Development of a Nomogram Prediction Model for Early Post-Stroke Seizures in Acute Stroke Patients Using LASSO Regression

Background and Purpose: Early post-stroke seizures (EPSS) are a common and serious complication in stroke patients. Early identification of individuals at high risk for EPSS is critical for optimizing management strategies and improving clinical outcomes. This study aimed to develop an individualized, visual predictive model based on a large multicenter dataset to accurately predict the risk of EPSS. Methods: We conducted a retrospective multicenter cohort study. Least absolute shrinkage and selection operator (LASSO) regression and multivariate logistic regression were used to identify independent predictors significantly associated with EPSS. A nomogram was then constructed incorporating these predictors. The model's performance was evaluated using receiver operating characteristic (ROC) curves, calibration plots, and decision curve analysis (DCA). Internal and external validation cohorts were used to assess the model's discrimination, calibration, and clinical utility. Results: Four independent predictors were identified: temporal lobe involvement, frontal lobe involvement, National Institutes of Health Stroke Scale (NIHSS) score at admission, and white blood cell (WBC) count. The nomogram demonstrated good predictive performance in both training and validation cohorts, with high area under the ROC curve (AUC), excellent calibration, and favorable clinical utility shown by DCA. Temporal and frontal lobe involvement significantly increased EPSS risk, while NIHSS score and WBC count were also important contributors. Conclusions: We developed a clinically practical and visual nomogram model incorporating imaging and laboratory parameters to aid in early identification of high-risk EPSS patients, facilitating personalized prevention and treatment strategies to improve prognosis and quality of life.