Over 30 million people in the United States have been diagnosed with type 2 diabetes [1]. Poor glycemic control, defined as hemoglobin A_1c (HbA_1c) > 8% (64 mmol/mol) [2], in patients with type 2 diabetes is a risk factor for the development of diabetes-related complications including retinopathy, neuropathy, heart disease, stroke, blindness, kidney failure, and lower limb amputations [3]. Costs for diabetes care are high and rising [4,5]. Within the United States, total costs have been estimated at US $465.2 billion, including morbidity, mortality, and medical costs [6]. Glycemic control is the primary therapeutic objective for the prevention of diabetes-related complications [7].

Diabetes is a nationwide epidemic, though difficulty managing this complex, chronic condition varies across the United States [8]. Management is markedly more difficult in rural communities with limited access to health information and specialty care [9,10]. Diabetes is nearly 10% more prevalent in rural than in urban areas, likely owing to greater risk factors including lower income, older age, and higher body mass index [11]. In addition, individuals living with diabetes in rural areas face numerous barriers (limited availability of diabetes education [12], reduced cell phone coverage and internet access [13], transportation barriers, and lengthy travel distances [14,15]), preventing patients from accessing health care [16]. The Veterans’ Affairs (VA) Office of Rural Health estimates nearly 5 million veterans live in rural areas where access to care can be difficult [17], and that almost 40% of Veterans Health Administration (VHA) patients with diabetes live in rural areas [18].

Diabetes self-management behaviors (eg, medication adherence, diet, physical activity, and monitoring blood glucose levels [19]) are consistently linked to achieving glycemic control. Accessible communication, via face-to-face visits or technology, with providers is essential to foster patients’ disease self-management [20]. Access to diabetes self-management education and ongoing support can be improved by using digital health solutions [21]. Previous research highlights the benefits of using web-based patient portals, suggesting that increased access to information and support may engage patients in the management of their disease and improve health outcomes [22,23]. Considering the access challenges rural patients face, virtual care services may be even more critical in this population for effective diabetes self-management. Features such as secure messaging (SM) in the VHA web-based patient portal My HealtheVet (MHV), are fundamental to the goal of increasing access to care. SM can be used in lieu of telephone or in-person visits, or to provide additional opportunities for patient-provider communication between visits. Previous coding of SM content revealed wide variety in how it is used, including self-management behaviors such as medication renewal/refill requests, scheduling, referrals, and discussing medication or health issues [24,25].

Research to date suggests that SM use is associated with higher odds of meeting HbA_1c control targets, with increased odds of control for every additional message sent per year [26], and with more years of use [27]. SM use may support improved diabetes self-management, though the exact mechanism among these 3 constructs has not been established. It is also unclear to what extent patient characteristics, such as where they live, may play a role in the effectiveness of SM. SM is potentially more beneficial for rural patients with reduced access to in-person care, though it is also possible that it may be less helpful or accessible for those in rural areas with more limited internet access [28-30].

This study examined and compared the benefits of sustained SM use for rural and urban patients with diabetes. Rural patients with diabetes are less likely to engage in self-management behaviors, have worse glycemic control, and more limited access to health care. Therefore, they may depend more on accessible communication to help manage their disease. This study uses a framework that was initially developed to evaluate how the BlueButton within the MHV patient portal can support key stakeholder (eg, patients’) experiences, processes of care (eg, patient-team communication, self-management, and care coordination), and health outcomes, and understanding how contextual characteristics (eg, environment or setting in which patients seek and receive health care) shape use of the technology [31]. We have adapted this framework to evaluate other MHV features including SM.

This study had 3 objectives. We sought to investigate whether diabetes self-management mediates the relationship between SM use and glycemic control (objective 1). Additionally, we sought to understand if this mediation was conditional on the patient’s environment (eg, where the patient lived; objective 2). Finally, we wanted to understand how patients are using SM for diabetes management (objective 3).

This retrospective observational, cohort, sequential, explanatory, mixed methods (QUAN qual) study included US veterans living with type 2 diabetes. Table 1 specifies the timeline and sources of sampling and data collection. All participants experienced uncontrolled diabetes in 2012 (defined as mean HbA_1c>8.0% and less than 25% of the year with an HbA_1c<8.0%). All participants were sustained users of MHV between 2013 and 2017, defined as having used the portal repeatedly (used prescription refills, viewed or downloaded their health information, and used SM at least twice a year for 2 years between 2013 and 2015) and recently (sent at least 4 SMs between January 2016 and June 2017). Seeking a diverse sample of users who were either in good or poor control of their HbA_1c, we randomly selected a sample of 500 patients who had achieved good HbA_1c control in 2016 (defined as mean HbA_1c<8.0% for 75% of the year or more) and 500 who remained in poor HbA_1c control in 2016 (defined as mean HbA_1c>8.0% for 75% of the year or more). We mailed the randomly selected participants (N=1000) surveys in November 2017, and an additional 200 surveys at the beginning of 2018.

Quantitative methods were used to examine the associations among SM use in 2017, diabetes self-management between November 2017 and February 2018, mean glycemic control in 2018, and differences between rural and urban patients. Data on patients’ use of the MHV patient portal, their glycemic control, in-person health care utilization, and demographic variables were obtained from the VHA Corporate Data Warehouse (CDW) and merged with survey responses.

Qualitative methods were used to further understand how participants were using SM for diabetes self-management. Purposeful sampling was used to identify 40 survey respondents to participate in semistructured interviews about their diabetes management and technology use. In the survey, participants were asked an open-ended survey question, “Can you tell us about an ‘A-Ha!’ Moment when you realized you could use the MHV portal to better manage your diabetes?” We selected interviewees to represent a variety of responses to this and other survey items about MHV use, including those who used a variety of MHV portal features, those with controlled and uncontrolled diabetes, urban and rural patients, and those with or without comorbid mental health diagnoses. Women and minority veterans were oversampled to broaden the representation of patient demographics. More details regarding our survey sampling methodology [25] and qualitative sampling methodology [32] are available elsewhere. This study was approved by the local institutional review board.

We performed 2-tailed t tests, chi-square tests, and correlation analyses to examine differences between rural and urban participants and relationships between covariates and model measures. Moderated mediation was used to address the first 2 study objectives. Moderated mediation (Figure 1) estimates the indirect effect (SM use on A_1c%TIC through diabetes self-management; research objective 1), and whether this indirect effect is conditional on values of a moderator (rurality; research objective 2). Analyses were conducted using Hayes’ PROCESS model in the SAS Enterprise Guide [37]. Moderation of the mediation model by rurality was assessed by calculating the index of moderated mediation [38] between rurality and the indirect effect between SM months in 2017 and A_1c%TIC in 2018. The index of moderated mediation with a dichotomous moderator is defined as the difference in the indirect effects, or mediated effects, between the 2 levels of the moderator (rural and urban). The test of this index is assessed by generating a bootstrap 95% CI of the difference in indirect effects across moderator groups. Effects were considered significant if the 95% CI did not include 0.00 (P<.05). Qualitative analysis of the SM was used to further understand the nature of the secure message content and patient perceptions of SM (research objective 3).

More months using SM was significantly and positively associated with greater diabetes self-management (B=0.12, 95% CI 0.033-0.212; P=.007; a in Figure 1). Rurality influenced the strength of the relationship between SM use and diabetes self-management (B=–0.08, 95% CI –0.138 to –0.026; P=.005). When we examined the conditional effects of SM on diabetes self-management for rurality, there was a trend to a significant positive relationship between SM and diabetes self-management for rural patients (B=0.04, 95% CI –0.001 to 0.083; P=.06) and a trend toward a negative relationship between SM and diabetes self-management among urban patients (B=–0.04, 95% CI –0.080 to 0.002; P=.06).

Patients who reported greater diabetes self-management had significantly higher A_1c%TIC (ie, more time in control of their diabetes throughout the year; B=10.38, 95% CI 5.539-15.217; P<.001; b in Figure 1). There was no direct effect of SM use on A_1c%TIC (B=0.09, 95% CI –1.463 to 1.651; P=.91; c’ in Figure 1). However, there was a conditional indirect effect between SM use and A_1c%TIC, via diabetes self-management for rural patients (B=0.42, 95% CI 0.004-0.927; Table 3 and ab^Rural in Figure 1). This conditional indirect effect represents the change in A_1c%TIC for every month of SM use, mediated by diabetes self-management. Among urban patients, there was no indirect effect between SM use and A_1c%TIC via self-management (B=–0.42, 95% CI –1.039 to 0.056; ab^Urban in Figure 1). The index of moderated mediation (ie, the difference between rural and urban indirect effects) was significant (index=–0.85, 95% CI –1.64 to –0.23).

This study modeled SM use in 2017 and A_1c%TIC in 2018. Had we examined both SM use and glycemic control in the same year, we would have risked potentially having some participants with SM data toward the end of the year and HbA_1c measurements in the beginning of the year. These data would not be consistent with the hypothesized temporal nature of the analysis. However, as sensitivity analysis, we compared SM use in 2017 and 2018. SM use in 2017 and 2018 were significantly correlated (r=0.53, P<.001). Additionally, we ran the moderated mediation model using both SM use and A_1c%TIC in 2018. A similar pattern of results occurred in a moderated mediation analysis that examined both SM use and A_1c%TIC simultaneously in 2018. Further information is included in Multimedia Appendix 1.

Qualitative analysis of the SM content revealed that significantly more rural participants (77%, n=177) discussed diabetes-related content in at least one SM thread than urban participants (67%, n=146; P=.01). There were no other significant differences between the proportion of urban and rural participants who engaged in at least one thread related to other health topics codes. Semistructured interviews with a subset of survey respondents further expanded on how rural patients perceived SM and were using SM (Table 4). Patients consistently expressed how SM helped them communicate with their clinical teams. Rural patients indicated that SM was a convenient tool to support tasks pertinent to effective diabetes self-management. For example, one patient reported that SM was a more reliable form of communication than through a cell phone to set up appointments or medication renewal requests. Patients also indicated they were able to use SM to communicate their diabetes-related equipment needs with their clinical team. Patients also reported that SM allowed them to communicate with various members of their clinical team.

In a population of veterans with diabetes, we examined the relationship between the use of SM and percent time in glycemic control, whether diabetes self-management behaviors mediated this relationship, and if the use of SM is beneficial for those both living in urban and rural areas. This study leveraged mixed methods to quantify these relationships through a moderated mediation analysis, examine how patients with diabetes use SM through a message content analysis, and learn from patients through qualitative interviews. Moderated mediation analysis revealed that the relationship between the use of SM, diabetes self-management, and A_1c%TIC was influenced by rurality. Among rural patients, increased use of SM was associated with a higher A_1c%TIC through diabetes self-management. The mediation of SM and A_1c%TIC through diabetes self-management was not found among urban patients. This finding does not indicate that SM is not necessarily beneficial for urban patients; rather, it indicates that SM may help support rural patients’ diabetes self-management efforts to a greater extent than among urban patients. In addition to the challenges of effective diabetes management, rural patients face additional barriers including limited access to diabetes education and clinical services, limited cell phone coverage and internet access, limited transportation, and long travel distances [39]. It is possible that the enhanced clinical access afforded by SM may not influence self-management among urban patients who do not face the same access barriers as their rural counterparts [10]. SM offers rural patients means to overcome many of these barriers.

Our quantitative analysis included all SM communication (ie, not just diabetes-specific SM) as many different subjects, such as messages about hypertension or physical activity, are likely to be helpful for diabetes management. We used qualitative analyses to further explore the ways in which rural patients leverage SM for diabetes self-management. Rural patients were more likely than their urban counterparts to communicate via SM with their health care team about diabetes-related content, which may be associated with more effective diabetes management efforts. While messages about other health topics may be just as important for diabetes management, there were no significant differences in the frequency in which these other health topics were discussed between rural and urban patients. Additionally, participant interviews revealed insights into some of the benefits SM affords rural participants, such as SM being a more reliable and convenient means to communicate with various members of their clinical team to engage in activities important for diabetes management (eg, appointment requests, medication renewals, and equipment requests).

This relationship between increased health care team access and greater self-management aligns with previous research; a systematic review evaluating technology-enabled diabetes self-management support concluded that 2-way communication between the patient and clinical team was an essential component for improved HbA_1c [40]. Patients who use web-based portals and SM can communicate with their team more regularly, as needed, and potentially reduce the need for in-person visits. Reports on the relationship between SM and in-person health care utilization are inconsistent. For example, we found that greater use of SM was positively associated with more in-person health care utilization, whereas other recent work has found that use of SM was associated with a decrease in in-person utilization [41]. It is difficult to disentangle if patients are using SM in place of in-person care, or if they are using SM because of an upcoming or recent in-person visit (eg, following up on a new medication). Owing to this potential confounder, we included in-person primary care visits as a covariate in our model to control for health care utilization and possible confounding by indication.

More consistent use of SM, particularly SM related to diabetes, can help overcome commonly reported regional disparities in diabetes self-management and glycemic control. Despite the benefits of SM for diabetes self-management and glycemic control in rural veterans with diabetes, rural patients are less likely to manage personal health information on the internet or communicate through the internet with their providers [30]. External support from a patient’s clinical team has been identified as a key facilitator of diabetes self-management [39], though such support is less available for patients with limited access to in-person visits. Fortunately, virtual modalities such as web-based patient portals and features including SM can provide easily accessible support for effective diabetes self-management. It is critical to identify methods that will promote patients’ use of web-based portals for better chronic disease management. Technology-based approaches and interventions are widely accepted for promoting diabetes self-management in rural communities [42]. Additionally, we previously found that as little as one team-initiated secure message was significantly associated with better diabetes self-management [25]. Providers may find that encouraging patients, particularly rural patients, to use SM may significantly improve their diabetes self-management and outcomes.

SM has the potential to reach an ever-increasing number of patients. As of July 2021, 3.7 million veterans (more than half of active VA patients) were registered portal users, of which 1.4 million were active users of SM. Increasing SM use can be considered a high-reach, light-touch intervention with the potential to improve population health. Understanding the benefits of modalities that can provide more accessible diabetes self-management support not only has implications for rural patients who typically face barriers accessing in-person health care owing to long travel distances—these findings also support the value of encouraging SM use when in-person visits are not feasible. During the COVID-19 pandemic, VA facilities were directed to convert in-person to virtual care whenever clinically appropriate [43] and for rural patients in particular [44]. Use of SM can help maintain patient-provider communication and support disease self-management when patients cannot access in-person care. Emerging evidence suggests that disparities in rural patients’ access to telemedicine, including video visits and portals, have persisted despite dramatic increases in adoption [45]. Our findings suggest that efforts to reduce these disparities are important not only to improve equity but also to support improved outcomes.

This study has some limitations. For one, this sample purposively surveyed patients who were both recent and repeated users of patient portals; it does not speak to the potential benefit of SM in those who have never used portals. Those who responded to our study may, as a group, have had better self-management than the average patient with diabetes. Indeed, our sample scored higher on the DSMQ than other populations, though not outside the SD [35]. Similarly, as is common in many US Veteran studies, our sample size was mostly male, which limits the potential generalizability of these findings to females and non-Veterans.

The current analysis examined self-management as a composite score. Future research may examine the relationship between SM use and various self-management behaviors, and if certain self-management behaviors are more important in the relationship between SM use and A_1c%TIC. Finally, another limitation is the cross-sectional and observational nature of the study. Our mediation model allows us to begin to think about the causal nature of these relationships. Future studies might benefit from interventional designs that examine changes to diabetes self-management and glycemic control after initiating SM use compared to a sample who have never used SM.

On average, patients with diabetes who live in rural areas are disproportionately affected by diabetes, in part owing to their limited access to health care. Among rural patients, greater use of SM was associated with better diabetes self-management, which was associated with better glycemic control. This was not observed among urban patients. Rural patients with diabetes may benefit significantly from using SM to support their diabetes self-management and diabetes-related outcomes. Encouraging patients to ask questions between visits, or reaching out to them directly via SM, are examples of light-touch interventions with potential to improve outcomes for millions of patients with diabetes who lack ready access to in-person care.