Development and Validation of a Machine Learning Model Predicting Arteriovenous Fistula Failure in a Large Network of Dialysis Clinics

Abstract

Background: Vascular access surveillance of dialysis patients is a challenging task for clinicians. We derived and validated an arteriovenous fistula failure model (AVF-FM) based on machine learning. Methods: The AVF-FM is an XG-Boost algorithm aimed at predicting AVF failure within three months among in-centre dialysis patients. The model was trained in the derivation set (70% of initial cohort) by exploiting the information routinely collected in the Nephrocare European Clinical Database (EuCliD®). Model performance was tested by concordance statistic and calibration charts in the remaining 30% of records. Features importance was computed using the SHAP method. Results: We included 13,369 patients, overall. The Area Under the ROC Curve (AUC-ROC) of AVF-FM was 0.80 (95% CI 0.79–0.81). Model calibration showed excellent representation of observed failure risk. Variables associated with the greatest impact on risk estimates were previous history of AVF complications, followed by access recirculation and other functional parameters including metrics describing temporal pattern of dialysis dose, blood flow, dynamic venous and arterial pressures. Conclusions: The AVF-FM achieved good discrimination and calibration properties by combining routinely collected clinical and sensor data that require no additional effort by healthcare staff. Therefore, it can potentially enable risk-based personalization of AVF surveillance strategies.

Construction and validation of a nomogram to predict mortality risk in HIV/AIDS patients undergoing maintenance hemodialysis

“…Grace Fet, et al constructed an automated lung ultrasound image assessment using artifcial intelligence to identify uid overload in dialysis patients [50] , in order to provide a scienti c support for clinicians evaluating the liquid load of hemodialysis patients, so that they can help patients achieve higher dialysis quality. Ricardo Peralta, et al [51] established a machine learning model predicting arteriovenous stula failure, which had the ability to help vascular access doctors predict the rate of hemodialysis patients encountering with arteriovenous stula failure in early stage, as a result they could conduct an alternative plan in advance to solve the problem of vascular access of patients. However, there was no researches predicting the mortality risk of HIV/AIDS patients undergoing hemodialysis at present, and all the studies currently only focused on the mortality risk of HIV/AIDS patients [24] and hemodialysis patients [25] , the clinical application value of which were limited.…”

Surgical planning of arteriovenous fistulae in routine clinical practice: A machine learning predictive tool

“…However, in contrast to the present study, their model was limited to radio-cephalic AVFs only. Finally, Peralta et al 26 apply ML for predicting AVF failure within 3 months, based on routinely recorded clinical information of HD patients. Their work, which would enable risk-based personalization of AVF surveillance strategies, is therefore complementary to ours, which aims to support VA surgical planning.…”

From electronic health records to clinical management systems: how the digital transformation can support healthcare services

“…Use of ACM has been shown to strongly improve hemoglobin target achievement and reduce drug utilization by 25% to 50% [ 27 , 29 ]. New AI modules, such as the Cardiovascular Literature-Based Risk Algorithm (CALIBRA [ 44 ]), the Prognostic Reasoning System for CKD Progression (PROGRES-CKD [ 45 ]) and the Arteriovenous Fistula Failure model (FFM [ 46 ]) to name a few, have been more recently developed using EuCliD data and are currently under evaluation before further clinical implementation.…”