Objective: The primary objective of this study is to develop and validate a high-risk model for Arteriovenous Fistula Thrombosis (AVFT) in patients undergoing autogenous arteriovenous fistula surgery for hemodialysis.
Methods: Retrospectively, we collected general information, clinical characteristics, laboratory examinations, and dialysis-related factors from a cohort of 1465 patients who received continuous arteriovenous fistula surgery at the Hemodialysis Access Center of Sichuan Provincial People's Hospital between January 2019 and June 2022. The patients were randomly divided into a training set and a validation set in a 2:1 ratio. The training set was utilized to select AVFT-related features using LASSO regression. A predictive model was constructed using logistic regression analysis, and its performance was assessed in the validation set.
Results: Through LASSO regression, we initially identified 13 candidate factors. Subsequently, based on the Akaike Information Criterion (AIC) principle, the following factors were selected to construct the AVFT prediction model: monocytes_ratio, Fistula blood velocity, cystatin-c, homocysteine, parathormone, artery_dysfunction, C-reactive protein, fibrinogen, and d-dimer. The discrimination C-index of the model in the training set was 0.8767. For this training set, the sensitivity was 48.05% and the specificity was 96.84%. In the validation set, the model's discrimination C-index, as evaluated by the ROC curve analysis, was 0.7888. The sensitivity was 14.29%, and the specificity was 97.04%. We assessed the calibration of the model using calibration curves, obtaining a maximum absolute difference of Emax = 0.205 and an average absolute difference of Eave = 0.032. Furthermore, we evaluated calibration and accuracy using the Spiegelhalter Z-test, yielding an S:P ratio of 0.704.
Conclusion: AVFT is a multifactorial outcome influenced by factors such as injury, inflammatory factors, blood glucose levels, blood velocity, coagulation, electrolyte metabolism, and vascular endothelial function.
Keywords: Hemodialysis; Lasso regression; arteriovenous fistula thrombosis; prediction model.