Development and validation of a risk prediction model related to inflammatory and nutritional indexes for postoperative pulmonary infection after radical colorectal cancer surgery

Objective: To develop and validate a risk prediction model related to inflammatory and nutritional indexes for postoperative pulmonary infection (POI) after radical colorectal cancer (CRC) surgery.

Design: Cross-sectional study.

Participants: This study analysed 866 CRC patients after radical surgery at a tertiary hospital in China.

Methods: Univariable and multivariable logistic regression (LR) analyses were used to explore influence factors of POI. Predictive models were constructed using LR, random forest, support vector machine, K-nearest neighbours, naive Bayes and XGBoost. The LR model was used to generate a nomogram for POI prediction. The discrimination and calibration of the nomogram were assessed using receiver operating characteristic (ROC) curves and calibration curves. The contributions of inflammatory and nutritional indexes to the nomogram were evaluated through Net Reclassification Improvement and integrated discrimination improvement, while clinical practicability was assessed using decision curve analysis.

Main outcome measures: POI during hospitalisation.

Results: Independent factors identified from multivariable LR for prediction POI included age, respiratory disease, Systemic Inflammation Response Index, albumin-to-globulin ratio, operative method and operative duration. The LR model demonstrated the best performance, with an area under the ROC curve of 0.773 (95% CI: 0.674 to 0.872). The nomogram has good differentiation ability, calibration and net benefit. Incorporating inflammatory and nutritional indexes into the nomogram enhanced predictive value compared with models excluding either factor.

Conclusion: The nomogram related to inflammatory and nutritional indexes may represent a promising tool for predicting POI after radical surgery in CRC patients.