Prediction of Early Hospital Admission (≤24 Hours) After Stroke Using Machine Learning and Deep Learning: Multicenter Study From China.

backgroundTimely hospital admission is a prerequisite for effective acute stroke management, yet a substantial proportion of patients fail to reach medical facilities within the optimal therapeutic window. Existing prediction models often lack temporal robustness and clinical interpretability, limiting their utility in real-world, evolving health care systems.

objectiveThis study aimed to develop and temporally validate machine learning and deep learning models using multicenter clinical data to predict early hospital admission (≤24 h) after acute stroke.

methodsIn this multicenter retrospective study, we analyzed routinely collected electronic medical record data from 1327 patients across 6 hospitals in China. We developed and compared 6 predictive models: logistic regression, support vector machine, random forest, multilayer perceptron (MLP), convolutional neural network, and long short-term memory, for early admission (≤24 h from symptom onset). Model training was performed on a train set (2019-2022), followed by independent temporal validation on a testing set (2023-2025). Model prediction performance was evaluated using discrimination metrics, sensitivity, and robustness under temporal distribution shift. Model interpretability was assessed using Shapley additive explanations.

resultsA total of 1327 patients were included, of whom 821 were assigned to the train set and 506 to the independent temporal testing set. Among the 6 candidate models, the MLP showed the best overall performance in the independent temporal testing set, achieving an area under the receiver operating characteristic curve of 0.9020 (95% CI 0.8718-0.9283), sensitivity of 91.5%, specificity of 75.6%, and F

conclusionsIn this multicenter Chinese cohort, the MLP showed favorable temporal performance for predicting early hospital admission after stroke. The model may support future risk stratification and targeted public health interventions, although further external validation and calibration refinement are needed before deployment-oriented use.