developed and compared. Model performance was evaluated by discrimination, calibration, and decision curve analysis, and its interpretability was assessed using SHAP and LIME. A sensitivity analysis was conducted by retraining all six models after excluding baseline SOFA.Results
A total of 1,729 patients were included, of whom 717 developed sepsis during ICU stay. The final model retained 13 predictors, including WBC, SOFA score, SpO₂, glucose, chloride, pneumonia, aspirin use, mannitol use, vasopressor use, mechanical ventilation, RBC, sedative use, and GCS score. XGBoost showed the most balanced overall performance, with AUCs of 0.794 in the internal test set and 0.820, 0.837, and 0.796 in the MIMIC-III, eICU, and independent clinical validation cohorts, respectively. SHAP analysis identified SOFA score, vasopressor use, glucose, pneumonia, and sedative exposure as major contributors to model predictions, while LIME supported individual-level interpretability. A web-based calculator was implemented for individualized risk prediction. Sensitivity analysis showed that XGBoost retained discrimination after SOFA exclusion, with no significant AUC changes across the external cohorts.Conclusion
This study developed and externally validated an interpretable XGBoost-based model for predicting sepsis risk during ICU stay among patients after intracranial hematoma evacuation. The model showed generally consistent AUC-based discrimination across multiple validation cohorts and may support individualized risk stratification. Further prospective validation and clinical impact assessment are required before routine clinical implementation.
Related
- Development and internal validation of a machine learning prediction model for postoperative hemorrhage following modified radical mastectomy for breast cancer
- Predictors of Postoperative Neurological Outcome After Surgical Evacuation of Traumatic Acute Subdural Hematoma: A Retrospective Cohort Study
- Temporal overlap of intraoperative hypotension and hypocapnia and postoperative outcomes after major open intracranial surgery: a time-weighted dual-hit cohort study
- War exposure, evacuation, and psychological and behavioural outcomes following a large-scale traumatic event.
- Intracranial metastasis following intrahepatic remission in primary hepatic small cell neuroendocrine carcinoma: a case report
BibTeX
@article{yu2026interpretable,
title = {{Interpretable machine learning for postoperative sepsis prediction in ICU patients following intracranial hematoma evacuation.}},
author = {Shaoyang Yu and Yun Liu and Lei Guo and Jinchang He and Yuyuan Ge and Wan Yu and Weijian Jia and Kexiang Xiong and Zixiang Wang and Wei Gao and Yong Zhang},
journal = {Frontiers in cellular and infection microbiology},
year = {2026},
month = sep,
doi = {10.3389/fcimb.2026.1895258},
url = {https://europepmc.org/article/MED/42839922},
note = {CC BY},
}
RIS
TY - JOUR
TI - Interpretable machine learning for postoperative sepsis prediction in ICU patients following intracranial hematoma evacuation.
AU - Shaoyang Yu
AU - Yun Liu
AU - Lei Guo
AU - Jinchang He
AU - Yuyuan Ge
AU - Wan Yu
AU - Weijian Jia
AU - Kexiang Xiong
AU - Zixiang Wang
AU - Wei Gao
AU - Yong Zhang
PY - 2026
DA - 2026/09/22
JO - Frontiers in cellular and infection microbiology
DO - 10.3389/fcimb.2026.1895258
UR - https://europepmc.org/article/MED/42839922
AB - Background Sepsis is a prevalent and life-threatening complication in ICU patients after intracranial hematoma evacuation. This study aimed to develop and externally validate an interpretable machine learning model for predicting sepsis risk in this high-risk neurocritical care population.Methods A multicenter retrospective study was designed. The MIMIC-IV was based on for model development and internal testing; MIMIC-III, eICU and an independent clinical cohort from Jiangsu Provincial Hospital of Integrated Traditional Chinese and Western Medicine for external validation. Sepsis during ICU stay was defined according to the operationalized Sepsis-3 criteria. Early ICU variables were selected using Boruta and LASSO, and six machine learning algorithms were developed and compared. Model performance was evaluated by discrimination, calibration, and decision curve analysis, and its interpretability was assessed using SHAP and LIME. A sensitivity analysis was conducted by retraining all six models after excluding baseline SOFA.Results A total of 1,729 patients were included, of whom 717 developed sepsis during ICU stay. The final model retained 13 predictors, including WBC, SOFA score, SpO₂, glucose, chloride, pneumonia, aspirin use, mannitol use, vasopressor use, mechanical ventilation, RBC, sedative use, and GCS score. XGBoost showed the most balanced overall performance, with AUCs of 0.794 in the internal test set and 0.820, 0.837, and 0.796 in the MIMIC-III, eICU, and independent clinical validation cohorts, respectively. SHAP analysis identified SOFA score, vasopressor use, glucose, pneumonia, and sedative exposure as major contributors to model predictions, while LIME supported individual-level interpretability. A web-based calculator was implemented for individualized risk prediction. Sensitivity analysis showed that XGBoost retained discrimination after SOFA exclusion, with no significant AUC changes across the external cohorts.Conclusion This study developed and externally validated an interpretable XGBoost-based model for predicting sepsis risk during ICU stay among patients after intracranial hematoma evacuation. The model showed generally consistent AUC-based discrimination across multiple validation cohorts and may support individualized risk stratification. Further prospective validation and clinical impact assessment are required before routine clinical implementation.
SN - 2235-2988
ER -