Development and Temporal Validation of an Interpretable Perioperative Nomogram for Prolonged Mechanical Ventilation in Older Adults Undergoing Cardiac Surgery

Background: Prolonged mechanical ventilation (PMV) after cardiac surgery reflects delayed early recovery and is associated with increased perioperative resource use. Older adults are especially vulnerable, yet practical models that jointly capture preoperative reserve and intraoperative burden remain limited. Methods: We performed a single-center retrospective cohort study of 1632 consecutive patients aged ≥60 years who underwent cardiac surgery at Nanjing First Hospital between 25 January 2022 and 8 March 2024. PMV was defined as postoperative mechanical ventilation lasting more than 24 h. Patients treated in 2022 formed the development cohort and were stratified by PMV status into a training set and an internal validation set at a 7:3 ratio. Patients treated from 1 January 2023 to 8 March 2024 comprised the temporal validation cohort. Candidate predictors available before or during surgery were entered into least absolute shrinkage and selection operator (LASSO) regression with 10-fold cross-validation. Predictors with non-zero LASSO coefficients were then refitted in a multiple logistic regression model and presented as a nomogram. Model performance was evaluated using receiver operating characteristic curves, calibration curves, Brier scores, Hosmer–Lemeshow testing, and decision curve analysis. Results: Among 1632 eligible patients, 346 (21.2%) developed PMV. The training, internal validation, and temporal validation sets included 484, 208, and 940 patients, with PMV incidences of 22.3%, 22.6%, and 20.3%, respectively. LASSO retained six predictors: age, preoperative creatinine, preoperative left ventricular ejection fraction, European System for Cardiac Operative Risk Evaluation II (EuroSCORE II) score, operation time, and non-valve surgery. In the final logistic model, older age, higher creatinine, higher EuroSCORE II score, and longer operation time were associated with an increased risk of PMV. In contrast, higher preoperative left ventricular ejection fraction was associated with decreased risk. The nomogram showed good discrimination in the training, internal validation, and temporal validation cohorts, with areas under the curve of 0.873 (95% confidence interval [CI], 0.835–0.911), 0.893 (95% CI, 0.844–0.942), and 0.878 (95% CI, 0.849–0.906), respectively. The corresponding Brier scores were 0.100, 0.115, and 0.098. Conclusions: This LASSO-derived perioperative nomogram estimates PMV risk in older adults undergoing cardiac surgery by integrating baseline vulnerability, cardio–renal reserve, and operative burden. The model is interpretable, relies on routinely available variables, and may facilitate preoperative counseling, end-of-surgery risk updating, intensive care unit resource planning, and postoperative extubation strategy. Multicenter prospective validation is needed before widespread implementation.