Risk factors analysis and prediction model construction of perioperative gram-negative bacterial infection in patients with intestinal fistula.

The purpose of this research is to explore the factors affecting perioperative gram-negative bacterial infection in intestinal fistula patients and to construct a predictive model. Between January 2022 and June 2024, 223 individuals suffering from intestinal fistula and undergoing surgical intervention at our medical facility were chosen and allocated randomly into a training cohort (n = 156) and a validation cohort (n = 67). The determinants of perioperative gram-negative bacterial infection in these intestinal fistula patients were examined utilizing both lasso regression and logistic regression methodologies. Employing the R programming language, a linear chart model aimed at forecasting infection risk was constructed. The accuracy of this predictive model was gauged through the application of the receiver operating characteristic curve and the Hosmer-Lemeshow goodness of fit test. Furthermore, a clinical decision curve analysis was conducted to assess the practical applicability of the nomogram in a clinical context. A total of 192 gram-negative bacterial strains were identified, with 175 being multidrug-resistant organisms, accounting for 91.15%. The most prevalent was Klebsiella pneumoniae, with 47 strains (24.48%), followed by Escherichia coli at 41 strains (21.35%). Multivariate logistic regression analysis revealed that undergoing a secondary surgery during hospitalization (odds ratio [OR] = 16.97, 95% confidence interval [CI]: 1.70-168.99, P = .016), extended hospital stays (OR = 1.03, 95% CI: 1.01-1.05, P = .005), and elevated postoperative white blood cell levels (OR = 1.29, 95% CI: 1.14-1.45, P < .001) were independent risk factors for infection. Conversely, a higher postoperative albumin level (OR = 0.87, 95% CI: 0.80-0.94, P < .001) served as an independent protective factor against infection. The receiver operating characteristic curve demonstrated that the model possesses excellent discriminatory power. Additionally, the Hosmer-Lemeshow test verified the consistency of predicted and observed probabilities. Furthermore, the decision curve analysis curve indicated that the nomogram model holds significant clinical value. The nomogram model, based on whether a secondary surgery is performed, hospital stays, postoperative white blood cell count, and postoperative albumin level, exhibits high predictive performance for the occurrence of perioperative gram-negative bacterial infection in patients with intestinal fistula. This model holds significant importance for clinical prediction, guiding antibiotic usage, and postoperative care.