Purpose: To develop machine learning models using national registry data for predicting factors associated with favourable outcomes in kidney transplantation using supra-marginal (DRI > 1.5) deceased donor kidneys.
Methods: A retrospective cohort study was conducted using UK Transplant Registry data from 2000–2019, administered by NHS Blood and Transplant. The study included 6,254 adult recipients of first kidney-alone transplants from very supra-marginal deceased donors (DRI≥1.5). The main outcome measures were death-censored graft failure and patient mortality. Comprehensive recipient, donor, and transplant characteristics were used as predictors. Bayesian neural networks, gradient boosting machines, random forest, and SMOTE-balanced bagging classifiers were tuned using Bayesian optimisation. Cox regression, competing risk analysis, and calibration plots were also employed.
Results: The overall 5-year graft survival was 81%. The random forest model demonstrated excellent predictive performance for graft failure (AUC 0.88, 95% CI 0.87–0.89; RMSE 0.29). Key determinants of poor survival included recipient age >75 years (SHR 1.02, 95% CI 1.01–1.03), recipient BMI >30 (SHR 1.04, 95% CI 1.02-1.07), HLA mismatches >4 (SHR 1.09, 95% CI 1.01–1.17), donor creatinine >120 mmol/L (SHR 1.002, 95% CI 1.001–1.003), and rejection within 3 months (SHR 1.56, 95% CI 1.32-1.85). Prolonged cold ischemia time >14 hours (SHR 1.01, 95% CI 1.007-1.015) was also detrimental.
Conclusions: Supra-marginal deceased donor kidneys can achieve excellent 5-year outcomes with careful recipient selection. Machine learning models accurately predicted factors associated with successful transplantation, providing valuable insights for clinical decision-making and allocation strategies. These findings can help optimise the utilisation of supra-marginal deceased donor kidneys and improve overall transplantation outcomes.