Development and Validation of a Predictive Model for Prolonged Mechanical Ventilation After Heart Valve Surgery

Background: Prolonged mechanical ventilation (PMV) is a common and serious complication after heart valve surgery, associated with increased morbidity, mortality, and healthcare resource utilization. Although several predictive models exist, many are limited by population homogeneity or lack of intraoperative variables. This study aimed to develop and validate a practical predictive model for PMV risk stratification to facilitate early intervention and optimize resource allocation. Methods: This was a retrospective study of adult patients who underwent elective heart valve surgery between January 2013 and January 2023. Patients from Center A were randomly assigned to a training cohort (n = 349) or an internal validation cohort (n = 149, with a 7:3 ratio). PMV was defined as mechanical ventilation lasting more than 48 hours postoperatively. Preoperative, intraoperative, and early postoperative variables were analyzed. Univariate and multivariate logistic regression analyses were used to identify independent predictors in the training cohort. A predictive nomogram was subsequently developed. Model performance was evaluated using discrimination (area under the receiver operating characteristic (AUROC) curve), calibration (calibration plots, Hosmer–Lemeshow test), and clinical utility (decision curve analysis (DCA) and clinical impact curve (CIC)). Results: Data were analyzed from 498 patients (training: n = 349; internal validation: n = 149). The incidence of PMV was 32.7% in the training cohort. Multivariate analysis identified six independent predictors: age (per 1-year increase), body mass index (per 1 kg/m2 increase), chronic obstructive pulmonary disease severity (per 1-grade increase), forced expiratory volume in 1 s (per 1% increase), left ventricular ejection fraction (per 1% increase), and cardiopulmonary bypass time (per 10 minute increase). The nomogram demonstrated strong discrimination, with area under the curve (AUC) values of 0.847 (95% confidence interval (CI), 0.798–0.882) in training and 0.891 (95% CI, 0.858–0.927) in internal validation. Calibration was good across cohorts (Hosmer–Lemeshow p > 0.05). The DCA and CIC indicated that the model provided meaningful clinical benefit compared with treating all or no patients when the predicted probability threshold ranged from 40% to 100%. Conclusion: PMV was associated with higher in-hospital mortality, increased healthcare resource utilization, and reduced long-term survival. The proposed predictive model may aid in optimizing perioperative management, thereby improving outcomes and reducing costs.