Assessment of ownership of smart devices and the acceptability of digital health data sharing

We conducted a survey study among a large and diverse sample of patients from DUHS, called Duke Health Listens (DHL)²⁷. The survey consisted of 14 questions (Supplementary Document 1) in total. The survey contained single and multiple choice questions to quantify the ownership and usage of smart devices, frequency of usage, reasons for usage or reason for not owning a wearable, willingness to participate in digital health studies, willingness to share data collected by smart devices for research purposes, and type of data the respondents are comfortable sharing and why. We further queried the highest level of education and employment status. Additional data made available by the DHL group included demographic information about respondents, including gender, age group/generation, race/ethnicity, County, and State (North Carolina/ Virginia). The survey was sent out to all DHL patient advisory group members (N = 3021) between January 18–30, 2022; 1368 responded (45% response rate). Table 1 describes the characteristics of the survey respondents. A total of 871 (64%) were female, 826 (60%) identified as white, 390 (29%) identified as Black, 60 (4%) identified as Asian, 78 (6%) identified as Hispanic, and about half of the respondents (52%) were age 58 and above.

Table 1 Respondent Demographics and Characteristics

Table of Contents

Smart device ownership

Of the 1368 respondents, 1343 (98%) owned a smartphone, with 894 (66.6%) owning an iPhone and 446 (33.2%) owning an Android device. Smartphone ownership in this population was higher than that reported by both the American Community Survey (2018)²⁸ and Pew Research Center (2021)¹, in which 84% and 85% of respondents reported owning smartphones, respectively. Supplementary Table 1 shows the relative comparison of smartphone and wearable ownership across different demographics between the Pew Research Study^1,29 and our study. We found that younger people tended to have higher smartphone ownership than older people, and employed people tended to have higher ownership than those who were retired. While we observed statistically significant differences in smartphone ownership across age (X²(4, N = 1368) = 68, p < 0.0001) and employment status (X²(5, N = 1368) = 28.3, p < 0.0001) (Fig. 1 and Table 2), no substantial differences were observed across gender, race/ethnicity, and education level. Specifically, there was significantly lower ownership in people age 77+ (89%) as compared with younger age groups (age 26–41: 100%; age 42–57: 100%; age 58–76: 98%) (post hoc pairwise comparison with Benjamini Hochberg multiple hypothesis correction p < 0.0001). Further, smartphone ownership among retirees (95%), although high in general, was significantly lower than ownership among full-time employees (99.5%) (p < 0.001).

Table 2 Smart Device Ownership Across Different Demographics

In terms of the type of smartphone ownership, there were statistically significant differences between individuals who own an iPhone vs an Android device across race/ethnicity (X²(3, N = 1263) = 11.1, p = 0.012), education (X²(3, N = 1320) = 33.5, p < 0.0001) and employment status (X²(5, N = 1323) = 37.5, p < 0.0001). Specifically, we found significantly higher (p < 0.01) iPhone ownership in White vs. Black individuals (70% vs. 60%), those with college or graduate degrees vs. those with high school degrees or some college (69–72% vs. 50–53%), and those with full-time and part-time employment vs. their demographic counterparts. On the contrary, no significant difference was seen in the type of smartphone ownership across gender and age.

Nearly 60% of survey respondents (802 out of 1368) owned a wearable device, which is higher than the numbers reported by Pew Research Center²⁹, National Cancer Institute’s Health Information National Trends Survey (HINTS)²⁴ and Statista², in which 21%, 30%, and 31% of respondents owned wearable devices, respectively. The most-owned wearable device was the Apple Watch (44.14%, N = 354), followed by the Fitbit (42.02%, N = 337). Females, younger generations, people with college and graduate degrees, and people with full-time employment had higher ownership compared with their demographic counterparts (Table 2). Specifically, females had a significantly higher (p < 0.01) wearable ownership than males (62% vs. 53%), individuals with college (60%) or graduate degrees (62%) had a significantly higher (p < 0.01) wearable ownership than individuals with high school degrees (41%), and individuals with full-time employment had a significantly higher (p < 0.001) wearable ownership (64%) compared to retired individuals (52%). Wearable ownership was inversely related to age, with those aged 77+ (39%) having significantly lower (p < 0.01) wearable ownership as compared with younger age groups (age 18–25: 71%, age 26–41: 69%, age 42–57: 65%, and age 58–76: 54%). Further, those aged 58–76 had significantly lower (p < 0.01) wearable ownership than those aged 42–57 (54% vs. 65%). No significant differences were found in wearable ownership by race/ethnicity, which is similar to the results reported by Chandrasekaran et al.²⁴ on the 2019 HINTS data. Details of the statistical test results supporting these differences are, for gender, X²(1, N = 1349) = 9, p < 0.01); age (X²(4, N = 1368) = 44, p < 0.0001); education (X²(3, N = 1365) = 13, p < 0.01); employment (X²(5, N = 1368) = 18.4, p = 0.0024). Taken together, these findings may help to explain the consistent demographic differences in participation seen in large-scale BYOD digital health studies^9,30,31.

Smart device usage

Understanding why and how people use smartphones and wearables can shed light on how easy it may be for researchers and clinicians to obtain physiological and behavioral RWD that would be sufficient for drawing robust research conclusions or effectively implementing health interventions for patients.

Of the 1343 respondents who own a smartphone, 792 (59%) reported using their smartphones for health and fitness tracking. While there were significant differences (p < 0.01) in smartphone usage for health and fitness tracking across age, education, and employment status, no differences were seen across gender and race. A higher percentage of individuals in younger age groups, with college and graduate degrees, and full-time employees reported using their smartphones for health and fitness tracking as compared with their demographic counterparts. Frequencies of smartphone usage for activity tracking during both weekdays and weekends were similar (p > 0.05) across gender, age, race, education, and employment, with the majority of people using their smartphone for activity tracking evenly across all weekdays and weekends (Fig. 2a and b).

**Fig. 2: Frequency of usage of smart devices.**

The primary reason for owning a wearable device across all demographic groups was fitness and workout monitoring (heart rate, step tracking, jogging, etc.) (65%, N = 522) (Fig. 3). This primary reason for ownership varied significantly (p < 0.05) across age, race, and education. The second most common primary reason was sending and receiving phone calls, emails, texts, and messages (36%, N = 291). For the majority of respondents, the most common secondary reason for owning a wearable device was health tracking (blood oxygen, heart rhythm, women’s health functions such as ovulation tracking, etc.) (34%, N = 272), followed by sleep monitoring (33%, N = 265). Only a small number of people reported owning wearables primarily for fashion (4.5%, N = 36), and the majority of those who did are Apple Watch owners (69%, N = 25). These results highlight that the data of most interest in digital health studies and clinical care, including fitness and workout monitoring, health tracking, and sleep tracking, are likely to be present in BYOD participants as these functionalities are the primary motivation for users owning and using these devices in the first place.

**Fig. 3: Why do people use wearables?.**

In addition to why people do own wearable devices, it is also very important to understand why people don’t own wearable devices. Such information may highlight factors that can impact the representation of different demographics in a study population, particularly with BYOD study designs, but perhaps even extending beyond to studies that supply devices to participants, for example, if there were discomfort with potential surveillance. Here, we found that the top three reasons for not owning wearable devices were their cost (31%, N = 178), lack of interest in tracking (22.4% N = 127), and no particular reason (22.08%, N = 125) (Fig. 4). The cost of the device was the most important factor for not owning wearable devices across all demographic factors, except gender, where this factor was more important to females as compared with males (37% (N = 123) vs 23% (N = 52), respectively, p < 0.001). This observation points to BYOD designs as potentially exacerbating unequal representation in study populations. The lack of interest in tracking was the second most important factor for not owning wearables across all demographic factors except race and ethnicity, where this reason for not owning a device was more prevalent in White, Asian, and Hispanic as compared to Black individuals (26%, 29%, 24% vs. 11% respectively, p < 0.001). This is important to note as intrinsic motivation for tracking may lead to increased adherence to device wear. On the other end, privacy concerns were the fifth most important factor overall (12%, N = 68) for not owning wearables, which varied significantly (p < 0.05) across age, race, education, and employment. Particularly, privacy concerns were more important for not owning wearables for White than Black respondents (15% vs. 4%, p < 0.01), for respondents with graduate and college degrees than respondents with some college experience with no degrees (17% and 14% vs. 2%, respectively, p < 0.05), and for respondents with full-time employment than retired respondents (18% vs. 6%, p < 0.01). Such factors driving people’s decisions not to own wearable devices can shed light on both why data may be nonrepresentative and what changes could be made in existing research and care ecosystems to address concerns to improve equitability.

**Fig. 4: Reason for not owning wearable devices.**

Understanding how people use wearables may improve our understanding of the availability and quality of wearable data that can be used for research or healthcare purposes. The context of data collection matters, for example, if people only wear their devices while exercising vs. wear their devices all the time, the average heart rate and activity values would be dramatically different. Furthermore, the circumstances of data collection can affect data availability and accuracy. For example, some devices only collect heart rate variability during sleep. In our study, the majority (>50%) of wearable device owners reported using their device(s) on all weekdays and weekends (Fig. 2a and b), which is similar to the HINTS data (where 47% and 25% of the respondents used wearable every day and “almost every day”, respectively)²⁴. However, the frequency of usage over weekdays and weekends varied significantly (p < 0.05) across gender, race, and employment. Although a higher percentage of females own wearable devices than males (62% vs. 53%), self-reported wear time was found to be higher in males than females (82% vs. 71% for all weekdays and 78% vs. 66% for all weekends). This is in contradiction to similar research by Pew, where women were more likely than men to say they regularly use their devices (25% vs. 18%)²⁹. Also different from the Pew study, we found that the frequency of wearable usage varies across racial and ethnic groups, with a higher percentage of Hispanic people reporting irregular wear times compared to White, Asian, and Black individuals (55% vs. 78%, 78%, and 71%, respectively, for all weekdays and 55% vs. 73%, 69%, and 66% respectively, for all weekends). In contrast, Pew observed higher self-reported regular wear time in Hispanics as compared to Whites and Blacks (26% vs. 20% and 23% respectively)²⁹.

Wearables are increasingly being used for studying sleep patterns and behaviors, given that sleep influences overall health. However, device wear during sleep is less prevalent due to factors like comfort and removal for charging, which impacts device use for overall health monitoring. Evaluating device usage during the daytime and nighttime can reveal important insights into how digital health studies can be better designed to monitor sleep and other digital biomarkers collected during sleep. Of wearable device owners, 42.3% reported wearing their devices all the time (day and night), followed by daytime only (42.1%), irregular times (7%), during workout only (3%), and nighttime only (1%) (Fig. 2c). These proportions varied significantly (p < 0.05) across gender, race, education, and employment. Upon further investigation across exclusively Fitbit and Apple Watch owners, we found that the majority of Apple Watch owners (63%) use their device during the daytime only, whereas the majority of Fitbit owners (64%) use their device during both the day and nighttime. This observation might relate to the more frequent need for charging for the Apple Watch compared to Fitbit, making Apple Watch users more likely to remove their devices during sleep for charging. Researchers and clinicians may consider such factors when designing and conducting digital health studies involving sleep tracking and other digital biomarkers measured at night and/or during sleep.

Willingness to participate in digital health studies and to share personal digital data

Willingness to share data is a key factor in understanding which data types are likely to be available for research or clinical purposes. Lack of available information regarding factors related to data sharing can result in biased study data or inequitable clinical practices that serve some, but not all, people. In our study, among the respondents owning smart devices (N = 1345), 50%, 32%, and 18% responded “Yes”, “Maybe”, and “No”, respectively, on their willingness to share the personal data collected by their smart devices for research purposes (Table 3). Of the 1102 people who are or may be willing to share their data, 702 (64%) own wearable devices. Willingness to share activity data for future research among wearable owners is slightly higher in our study (57% Yes, 30% Maybe, and 12% No) than previously reported values by Pew Research Center (53% acceptable, 18% not sure, and 29% unacceptable)²⁹, Seltzer et al. (where ~40% of respondents were willing to share wearable data immediately, and ~75% agreed to donate wearable data after death)¹⁸, and Hirst et al. (where roughly one-third of the respondents were willing to share smartphone and wearable data for health research)²¹. Willingness to share wearable data among wearable owners in our study is comparable to previously reported values in the 2019 HINTS data as reported by Rising et al.²², where 82% and 70% of respondents were willing to share wearable data with healthcare providers and family or friends, respectively. However, the HINTS study particularly asked about sharing wearable health data with healthcare providers and family or friends, whereas our study focused on willingness to share wearable activity data for research purposes. In our study, participants’ willingness to share their personal data for research purposes varied significantly across age and retirement status– younger generations are more willing and retired individuals are less willing to share their personal data than their demographic counterparts (X²(8, N = 1344) = 28.6, p = 0.0004) and employment (X²(10, N = 1345) = 21.5, p = 0.018). These findings are similar to those of Pew²⁹, which demonstrated a higher acceptance of wearable data sharing with medical researchers in younger generations than older generations (47% vs 35%), although the overall acceptability for all age groups is higher in our study.

Table 3 Willingness to participate in future research studies involving sharing activity data from smart devices

Of participants who are willing to or would consider sharing their personal data for research purposes (N = 1102), most are willing to share fitness and workout monitoring data (heart rate, step tracking, jogging, etc.) (69% Yes, 22% Maybe, and 9% No) (Fig. 5). Conversely, the data type that was the least amenable to being shared was self-reported measures from health and fitness apps (e.g., mindfulness and mood, water intake, food logs, women’s health, etc.) with responses of 50% Yes, 31% Maybe, and 19% No. Comfort around sharing different data types for research purposes varied (p < 0.05) by gender, age, education, and employment. For example, females were more willing to share fitness and workout monitoring data and self-reported measures from health and fitness apps than their male counterparts (92% vs. 89% and 84% vs. 76%, respectively). Overall, data sharing acceptability decreases with age across all data types. Similarly, retired individuals are less comfortable sharing their personal data compared to full-time and part-time employees. On the contrary, progressing education has demonstrated higher comfort in sharing all data types. There were no differences in the comfort of sharing these data types across races (p > 0.05).