Prediction models of surgical site infection after gastrointestinal surgery: a nationwide prospective cohort study.

Abstract

This study aimed to construct and validate a clinical prediction model for surgical site infection (SSI) risk 30 days after gastrointestinal surgery. This multicentre study involving 57 units conducted a 30-day postoperative follow-up of 17353 patients who underwent gastrointestinal surgery at the unit from 1 March 2021 to 28 February 2022. The authors collected a series of hospitalisation data, including demographic data, preoperative preparation, intraoperative procedures and postoperative care. The main outcome variable was SSI, defined according to the Centres for Disease Control and Prevention guidelines. This study used the least absolute shrinkage and selection operator (LASSO) algorithm to screen predictive variables and construct a prediction model. The receiver operating characteristic curve, calibration and clinical decision curves were used to evaluate the prediction performance of the prediction model. Overall, 17353 patients were included in this study, and the incidence of SSI was 1.6%. The univariate analysis combined with LASSO analysis showed that 20 variables, namely, chronic liver disease, chronic kidney disease, steroid use, smoking history, C-reactive protein, blood urea nitrogen, creatinine, albumin, blood glucose, bowel preparation, surgical antibiotic prophylaxis, appendix surgery, colon surgery, approach, incision type, colostomy/ileostomy at the start of the surgery, colostomy/ileostomy at the end of the surgery, length of incision, surgical duration and blood loss were identified as predictors of SSI occurrence ( P <0.05). The area under the curve values of the model in the train and test groups were 0.7778 and 0.7868, respectively. The calibration curve and Hosmer-Lemeshow test results demonstrated that the model-predicted and actual risks were in good agreement, and the model forecast accuracy was high. The risk assessment system constructed in this study has good differentiation, calibration and clinical benefits and can be used as a reference tool for predicting SSI risk in patients.