Overweight and obese adults with type 2 diabetes mellitus (N=260) were recruited from Forsyth County, NC, USA to participate in a study of lifestyle modifications for effective risk factor control and the prevention of disease complications—the Lifestyle Intervention for the Treatment of Diabetes study (LIFT Diabetes). Participants were recruited between 2013 and 2015 using a variety of approaches, including mailing potentially eligible adults identified in the electronic medical records system, direct referrals (primarily from health providers), advertisements, and community health events. Details regarding recruitment have been published [15]. The participants, primarily low-income and minority patients, were randomly assigned to 1 of 2 diabetes education groups: 1 consisted of a weekly intensive group-based lifestyle intervention promoting weight loss in a community setting and the other consisted of monthly diabetes self-management education resources delivered in the clinical setting. A more detailed description of the LIFT Diabetes protocol and design can be found in Katula et al [16]. At the baseline visit, all participants completed a survey, which had been used previously in the Action for Health in Diabetes (Look AHEAD) study (which also involved a lifestyle intervention for adults with diabetes) [17]. This technology use survey posed participants multiple questions regarding their current access to specific technologies, use of specific functions, and frequency of use. Additionally, the survey evaluated how participants would prefer to receive study information and whether they would be willing to complete future questionnaires online. The study reported here is an analysis of participants’ responses to this survey. The study design, methodology, and data collection protocols were approved by the Wake Forest Institutional Review Board, and written informed consent was obtained from participants. We handled all deidentified data in the statistical package Stata/IC version 14.1 (StataCorp LLC).

We compared frequencies and means of demographic characteristics, health outcomes, and survey responses through Pearson chi-square or Fisher exact analysis and t tests. To understand the racial disparities within the sample, we stratified overall demographic characteristics by racial/ethnic group. Due to the small number of nonblack minorities (n=17), as well as their similarity in health and demographic trends to black participants, we grouped all minority participants together for analysis. Due to the assumption that a technology-based intervention would require participants to engage with a device at least once a week, we collapsed all survey responses indicating frequency of use into 2 groups (at least once a week vs less than once a week). Following a description of the sample means and demographics, we used logistic regression as a means of understanding the relative impact of each demographic characteristic on survey responses. To limit the number of missing values in the regressions, questions that participants were prompted to skip after answering “no” to the previous question were recoded as “no” rather than “missing.” For example, if a participant did not own a mobile phone, in the subsequent question regarding smartphone ownership, their answer was coded as “no” rather than “missing.” Outcome variables used were ownership of a computer and willingness to complete future study questionnaires online.

We collected demographic information, health characteristics, and survey responses from each of the 260 LIFT Diabetes participants; therefore, we included all study participants in the descriptive statistics. However, due to missing information regarding employment, smartphone ownership, and text messaging, the final regression sample for ownership of a computer consisted of 257 participants, and the regression sample for willingness to complete future surveys online consisted of 208 participants.

Table 1 shows the sample population’s demographic and baseline health characteristics. Of the ethnic minority participants, 88.0% were black (125/142), 4.3% were Hispanic (n=6), and 7.7% were other (n=11). Although 22.3% (58/260) of the study population declined to report annual income, most of the respondents (61.4%) reported an annual income of US $49,999 or less. Minority participants were much more likely to have a lower income than white participants (P<.001). Education followed similar, though not significant, trends between racial/ethnic groups, with minority participants typically achieving lower educational attainment. Overall, however, most of the study population (n=207, 79.6%) achieved greater than high school education. Of the population health characteristics, we found significant racial/ethnic differences in mean age (P<.001), body mass index (BMI; P=.002), diastolic blood pressure (P<.001), and HbA_1c (P=.003), suggesting a slightly worse baseline health profile among minority participants.

Table 2 shows technology access and use variables. Minority participants were significantly less likely to own a home computer (P<.001) and have email access at home (P=.03). To assess the participants’ ability and willingness to engage in future technology-based interventions, we correlated survey responses to demographic, health, and access variables (Table 3). Those who owned a home computer were more likely to have a higher income (P<.001), have higher educational attainment (P<.001), be employed full time (P=.01), and own a smartphone (P=.001). As expected, those with higher education and income were also significantly more likely to indicate a willingness to participate in technology-based surveys (P ≤.001). Similarly, those who were not employed full time were less likely to indicate a willingness to complete future study questionnaires online (P=.01). Each of the access variables also played a significant role in participants’ responses (P ≤.05). Of those who were not willing to complete online questionnaires, 52% (17/33) owned a home computer and 46% (15/33) had access to the internet at home.

Table 4 presents the results of the regression analysis for owning a home computer and willingness to complete questionnaires online, when adjusted for demographic characteristics (age, sex, race/ethnicity, education, employment, family size, marital status) and health status (duration of diabetes, smoking status). Minority participants had significantly lower odds of owning a home computer (adjusted OR 0.26, 95% CI 0.10-0.67; P=.01). Those with a high school education or General Education Development also had lower odds of owning a home computer (adjusted OR 0.25, 95% CI 0.10-0.60; P=.002), while those who owned a smartphone had 3.11 higher odds of owning a computer (adjusted OR 3.11, 95% CI 1.37-7.08; P=.01). Education also played a role in willingness to complete questionnaires online, as those with a below–high school education had significantly lower odds of responding “yes” (adjusted OR 0.45, 95% CI 0.01-0.34; P=.003). Additionally, those who owned a home computer were more likely to be willing to complete questionnaires online (adjusted OR 3.87, 95% CI 1.14-13.2; P=.03).

The objective of this study was to determine both access to and use of technology among an ethnic minority population to assess the correlates to use of information technology among adults enrolled in a diabetes study. Broadly, at proportions consistent with national averages, most of the sample owned a home computer and mobile phone and had access to email and the internet [1]. However, significant dissimilarity between racial/ethnic groups was evident in demographics, health, and technology access. The racial disparity in income was reflected in the strong correlation between income and both ownership of a home computer and willingness to complete questionnaires online. The associated racial disparity in owning a home computer persisted even after controlling for numerous demographic characteristics, suggesting that socioeconomic status cannot fully explain the digital divide along racial lines. Subsequent analysis demonstrated that willingness to complete questionnaires online depends heavily on home computer access, which parallels a recent finding that home internet access drives patients’ willingness to use technology for glycemic monitoring [18]. Together, these results indicate that access to home technology is critical to the advancement of Web-based interventions, yet a significant racial discrepancy in access limits practical, translational implementation among a minority population. Regardless of literature citing effectiveness, the potential success of a technology-based program is irrelevant without sufficient access [5,7,8]. Yet the results also indicate that access alone may not be the sole barrier to such an intervention. Approximately 50% of the study sample indicated that they had access to both a home computer and the internet yet indicated “no” when asked about completing study questionnaires online. While it is perhaps surprising that those with home access to technology would indicate an unwillingness to participate in a technology-based intervention, these findings suggest that, while access is a critical determinant for any future intervention, purely supplying resources may not be sufficient. Whether this is an issue of familiarity with using technology or a deeper discomfort with nontraditional clinical settings, it is evident that initiating a Web-based protocol among minority patients with diabetes may fail to capture a meaningful portion of the population.

The significant relationship between educational attainment and an interest in online platforms presented here parallels previous discussions of education and telecommunication [18,19]. While higher education typically correlates to higher income, concurrent relatively low income and relatively high education seen in this study may not be unrealistic. Dray-Spira et al suggested that diabetes may impair patients’ ability to maintain the standard of employment associated with higher education, or that diabetes-related disability results in significant work loss followed by termination of employment at higher levels [20]. These possibilities are bolstered by the low level of full-time employment in this study (116/260, 44.6%) and broader consideration that work disability days are significantly higher for employees with diabetes [21,22]. Therefore, significant diabetes-related work disability may have exacerbated the disparity in education status and income seen in the study population.

Although health characteristics are unlikely to influence willingness to participate in Web-based programs, it is possible that underlying health problems could influence earning capacity and thus also access to technology. However, the health technology access and health technology use correlations explored were not statistically significant (data not shown). Despite this negative finding, the health metrics described reinforce previously observed racial disparities in health outcomes [23,24]. Prior research demonstrated that these disparities cannot be adequately accounted for by childhood socioeconomic status, adult income disparities, or health behaviors, but rather by the influence of allostatic load, exposure to discrimination, and decreased social capital [24,25]. While we did not explore these social variables in this study, we did find the expected trend that minority participants with diabetes exhibited higher BMI, diastolic blood pressure, and HbA_1c than did white participants. Therefore, while the sample was geographically restricted, the observed health patterns are generally consistent with other research studies.

This investigation was limited by its reliance on a single survey item—willingness to complete questionnaires online—as a proxy for participants’ willingness to engage in health coaching or health metric tracking through Web-based technology. A more robust analysis would be facilitated by additional, specific questions focusing on the issue of telecommunication in diabetes management. Demographically, the study had relatively few black men and generally had few nonblack minority participants, which necessitated grouping these participants into a general minority classification, and this therefore may have obscured distinct minority group responses. Finally, the study was restricted to a population of patients with diabetes already under a physician’s care at the time of enrollment and who were willing to be randomly assigned to a clinical trial, which may indicate higher socioeconomic position and therefore limits the generalizability of the findings to a broader minority population.

This study established demographic characteristics, health profiles, and access to technology within an ethnic minority population in the southeastern United States. Research on chronic disease management—specifically diabetes—and clinical practice have demonstrated the effectiveness of intensive behavioral and lifestyle interventions in reducing the risk of disease complications [26-28]. Smartphone apps, telecommunication, and other mobile technologies have been proposed as efficient and effective alternatives [4,29-31]. However, these findings have yet to be reconciled with the results presented above—that minority patients of lower socioeconomic status lack both access to and familiarity with certain computer technologies—which limits the possibility for a translational intervention. While advancing diabetes research to address health disparities will require an innovative approach, the argument for mobile technology is not well supported at this time. In addition to focusing intervention efforts in other areas, studying minority patients’ perceptions of technology in the clinical setting may provide a much-needed perspective and inform the use of Web-based apps when technology becomes a viable approach. Here we highlighted the limited feasibility of introducing mobile technology to reduce health disparities among those with diabetes. However, as access to technology increases with time, future studies should investigate users’ perceptions of data safety and privacy, the cost of data plans associated with mHealth tools, and barriers to using personal technology in the clinical setting, aside from resource deprivation.