Advances in wireless sensors and digital technologies have led to a proliferation of wearable health care devices with which users can examine, monitor, and track their physiological conditions. Wearable health care devices are autonomous, noninvasive, wearable equipment with embedded sensors to collect a variety of physiological health information [1]. These devices range from the popular fitness trackers (eg, Fitbit, AppleWatch, Samsung, Galaxy Fit) that collect data on physical activities such as number of steps taken, calories burned, sleep duration, and heart rate to more sophisticated devices that can collect information on blood pressure, glucose levels, and oxygen levels. The collected health data can be transmitted to smartphones or other computer-aided applications to store and analyze to provide appropriate health interventions. Fueled by increased popularity, the use of wearables has significantly increased in recent years. According to estimates, the market for wearable health care devices in 2018 was US $24.57 billion and was slated to grow 24.7% annually to US $139.35 billion by 2026 [2].

Wearable health care devices offer several potential benefits to users. First, they offer a convenient way to monitor, store, and share health information in real-time. Second, wearables provide feedback to users to make appropriate changes to their daily routines or behavior [3,4]. Third, wearables can facilitate remote patient monitoring and provide proactive and faster data access to physicians, resulting in improved health outcomes [4,5] and reduced number of physician visits [6]. Fourth, these devices can be particularly useful for patients with chronic conditions [7], patients with cardiovascular risks [8], and elderly populations [9]. Therefore, the use of wearable devices has the potential to significantly improve health care delivery and reduce costs.

Despite potential benefits, significant challenges remain to the widespread adoption and use of wearable health care devices [10-12]. The ability of these devices to track, store, and transmit patients’ health information raises questions about data security and privacy [13-15]. In addition, the design, accuracy, and reliability of wearables have also been a major concern [16,17]. Concerns have also been raised about the accuracy of data gathered by wearables in people of color [18]. Technology acceptance of new wearable devices remains another significant barrier [19]. Despite the forecasted growth, use of these devices has reportedly slowed [20]. Market studies have pointed out a gradual decline in the use of these devices as well as abandonment within a few months of purchase [21,22].

Extant research on the use of wearables has found these devices to increase physical activity, increase energy expenditure, and reduce sedentary behavior [23]. Wearables offer an easy way to obtain large amounts of real-time health data that can be useful for clinicians [24]. Health data obtained from wearables, when combined with sophisticated machine learning algorithms, have helped develop predictive models that can greatly improve health delivery [25-27]. If effectively used, wearables can greatly help in the management of several conditions including rheumatic and musculoskeletal diseases [28,29], chronic pain management [30], and cardiovascular problems [31]. To effectively realize the potential benefits of wearable health care devices, a solid understanding of the factors related to their adoption and use is warranted. Most extant studies have largely examined the intention to adopt wearable devices [32-36] rather than the actual usage. Further, these studies have focused on a limited set of antecedents driven by the technology adoption lens. Hence, we know very little about US adults' use of wearables and key influential factors affecting usage. Addressing this gap, we build on an emergent stream of studies to examine the key factors affecting the use of wearable health care devices by US adults.

The main objective of this study was to examine the use of wearable health care devices and the key predictors of wearable use by US adults. We examine predictors related to individual health [37], technology self-efficacy [36,38], personal demographics [32], and attitudes towards fitness or exercise [35] as well as their associations with the use of wearable health care devices. Our research model, with all the predictor variables, is shown in Figure 1.

The dataset for this study comes from the National Cancer Institute’s Health Information National Trends Survey (HINTS)-5, Cycle 3, with data collected from January 2019 to April 2019 through self-administered mailed questionnaires and a web-based pilot. HINTS is a nationally representative survey that includes US adults ≥18 years of age in civilian, noninstitutionalized settings. This survey collects data on US adults’ need, access, and use of information related to health and health care and the related behaviors, perceptions, and knowledge. It uses a stratified sampling method defined by areas with high concentrations of minorities and areas with low concentrations of minorities. Survey invitees for both mailed questionnaires and the web pilot involved an initial mailing of the questionnaire, followed by a reminder postcard and up to two additional mailings of the questionnaire as needed for nonrespondents. More details on the survey and survey methodology can be found elsewhere [39].

Since HINTS used probability sampling to improve representation of specific groups, our analysis applied weights to calculate US population estimates and standard errors. These sampling weights were needed to ensure unbiased estimations. Sampling weights are inverse probabilities of selection but are modified based on census or population information and further adjusted for nonresponses, so that the weighted totals of poststratification variables match the population totals. Weight adjustment accounted for nonresponse and known population totals based on the 2017 American Community Survey of the US Census Bureau on age, gender, education, marital status, race, ethnicity, and census region. We used the jackknife approach to compute replication weights [40]. Jackknife is a popular resampling approach that creates many replicate samples taken from the original sample. Each replicate sample provides an estimate of the variable of interest, and the variability among the estimates from multiple replicate samples are used to estimate the variance of the variable of interest [41].

For data analysis, we included all respondents who responded to the question about their use of an electronic health care device to track or monitor their health or physical activity in the past 12 months. Of the 5380 overall respondents, 4551 had indicated if they used (yes/no) a wearable health care device and were included in our analysis. STATA 16.1 software was used to perform the statistical analyses. We compared the demographics of all respondents with our sample that had answered the question about the use of wearable devices and did not detect any significant differences [42].

The main variable of interest was the use of a wearable health care device. This was assessed as a binary variable (yes/no) that asked respondents to indicate their use of an electronic wearable device to monitor or track health or activity in the past 12 months. In addition, we also explored the responses about frequency of wearable health care device use in the past month (daily, almost daily, 1-2 times a week, <1 time a week, never used) and the respondents’ willingness to share the data from wearables with a provider (yes/no). Data from several survey items were included to capture survey respondents’ self-reported characteristics: sociodemographic (age, gender, race/ethnicity, education, marital status, annual household income), health-related (general health, presence of any chronic conditions, frequency of provider visits, self-perception about weight), and technology-related (technology self-efficacy) variables. General health was assessed using a question for which respondents rated their health on a Likert scale (rated as 1-5) as being poor, fair, good, very good, or excellent. Presence of chronic conditions was coded as a binary variable (0/1) if the respondents indicated they had at least one of the following conditions: diabetes or high blood sugar; high blood pressure or hypertension; heart condition such as heart attack, angina, or congestive heart failure; or a chronic lung disease such as asthma, emphysema, or chronic bronchitis. Frequency of provider visits was captured using a Likert scale based on the number of times the respondents visited a health provider in the past 12 months. Self-perception about weight was assessed on a Likert scale (rated as 1-5) as underweight, slightly underweight, just about the right weight, slightly overweight, or overweight.

As people interested in physical fitness and exercise are likely to be drawn towards a wearable health care device, another study variable included attitude towards exercise (to what extent do you enjoy exercising?). This was captured using a Likert scale (rated as 1-4), with the following options: not at all, a little, some, or a lot. Technology self-efficacy was captured using an additive score (ranging from 0 to 6 points) that was captured using questions that asked participants if they used a computer, smartphone, or electronic means for performing 6 tasks: (1) look for health information, (2) purchase medicines or vitamins online, (3) communicate with a provider using email or the internet, (4) track health care charges, (5) look up medical test results, and (6) make appointments with a provider.

We first conducted a descriptive analysis on our data sample. To assess the relationship between the use of wearable health care devices and sociodemographic or health-related categorical predictor variables, crosstab tables were generated and tested using the Wald chi-square test. All results were weighted to give US population level inferences using a standard weighting approach that was specifically developed for the HINTS dataset.

Given that our main variable of interest (ie, use of a wearable health care device) was a binary variable and our predictor variables were a mix of categorical and continuous variables, logistic regression analysis was chosen to assess the influence of sociodemographic variables, health-related predictors, and technology self-efficacy on the use of a wearable health care device. We computed adjusted odds ratios (ORs) and 95% CI estimates for the predictors and used a cutoff of P<.05 to determine the statistical significance of all our analyses. In accordance with the complex survey design, weights with jackknife replications were applied in our analyses.

Little is known about the adoption and use of wearable health care devices and the predictors of adoption and use among US adults. Our work includes a broad US population and moves beyond reported intentions to use wearables to actual use of these devices. Future work should attempt to explore more predictors of the frequency of wearable usage and the actual sharing of wearable data with providers. Our study has many limitations. Although the HINTS survey used a national sample and involved stratified selection to improve the responses of subgroups, the data had responses from both mailed questionnaires and a web-based pilot, which included an incentive bonus. Web pilot responses are likely to be from adults who are more comfortable using technology and the internet, and this could possibly introduce bias. Moreover, the self-perception questions are subjective and could be interpreted differently by participants.

The potential of wearable health care devices is under-realized, with less than one-third of US adults actively using these devices. With only younger, healthier, wealthier, more educated, technoliterate adults using wearables, other groups have been left behind. To effectively capitalize on the growing popularity of wearable health care devices to improve health care delivery and outcomes, more concentrated efforts by clinicians, device makers, and health care leaders are needed.