Abstract:  Objective  To analyze the risk factors for poor maturation of autologous arteriovenous fistula (AVF) and construct a decision tree prediction model based on the risk factors.  Methods  A total of 648 maintenance hemodialysis patients undergoing AVF creation surgery in the Department of Nephrology, the First Affiliated Hospital of Chongqing Medical University from October 2023 to October 2024 were retrospectively reviewed. Their general data were analyzed. The diagnostic efficacy of the decision tree prediction model was analyzed. Statistical processing was performed using SPSS 26.0 and SPSS Modeler 18.0 software.  Result  Among the 648 patients undergoing AVF creation surgery, 145 had poor AVF maturation with an incidence rate of 22.38%. BMI ＞28 kg/m² (OR=1.863, 95% CI: 1.252～2.771, P=0.002), hypertension (OR=2.033, 95% CI: 1.362～3.035, P= 0.001), smoking (OR=2.134, 95% CI: 1.432～3.181, P<0.001), radial artery diameter <2 mm (OR=1.814, 95% CI: 1.217～2.704, P=0.003), brachial artery blood flow <650 mL/min (OR=2.086, 95% CI: 1.398～3.113, P<0.001), parathyroid hormone >65ng/L (OR=1.708, 95% CI: 1.149～2.539, P=0.008), and no calcium channel blocker medication (OR=1.707, 95% CI: 1.145～2.547, P=0.009) are independent risk factors for poor AVF maturation (P<0.05). The decision tree model selected brachial artery blood flow, BMI, smoking, calcium channel blocker medication, parathyroid hormone, hypertension, and radial artery diameter as the nodes in the model. Among these nodes, brachial artery blood flow was in the first layer, showing the strongest correlation with AVF immaturity. The AUC value for validating the decision tree model was 0.822 (95% CI: 0.790～0.854).  Conclusion  The decision tree prediction model constructed in this study has a higher clinical applicability. It can comprehensively analyze the impact of a specific predictor variable in different subgroups, offering a useful reference for predicting AVF immaturity.

Key words: Arteriovenous fistula, Decision tree algorithm, Risk factors