Early Risk Stratification of Chronic Kidney Disease and Dialysis Requirement Using Clinical Data

Abstract

Chronic Kidney Disease (CKD) is a long-term condition which is usually not diagnosed until there is significant renal damage, which can lead to dialysis and other serious complications. Thus, there is a need for early identification of high risk patients for better clinical outcome and disease burden reduction. This study introduces a machine learning-based approach for early risk-stratification of CKD and predicting the dialysis requirement from routine clinical data. The framework consists of two parts: demographic data, laboratory biomarkers and comorbidity data, followed by data preprocessing, exploratory analysis, development of predictive models and comparison of their performance. Several supervised machine learning models were developed and evaluated with the common performance metrics such as accuracy, precision, recall, F1-score and ROC-AUC. The most reliable predictive model was identified to differentiate the cases of CKD and to estimate dialysis requirement, when compared with the other models. The study results showed that machine learning approaches could be beneficial for early clinical decision making, as they can provide timely assessment of risk, better patient management and enable proactive intervention in patients at risk of CKD progression.