Incidence, risk factors, and predictive modeling of pulmonary infection after high-risk surgery for lung cancer: a retrospective case-control study.

Background: The hierarchical operation management system is one of the core medical systems. Graded management based on the degree of surgical risk, difficulty, resource consumption, and ethical risks can help ensure the quality and safety of the surgery. With the progress of medical technology and the continuous development of medical standards, the proportion of lung cancer patients who underwent high-risk surgery was increasing rapidly. The purpose of this study is to explore the incidence, risk factors, and prediction models of pulmonary infection after high-risk surgery for lung cancer based on machine learning algorithms. Methods: This study included individuals who underwent lung cancer high-risk surgery at Shanghai Chest Hospital, Shanghai Jiao Tong University School of Medicine from January 2021 to December 2023. Five machine learning algorithms including least absolute shrinkage and selection operator (LASSO)-assisted logistic regression (LR), artificial neural network (ANN), support vector machine (SVM), random forest (RF), and eXtreme gradient boosting (XGB) were adopted to explore risk factors and prediction models of pulmonary infection after high-risk surgery for lung cancer. Results: A cohort of 2,650 patients were eligible for the study after application of the exclusion criteria, with an overall incidence of postoperative pulmonary infection at 9.66% (256/2,650). LASSO regression screened out eight characteristic variables including daily smoking, history of diabetes, diffusing capacity of the lung for carbon monoxide percentage of predicted (DLCO%Pred), airway resistance percentage of predicted (Raw%Pred), maximum tumor diameter, perioperative oral nutritional supplements (ONS) supplement, postoperative urinary catheter, and pleural adhesion degree. The risk prediction model of postoperative pulmonary infection was constructed using these eight clinical features. The area under the curve (AUC) range of the five models was 0.893-0.936. The XGB model outperformed the others, with an AUC of 0.936 [95% confidence interval (CI): 0.923-0.949]. The LR model had an AUC of 0.927 (95% CI: 0.921-0.939), second only to the XGB model, which was converted into a nomogram for model visualization. Conclusions: The establishment of a risk prediction model based on machine learning can help clinical nursing staff identify high-risk patients for pulmonary infection after lung cancer high-risk surgery. The nomogram is expected to be an effective tool for nursing staff to manage the risk of pulmonary infection after lung cancer high-risk surgery.