ASN's Mission

ASN leads the fight to prevent, treat, and cure kidney diseases throughout the world by educating health professionals and scientists, advancing research and innovation, communicating new knowledge, and advocating for the highest quality care for patients.

learn more

Contact ASN

1401 H St, NW, Ste 900, Washington, DC 20005


The Latest on Twitter

Kidney Week

Abstract: TH-PO1117

Predicting Health Outcomes for Elderly Renal Transplant Recipients with Machine Learning

Session Information

Category: Transplantation

  • 1902 Transplantation: Clinical


  • Fu, Rui, University of Toronto, Toronto, Ontario, Canada
  • Mitsakakis, Nicholas, Toronto General Hospital Research Institute, THETA Collaborative, Toronto, Ontario, Canada
  • Luo, Kai, University of Toronto, Toronto, Ontario, Canada
  • Coyte, Peter C., University of Toronto, Toronto, Ontario, Canada

Application of machine learning to nephrology research has been scarce. In this study, we demonstrated the use of classification algorithms in predicting all-cause death at three-year among elderly deceased-donor renal transplant recipients.


This is a retrospective, population-based, cohort study of all cases of deceased-donor renal transplants performed in Ontario, Canada from March 31, 2002 to April 1, 2013. Recipients aged over 70 years were followed up until death or to April 1, 2016. Bootstrap-aggregating classification tree and K-Nearest Neighbors (KNN) were used to train a predictive model for death at three-year post-transplant. Patient-level attributes at the time of transplantation, including demographic characteristics, lab results, transplant information, comorbidities, and pre-transplant health care utilization, were examined as potential determinants of post-transplant death. A ratio of 3:2 was used to construct training and testing sets. Synthetic Minority Oversampling Technique was applied to generate artificial positive cases (death) and under-sample negative cases (alive) in the training set to reduce bias. Models were trained and tuned using ten-fold cross-validation on the training set and tested on the specificity and sensitivity of prediction using the testing set.


Among 275 elderly transplant recipients, the majority (n=271, 98.5%) were transplanted at 71-80 years and four (1.5%) were older than 80 years. Death occurred in 52 (18.9%) cases at three-year post-transplant. Before sampling, classification tree and KNN had test sensitivity of 0.11 (95% confidence interval [CI], 0.01-0.33) and 0.07 (95% CI, 0-0.18), respectively, while both achieving 0.95 (95% CI, 0.88-0.98) specificity. After sampling, classification tree and KNN achieved test sensitivity of 0.21 (95% CI, 0.06-0.46) and 0.26 (95% CI, 0.03-0.50), respectively, as well as test specificity of 0.89 (95% CI, 0.81-0.95) and 0.84 (95% CI, 0.74-0.90), respectively.


Our findings add to the growing body of knowledge aimed at improving the performance of risk calculators (e.g., iChoose Kidney) that help patients and families to make informed decisions in renal care. Furthermore, our study confirmed the strength of machine learning techniques in population-based nephrology research despite our limited sample size and the rarity of the outcomes assessed.