Chronic disease risk monitoring based on an innovative predictive modelling framework

Smart watches / Fitness bands aim to capture the different vital signs, such as heart rate, energy expenditure and sleep patterns of the users which can be immensely useful for monitoring and prediction of the overall wellbeing of the user. Further, detection of early signs of the diseases at geospatial level can help in promoting evidence based health policies and proper disease management strategies to be formulated beforehand. In this paper, a novel unified predictive modelling framework is proposed, which can perform in both static and low velocity, big data clinical settings from EHRs, as well as high velocity, dynamic, streaming big data settings captured from personal wearable devices, such as smart watches and fitness bands. In this paper, we report the results of the platform implementation of the framework for static/low velocity settings from the electronic health records and hospital databases, with the experimental validation of the proposed framework, for two publicly available cardiovascular disease datasets (the NHANES dataset, and the Framingham Heart Study CHS dataset), showing promising outcomes, in terms of performance of different predictive modelling algorithms for prediction of disease status.