A Nomogram Model for Early Prediction of Postoperative Intra-Abdominal Infection in Colorectal Cancer patients Based on Perioperative Clinical Variables

Purpose To establish a predictive nomogram model for Postoperative Intra-Abdominal Infection (PIAI) following Colorectal Cancer (CRC) surgery using perioperative clinical variables, thereby facilitating early identification of high-risk patients and enhance postoperative management. Method This retrospective cohort study included colorectal cancer patients undergoing surgery from 2022 to 2024 at a single center. Perioperative clinical and laboratory data, along with blood glucose levels from the day of surgery to postoperative day 3, were collected. PIAI was defined according to the Centers for Disease Control and Prevention (CDC) criteria. Blood glucose trajectories were identified using latent class mixed modeling(LCMM). LASSO and logistic regression analyses were used to select risk factors for PIAI. A predictive nomogram was constructed and internally validated by calibration curve, ROC curve analysis (AUC), decision curve analysis (DCA), and clinical impact curves (CIC). Result A total of 197 patients' data were collected, and 163 patients were finally included in the study. The incidence of PIAI in the cohort was 17.8%. Compared with patients without PIAI, those who developed infection had significantly higher rates of NRS2002 ≥ 3 (62.1% vs. 38.1%, P = 0.018), PGSGA ≥ 4 (72.4% vs. 41.0%, P  = 0.002), ASA grade ≥ 3 (17.2% vs 6.0%, P  = 0.042) and preoperative antibiotic use (10.3% vs 1.5%, P  = 0.012), as well as greater intraoperative blood loss (97.9 ± 87.9 mL vs 49.8 ± 40.8 mL, P  = 0.001) and higher Creactive protein levels on postoperative day 1 (42.7 ± 35.2 mg/L vs 25.8 ± 22.4 mg/L, P  = 0.007) and day 3 (100.8 ± 65.7 mg/L vs. 50.7 ± 39.8 mg/L, P  < 0.001). The LCMM model classified postoperative blood glucose trajectories into high and lowglucose groups, with the highglucose group demonstrating a significantly higher infection rate (38.1% vs. 14.8%, P  = 0.009). Following further selection, antibiotic use before surgery (HR = 9.292, 95%CI: 1.062–81.320, P  = 0.044), blood loss (HR = 1.011, 95%CI: 1.001–1.021, P  = 0.029), and POD3 CRP (HR = 1.014, 95%CI: 1.004–1.025, P  = 0.006) were incorporated into the prediction model.The AUROC values of the model was 0.8111. The calibration curve, DCA, and CIC demonstrated the favorable clinical applicability of the models. Conclusion This study established a concise and clinically applicable nomogram for the early prediction of PIAI in CRC patients, incorporating preoperative antibiotic use, intraoperative blood loss, and POD3 CRP as independent predictors. The model demonstrated favorable discrimination and calibration. Furthermore, while not included in the final model, LCMM of postoperative glucose trajectories provided a novel perspective for future research on metabolic patterns and infection risk.