Prediction of Personal Cardiovascular Risk using Machine Learning for Smartphone Applications

Cardiovascular disease is a major global health burden. Machine learning may be used on big data from national surveys to develop models that predict various cardiovascular risk factors. We used machine learning to evaluate and compare generalized linear, stochastic gradient boosting, random forest, and neural network model performance on predicting cardiovascular risk factors, such as hypertension, body mass index, and total cholesterol level on 5,992 adults in the US National Health and Nutrition Examination Survey (NHANES). The highest accuracy of 73% was found for predicting hypertension status, using a random forest model on a combination of demographic, diet and physical activity behavior, and mental state predictor variables. We demonstrate the use of the machine learning model through the development of an Application Programming Interface (API), which is called by a mHealth smartphone app and web interface. This work has promise for future intervention studies that assess how users respond to feedback on cardiovascular risk predictions, and which could evaluate improvements in costeffective cardiovascular healthcare.