Dataset Information

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.

SUBMITTER: Peralta R

PROVIDER: S-EPMC8656573 | biostudies-literature | 2021 Nov

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

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

Peralta Ricardo R Garbelli Mario M Bellocchio Francesco F Ponce Pedro P Stuard Stefano S Lodigiani Maddalena M Fazendeiro Matos João J Ribeiro Raquel R Nikam Milind M Botler Max M Schumacher Erik E Brancaccio Diego D Neri Luca L

International journal of environmental research and public health 20211124 23

<b>Background:</b> 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. <b>Methods:</b> 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 ...[more]

PMID: 34886080

Similar Datasets

Project description:ObjectiveCurrent guidelines suggest that arteriovenous fistula (AVF) is associated with survival advantage over arteriovenous graft (AVG). However, AVFs often require months to become functional, increasing tunneled dialysis catheter (TDC) use, which can erode the benefit of an AVF. We sought to compare survival in patients with end-stage renal disease after creation of an AVF or AVG in patients starting hemodialysis (HD) with a TDC and to identify patient populations that may benefit from preferential use of AVG over AVF.MethodsUsing U.S. Renal Data System databases, we identified incident HD patients in 2005 through 2008 and observed them through 2008. Initial access type and clinical variables including albumin levels were assessed using U.S. Renal Data System data collection forms. Attempts at AVF and AVG creation in patients who started HD through a TDC were identified by Current Procedural Terminology codes. We accounted for the effect of changes in access type by truncating follow-up when an additional AVF or AVG was performed. Survival curves were then constructed, and log-rank tests were used for pairwise survival comparisons, stratified by age. Multivariate analysis was performed with Cox proportional hazards regressions; variables were chosen using stepwise elimination. An interaction of access type and albumin level was detected, and Cox models using differing thresholds for albumin level were constructed. The primary outcome was survival.ResultsAmong the 138,245 patients who started with a TDC and had complete records amenable for analysis, 22.8% underwent AVF creation (mean age ± standard deviation, 68.9 ± 12.5 years; 27.8% mortality at 1 year) and 7.6% underwent AVG placement (70.2 ± 12.0 years; 28.2% mortality) within 3 months of HD initiation; 69.6% remained with a TDC (63.2 ± 15.4 years; 33.8% mortality). In adjusted Cox proportional hazards regression, AVF creation is equivalent to AVG placement in terms of survival (hazard ratio [HR], 0.98; 95% confidence interval [CI], 0.93-1.02; P = .349). AVG placement is superior to continued TDC use (HR, 1.54; 95% CI, 1.48-1.61; P < .001). In patients older than 80 years with albumin levels >4.0 g/dL, AVF creation is associated with higher mortality hazard compared with AVG creation (HR, 1.22; 95% CI, 1.04-1.43; P = .013).ConclusionsFor patients who start HD through a TDC, placement of an AVF and AVG is associated with similar mortality hazard. Further study is necessary to determine the ideal access for patients in whom the survival advantage of an AVF over an AVG is uncertain.

Dataset Information

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

Publications

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

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets