Design and implementation of an intelligent clinical decision support system for diagnosis and prediction of chronic kidney disease

Introduction: Chronic kidney disease (CKD) is one of the most important public health concerns worldwide. The steady increase in the number of people with End-stage renal disease (ESRD) needing a kidney transplant to survive and incur high costs, highlights early diagnosis and treatment of the disease. This study aimed to design a Clinical Decision Support System (CDSS) for diagnosing CKD and predicting the advanced stage to achieve better management and treatment of the disease. Materials and Methods: In this retrospective and developmental study, we studied the records of 600 suspected CKD cases with 22 variables referred to ShahidLabbafinejad Hospital in Tehran from 2019 to 2020. Data mining algorithms such as Naïve Bayesian, Random Forest, Multilayer Perceptron neural network, and J-48 decision tree were developed based on extracted variables. Then the recital of selected models was compared by some performance indices and 10-fold cross-validation. Finally, the most appropriate prediction model in terms of performance was implemented using the C # programming language. Results: Random Forest classification algorithm with an accuracy of 99.8% and 88.66%, specificity of 100% and 93.8%, the sensitivity of 99.75% and 88.7%, f-measure of 99.8% and 88.7%, kappa score of 99.4% and 82.73%, and ROC of 100% and 90.52% was identified as the best data mining model for CKD diagnosis and prediction respectively. Conclusion: The developed MC-DMK system based random Forestcan be used practically in clinical settings.