Liao Xiaoying, Li Chunhua, Liu Qunyan, Xia Wang, Liu Zhenglin, Zhu Jiamao, Hu Wei, Hong Qionghua
Shangrao People's Hospital, Shangrao, China.
School of Nursing, Jinzhou Medical University, Jinzhou, China.
Front Nutr. 2025 Jun 25;12:1584717. doi: 10.3389/fnut.2025.1584717. eCollection 2025.
Enteral Nutrition-Associated Diarrhea (ENAD) is a common complication in critically ill patients, significantly impacting clinical outcomes. Accurately predicting the risk of ENAD is crucial for early intervention and improving patient care.
This study aims to develop and validate a machine learning (ML)-based risk prediction model for Enteral Nutrition-Associated Diarrhea (ENAD) in ICU patients, and explore its application in nursing practice.
This study was conducted from January 2023 to October 2024 in the Comprehensive Intensive Care Unit (ICU) of a tertiary hospital in China, retrospectively analyzing data from ICU patients receiving enteral nutrition. LASSO regression was used for feature selection, and 9 machine learning (ML) algorithms were evaluated. Model performance was assessed using metrics such as the area under the receiver operating characteristic curve (AUC). The SHapley Additive exPlanation (SHAP) method was employed to interpret feature importance and determine the final model.
Among the 9 ML models, the random forest (RF) model demonstrated the highest discriminative ability, achieving an AUC (95% CI) of 0.777 (0.702-0.830). After dimensionality reduction based on feature importance analysis, a simplified and interpretable RF model with 12 key predictors was established, yielding an AUC (95% CI) of 0.754 (0.685-0.823).
The RF-based predictive model developed in this study provides a reliable and interpretable tool for identifying the risk of ENAD in ICU patients, contributing to targeted nursing interventions and improved patient outcomes. The research highlights the potential of machine learning in enhancing clinical decision-making and personalized care.
肠内营养相关性腹泻（ENAD）是危重症患者常见的并发症，对临床结局有显著影响。准确预测ENAD风险对于早期干预和改善患者护理至关重要。
本研究旨在开发并验证一种基于机器学习（ML）的ICU患者肠内营养相关性腹泻（ENAD）风险预测模型，并探索其在护理实践中的应用。
本研究于2023年1月至2024年10月在中国一家三级医院的综合重症监护病房（ICU）进行，回顾性分析接受肠内营养的ICU患者的数据。采用LASSO回归进行特征选择，并评估了9种机器学习（ML）算法。使用受试者操作特征曲线下面积（AUC）等指标评估模型性能。采用SHapley加法解释（SHAP）方法解释特征重要性并确定最终模型。
在9个ML模型中，随机森林（RF）模型表现出最高的判别能力，AUC（95%CI）为0.777（0.702-0.830）。基于特征重要性分析进行降维后，建立了一个具有12个关键预测因子的简化且可解释的RF模型，AUC（95%CI）为0.754（0.685-0.823）。
本研究开发的基于RF的预测模型为识别ICU患者ENAD风险提供了一种可靠且可解释的工具，有助于进行针对性的护理干预并改善患者结局。该研究突出了机器学习在增强临床决策和个性化护理方面的潜力。
