SP3.7 Regional Outcomes of Paediatric Renal Transplantation in the UK. A Machine Learning Analysis: An Epidemiology In Transplant Study Group (ETrSG) Initiative

Abstract

Abstract Purpose <25%of UK children commencing RRT receive a pre-emptive transplant in the UK. Paediatric patients on the waiting list have increased over the last 5 years. This study assesses the paediatric renal transplant rates in the UK since the year 2000 and the regional graft outcomes of paediatric renal transplant in the UK and determine the predictors of survival at 1, 3, and 5 years. Methods We harmonized the NHSBT-Data to shortlist the paediatric renal transplants performed in the UK from the year 2000-2019. The Predictors of Transplant outcomes were evaluated by five classifiers including logistic regression, SVM, random forest, K-Nearest neighbour matching, and adaptive boosting. Random Forest Model had the best performance validated by RMSE. Rattle R/ JASP was used to derive Variables of importance. Survival outcomes and predictors data mined from MLA were further mined with Cox-Regression. Results A total of 2610 paediatric transplants were reported to NHSBT (the year 2000-2019). London region did the maximum transplants (N= 854) followed by Northern & Yorkshire (N=347) and North-West region (N= 340). High-Volume center doing living donors had considerably better survival compared to others. On Regression-analysis, the better graft survival was correlated to recipient age group >8 years (p <0.001), CIT <12 hr (p= 0.02), Transplant year -2010 onwards (p <0.01), ethnically matched donor-recipient (p= 0.01), HLA NHSBT group 1 and 2 (p=0.04) and high volume centers (> 15 paediatric transplants/annually). Conclusions Superior outcomes were reported from high-volume centers and transplants done after the year 2010. Immunologically well-matched kidneys and low Cold-Ischemia-time (CIT) predict better survival.