Abstract: FR-PO489
Comparative Prognostic Accuracy of Vascular Access Flow and Artificial Intelligence (AI)-Based Arteriovenous Fistula (AVF) Failure Risk Score
Session Information
- Dialysis: Vascular Access
November 03, 2023 | Location: Exhibit Hall, Pennsylvania Convention Center
Abstract Time: 10:00 AM - 12:00 PM
Category: Dialysis
- 803 Dialysis: Vascular Access
Authors
- Bellocchio, Francesco, Fresenius Medical Care Italia SpA, Palazzo Pignano, Lombardia, Italy
- Garbelli, Mario, Fresenius Medical Care Italia SpA, Palazzo Pignano, Lombardia, Italy
- Stuard, Stefano, Fresenius Medical Care Italia SpA, Palazzo Pignano, Lombardia, Italy
- Nikam, Milind, Fresenius Medical Care Asia Pacific Ltd, Hong Kong, Hong Kong
- Usvyat, Len A., Fresenius Medical Care, Waltham, Massachusetts, United States
- Neri, Luca, Fresenius Medical Care Italia SpA, Palazzo Pignano, Lombardia, Italy
Background
Technical surveillance based on vascular access blood flow (Qa) measurement it is time consuming, requires training, and specific equipment. We developed a risk score which accurately and reproducibly predicts AVF failure within 3 months. The risk score may offer a cheap and automated alternative to Qa measurement. We compared the prognostic accuracy of Qa measurement and AVF failure risk score.
Methods
We included 8,969 AVF Qa measurements from 1,116 dialysis patients in 4 European countries. We used the AVF failure risk score to classify patients into 3 classes (low:<10%, moderate: 10%-60%, high:>60%) representing AVF failure risk within 3 months. Qa was measured with the thermodilution method and classified in very low (<525 ml/min), low (525-925 ml/min), normal (>925 ml/min). We computed the incidence of observed AVF failures by Qa levels and AVF risk classes to assess the overlap between the two metrics.
Results
A large share of patients was classified in the low-risk class by the AVF failure risk model (44%) whereas a tiny proportion were considered high risk (0.5%). We observed 907 AVF failures (10.1 %). There was a weak correlation between Qa and AVF failure risk score (p>0.05; Figure 1). The AVF failure risk score showed stronger association with AVF prognosis over a 3-month horizon compared to Qa measurement (p<0.05; Table 1).
Conclusion
Our study suggests that AVF failure risk may represents a valuable, cheap, automated alternative to Qa measurement for AVF technical surveillance.
| AVF Failure Risk | |||
| Qa | low: <10 % | High: 10-60 % | Very high: >60 % |
| very low: <525 ml/min | 4% (58) | 21% (373) | 83% (20) |
| low: 525 - 925 ml/min | 5% (46) | 17% (245) | 67% (6) |
| normal: >925 ml/min | 2% (31) | 8% (127) | 33% (1) |
Incidence of AVF failures (“number of failures”/”number of Qa measurements” × 100) stratified by Qa and AVF failure risk levels.
Percentage of Qa measures for each AVF failure risk class.