About PREDICT-AVF
PREDICT-AVF brings together prediction tools for radiocephalic arteriovenous fistula (RC-AVF) outcomes. They were developed from pooled patient-level data from the PATENCY-1 and PATENCY-2 randomized trials (2014–2019; 914 patients at 39 centers in the United States and Canada).
How the calculators work
Every prediction is computed in your browser; nothing you enter is sent anywhere. The published models were exported once, as just the parameters a prediction needs, and are never retrained or modified:
| Calculator | Model | Exported parameters |
|---|---|---|
| Unassisted use | Logistic regression with lasso selection | Coefficients, plus the mean and SD used to standardize flow and diameter |
| Recurrent interventions: early and late ultrasound | Cox proportional hazards models for recurrent events | Coefficients, centering means and the baseline cumulative hazard |
| Recurrent interventions: baseline | Random survival forest (500 trees, randomForestSRC 3.2.1) | Each tree’s split rules, and each terminal node’s cumulative hazard on a 10-day grid |
Automated tests check that this website reproduces the original R models to within 10−9 on hundreds of input combinations. The original model objects are never published, because they contain patient-level trial data.
The interventions calculators show the expected cumulative number of interventions by each day of the first year, as estimated by each published model.
How to cite
- Heindel P, Dey T, Feliz JD, et al. Predicting radiocephalic arteriovenous fistula success with machine learning. npj Digit Med. 2022;5:160. doi:10.1038/s41746-022-00710-w
- Heindel P, Dey T, Fitzgibbon JJ, et al. Predicting recurrent interventions after radiocephalic arteriovenous fistula creation with machine learning and the PREDICT-AVF web app. J Vasc Access. 2025;26(1):202–210. doi:10.1177/11297298231203356
- Fitzgibbon JJ, Ruan M, Heindel P, et al. Predicting long-term patency of radiocephalic arteriovenous fistulas with machine learning and the PREDICT-AVF web app. Sci Rep. 2025;15:19203. doi:10.1038/s41598-025-04310-y
Disclaimer
For educational and research use only. PREDICT-AVF provides statistical estimates from published research models. Nothing on this website is medical advice. It should not be used to make a diagnosis or to replace or overrule a qualified health care provider’s judgment. We strongly encourage users to consult a qualified health care professional for answers to personal questions.
Intended use and limitations.
- The models were developed and internally validated using data from adults with advanced chronic kidney disease who underwent new radiocephalic AVF creation in the PATENCY-1 and PATENCY-2 trials (2014–2019) at centers in the United States and Canada. Estimates may not apply to patients, access types (such as upper-arm fistulas or grafts) or practice settings that differ from that population.
- Predictions are average estimates for patients with similar characteristics. They are not certainties for any individual patient, and the uncertainty around them is not shown.
- Inputs are limited to the ranges used in the published calculators. Model performance and limitations are described in the original publications.
- This website is not a medical device and has not been reviewed, cleared or approved by the U.S. Food and Drug Administration or any other regulatory authority.
Privacy. Predictions are computed entirely in your web browser. The values you enter are not sent to or stored by this website. Linked external sites, such as the long-term patency calculator, are governed by their own terms and policies.
No warranty; limitation of liability. This website and its content are provided “as is”, without warranties of any kind, express or implied, including warranties of accuracy, completeness or fitness for a particular purpose. The authors and their affiliated institutions are not liable or responsible for any advice, course of treatment, diagnosis or any other information, services or product obtained through this site. Using this website does not create a physician–patient relationship, and its content is not a recommendation or endorsement by any doctor or health care provider for any individual patient.
Copyright. Material on this website is protected by copyright law. Unless otherwise stated, users may use information from this site for personal, non-commercial use only, provided they identify the source of the material and include a statement that the materials are protected by copyright law. Permission to reprint or otherwise reproduce any document or program in whole or in part is prohibited unless prior written consent is obtained from the copyright owner.