Integrating Vascular, Neurological, and Metabolic Factors into a Validated Nomogram to Predict 90‑day Outcomes after Mechanical Thrombectomy for Acute Ischemic Stroke

Background Precise forecasting of functional outcomes following mechanical thrombectomy (MT) for acute ischemic stroke (AIS) is essential for tailored patient care. However, many existing models have methodological limitations, limited generalizability, or poor interpretability, which hinder their clinical adoption. Our study focused on creating and validating a clinically useful nomogram to predict 90-day outcomes following MT in AIS patients. Methods This retrospective study analyzed data from AIS patients treated with MT at Zhaoqing First People’s Hospital (Jan 2019–Dec 2022). Patients were allocated into a training cohort (n = 115) and a validation cohort (n = 48) in a 7:3 ratio using a computer-generated random sequence. Predictive factors were identified using Least Absolute Shrinkage and Selection Operator (LASSO) regression and validated with multivariable logistic regression. A nomogram was created and assessed for its discrimination ability using the area under the receiver operating characteristic curve (AUC), calibration, and clinical utility through decision curve analysis (DCA). The model's performance was evaluated against a support vector machine (SVM) classifier, with Shapley Additive exPlanations (SHAP) employed to analyze feature contributions and their directionality. Results identified four independent predictors of a poor 90-day functional outcome (mRS ≥ 3). Anterior cerebral artery involvement reduced the risk of poor outcome (OR = 0.269, 95% CI: 0.130–0.557, P < 0.001), while higher admission NIHSS score (OR = 1.069, 95% CI: 1.009–1.133, P = 0.023), elevated fasting blood glucose (OR = 1.333, 95% CI: 1.089–1.632, P = 0.005), and lower serum albumin (OR = 0.845, 95% CI: 0.715–0.999, P = 0.049) raised the risk. The nomogram exhibited robust discrimination (training AUC 0.810; validation AUC 0.731), precise calibration, and notable clinical net benefit. Similarly, the SVM model showed comparable predictive accuracy (training AUC: 0.798; validation AUC: 0.720), while SHAP analysis confirmed the biological plausibility of the selected features, consistent with the logistic regression results. Conclusion In light of these results, we present a validated, interpretable nomogram that effectively stratifies 90-day post-MT risk by incorporating vascular, neurological, and metabolic factors. This tool balances predictive performance with clinical transparency, potentially aiding individualized patient management and shared decision-making. External validation is recommended to confirm its broad applicability.