LASSO-based nomogram and machine learning models for predicting 30-day nasogastric tube dependence after acute...

Nasogastric tube (NGT) feeding is a common method for providing enteral nutrition to patients with acute ischemic stroke (AIS) and dysphagia. However, prolonged NGT dependence contributes to adverse clinical consequences. Early identification of patients who may develop persistent NGT dependence remains challenging.

A total of 852 AIS patients requiring NGT were included and allocated to the training (n = 596) and internal validation (n = 256) sets. Predictor selection utilized least absolute shrinkage and selection operator (LASSO) regression, the eXtreme Gradient Boosting (XGBoost) algorithm, and multivariate logistic regression. A nomogram was then constructed and internally and externally validated. Three machine learning algorithms-Gradient Boosting Machine (GBM), Support Vector Machine (SVM), and Regularized Discriminant Analysis (RDA) were also developed for comparison. Model performance was evaluated using the Area Under the Curve (AUC), calibration plots, and Decision Curve Analysis (DCA).

Four key independent predictors were retained: Diabetes Mellitus (DM), age, admission National Institutes of Health Stroke Scale (NIHSS) score, and exclusive NGT feeding. The model demonstrated strong discrimination (AUC = 0.924 in training, 0.920 in internal validation, and 0.935 in external validation), good calibration, and favorable clinical utility in DCA. Among the machine learning models, GBM demonstrated the highest accuracy, with an AUC of 0.918. The model confirmed age and NIHSS score as the most influential predictors, followed by exclusive NGT feeding and DM.

The developed nomogram provides an effective approach for predicting 30-day NGT dependence in AIS patients, enabling timely risk stratification and individualized clinical management.