O031 Machine learning models in renal transplantation: a systematic review and meta-analysis of predictive performance in graft outcomes. (22nd July 2022)

• Record Type: Journal Article
• Authors: Ravindhran, B; Chandak, P; Haroon, U; Kundalia, K; Ameen, T; Woochan, H; Zakri, RH
• Is Part Of: British journal of surgery. Volume 109(2022)Supplement 4
• Publication Date: 2022-07-22
• Subjects: Surgery -- Periodicals; 617.005
• DOI: 10.1093/bjs/znac242.031
• Languages: English
• ISSNs: 0007-1323

Abstract

Abstract: Introduction: Kidney transplantation (KT) is currently the renal replacement therapy of choice for most patients with end-stage kidney disease. Despite many advancements, the variations in outcome and frequent occurrence of graft failure continue to pose important clinical and research challenges. The aim of this study was to carry out a systematic review of the current application of Machine Learning (ML) models in KT and perform a meta-analysis of these models in the prediction of graft outcomes. Methods: This review was registered with the PROSPERO database (CRD42021247469) and all peer reviewed and preprint original articles that reported the sensitivity and specificity of AI-based models were included in the meta-analysis. Data were analysed using MetaDTA, , an interactive online software for meta-analysis of diagnostic studies. The diagnostic performance of the meta-analysis was assessed by a summary receiver operating characteristics (sROC) plot. Results: 38 studies met the inclusion criteria for the review and 12 studies met the inclusion criteria for the meta-analysis. The most common models used were artificial neural networks, decision trees and Bayesian belief networks. Seven studies compared the predictive performance of ML models with traditional regression methods. The summary sensitivity and specificity of ML-based models were 0.84 (95% CI, 0.72–0.91) and 0.68 (95% CI, 0.57–0.77), respectively. The area under the SROC for all the available evidence was 0.83. The Diagnostic Odds Ratio of ML models was 11.19 (95% CI 6.66–18.75). Conclusion: Our study shows that ML models can accurately predict outcomes following KT by the integration of the vast amounts of available clinical data. Take-home message: This study confirms the superior ability of ML Models in handling complex relationships between large datasets, features and outcomes, which has definitely led to improved precision and accuracy of outcomes.