Development and External Validation of a Machine Learning Model to Predict Post-Implantation Syndrome After Thoracic Endovascular Aortic Repair in Patients With Type B Aortic Dissection

Objective Post-implantation syndrome (PIS) is a common acute inflammatory complication after thoracic endovascular aortic repair (TEVAR) for type B aortic dissection. However, effective tools for the early identification of high-risk patients are still lacking. This study therefore developed and validated a machine learning–based risk prediction model for PIS after TEVAR. Methods In this two-center retrospective cohort study, 676 patients who underwent TEVAR between 2021 and 2025 formed the development cohort, and an independent external cohort of 198 patients (2023–2025) was used for external validation. Predictors were selected through a four-method voting procedure; eight algorithms were compared under stratified 10-fold cross-validation and evaluated with a composite score on the internal validation set. Model performance was assessed by discrimination, calibration, and decision curve analysis; the model was interpreted with SHAP and deployed as a nomogram and a web-based calculator to facilitate clinical application. Results Nine predictors were selected. Among the eight machine learning models built with these predictors, ridge logistic regression ranked highest on the composite score and was selected as the optimal model, with a pooled out-of-fold area under the receiver operating characteristic curve (AUC) of 0.811. The AUC was 0.784 in internal validation and 0.801 in external validation; the Brier score was 0.097 in both cohorts, and the calibration slopes were 0.98 and 1.13, respectively. Conclusions A ridge logistic regression model, integrating disease severity, inflammatory status, and perioperative factors, was developed and externally validated for predicting PIS after TEVAR. The model showed good discrimination and calibration, and was deployed as a nomogram and web-based calculator to facilitate three-tier risk stratification and perioperative decision-making. Clinical trial registration Not applicable, as this study is a retrospective cohort analysis and does not require registration.