Development and internal validation of a dynamic NIHSS-based prognostic model for predicting 90-day poor functional outcome after acute ischemic stroke in an extremely high-altitude region

Objective To evaluate the incremental prognostic value of adding the discharge National Institutes of Health Stroke Scale (NIHSS) score to admission clinical information for predicting 90-day poor functional outcome after acute ischemic stroke. Methods This single-center retrospective cohort study consecutively included patients with acute ischemic stroke admitted to the Department of Neurology, Xizang Autonomous Region People’s Hospital, from January 2019 to February 2026. Poor functional outcome was defined as a 90-day modified Rankin Scale score of 3–6. The admission-based model included age, sex, admission NIHSS score, systolic blood pressure, blood glucose, onset-to-door time, and atrial fibrillation. The dynamically updated model additionally included discharge NIHSS score. L2-penalized binary logistic regression was used, with the penalty parameter selected by 5-fold stratified cross-validation and internal validation performed using 1,000 bootstrap resamples. Performance was evaluated using AUC, Brier score, calibration measures, and decision curve analysis. Results Among 171 patients, 52 (30.4%) had poor functional outcomes. Apparent and optimism-corrected AUCs were 0.867 and 0.847 for the admission-based model and 0.909 and 0.893 for the dynamically updated model, respectively. The apparent AUC increased by 0.042 (paired bootstrap 95% CI, 0.013–0.075). Optimism-corrected Brier scores were 0.144 and 0.119, with calibration slopes of 1.167 and 1.157, respectively. Conclusion Adding discharge NIHSS score improved discrimination and reduced overall prediction error, providing incremental prognostic information and supporting dynamic risk reassessment during hospitalization. External validation in larger, multicenter, prospective cohorts is required.