Standing on the Threshold of Wearables and Research
Recently, the All of Us program announced the launch of its Fitbit Bring-Your-Own-Device (BYOD) project. Funded through the National Institutes of Health (NIH), All of Us is a key element of the Precision Medicine Initiative (PMI) launched in 2016 and seeks to build a voluntary, national research participant group of one million individuals. To date, research participants have been asked to provide information through surveys, electronic health records (EHRs), and biosamples including blood and saliva.
As researchers with experience evaluating and using consumer wearable device data in studies, we are excited that All of Us is moving into the collection of data through digital technologies. RTI has been pioneering the use of wearables in clinical and public health research for several years. In addition to conducting numerous studies that incorporate these devices, we have contributed to the foundational literature on wearable device performance, data quality, research methods, and data management. The NIH’s decision to use wearable device data in the All of Us program raises the profile of the use of these data in research. However, we haven’t quite reached a point in which this data is valid, reliable, and standardized across different devices.
In the first ever systematic review of consumer-wearable activity tracking devices published in 2015, we summarized the evidence for wearables’ validity, reliability, and ability to estimate steps, distance, physical activity, energy expenditure, and sleep. We found that some data collected by consumer wearables (like physical activity) had a higher correlation with research-grade tracking devices than others (such as sleep) and that some devices (Fitbit) were more accurate than others (RIP, Jawbone). The conclusion at that time: wearable-devices had promise for reporting certain measures but had a long road to travel before they could be reliably used for research.
Fast forward four years. As of January 2019, there are now 370 registered clinical trials and 682 publications specific to Fitbit devices. But that’s not all: there are around 431 wearable devices available to consumers from 266 different vendors, and researchers from across a variety of unrelated disciplines are all trying get a handle on how to collect, process, and make sense of what frequently seems completely overwhelming and confusing.
These experiences highlight the opportunities that consumer wearable devices provide for observing human behavior. These devices allow researchers to collect passive objective data on measures like physical activity and sleep, replacing self-reported data and all the biases that entailed.These observational data have the capacity to augment and impact clinical care and research with repeated and reliable measures to study the numerous conditions affected by variations in physical activity, sedentary time, and sleep whereas their use for behavioral change at scale has yet to be fully developed.
While our current work is aligned with this observational approach, we fully expect that as these devices improve, their role in improving health behaviors will increase over time.
The All of Us BYOD program is just in the beginning stage of using sensors for passive, longitudinal data collection from increasingly large and dynamic human cohorts. While using wearables for observational purposes seems like the right place to start, we’re also continuing our work on novel digital biomarkers of stress and ramping up new work for 2019 that pushes these methods toward using sensor data for presymptomatic disease detection in humans.
We have reason to believe that we are indeed standing on the threshold of using valid, reliable wearable device data in research. We invite you to join us. You can find more information on our Wearables Research here.