Risk Factors and Prediction Models for 1-Year Recurrence of Ischemic Stroke in County-Level Regions of China

Background Ischemic stroke (IS) recurrence poses a significant health burden in China, particularly in county regions where 1-year recurrence rates exceed urban rates. Despite known risk factors like hypertension and diabetes, rural populations face challenges from socioeconomic disparities and limited healthcare resources. Current models, derived from urban tertiary hospitals, lack applicability to resource-constrained institutions and require tailored analysis of county-level populations. This study aims to identify risk factors and develop a predictive model for 1-year IS recurrence in county regions using statistical methods and machine learning with local clinical data. ​Methods​ A retrospective cohort study analyzed 984 patients with IS admitted to Dujiangyan People’s Hospital from 2017 to 2019 and followed up with all patients. Data were systematically collected from electronic medical records. Multivariable logistic regression was used to identify independent risk factors, and five machine learning algorithms (Decision Tree, Random Forest, Gradient Boosting, K-Nearest Neighbors and Neural Network) were optimized using 5-fold cross-validation. The model performance was evaluated using the AUC, accuracy, and specificity. ​Results​ Among 984 patients with IS, 726 (73.78%) experienced 1-year recurrence. Multivariate logistics analysis found that independent predictive factors include NIHSS (5-15: OR: 2.739, 95% CI 1.926-3.894, 16-20: OR: 5.027, 95% CI 2.266-11.152 );LDL (OR: 2.25, 95% CI 1.471-2.789); Time from Onset to Hospitalization (OR: 1.741, 95% CI 1.262-2.413); MRI Infarct Site (OR: 1.576, 95% CI 1.046-2.377). Among the five predictive models, the Random Forest model algorithm achieved optimal performance (AUC=0.969, specificity=90%, accuracy=93%). The key predictors were NIHSS score, LDL-C level, MRI lesion patterns, Time from Onset to Hospitalization, and CT lesion location. ​ Conclusion​s This study found NIHSS score, LDL level, and MRI infarct site and time from onset to hospitalization were identified as critical risk factors for IS recurrence in county-level regions within 1 year. The high-performance Random Forest model of stroke recurrence prediction based on clinical data available in county-level healthcare institutions has good clinical application value.