Healthcare predictive analytics for disease progression: a longitudinal data fusion approach

Abstract

Healthcare predictive analytics using electronic health records (EHR) offers a promising direction to address the challenging tasks of health assessment. It is highly important to precisely predict the potential disease progression based on the knowledge in the EHR data for chronic disease care. In this paper, we utilize a novel longitudinal data fusion approach to model the disease progression for chronic disease care. Different from the conventional method using only initial or static clinical data to model the disease progression for current time prediction, we design a temporal regularization term to maintain the temporal successivity of data from different time points and simultaneously analyze data from data source level and feature level based on a sparse regularization regression approach. We examine our approach through extensive experiments on the medical data provided by the Alzheimer’s Disease Neuroimaging Initiative (ADNI). The results show that the proposed approach is more useful to simulate and predict the disease progression compared with the existing methods.