Guideline-concordant Arteriovenous Fistula Care with the PREDICT-AVF Web App

The aim of this study was to develop a tool for prediction of arteriovenous fistula (AVF) outcomes and guide shared decision-making about access appropriateness and futility in the context of recent Kidney Disease Outcomes Quality Initiative (KDOQI) guidelines. KDOQI guidelines discourage ongoing hemodialysis access salvage attempts after two interventions prior to successful use or three interventions per year overall. The data source was the international multicenter PATENCY-1 and PATENCY-2 randomized trials, which enrolled patients undergoing new radiocephalic AVF creation. The prediction outcome was the number of surgical or endovascular interventions required at 1 year. Cox, random survival forest, pooled logistic, and elastic net recurrent time-to-event machine learning prediction models were built using a combination of baseline characteristics and postoperative ultrasound measurements. Models were trained with 10-fold cross-validation hyperparameter selection within a randomly split training dataset. Discrimination and calibration performance was assessed on a holdout testing dataset. An interactive web application was created, which generates patient-specific predictions contextualized with the KDOQI guidelines. The cohort included 914 patients; mean age was 57 years (standard deviation, 13 years); 22% were female. Radiocephalic AVFs were created at the snuffbox (2%), wrist (74%), or forearm (24%). Patients underwent an estimated 1.04 (95% confidence interval [CI], 0.94-1.13) interventions in the first year. The Cox models using either 4- to 6-week or 12-week postoperative ultrasound information alone had the best discrimination performance, with areas under the receiver operating characteristic curve of 0.75 (95% CI, 0.68-0.82) and 0.77 (95% CI, 0.71-0.83) at 1 year. The random survival forest model performed best when restricting to only baseline characteristics (Table I). The interactive web application is deployed at https://predict-avf.com (Fig 1). The PREDICT-AVF web application can guide patient counseling and guideline-concordant shared decision-making as part of the KDOQI patient-centered end-stage kidney disease life plan in both preoperative and postoperative settings. Future versions of PREDICT-AVF will include maturation and access patency in a variety of access configurations.Table 1Discrimination performance of each model and feature set, measured by the area under the receiver operating characteristic curve (AUROC [ie, C-statistic]) at 1 year postoperativelyFeaturesCoxRandom survival forestPooled logistic regressionElastic netBaseline0.704 (0.639-0.769)0.715 (0.647-0.783)a0.696 (0.631-0.762)0.703 (0.638-0.769)Baseline + 4- to 6-week US0.710 (0.635-0.785)0.747 (0.676-0.818)a0.707 (0.632-0.782)0.708 (0.633-0.783)Baseline + 12-week US0.731 (0.664-0.798)0.776 (0.712-0.841)a0.725 (0.657-0.793)0.738 (0.671-0.804)4- to 6-week US0.747 (0.678-0.816)a0.722 (0.647-0.797)0.743 (0.674-0.813)0.745 (0.675-0.814)12-week US0.770 (0.708-0.831)a0.736 (0.669-0.803)0.761 (0.699-0.824)0.763 (0.700-0.825)US, Ultrasound.Perfect discrimination corresponds to an AUROC of 1.Feature sets included baseline (age, sex, smoking status, diabetes, heart failure, renal transplant history, anticoagulation use, antiplatelet use, statin use, hemodialysis status at the time of arteriovenous fistula creation, central venous catheter history, number of prior failed accesses, vein and artery diameter, anastomotic suture technique, access location, and access side), baseline with the addition of postoperative ultrasound measurements (flow volume, access diameter, and luminal stenosis), and ultrasound measurements alone.aBest discrimination performance within each feature set. Open table in a new tab