Prediction models for postoperative pulmonary complications in intensive care unit patients after noncardiac thoracic...

Postoperative pulmonary complication (PPC) is a leading cause of mortality and poor outcomes in postoperative patients. No studies have enrolled intensive care unit (ICU) patients after noncardiac thoracic surgery, and effective prediction models for PPC have not been developed. This study aimed to explore the incidence and risk factors and construct prediction models for PPC in these patients.

This study retrospectively recruited patients admitted to the ICU after noncardiac thoracic surgery at West China Hospital, Sichuan University, from July 2019 to December 2022. The patients were randomly divided into a development cohort and a validation cohort at a 70% versus 30% ratio. The preoperative, intraoperative and postoperative variables during the ICU stay were compared. Univariate and multivariate logistic regression analyses were applied to identify candidate predictors, establish prediction models, and compare the accuracy of the models with that of reported risk models.

A total of 475 ICU patients were enrolled after noncardiac thoracic surgery (median age, 58; 72% male). At least one PPC occurred in 171 patients (36.0%), and the most common PPC was pneumonia (153/475, 32.21%). PPC significantly increased the duration of mechanical ventilation (p < 0.001), length of ICU stay (p < 0.001), length of hospital stay (LOS) (p < 0.001), and rate of reintubation (p = 0.047) in ICU patients. Seven risk factors were identified, and then the prediction nomograms for PPC were constructed. At ICU admission, the area under the curve (AUC) was 0.766, with a sensitivity of 0.71 and specificity of 0.60; after extubation, the AUC was 0.841, with a sensitivity of 0.75 and specificity of 0.83. The models showed robust discrimination in both the development cohort and the validation cohort, and they were well calibrated and more accurate than reported risk models.

ICU patients who underwent noncardiac thoracic surgery were at high risk of developing PPCs. Prediction nomograms were constructed and they were more accurate than reported risk models, with excellent sensitivity and specificity. Moreover, these findings could help assess individual PPC risk and enhance postoperative management of patients.