Abstract: TH-PO335
Algorithmic Estimation of Vascular Access Dysfunction from Serial Bruit Recordings
Session Information
- Vascular Access - I
November 07, 2019 | Location: Exhibit Hall, Walter E. Washington Convention Center
Abstract Time: 10:00 AM - 12:00 PM
Category: Dialysis
- 704 Dialysis: Vascular Access
Authors
- Desai, Niraj, Case Western Reserve University School of Medicine, Cleveland, Ohio, United States
- Majerus, Steve, Louis Stokes Cleveland VA Medical Center, Cleveland, Ohio, United States
Background
Hemodialysis vascular access dysfunction accounts for a large fraction of hospital visits for dialysis patients and has significant implications on cost of care, morbidity and mortality. This study examined if mathematical analysis of serially-recorded bruits could detect changes in bruit pitch indicative of imminent vascular access dysfunction.
Methods
An algorithm was developed to detect bruit pitch shifts between recording locations along the vascular access. The algorithm produced a single value for each location which monotonically increased with bruit pitch caused by turbulent blood flow. Algorithm performance was validated on a phantom vascular access at flows of 80-1,200 mL/min and stenoses ranging from 10-85%.
Prognostic value in predicting vascular access dysfunction was tested longitudinally in 24 patients at an outpatient Veteran’s Affairs HD unit. Bruit recordings were made using a digital stethoscope (Littman 3200) 1-4 times per month just prior to HD. In each evaluation, 10-s bruit recordings were obtained at 5-9 serial locations, 3 cm apart along an upper extremity arteriovenous fistula (AVF) or graft (AVG). Recordings were made from the arterial anastomosis to beyond the cannulation zone (for AVF) or to the venous anastomosis (for AVG).
True positive detection of access dysfunction was counted when the algorithm value exceeded a fixed threshold at one or more recording sites and when clinical evidence of access dysfunction occurred within 30 days after the recording date. Evidence of access dysfunction included prolonged post-HD bleeding or access rupture, access pain, radiologic intervention, non-functional access, or unexplained KT/V below 1.3.
Results
3,441 bruits were recorded for 24 Veterans (age 67 years, HD vintage 51 months, 96% male, 83% Black) over an average of 13 months per patient. 71 instances of access dysfunction were observed in 18 patients (median 3 per patient). Use of the algorithm showed 89% sensitivity, 83% specificity, and 92% accuracy at predicting access dysfunction up to 30 days in advance.
Conclusion
Prospective use of serial bruit recordings and algorithmic analysis positively identified at-risk patients up to 30 days before clinical evidence of access dysfunction. This technique may help prevent emergency interventions or hospital admissions for access dysfunction.