Prediction of risk factors and electrocardiographic changes in chronic kidney disease patients.

OBJECTIVES: Chronic kidney disease (CKD) is a global health issue with significant morbidity and mortality, particularly due to cardiovascular events. Early identification and management of risk factors are crucial to prevent CKD progression and complications. CKD is heterogeneous with diverse etiologies and presentations, generalizing across populations challenging. This study aims to develop accurate predictive models for cardiovascular events in CKD patients. METHODS: Biosensors capture key parameters, including SpO2 (Oxygen saturation), PR (Pulse rate), Pi (Perfusion index), RRp (Respiration rate), and PVi (Pleth variability index), enabling comprehensive evaluation of physiological dynamics in CKD patients. Stacked Auto-Encoders (SAEs) are applied for diagnostics. Genetic risk score (GRS) and nongenetic risk score (NGRS) models are developed using natural logarithms of odds ratios (OR) of risk factors. RESULTS: The models integrate properties of each factor with weighted contributions to create predictive models for CKD. A novel machine learning technique incorporates automatic machine learning (AutoML). CONCLUSIONS: The models integrate properties of each factor with weighted contributions to create predictive models for CKD. A novel machine learning technique incorporates automatic machine learning (AutoML).