Abstract
Background/Objectives: We aimed to construct a nomogram model for predicting arteriovenous fistula dysfunction risk and to conduct internal validation. Methods: The clinical data of 335 patients from the 8th Affiliated Hospital of Sun Yat-Sen University, collected from January 2019 to January 2024, were retrospectively analyzed. Among these patients, 103 were assigned to the arteriovenous fistula (AVF) dysfunction group, while 232 were in the non-dysfunction group. In this study, we first identified risk factors for AVF dysfunction using univariate and logistic regression analyses, and then constructed a prediction model by resampling the data. The model's performance was evaluated using the C-index, ROC curve, calibration plot, and decision curve analysis, confirming its strong predictive ability and clinical value. Results: The results indicated that post-dialysis hypotension, abnormal fibrinogen levels, platelet abnormalities, total cholesterol levels, and diabetes mellitus emerged as independent risk factors for AVF dysfunction in MHD patients; however, total protein levels were a protective factor for AVF dysfunction. The model's performance was assessed using the receiver operating characteristic (ROC) curve, the Hosmer-Lemeshow test, and the calibration curve. The ROC curve results demonstrated that the area under the curve (AUC) for the training set was 0.852 (0.799-0.904), while that for the validation set was 0.810 (0.715-0.906), indicating good calibration. The decision curve analysis revealed that the predictive nomogram was clinically useful when the threshold for intervention was set between a 15% and 78% probability of dysfunction. Conclusions: The nomogram prediction model constructed in this study can be used to predict the risk of autogenous arteriovenous fistula dysfunction in hemodialysis patients.