Construction of a prediction model for arteriovenous fistula maturation failure in hemodialysis patients based on machine learning algorithms

WANG Gaoyuan; LI Mingfeng

doi:10.12173/j.issn.1004-5511.202507122

Abstract：Objective To investigate the influencing factors of arteriovenous fistula(AVF)maturation failure in hemodialysis patients,and to construct a risk prediction model for AVF maturation failure based on machine learning algorithms.Methods Patients who underwent initial AVF formation at Kaifeng People's Hospital from January 2021 to January 2025 were selected as the study subjects.Relevant medical records were collected,and patients were divided into a normal group and a maturation failure group based on whether maturation failure occurred within 12 weeks after AVF formation.The patients were then split into training and validation sets in a 7∶3 ratio.Logistic regression,extreme gradient boosting(XGBoost),random forest(RF),and gradient boosting machine(GBM)algorithms were used to identify factors influencing AVF maturation failure.A predictive nomogram,calibration curve,and receiver operating characteristic curve were plotted.Results A total of 260 patients undergoing initial AVF formation were included,with 143 in the normal group and 117 in the maturation failure group,182 in the training set and 78 in the validation set.Multivariate Logistic regression results indicated that age and ALB were independent influencing factors.The XGBoost,RF,and GBM algorithms identified 20,21,and 19 risk factors for AVF maturation failure,respectively.Age,ALB,GLU,and Hb were common variables across all models.The nomogram model constructed based on these variables demonstrated high predictive accuracy the AUCs of the training set and validation set were 0.881[95％CI(0.829,0.933)]and 0.912[95％CI(0.838,0.986)],respectively.Conclusion Age,ALB,Hb,and GLU are the key factors affecting AVF maturation failure in hemodialysis patients.The risk prediction model developed using these indicators exhibits excellent predictive performance and may assist clinicians in selecting vascular access strategies and preventing complications.