Abstract
This study aims to develop and validate a prediction model for nutritional risk in Post-Stroke Dysphagia (PSD). This study retrospectively analyzed data on stroke patients with dysphagia from January 2022 to December 2023. A stepwise logistic regression model was used to construct the prediction model, and internal validation was performed using the bootstrap resampling (1000 iterations), the nomogram was developed for clinical applications. The final prediction model incorporated the following factors: age, marital status, mechanical ventilation, dysphagia treatment, fasting duration, atrial fibrillation, oral care frequency, serum potassium levels, and National Institute of Health stroke scale (NIHSS) score. The model demonstrated strong discriminatory power, with area under the ROC curve (AUC) values of 0.916 in the development set and 0.878 in the validation set. Calibration curves and the Hosmer-Lemeshow (H-L) test further confirmed the strong correlation between predicted and observed nutritional risks.The prediction model developed in this study exhibits high accuracy, consistency, and practical applicability, making it a valuable tool for predicting nutritional risk in PSD patients. The code library: https://osf.io/p3hjm.