Measurement and data validity
Good measurement properties of wearable device data are critical to successful adoption by patients and providers. The validity and reliability of measurements obtained from wearable devices and lack of standardization of devices continue to be of concern . For example, it has been documented that the FitBit One® has good validity and reliability for step count, but not for distance travelled . The Fitbit® and Fitbit Ultra® underestimate energy expenditure . One study of eight different activity monitors (BodyMedia FIT®, Fitbit Zip®, Fitbit One®, Jawbone Up®, ActiGraph®, DirectLife®, NikeFuel Band®, and Basis B1 Band®) compared to a portable metabolic system (i.e., Oxycon Mobile®) found a mean percent error range from 9% to 24%, with BodyMedia FIT® on the low end and the Basis B1 Band® on the high end . In addition, measurements are further confounded as wearable devices used for passive monitoring do not capture all free-living physical activity. For example, our observations of PatientsLikeMe (PLM) Fitbit® users found erroneous characterization of true activity due to infrequent or sporadic use and compromised accuracy due to predominant hand motion activities (playing the drums and cooking), driving or cycling, and long or short stride length. With technological advancement, it is expected that manufactures will respond to these limitations. Improved portability and ease of use, such as that offered in newly developed smart watches, could facilitate routine data gathering . Physical activity estimations can be enhanced if there is an accounting of a range of indoor and outdoor activities, different walking speeds, and types of ambulation (walking, jogging, running) .
User experience and engagement
Factors in long-term maintenance of wearable device usage include aesthetics, ease of set-up, lifestyle compatibility, and a clear value proposition to the user . In a usability study of activity trackers, patients with chronic obstructive pulmonary disease were asked to rate 16 aspects of usability of both the Fitbit® and a wearable device called the Physical Activity Monitor . The following concerns were identified by patients: i) technical difficulties (e.g., installation of device or software), ii) intention/willingness to monitor activity (e.g., willingness to use an activity monitor, recommend to others), iii) opinions towards wearing the device (e.g., pleasant, frightening, or frustrating to wear), and iv) the general attractiveness of the device. In contrast, the comfort in attaching or wearing the device, as well as the usefulness of activity monitoring, were rated high on the usability scale. It should be noted that some of these usability issues can be addressed by alterations to the devices (e.g., attention to fashion, good data visualization), whereas individual characteristics (e.g., negative expectations of use, inconsistent monitor use, lack of activity planning) may require a greater integration of engagement and behavior change principles.
In addition to usability factors, health literacy is of prime importance to maximize the benefit of consumer-accessible technology that may be utilized outside of clinical settings. According to the National Assessment of Adult Literacy , only 12% of US adults have proficient health literacy and over a third would have difficulty with following directions on a prescription drug label or maintaining a childhood immunization schedule based on a standard chart. In addition, limitations in numeracy, the ability to use numbers in daily living, are common in the US population . These limitations may make it difficult to comprehend and utilize personal data. Will patients understand visualized data, attribute meaning to the ‘findings’, and make behavioral changes if necessary? Wearable devices capture potentially useful and clinically impactful data, but these devices will gain wider adoption when patients find devices usable and their data comprehensible.
There is great variability in the effective use of behavioral science principles in order to effect behavior change. There are three key components that are critical to long-term engagement with wearables: i) habit formation (setting cues, routines, and rewards), ii) social motivation (sharing or competing for goals with others), and iii) goal reinforcement feedback to monitor personal progress . BJ Fogg discusses the importance of computers as ‘persuasive technology’, with an ability to enhance self-efficacy, provide tailored information, trigger decision-making, and help people reduce barriers that impede target behaviors . A recent review of behavioral techniques in activity monitors found that goal-setting, behavioral goal review, behavior feedback, self-monitoring, and rewards were generally included in popular monitors, but that other important components, such as problem-solving, behavioral instruction, and commitment strategies, were rare . Overall, this review found that these devices included a range of 5 to 10 of 14 total potentially effective research-based techniques.
At present, the mechanisms of action are unclear. Do patients develop greater insight by viewing data about their activity levels? Does self-monitoring encourage a greater generalized sensitivity to one’s daily health status? Perhaps the act of tracking can itself lead to behavior change. It has long been known that simple self-monitoring can affect one’s self-evaluation and lead to self-administered consequences that affect response frequency . Some wearable manufacturers attempt to leverage social media in the form of data sharing, but many consumers reject this idea . The assumption that user measurement and feedback of data are sufficient in the success of these devices is dubious. The majority of consumer-accessible wearable devices are not part of a formal disease management program with a clearly reported time of use period and specific goals. The inclusion of effective behavior change components in a programmatic manner may drive positive outcomes more than wearable use alone .
Privacy and safety
Many patients understand the value but state concerns about open-data sharing. Patients are increasingly amenable to sharing their data with peers . PLM, in conjunction with the Institute of Medicine, conducted a survey of 2,125 PLM members and found that 94% of responders are willing to share their health information on social media if it helps doctors improve care . In addition, 94% of responders would also be willing to share their health information on social media if it would help other patients like them, and 92% of responders would be willing to share information to help researchers learn more about their disease . The 2014 State of the Internet of Things Study found that more than half of consumers are willing to share their wearable data with physicians .
On the other hand, a majority of those surveyed in the PLM/Institute of Medicine study agree that shared data can be used in negative ways ; 80% of consumers express privacy concerns about personal data-sharing . This finding echoes a California Institute for Telecommunications and Information Technology survey, which found that 90% of respondents are seeking data anonymity . These concerns are realistic, especially in an era of GPS technology, which could potentially reveal sensitive personal activities. Could such data open up health insurance subscribers with low activity levels to higher premiums?
Because of the issues related to the use and reuse of data in studies, the nature of informed consent is evolving . Research suggests that individual openness to sharing will be dependent on the nature, use, user(s), legal protections, and potential compensation associated with the data [19,33]. Taken together, it is clear that greater transparency by device companies and researchers will be critical to patient engagement with these technologies.
Care delivery and integration
Wearable adoption by patients in their health care will also be driven by their belief that such usage will have an impact on their care experience. How does the introduction of wearable devices that monitor activity impact treatment selection, treatment effectiveness, management of disease, and patient-doctor communication? Personal data collection offers the potential for greater patient-provider collaborations, but will require patient (and provider) confidence about the usefulness of the data in treatment planning. As seen with the rollout of electronic health records, providers may not welcome yet another data source due to concerns about additional pressures on their time, reimbursement, and workflow .
Wearable devices may provide insight into the progression and impact of illnesses and may provide insights with conditions in which activity levels or movement may be compromised, e.g., multiple sclerosis, depression, rheumatoid arthritis, pain, and chronic obstructive pulmonary disease. Does the use of this data affect the understanding of disease trajectory, symptom severity, and progression? Can this data ultimately affect disease-related quality of life and treatment effectiveness? Ultimately, patients and their providers will more likely consider wearable data if it impacts outcomes.
As physical activity tracking research evolves, the challenges in clinical measurement, adherence, privacy, and clinical integration need to be addressed before these devices are broadly adopted as clinical and self-management tools. The following research and clinical directions are being adopted by PLM and other organizations. First, the measurement need perceived by patients will define the creation and use of the appropriate wearable sensor systems. Rather than focus on retrofitting fitness wearables for medical applications, we believe that enabling patients to voice their needs will encourage the development of devices more attuned to chronic health needs. This will place greater emphasis on a user’s experience in tracking their chronic disease and potentially improve overall user engagement. Second, the data from wearable activity trackers will be increasingly understood in relation to established disease-specific clinical measures and assessments. For example, activity and mobility wearable data in multiple sclerosis will be understood in the context of measures such as the 6-Minute Walk Test . This paves the way for provider acceptance of the data in clinical encounters. Device usage becomes part of the ‘prescription’ and patients share data as part of a treatment plan . Third, as wearable sensors aim for increased sophistication, validation best practices will need to be established and new alliances will emerge. Project HoneyBee, for example, is evaluating the cost effectiveness of consumer physiological monitoring with heart disease, chronic obstructive pulmonary disease, atrial fibrillation, mobility, gait monitoring in hydrocephalus, and diabetes . The Health Data Exploration project, with support from the Robert Wood Johnson Foundation , is bringing together researchers and industry to address key issues such as methodology, ethics, intellectual property, and intersections between data types. Fourth, wearables will be regarded as facilitators rather than drivers of change in health behavior . The development of chronic disease behavior change programs that utilize activity and other measurements offer a more organized and engaging experience than use of the device alone.