郝光伟

Abstract Background: Early postoperative mortality remains a major clinical challenge in left ventricular assist device implantation. Accurate preoperative identification of high-risk patients could optimize candidate selection and improve perioperative management. Methods: From February 2021 to March 2025, 60 patients who underwent LVAD implantation at Zhongshan Hospital, Fudan University were included in the present study. Early mortality was defined as death within 2 months postoperatively, a threshold derived from survival analysis showing a distinct early hazard phase. Preoperative demographic, echocardiographic, hemodynamic, and laboratory variables were evaluated. Predictor selection used least absolute shrinkage and selection operator (LASSO) logistic regression with 10-fold cross-validation, followed by multivariable Firth logistic regression. Internal validation employed bootstrap resampling (1,000 iterations). Results: The mean age was 55 ± 10 years, and 50 (83%) patients had dilated cardiomyopathy. Kaplan–Meier analysis indicated an early hazard phase, with all deaths occurring by ~ 1.7 months after implantation. LASSO selected age, MELD-XI, passive cardiac index (PasCI), and mixed venous oxygen saturation (SvO₂) for multivariable modeling. In the final Firth model, lower SvO₂ (odds ratio [OR] 0.92; 95% CI 0.83–1.00; p = 0.048) and older age (OR 1.13; 95% CI 1.00–1.35; p = 0.059) were associated with higher early mortality. Internal validation yielded an optimism-adjusted AUC of 0.84. Conclusions: In this exploratory study, lower preoperative SvO₂ and older age were the principal predictors of early mortality after LVAD implantation. Our internally validated risk assessment model supports preoperative risk prediction and stratification; however, confirmation in larger, multicenter cohorts is warranted before broader adoption. (246/250)