A comprehensive feature importance analysis of surgical site infection following colorectal cancer surgery

Surgical site infection (SSI) after colorectal cancer (CRC) surgery is still a significant healthcare issue. This study aimed to analyze risk factor associated with SSI. A total of 528 consecutive CRC patients who underwent curative resections between September 2017 and August 2024 were analyzed. Receiver operating characteristic (ROC) curves and machine learning (ML) models of Random Forest, K-Nearest Neighbor, Decision Tree, and XG Boost were employed for risk factor evaluation. SSI rate was 17.6%. SSI development varied significantly by tumor location, surgical technique, stoma formation, and neoadjuvant therapy. Furthermore, the preoperative values of white blood cell (WBC) count ≤ 5 (10³/mcL), lymphocyte count ≤ 1.32 (10³/mcL), polymorphonuclear (PMN) count ≤ 3.3 (10³/mcL), platelet distribution width (PDW) ≤ 11.2 (fL), and serum creatinine ≤ 0.9 (mg/dL) were associated with SSI development. The results of the ML models also showed that PDW, PMN count, WBC count, lymphocyte count, RDW, serum urea, tumor location, surgery duration, surgical technique, and serum creatinine were the most important factors affecting SSI development. Development of SSI following CRC surgery was associated with several risk factors, including patients' characteristics and perioperative conditions.