Development a clinical decision support system for the diagnosis, prediction and management of chronic renal failure

Background:Chronic kidney disease is one of the most important public health concerns around the world. It is defined as a general term for heterogeneity disorders that affects the structure and function of the kidneys. The purpose of this paper is to develop a clinical decision support system for the diagnosis, prediction and management of chronic renal failure.Material and Methods:This is an applied-development study. In the present study, the records of inpatient and outpatient due to renal disorders in kidney transplantation and dialysis centers of Tehran were investigated from 2015-11-07 to 2016-05-09, then clinical and laboratory data of patients were extracted while respecting the principles of confidentiality and privacy. 635 medical records were investigated, but data from 400 people were extracted due to missing data in the files. To develop the system, artificial neural networks, decision tree (j48 algorithm and random forest algorithm) and Naive Bayes algorithm were used in the context of MATLAB and WEKA programs on data extracted from medical records. In order to evaluate the algorithms, some evaluation criteria such as specificity, sensitivity, accuracy, Kappa and F-measure were used. These evaluation criteria are based on confusion matrix. Finally, the C # programming language was used for the final implementation of the system.Results:The results showed that the random forest algorithm in the first phase, which is the diagnosis of chronic renal failure, had better results than other algorithm in specificity (98.1%), sensitivity (99.1%), accuracy (99.1%), F-measure (99%) and kappa criteria (99%). The results of the second phase, which are related to the prediction of the level of renal failure, also showed that the random forest algorithm had the best results in specificity (99.1%), sensitivity (84.1%), accuracy (84.1%), F-measure (83.1%) and kappa criteria (79.1%).Conclusion:Based on the results obtained from the implementation of data mining techniques and that the proposed system is a system for diagnosis and prediction and management, achieving a sensitivity of 99% in the first phase means that almost all chronic renal failure patients are correctly diagnosed. It should be noted that the sensitivity of 84% in the second phase indicates that system is able to predict the level of renal failure at high percentage. Therefore, the proposed system can be used and evaluated in practice as a clinical decision-making system

Keywords:

Clinical Decision Support System , Chronic Kidney Failure , Data Mining Methods , Random Forrest Algorithm