Machine Learning Model to Predict Graft Rejection After Kidney Transplantation.

BACKGROUND: There are few predictive studies about early posttransplant outcomes taking into account baseline and posttransplant variables. The objective of this study was to create a predictive model for 30-day graft rejection using machine learning techniques. METHODS: Retrospective study with 1255 patients undergoing transplant from living and deceased donors at a tertiary health service in Brazil. Recipient, donor, transplantation, and postoperative period data were collected from physical and electronic records. We split the data into derivation (training) and validation (test) datasets. Five supervised machine learning algorithms were developed with this subset of variables in the training set: Simple Logistic Regression, Lasso, Multilayer Perceptron, XGBoost, and Light GBM. RESULTS: There were 147 (12.48%) cases of graft rejection within 30 days of transplantation. The best model was XGBoost (accuracy, 0.839; receiver operating characteristic area under the curve, 0.715; precision, 0.900). The model showed that deceased donor transplantation, glomerulopathy as an underlying disease, and donor's use of vasoactive drugs had more than 20% importance as rejection risk factors. The variables with the greatest predictive values were thymoglobulin induction and delayed graft function. CONCLUSIONS: We fitted a machine learning model to predict 30-day graft rejection after kidney transplantation that reaches a higher accuracy and precision. Machine learning models could contribute to predicting kidney survival using nontraditional approaches.

The future is coming: promising perspectives regarding the use of machine learning in renal transplantation / O futuro está chegando: perspectivas promissoras sobre o uso de machine learning no transplante renal

Abstract Introduction: The prediction of post transplantation outcomes is clinically important and involves several problems. The current prediction models based on standard statistics are very complex, difficult to validate and do not provide accurate prediction. Machine learning, a statistical technique that allows the computer to make future predictions using previous experiences, is beginning to be used in order to solve these issues. In the field of kidney transplantation, computational forecasting use has been reported in prediction of chronic allograft rejection, delayed graft function, and graft survival. This paper describes machine learning principles and steps to make a prediction and performs a brief analysis of the most recent applications of its application in literature. Discussion: There is compelling evidence that machine learning approaches based on donor and recipient data are better in providing improved prognosis of graft outcomes than traditional analysis. The immediate expectations that emerge from this new prediction modelling technique are that it will generate better clinical decisions based on dynamic and local practice data and optimize organ allocation as well as post transplantation care management. Despite the promising results, there is no substantial number of studies yet to determine feasibility of its application in a clinical setting. Conclusion: The way we deal with storage data in electronic health records will radically change in the coming years and machine learning will be part of clinical daily routine, whether to predict clinical outcomes or suggest diagnosis based on institutional experience.

Resumo Introdução: A predição de resultados pós-transplante é clinicamente importante e envolve vários problemas. Os atuais modelos de previsão baseados em padrões estatísticos são muito complexos, difíceis de validar e não fornecem previsões precisas. Machine Learning, é uma técnica estatística que permite que o computador faça previsões futuras usando experiências anteriores, está começando a ser usada para resolver essas questões. No campo do transplante renal, o uso da previsão computacional foi relatado na predição de rejeição crônica de aloenxerto, função tardia do enxerto e sobrevida do enxerto. Este artigo descreve os princípios e etapas de machine learning para fazer uma previsão e realiza uma breve análise das aplicações mais recentes de seu uso na literatura. Discussão: Existem evidências convincentes de que as abordagens de machine learning baseadas nos dados do doador e do receptor são melhores para proporcionar melhor prognóstico dos resultados do enxerto do que a análise tradicional. As expectativas imediatas que emergem dessa nova técnica de modelagem de previsão são que ela gerará melhores decisões clínicas baseadas em dados de práticas dinâmicas e locais e aperfeiçoará a alocação de órgãos, bem como o gerenciamento de cuidados pós-transplante. Apesar dos resultados promissores, ainda não há um número substancial de estudos para determinar a viabilidade de sua aplicação em um cenário clínico. Conclusão: A forma como lidamos com dados de armazenamento em prontuários eletrônicos de saúde mudará radicalmente nos próximos anos e a machine learning fará parte da rotina clínica diária, seja para prever resultados clínicos ou sugerir um diagnóstico baseado na experiência institucional.

The future is coming: promising perspectives regarding the use of machine learning in renal transplantation.

INTRODUCTION: The prediction of post transplantation outcomes is clinically important and involves several problems. The current prediction models based on standard statistics are very complex, difficult to validate and do not provide accurate prediction. Machine learning, a statistical technique that allows the computer to make future predictions using previous experiences, is beginning to be used in order to solve these issues. In the field of kidney transplantation, computational forecasting use has been reported in prediction of chronic allograft rejection, delayed graft function, and graft survival. This paper describes machine learning principles and steps to make a prediction and performs a brief analysis of the most recent applications of its application in literature. DISCUSSION: There is compelling evidence that machine learning approaches based on donor and recipient data are better in providing improved prognosis of graft outcomes than traditional analysis. The immediate expectations that emerge from this new prediction modelling technique are that it will generate better clinical decisions based on dynamic and local practice data and optimize organ allocation as well as post transplantation care management. Despite the promising results, there is no substantial number of studies yet to determine feasibility of its application in a clinical setting. CONCLUSION: The way we deal with storage data in electronic health records will radically change in the coming years and machine learning will be part of clinical daily routine, whether to predict clinical outcomes or suggest diagnosis based on institutional experience.