Construction of machine learning-based risk prediction models for enteral nutrition intolerance in patients with sepsis
10.3760/cma.j.cn115682-20241104-06020
- VernacularTitle:基于机器学习构建脓毒症患者肠内营养不耐受风险预测模型
- Author:
Yali CHEN
1
;
Lei WANG
1
;
Ji LI
1
Author Information
1. 杭州师范大学附属医院重症医学科，杭州 310000
- Publication Type:Journal Article
- Keywords:
Sepsis;
Enteral nutrition intolerance;
Risk prediction model;
Machine learning
- From:
Chinese Journal of Modern Nursing
2025;31(20):2736-2741
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To construct risk prediction models for enteral nutrition intolerance in patients with sepsis based on machine learning, and conduct internal validation to evaluate the predictive performance of the model.Methods:Convenience sampling was used to select 320 sepsis patients admitted to the Intensive Care Unit of the Affiliated Hospital of Hangzhou Normal University from June 2022 to June 2024 for the study. Patients were randomly divided into a training set ( n=240) and a validation set ( n=80) in a 3∶1 ratio. Univariate analysis of variance and Binary Logistic regression analysis were used to explore the factors influencing enteral nutrition intolerance in patients with sepsis. Four machine learning algorithms, namely, categorical boosting (CatBoost), extreme gradient boosting (XGBoost), support vector machine (SVM), and random forest (RF), were used to construct the risk prediction models. The performance of the prediction models was assessed through cross-validation, and the optimal risk prediction model was selected based on the area under the receiver operating characteristic curve, accuracy, sensitivity, specificity, recall, precision, and F1 value, and the model was validated in validation set. Results:Of the 320 patients with sepsis, enteral nutrition intolerance occurred in 131 patients, with an incidence of 40.94%. Age, body mass index, Acute Physiology and Chronic Health EvaluationⅡ scores, mode of enteral nutrition infusion, time of enteral nutrition initiation, albumin, and intra-abdominal pressure were influencing factors for enteral nutrition intolerance in patients with sepsis. Among the four risk prediction models, CatBoost, XGBoost, SVM, and RF, the RF model showed the best performance.Conclusions:In this study, we construct risk prediction models for enteral nutrition intolerance in sepsis patients based on four machine learning algorithms, among which the overall performance of the RF model is superior, which is helpful for healthcare professionals to accurately identify the risk of enteral nutrition intolerance in sepsis patients and formulate interventions at an early stage.
