[Establishment and validation of a predictive model for postoperative pulmonary complications in elderly patients undergoing non-cardiac surgery]

Objective: To develop and validate a predictive model for postoperative pulmonary complications (PPCs) in elderly patients undergoing non-cardiac surgery. Methods: This retrospective study included 51 354 elderly patients over 65 years old who underwent non-cardiac surgery at the First Medical Center of Chinese PLA General Hospital from January 2009 to December 2018. The patients were randomly divided into a modeling group [n=41 084; 21 550 males, 19 534 females; age: 70 (67, 74) years] and an internal validation group [n=10 270; 5 458 males, 4 812 females; age: 70 (67, 74) years] at a ratio of 4∶1. Additionally, an external validation group of 14 378 patients [7 893 males, 6 845 females; age: 70 (67, 75) years] who underwent non-cardiac surgery at Henan Provincial People's Hospital between November 2014 and May 2022 was retrospectively included. Multivariate logistic regression were performed to identify factors associated with PPCs. A nomogram prediction model was constructed based on these factors and validated internally and externally. The model's performance and clinical applicability were assessed using receiver operating characteristic (ROC) curves, calibration curves, and decision curves. Results: Among the 51 354 elderly patients underwent general anesthesia for non-cardiac surgery, the incidence of PPCs was 17.5% (9 008/51 354). Multivariate logistic regression analysis reveals that anesthesia duration 130-<183 min (OR=1.858, 95%CI: 1.529-2.266), anesthesia duration 183-<250 min (OR=2.537, 95%CI: 2.079-3.108), anesthesia duration≥250 min(OR=3.533, 95%CI: 2.868-4.368), crystalloid infusion volume 1 400-<2 000 ml (OR=1.481, 95%CI: 1.204-1.829), crystalloid infusion volume 2 000-<9 000 ml (OR=1.776, 95%CI: 1.426-2.220), upper abdominal surgery (OR=1.658, 95%CI: 1.498-1.835), malignancy (OR=1.796, 95%CI: 1.606-2.012), fentanyl dosage 0.40-<0.55 mg (OR=1.404, 95%CI: 1.203-1.640), fentanyl dosage≥0.55 mg (OR=1.601, 95%CI: 1.386-1.854), prophylactic use of antibiotics (OR=7.897, 95%CI: 5.124-12.983), age (OR=1.039, 95%CI: 1.030-1.049), smoking (OR=1.124, 95%CI: 1.014-1.246), preoperative chest X-ray abnormalities (OR=2.139, 95%CI: 1.820-2.509) and intraoperative hypotension (OR=3.184, 95%CI: 2.120-4.795) were risk factors for PPCs, while elective surgery (OR=0.301, 95%CI: 0.220-0.417) was a protective factor. The nomogram model incorporating these factors had an area under the curve (AUC) of 0.757 (95%CI: 0.748-0.766, P=0.309) in the modeling group, 0.779 (95%CI: 0.760-0.796, P=0.171) in the internal validation group, and 0.778 (95%CI: 0.763-0.792, P<0.001) in the external validation group. Calibration curves and decision curves demonstrated good consistency and benefit of the model. Conclusion: The nomogram model which based on anesthesia duration, crystalloid infusion volume, upper abdominal surgery, malignancy, fentanyl dosage, prophylactic use of antibiotics, age, smoking, preoperative chest X-ray abnormalities, intraoperative hypotension and elective surgery provides strong predictive value and clinical utility for assessing the risk of PPCs in elderly patients undergoing non-cardiac surgery.