Development and Validation of a Risk Prediction Model for Multidrug-Resistant Organisms Infection in Diabetic Foot Ulcer Patients

Jinghang ZhangXuemei LiBai ChangYongmei LiBaocheng Chang^*

• NHC Key Laboratory of Hormones and Development, Tianjin Key Laboratory of Metabolic Diseases, Chu Hsien-I Memorial Hospital & Tianjin Institute of Endocrinology, Tianjin Medical University, Tianjin Key Laboratory of Metabolic Diseases, Tianjin Medical University, Tianjin, China

Objective: To develop and validate a nomogram for predicting the risk of multidrug-resistant organisms (MDROs) infection in diabetic foot ulcer (DFU) patients. Methods: 701 DFU patients were divided into training (491 cases) and validation (210 cases) sets (7:3 ratio). Multivariate logistic regression analysis was performed to identify the independent risk factors for MDRO infection in DFU patients. Two nomogram prediction models were developed based on the independent risk factors. The predictive efficacy of the prediction models was evaluated using the receiver operating characteristic (ROC) curve and calibration curve analysis. The decision curve analysis (DCA) was performed to evaluate the prediction model's performance during clinical application. Results: Multivariate logistic regression analysis identified previous antibiotic therapy, surgical therapy, ulcer size>4cm2, and CRP as independent risk factors. Two models were developed and validated based on the analysis. Model 1 included previous antibiotic therapy, surgical therapy, and ulcer size>4cm2. Model 2 added a further laboratory indicator to Model 1, such as CRP. In the training set, the AUC of the nomogram for Model 1and Model 2 was 0.763(95% CI 0.711-0.815) and 0.789 (95% CI 0.740-0.838), respectively, and 0.837 (95% CI 0.744-0.900) and 0.845 (95% CI 0.785-0.905) in the validation set. The Youden indexes for Models 1and 2 were 0.416 and 0.470 in the training set and 0.558 and 0.588 in the validation set, respectively. Notably, Model 2 showed This is a provisional file, not the final typeset article higher sensitivity and specificity. The calibration plot and Hosmer‒Lemeshow test for Model 1 and Model 2 indicated that the predicted probability had good consistency with the actual probability in both the training set (P=0.689 for Model 1 and P=0.139 for Model 2) and validation set (P=0.607 for Model 1and P=0.635 for Model 2). The DCA curve indicated that the models had good clinical utility. All models performed well for both discrimination and calibration. Conclusion: This study developed two nomogram models for predicting MDRO infection risk in DFU patients. Model 2, with superior predictive performance, enables early identification of high-risk patients. These models facilitate targeted interventions, potentially reducing MDRO complications and healthcare burdens.