Determining Association between Fatal Heart Failure and Chronic Kidney Disease: A Machine Learning Approach

Cardiorenal syndrome is a term that refers to a spectrum of heart and kidney disorders that demonstrate how a condition affecting one of the organs impairs the other. In this work, the association between two of the most persistent conditions: chronic kidney disease (CKD) and fatal heart failure (HF), was investigated using machine learning approaches. The research visualizes dependencies and identifies patterns in a subtype of cardiorenal syndrome with the primary goal of determining the risk of fatal heart failure in individuals with chronic kidney disease using state-of-the-art techniques. Firstly, heart failure and chronic kidney disease datasets were used for disease prediction with five classifiers: Random Forest (RF), XGBoost, CatBoost, Logistic Regression, and Support Vector Machine. The prediction accuracy for heart failure was between 70%-76%, and CKD was between 97%-99%. The top predicting models were random forest, XGBoost, and CatBoost classifiers. In the second stage, the feature importance scores of the best predictors were analyzed to gauge the relationship between the conditions. Numerous features of HF and CKD that were common and obtained high importance scores for the top classifiers were age, serum creatinine, serum sodium, and diabetes mellitus. Finally, a variety of visualization techniques were employed to acquire insight into the relevance of different features, resulting in medically sound findings. The analysis of the physiological attributes and their importance with the help of machine learning was aided in successfully reaffirming the medical findings of a crucial subtype of cardiorenal syndrome, associating fatal heart failure with chronic kidney disease.