Construction and Validation of a Risk Prediction Model for Aspiration in Stroke Patients with Dysphagia

Abstract Objective To analyze factors influencing aspiration in stroke patients with dysphagia, construct and validate a risk prediction model, and provide a reference for early identification and prevention of aspiration. Methods: A retrospective study design was used. Clinical data were collected from 951 stroke patients with dysphagia admitted to a tertiary Grade A hospital in Jinan from January 2024 to December 2024. Patients were randomly divided into a training set (n = 667) and a validation set (n = 284) at a ratio of 7:3. Multivariable logistic regression analysis was used to identify factors associated with aspiration, construct a risk prediction model, and evaluate its predictive performance. Results: Among the 667 patients in the training set, 206 experienced aspiration. Logistic regression analysis showed that stroke location, artificial airway, mechanical ventilation, history of aspiration, feeding method, food source, language disorder, sedative drugs, and albumin were factors associated with aspiration in stroke patients with dysphagia (P < 0.05). Based on these findings, a prediction model suitable for nurses to identify aspiration risk early in stroke patients with dysphagia was constructed. In the validation set, the optimal cut-off value was 0.368, with an accuracy of 89.9%, sensitivity of 90.1%, specificity of 89.7%, positive predictive value of 82.5%, and negative predictive value of 94.8%. The Hosmer-Lemeshow test showed χ²=5.430 (P = 0.608). Conclusion: The risk prediction model for aspiration in stroke patients with dysphagia showed good discrimination and calibration and may provide a reference for clinical nurses in predicting related risks.