Predictive Models of Arteriovenous Endovascular Fistula Deterioration in Maintenance Haemodialysis Patients: A Systematic Review

ABSTRACT Aims The aim of this study is to systematically evaluates of risk prediction models for loss of arteriovenous endovascular fistula function in maintenance haemodialysis patients to provide guidance for establishing and improving the model. Design The PRISMA guidelines guided this systematic review. The review was registered with PROSPERO (CRD:42023416964). Methods This systematic review was conducted in accordance with the PRISMA 2020 guidelines. A comprehensive search was conducted on various databases, including Pubmed, Embase, Cochrane Library, Web of Science, China Knowledge Network, Wanfang Database, Vipshop Journal Repository and Chinese Biomedical Literature Database. The objective of this extensive search was to identify studies pertaining to predictive models for the occurrence of autogenous arteriovenous fistula failure. The search period extended from the creation of the databases to 4 April 2023. Two reviewers independently reviewed the literature to ensure the reliability of the findings. Extraction of relevant data and analysis of the risk of bias and applicability of the included literature were conducted using the Risk of Bias Assessment Tool for Predictive Modelling Studies. Outcome data were reported by narrative synthesis. Results Eight studies were included in the analysis. One study incorporated both internal and external validation, two studies employed internal validation and one study utilized external validation. The multivariate models reported that age, diabetes and hypotension on dialysis were the independent predictors in common. It is noteworthy that all studies exhibited some degree of bias. However, the applicability of the findings was deemed adequate. Conclusion The prediction model for the occurrence of arteriovenous endovascular fistula failure in patients with maintenance haemodialysis has good applicability. However, the overall bias is high, and the model's methodology contains defects. To address these issues, further research is necessary to construct the model in accordance with the PROBAST tool. Healthcare professionals should intervene promptly in high‐risk patients with the aforementioned risk factors to minimize the incidence.