NIHSS and inflammatory biomarkers predict 90-day poor outcome in older patients with acute ischemic stroke.

Early prediction of functional outcomes following acute ischemic stroke (AIS) in older adults presents significant clinical challenges. Inflammation-related biomarkers may enhance clinical severity scores for early risk stratification. This single-center prospective observational study aimed to develop and internally validate an exploratory prediction model that integrates the admission National Institutes of Health Stroke Scale (NIHSS) score with serum inflammatory biomarkers to predict 90-day poor outcomes in older patients with AIS. Consecutive patients aged 60 years and older, with magnetic resonance imaging-confirmed AIS and blood sampling within 24 hours of stroke onset, were enrolled from May to December 2025. Patients with a history of previous strokes, active infections or autoimmune diseases, severe hepatic or renal dysfunction, incomplete data, or those who received intravenous thrombolysis or endovascular treatment were excluded from the study. Poor outcomes were defined as a modified Rankin Scale (mRS) score of 3 or higher at 90 days. Serum levels of prostaglandin-endoperoxide synthase 2 (PTGS2), matrix metalloproteinase-9 (MMP-9), and interleukin-17A (IL-17A) were measured using enzyme-linked immunosorbent assay (ELISA). Among the 171 patients included in the final analysis, 50 experienced poor outcomes, while 121 had favorable outcomes. In a multivariable logistic regression analysis adjusted for prespecified clinical covariates, admission NIHSS score, MMP-9, PTGS2, and IL-17A were found to be independently associated with poor outcomes and were retained in a simplified nomogram. The combined model demonstrated an apparent area under the receiver operating characteristic curve (AUC) of 0.907 (95% confidence interval [CI], 0.864-0.950) and a bootstrap-corrected concordance index (C-index) of 0.897. Calibration was acceptable, with a bootstrap-corrected calibration intercept of 0.0004 and a slope of 0.916. Additionally, decision curve analysis (DCA) indicated a potential net benefit across clinically plausible threshold probabilities. However, the model development and evaluation were conducted within a modest, selected, single-center cohort, and the number of poor outcome events was limited. Furthermore, the routine availability of inflammatory biomarker testing is lacking in many clinical settings, and the exclusion of reperfusion-treated patients restricts the model's generalizability. Consequently, this NIHSS- and biomarker-based model should be regarded as exploratory and hypothesis-generating. External validation in larger, multicenter, and more heterogeneous cohorts is essential before clinical application.