A novel machine learning approach generates personalized estimates of graft and patient survival to better inform older kidney transplant candidates.

Authors

Pierre-Luc Boivin, Jonathan Jalbert, Lévis Thériault, Yue Qi, Anastasiya Olek-Basanets, Héloise Cardinal

Published in

Frontiers in immunology. Volume 17. Pages 1881830. Epub Jul 08, 2026.

Abstract

An information most relevant to patients when deciding to accept a kidney offer is an estimate of its potential longevity. Using a novel machine learning approach, our aim was to develop a model that can output personalized curves estimating kidney graft and patient longevity to support clinical decision-making.
We performed a retrospective cohort study in recipients of a first deceased donor kidney transplant aged ≥60 years between 2000 and 2020, using the United Network for Organ Sharing dataset. Outcomes were overall graft survival and patient survival. Independent variables included the Kidney Donor Risk Index (KDRI) and recipient-related variables. Random survival forest models were trained on 70% and validated on 30% of the dataset.
The study cohort included 57, 280 patients, amongst whom 27, 244 (48%) experienced graft loss, while 25, 210 (44%) died during a median follow-up of 4.3 years. The most important variables in tree development for graft survival were recipient age, recipient diabetes, KDRI, recipient ethnicity, and time on dialysis. For patient survival, these variables were recipient age, recipient diabetes, time on dialysis, recipient cause of chronic kidney disease, and recipient HCV status. The difference in expected graft survival between low and high KDRI kidneys decreased when recipient characteristics, in particular age, were accounted for. KDRI had a negligible impact on mortality.
While implementation studies are needed, the individualized curves provided by our novel machine learning approach have the potential to support shared decision-making to a greater extent than the KDRI alone.

PMID:
42488683
Bibliographic data and abstract were imported from PubMed on 23 Jul 2026.

Read full publication at:
Please sign in to see all details.

Stats

1-terrible, 9-excellent. How would you rate this publication? Sign in in to submit your rating.

Post a comment

You need to be signed in to post comments. You can sign in here.

Comments

There are no comments yet.