Errors, Uncertainties, and Ambiguities in Wearable Health Monitoring Systems

Healthcare and diagnostics have recently undergone a paradigm shift with a greater focus on remote health monitoring through wearable technologies. Advances in miniaturized electronics, wireless communications, and big data analytics are all converging in this space to take health monitoring out of the clinic and into the home. However, while the exponential increase in wearable technologies is driving excitement in this field, such technologies have found limited success in clinical integration. While consumers might find a plethora of smart gadgets from watches to rings that can track activity and heart rate, little of this information is getting utilized by clinicians. This is in part due to the lack of transparency and perceived inaccuracy of wearable monitoring systems. We will address this limitation by characterizing errors in measured real-world health signals, accounting for errors in user-device interactions, and capturing uncertainties and ambiguities in decisions that will allow wearable sensors and underlying machine learning algorithms to provide more contextual and nuanced information for clinicians. This will help clinicians decide when and how to apply wearable data to clinical decisions.