Admission NIHSS Score and Diabetes as Independent Predictors of In-Hospital Early Neurological Improvement Following Mechanical Thrombectomy: A Retrospective Cohort Study

Chenyang ZhaoXihua LiYi HanXuefei RenYaxuan Sun^*

• Fifth Hospital of Shanxi Medical University, Taiyuan, China

Background: Accurate early prognostication in acute ischemic stroke (AIS) is essential for optimizing post-thrombectomy management strategies. However, the predictive utility of baseline clinical characteristics remains underexplored in real-world emergency settings. Objective: To identify independent clinical predictors of in-hospital neurological improvement following mechanical thrombectomy in AIS patients, with particular focus on admission NIHSS score and comorbid diabetes mellitus. Methods: In this retrospective single-center cohort study, 250 AIS patients who underwent emergency mechanical thrombectomy between January 2020 and December 2022 were analyzed. Patients were dichotomized according to an in-hospital early neurological improvement endpoint defined a priori as ENI-4 (decrease ≥4 points in NIHSS from admission to discharge). All analyses were repeated in sensitivity analyses using two alternative definitions: a clinician-adjudicated composite of in-hospital neurological improvement and discharge NIHSS ≤1/0. Logistic regression analyses were employed to determine independent predictors. Model performance was evaluated using ROC curve analysis, calibration plots, and nomogram construction. Results: Among the 250 patients, 196 (78.4%) showed neurological improvement during hospitalization. Multivariate logistic regression revealed that a lower admission NIHSS score (OR = 0.867, 95% CI: 0.810–0.927; p < 0.001) and absence of diabetes mellitus (OR = 0.357, 95% CI: 0.129–0.988; p = 0.047) were independently associated with favorable short-term outcomes. The final model demonstrated moderate discriminative ability (AUC = 0.711) and good calibration. Spline analysis demonstrated a non-linear NIHSS–outcome relationship, and decision-curve analysis showed positive net benefit across 10–30% thresholds.A nomogram based on the model was developed for bedside application.Using ENI-4 as the primary outcome, lower admission NIHSS and absence of diabetes remained independently associated with in-hospital neurological improvement in the multivariable model (NIHSS OR 0.867; diabetes OR 0.357). Conclusion: Lower NIHSS scores at presentation and non-diabetic status are independent predictors of early neurological improvement following thrombectomy. The internally validated model provides a clinically accessible tool for early risk stratification in AIS