Multiscale Regression for Chronic Kidney Disease via Age-Curvature and Electrolytes

This study investigates chronic kidney disease (CKD) through multiscale regression techniques, incorporating both binary logistic regression and multivariable linear regression to identify clinically relevant predictors. Using a dataset from the UC Irvine Machine Learning Repository, the analysis highlights the significant role of sodium and potassium levels—termed electrolyte signatures—in predicting CKD status. Additionally, a nonlinear age component is introduced to model its accelerating effect on blood pressure, a key comorbidity of CKD. Model comparison tests confirm the added predictive value of these features. The findings contribute to clinically interpretable models that support early diagnosis and personalized disease management.