As most of diabetes care occurs in outpatient settings and involves ongoing patient self-management, mobile health (mHealth) technologies may greatly improve diabetes management and health outcomes. mHealth involves the use of mobile devices to support continuous health monitoring and healthy behaviors [1]. Mobile devices include mobile phones and sensors that are worn, carried, placed in the physical environment, or accessed by individuals during normal daily activities [2]. These devices allow reporting of patient data such as blood glucose through a wireless glucometer, weight through a cellular-enabled scale, and physical activity through a wireless accelerometer in near real-time in the patient’s daily environment. Moreover, these data can be transmitted to clinicians and health systems and may lead to the development of precision health strategies [3].

According to the Pew Research center, more than 92% of US adults own a cell phone, and more than 77% own a smartphone [4]. Furthermore, 84% of low-income individuals in the United States now own a cell phone, and almost 70% own a smartphone. Low-income racial/ethnic minorities are actually more likely than low-income whites to own mobile devices and to use features such as SMS (short message service) text messaging or smartphone apps [4]. Research demonstrates that mHealth tools are useful for diabetes self-management [5,6] including collaborative decision making between providers and patients [7]. Thus, mHealth technologies have great potential to facilitate wide delivery of diabetes treatment and enhance the development of self-management tools for diverse populations.

The study protocol described in this manuscript extends current research by using multiple mHealth technologies to provide diabetes-related data to help patients and their clinicians better understand illness dynamics and develops personalized approaches through data visualization to improve health outcomes in type 2 diabetes mellitus (T2DM). Specifically, we describe methods for identifying precision health strategies to help patients self-manage using multiple types of self-generated diabetes-related data. We will also demonstrate how clinicians can help guide near real-time patient self-management by collecting and aggregating streams of health data from multiple mobile technologies and then creating a variety of data visualizations that we present to both participants and clinicians. Our goal is to understand what kinds of visualizations patients of varying backgrounds and clinicians need. Furthermore, we describe more about the need for understanding what kind of alerts patients might find useful to self-manage their diabetes. These alerts, or algorithms, will be developed from health data from the mobile technologies and associated electronic health record (EHR) data from patients.

The aims of this project are to examine the feasibility and utility of having patients self-monitor multiple types of diabetes-related data (blood glucose, weight, physical activity, medication adherence) using mHealth technologies (wireless glucometer, cellular-enabled body scale, wrist-worn accelerometer, and medication adherence SMS text message surveys). This will allow us to examine trajectories of diabetes-related variables (blood glucose, weight, physical activity, medication adherence) and challenges in self-management that patients face at points in time. We are also exploring the challenges and successes of patients self-managing diabetes using mHealth technologies through interviews and data visualizations. And finally, we are exploring clinicians’ perspectives and input on using these data to develop algorithms for software that will facilitate patient self-monitoring and meet near real-time self-management needs. Results from this study will be used for the integration of data from mHealth tools into EHRs and for developing new models of care delivery to support diabetes management.

The study protocol is supported by the adaptive leadership framework [8], which divides health challenges into 2 types: technical and adaptive [9,10]. Technical challenges such as how and when to measure blood sugar and what the appropriate medication dosages are to manage diabetes are addressed by the clinician with technical solutions. Adaptive challenges require the patient to adjust to new conditions and do the work of learning and making behavioral changes, for example, incorporating exercise into everyday life. Adaptive leadership is the work that clinicians do to help patients perform this work [11]. Using multiple types of self-generated diabetes self-monitoring data, clinicians can support patients in near real time to respond to adaptive challenges.

Figure 1 (adapted from National Institute of Nursing Research 1P30NR014139-01) shows how mobile technologies might help patients monitor behavioral, symptom, and biophysical trajectories. These trajectories, shown in the top half of the model, suggest these variables are dynamic. For example, when patients are able to overcome challenges in scheduling exercise and medication adherence, these behaviors increase, and their blood sugars are likely to decrease. The goal of adaptive leadership is for clinicians to address technical challenges, such as finding the appropriate medications and dosages, and also to help patients address adaptive challenges such as lifestyle changes by facilitating their adaptive work for self-management. Over time, as patients address the challenges of self-management and their adaptive work increases, the amount of technical work needed by clinicians will decrease. The trajectories of the signs and symptoms of T2DM, such as high blood sugar, will also decrease. The lower half of the model illustrates that mobile technology can provide important information to be used in the collaborative work-relationship to facilitate both the adaptive work of patients and the technical and adaptive approaches used by clinicians to support patients. According to this framework, patients and clinicians collaborate to monitor symptom dynamics and self-management techniques. Together, they assess adaptive challenges and plan the technical and adaptive work needed to help patients meet their diabetes goals such as weight loss, medication adherence, and so forth. We considered this framework when designing the protocol described below including the questions we ask both patients and providers.

We began with software development (Figure 2). For Phase 1, we selected consumer-friendly devices that patients could easily use for their diabetes self-management. We chose a wireless glucometer by iHealth, a wrist-worn accelerometer by Fitbit, and a cellular-enabled scale by BodyTrace. Using Prompt, a research platform designed to collect, analyze, and message patients about mHealth data [12], we programmed the ability to pull in data from these respective companies via their application programming interfaces (APIs). Every day, Prompt uses an authentication token to request data via each API for each participant in the study. Those data are collected over the course of the study and stored in a secure database. A study coordinator is also able to view data in aggregate that participants transmit over time. In addition, Prompt was programmed to send out a scheduled SMS text message with a survey link every 2 weeks. This link allowed participants to complete a short survey about their medication adherence via their phones in the Web-based Research Electronic Data Capture (REDCap) Web-based platform hosted at Duke University [13]. REDCap is a secure, Web-based application designed to support data capture for research studies, providing: 1) an intuitive interface for validated data entry; 2) audit trails for tracking data manipulation and export procedures; 3) automated export procedures for seamless data downloads to common statistical packages; and 4) procedures for importing data from external sources.

Following software development, we received institutional review board (IRB) approval for phase 2. In this phase, we are conducting a mixed methods explanatory sequential designed study [14] for which we recruited 60 adults (aged ≥18 years) with T2DM. This is a 2-phase design where quantitative data are collected and followed by qualitative data collection. Patients were recruited from a primary care clinic associated with an academic medical center in the southeastern United States and through local advertisements. Eligible participants met the following inclusion criteria: (1) aged ≥18 years, (2) able to speak and read English, (3) diagnosed with T2DM, (4) told by their primary care provider to monitor their blood sugar, (5) own and use a smartphone, (6) capable of giving informed consent, and (7) able to travel for study enrollment. Participants were excluded if they had a preexisting severe medical condition(s) that would interfere with study participation (eg, renal failure, severe orthopedic conditions or joint replacement scheduled within six months, paralysis, or cancer). A proactive effort was made to enroll patients at various stages of diabetes mellitus, with treatment regimens including both oral and injectable medications were included. We purposefully targeted patients with a range of diabetes severity and included at least 25 patients with a hemoglobin A_1c (HbA_1c)>7.5%. Patients were not required to have Wi-Fi or in-home internet; a smartphone with an internet connection was adequate. All other devices (eg, glucometer, cellular-enabled scale, and accelerometer) were provided to patients. Patients were excluded if they (1) had active dementia or psychiatric illness, (2) resided in a nursing home, or (3) were participating in another self-management study.

We used a study recruiter to identify patients by reviewing EHRs at primary care clinics. Eligible participants received a letter inviting them to participate. Those who were interested contacted the study recruiter and were screened over the phone. Baseline appointments were scheduled. We also used Web advertisements and posted flyers to recruit half of our participants from the local community.

No changes to eligibility criteria or outcomes changes were made during the study. We did change clinician focus groups to interviews. Interviewing emerged as a better opportunity to discuss various data visualizations and to receive feedback about workflow integration among diverse clinicians

All statistics will be calculated using SAS Version 9.3 (SAS Institute Inc.). Baseline demographic and clinical characteristics of enrolled participants are summarized by mean and SD for continuous variables and frequency and percentage for categorical variables. A significance level of 0.05 is considered as statistically significant for all tests.

Using the data collected in Prompt, we will create data visualizations using 2 different data visualization software packages (Tableau and R) and pilot their use with participants via a phone interview. We will use the interviews as an iterative process to discover how to present the visualizations back to patients and to refine the visualizations. We will start with plotting the actual data transmitted from each of the devices from individuals over the entire follow-up time to show patients overall trend in blood glucose, exercise level, and weight. For blood glucose data, we will add a band of ideal blood glucose range to help patients visualize if their measurements are within targeted range. For Fitbit data, we will plot both the average active steps daily and weekly over time and also plot the steps by weekday to explore if there are any cyclic and long-term drift trends in their exercise level. For weight data, similarly, we will plot both daily weight and weekly average to show overall trend. Further details on this approach are described elsewhere [24].

This study provides an overview of the methodology used in a study examining the feasibility of collecting near real-time data from mHealth technologies for patients with T2DM. Determining whether it is feasible for patients to use multiple mobile devices to self-manage their diabetes is an important first step in developing effective personalized care delivery strategies.

In addition to exploring the use of mobile technology to manage chronic illness, our approach has several strengths. First the use of mixed methods to address our aims is a strength of this study. The convergence of both quantitative and qualitative elements, data collection techniques, and analyses will facilitate a more in-depth and comprehensive approach to and understanding of diabetes self-management [27]. Another strength of our study is the use of data visualizations during participant interviews as an innovative approach to enhance discussion and gain insight into the challenges and successes participants experienced in diabetes self-management with mobile devices over the past 6 months. Presenting data this way could stimulate discussion and improve communication by graphically displaying patterns and trends [28-30].

We do acknowledge several limitations to this study. The first is the limited sample size, which does not allow us to draw statistical conclusions from the data. We are not testing the effect of any intervention, so cannot assess the impact on diabetes outcomes of using these technologies. Finally, we were not able to integrate these data into the EHR to assess pragmatic aspects of the use of the data at this time within the health system.

Our study is among the first to seek answers to the many questions related to integrating patient data from mobile devices into diabetes self-management care delivery models and EHRs. Questions such as how long patients will track multiple types of diabetes-related data, which strategies will best help patients self-manage their diabetes using self-generated data, and how clinicians might effectively guide patients to better manage their disease in near real time need to be addressed in the emerging era of digital health.