Wearable devices in cardiology

Abstract

Background. Wearable technology such as smartwatches has many potential uses in personal disease screening, health monitoring, and telemedicine. Smartwatch-derived biosignals can be used to measure individual health parameters (such as heart rate) of the user at that time. However, in many case, it is not known whether any combination of smartwatch-derived biosignals can be used in the context of disease screening or health outcome prediction. Objectives. To identify predictive variables derived from smartwatch data for predicting the presence of coronary artery disease and six-minute walk distance. Methods: Subjects with coronary computed tomography angiogram record were recruited to perform a modified six-minute walk test (6MWT) with continuous smartwatch heart rate monitoring during and after the test, as well as smartwatch electrocardiogram in the three defined time points – just before 6MWT, just after 6MWT and after recovery from exercise. Least absolute shrinkage and selection operator (LASSO) regression and random forest classifier or regressor were used to construct predictive models for the presence of significant coronary artery disease and six-minute walk distance. Results: None of the watch-derived variables was found to be predictive of the presence of significant coronary artery disease. For six-minute walk distance prediction, we found that chronotropic response and peak heart rate reached during 6MWT were significantly predictive of six-minute walk distance under the LASSO regression model and random forest regressor. Conclusions: The identified significant predictors chronotropic response and peak heart rate were consistent with the literature findings. Given that these two parameters can potentially be obtained by a smartwatch without doing a 6MWT, our findings open up new possibilities of using smartwatch-derived data to predict 6MWD, which itself is a predictor of a variety of cardiopulmonary health status.