An optimized machine learning framework for predicting intradialytic hypotension using indexes of chronic kidney disease-mineral and bone disorders

Abstract

Intradialytic hypotension (IDH) is the most common acute complication in hemodialysis (HD) sessions and is associated with increased morbidity and mortality in HD patients. To prevent the episode of IDH, it is critical to predict its occurrence. Chronic kidney disease-mineral and bone disorders (CKD-MBD) induce cardiac and vascular calcification, which impairs the compensatory mechanisms of blood pressure during HD. In this study, we proposed a feature selection framework called BSWEGWO_KELM to analyze 1940 records from 178 HD patients, which was based on an enhanced grey wolf optimization (GWO) algorithm and the kernel extreme learning machine (KELM). Then, global optimization experiments, together with feature selection experiments on public data sets and HD dataset, were performed to verify the effectiveness of the BSWEGWO_KELM method. The experimental results showed that the established BSWEGWO_KELM had the capability of screening out the key indicators such as dialysis vintage, mean arterial pressure (MAP), alkaline phosphatase (ALP), and intact parathyroid hormone (iPTH). Consequently, BSWEGWO_KELM can be applied as a practical and accurate method to predict IDH.