Risk score for the prediction of arteriovenous fistula maturation.

Abstract

The poor maturation rate of arteriovenous fistulas (AVFs) remains a challenge. We aim to develop a parsimonious risk score equations that incorporate patients' demographics, comorbidities, and vessel characteristics to assist the access surgeons in predicting the likelihood of AVF maturation before surgery. Data of AVFs created from 2015 to 2020 in National University Hospital Singapore was retrospectively obtained and used as the training dataset. Two prediction models were developed to predict: (1) the commencement of cannulation; and (2) functional maturation, by 3 months after creation. The models were internally validated using k-fold cross validation and externally validated on patients who had AVFs created in 2021 and 2022. The final models were converted to interpretable risk score equations for clinical usage. The training dataset included 474 AVF cases. Six and eight variables were selected for predicting: (1) the commencement of cannulation; and (2) functional maturation, respectively. The two models demonstrated satisfactory performance with C-statistics of 0.72 and 0.74, respectively. The C-statistics in internal validation were 0.7 and 0.69 for the two models, respectively. External validation was performed using the data of 220 different patients; the C-statistic was 0.7 in both models. The two risk scores developed demonstrated reasonable preoperative predicative values for AVF maturation. They are potentially useful in aiding clinicians to make decisions about AVF creation sites. Further evaluation of these prediction scores in a large cohort is required to reveal their clinical value.