Risk stratification tool for all surgical site infections after coronary artery bypass grafting

OBJECTIVE: To develop a risk score for surgical site infections (SSIs) after coronary artery bypass grafting (CABG). DESIGN: Retrospective study. SETTING: University hospital. PATIENTS: A derivation sample of 7,090 consecutive isolated or combined CABG patients and 2 validation samples (2,660 total patients). METHODS: Predictors of SSIs were identified by multivariable analyses from the derivation sample, and a risk stratification tool (additive and logistic) for all SSIs after CABG (acronym, ASSIST) was created. Accuracy of prediction was evaluated with C-statistic and compared 1:1 (using the Hanley-McNeil method) with most relevant risk scores for SSIs after CABG. Both internal (1,000 bootstrap replications) and external validation were performed. RESULTS: SSIs occurred in 724 (10.2%) cases and 2 models of ASSIST were created, including either baseline patient characteristics alone or combined with other perioperative factors. Female gender, body mass index >29.3 kg/m2, diabetes, chronic obstructive pulmonary disease, extracardiac arteriopathy, angina at rest, and nonelective surgical priority were predictors of SSIs common to both models, which outperformed (P < .0001) 6 specific risk scores (10 models) for SSIs after CABG. Although ASSIST performed differently in the 2 validation samples, in both, as well as in the derivation data set, the combined model outweighed (albeit not always significantly) the preoperative-only model, both for additive and logistic ASSIST. CONCLUSIONS: In the derivation data set, ASSIST outperformed specific risk scores in predicting SSIs after CABG. The combined model had a higher accuracy of prediction than the preoperative-only model both in the derivation and validation samples. Additive and logistic ASSIST showed equivalent performance.