Epidemiology of infections in kidney transplant recipients - data miner's approach.

Abstract

Infections remain a frequent complication following organ transplantation. Agents present within the general population remain common in recurrent infections among renal transplant recipients. Data mining methodology has become a promising source of information about patterns in the organ transplant recipient population. The aim of the study was to use data mining to describe the factors influencing single and recurrent infections in kidney transplant recipients. A group of 159 recipients who underwent kidney transplantation between 2005 and 2008 was analysed. RapidMiner and Statistica softwares were used to create decision tree models based on CART Quinlan and C&RT algorithms. There were 171 microbiologically confirmed episodes among 67 recipients (41%), and 191 separate species isolations were performed. Over 50% of the infected patients underwent two or more infectious episodes. Two classification decision tree models were created. The following features were enabled to differentiate the groups with single or recurrent infections: the duration of cold ischaemia, the post-transplant hospitalization period, the cause of chronic kidney disease and pathogens. The post-transplant hospitalization period and the length of cold ischaemia appear to be the principal parameters differentiating the subpopulations analysed. These coexisting factors, connected with recurrent infections in kidney transplant recipients, resemble a network which requires an advanced analysis to support the traditional statistics.