The incidence of type 1 diabetes (T1D) is rising [1], and T1D results in significant economic costs in the United States, with yearly medical expenditures estimated at approximately $7 billion with an additional $7 billion in lost wages [2]. T1D also significantly increases mortality, especially among those with above target hemoglobin A_1c (HbA_1c) levels [3]. Young adults are a population at unique risk, with only 14% of young adults aged 18 to 25 years meeting the target HbA_1c goal (HbA_1c≤7%) versus 30% of those over 30 years [4]. One in 4 young adults aged 18 to 35 years already have one or more medical complications related to their T1D, most commonly renal problems reflecting micro- or macroalbuminuria and/or retinopathy [5]. Furthermore, young adulthood is a critical developmental period when adult habits are formed as patients transition from parental involvement with diabetes management to independence in self-management of their T1D [6,7].

Despite the unique clinical needs of patients in this age group, few interventions have been tested for this high-risk population. A 2017 systematic review found 18 intervention studies for young adults with T1D [8]. Across studies, the most common intervention strategy (13/18, 72%) targeted engaging young adults with clinical services, an important goal, but unfortunately one that did not routinely result in improved glycemic control in most trials. Only 67% (12/18) of studies reported HbA_1c outcomes and, of these, only 2 were randomized, both showing no impact on HbA_1c. Since this review, several protocols and intervention development studies have been published [9,10]; several reported the results of uncontrolled studies with none showing impact on HbA_1c to date [11-13] and one showing significant effects on HbA_1c for continuous glucose monitor (CGM) use versus blood glucose meters [14]. These results highlight a major gap in and need for more rigorous research on effective ways to improve glycemic outcomes among young adults with T1D.

The proposed intervention model offers a multipronged self-regulation approach for targeting glycemic control that is tailored for young adults. The goal of the selected intervention components is to improve self-regulatory mechanisms [15] including self-monitoring, goal setting, self-efficacy about diabetes management, and problem-solving skills. These self-regulatory mechanisms promote improved T1D regimen adherence and HbA_1c [16-20]. The conceptual model in Figure 1 highlights the role of self-regulation as an intervention target leading to improved outcomes. To target self-regulation among youth with T1D, we developed a multicomponent intervention that includes (1) weekly diabetes device data upload and data review designed to promote healthy self-monitoring and goal-setting habits for diabetes management and provide feedback about goal attainment, (2) web-based human coaching to deliver motivational interviewing exercises and teach a structured problem-solving method, and (3) motivational incentives to enhance adherence.

A series of prior iterative studies developed and tested this intervention approach, including a randomized trial comparing a similar intervention to usual care for adolescents aged 13 to 17 years. Intervention youth had significantly lower HbA_1c levels at the end of the 6-month intervention (d=.45), and this effect was fully maintained indicating no weakening of the intervention effect at the 12-month follow-up (d=.44) [21]. Our new study adapted this earlier intervention to enhance both efficacy and disseminability. The primary modification to increase efficacy involves expanding the goal target from glucose checking to encompass the additional key self-management behavior of carbohydrate counting. Enhancements to promote dissemination (and scalability) include (1) modifying goals to encompass the full spectrum of diabetes devices, (2) automating the goal setting and feedback components via an app, and (3) recruiting young adults via social media. In addition, this study uses a factorial design to test the independent effects of incentives to promote consistent daily adherence to goals and web health coaching to teach effective problem solving focused on personalized barriers to self-management, providing better understanding of the intervention mechanisms.

Incentives may be an effective tool for increasing self-management behavior. Because interventions that support self-monitoring of diabetes management have shown limited effects on HbA_1c, incentives are designed to enhance the impact of the intervention on HbA_1c. Consistent with behavior economic theory [22], most daily adherence behaviors necessary to manage T1D do not result in immediate positive experiences. The benefits from consistent, daily adherence accrue over weeks, months, years, and decades of life. The use of immediate incentives for adherence is one way to increase the value of such behaviors in the present, providing an immediate reason to adhere. In the SweetGoals intervention, incentives target improvement in specific self-management behaviors, and such improvement is expected to improve glycemic control. Three studies [23-25] using incentives for glucose checks have shown significant positive effects, as have our prior studies [21,26,27].

This study will also test the impact of a coaching intervention focused on enhancing motivation and teaching problem-solving skills to promote long-term outcomes. There is evidence that motivational interviewing and instruction in problem-solving skills can improve medical adherence including in T1D [28-35]. Health coaches will teach these skills in brief web-based sessions in the context of device data review focused on actionable self-management targets. The curriculum begins with motivational exercises to guide selection of concerns that become the target of the problem-solving sessions. The health coach also facilitates engagement with self-monitoring and goal-setting habits using the Glooko data visualization platform and the SweetGoals app goal feedback, consistent with the supportive accountability model [36]. Providing such support increases the efficacy of digital interventions [37-42], and delivery by bachelor’s level coaches can be as effective as professional clinicians across diverse clinical targets [43,44]. The supportive accountability model emphasizes the key role that social presence (human coach) plays in setting clear expectations regarding adherence to the steps necessary to achieve a positive outcome (eg, glucose monitoring) and in supporting goal attainment via progress monitoring and feedback [36]. Our pilot results strongly support the long-term sustained efficacy of this coaching approach combined with incentives [21], but those results cannot inform the need for both of these distinct interventions. This new study will replicate those earlier results in a novel population of young adults and test the separate impact of incentives, coaching, and their interaction.

We will enroll 300 young adults with T1D, aged 19 to 25 years (target 50% female) who have HbA_1c≥8%. Young adults must use a Glooko compatible glucometer or CGM and may use either multiple daily injections (MDI) or continuous subcutaneous insulin infusion (CSII). See Textbox 1 for detailed inclusion and exclusion criteria.

Participants will be recruited using Facebook, Instagram, and Google ads. Best practices for ethical recruitment using social media platforms will be followed [45,46]. Breaks will occur between ad runs to reduce ad fatigue. The ads will prompt potential participants to download the SweetGoals app. Once they confirm that they are in the target age range, they will complete the study consent in the app. Consenting participants will complete a brief survey via the app to confirm eligibility. Research staff will follow up to complete the screening process, confirming device compatibility with Glooko and arranging for completion of the HbA_1c test via postal mail. Once eligibility is confirmed (mail-in HbA_1c≥8%), staff will help participants create their Glooko accounts and install the Glooko uploader and Glooko smartphone app via phone or a web video chat. Staff will then complete the randomization as described below. Figure 2 displays the screening, recruitment, and randomization process.

A fully powered 2×2 randomized factorial experiment (see Table 1) will be used to evaluate specific intervention components in terms of HbA_1c outcomes. Factorial designs are a highly efficient experimental approach for answering questions about the utility of multiple intervention components and their combinations. Data from these designs can be used to test the main effect of each intervention component as well as their interaction [47,48]. All participants will receive the SweetGoals app. Those assigned to the incentive group will receive incentives for meeting glucose monitoring and mealtime targets. Those assigned to the coaching group will receive web coaching in problem-solving skills focused on achieving better self-management and glycemic control. If assigned to incentives, participants will receive incentives weekly for 3 months, with gradually fading frequency over the next 3 months. If assigned to coaching, participants will meet with the coach weekly via web video for 3 months, with the frequency of those meetings also fading gradually over the next 3 months. The intervention period lasts for 6 months. Follow-up assessments will be completed at 6 months and 12 months after baseline.

Each component will take on 2 levels: yes or no. Note that the factorial design in Table 1 is not a 4-arm trial with each condition compared with a control or to each other. Instead, our interest is in tests of standard analysis of variance main effects and interactions [49]. These involve comparison of outcome means across multiple experimental conditions. For example, the main effect of the incentives component will be tested by comparing the mean of the outcome variable across the 2 conditions in which incentives will be delivered (ie, those in conditions 3 and 4; n=150 before attrition) versus the 2 conditions in which incentives will not be delivered (ie, those in conditions 1 and 2; n=150 before attrition). Hence, this main effect will be tested by comparing half of the sample (those offered incentives) versus the other half (those not offered incentives) in terms of the primary outcome. In this factorial experiment, each effect (including the interaction between the 2 components) will be estimated based on data from all 4 conditions (ie, the full sample) [48,50,51].

Figure 3 shows the study design. An online minimization program (MinimPy) will be used to assign participants [52]. Minimization assures similarity across intervention groups on multiple key covariates [52]. Differences between conditions will be minimized on gender, age, ethnicity (minority vs White), CSII versus MDI, CGM use, and HbA_1c. Follow-ups will occur at 6 months and 12 months. Compensation will be $25 for the baseline assessment, $50 for each follow-up assessment, and a supplement of $50 for completing both follow-ups. This study was approved by the Committee for the Protection of Human Subjects at Dartmouth College and is registered at ClinicalTrials.gov [NCT04646473].

All participants will receive SweetGoals, an app that sends messages about self-management goals, goal adherence, and encouragement about adherence in the upcoming week. Participants will receive messages on Sunday reminding them to upload their data and feedback detailing their goal achievement on Monday. They will also receive educational materials in a second message most weeks (eg, links to information about self-management of T1D). The app is programmed in MobileCoach [53-55], an open-source app platform that has the functionality to integrate our goal-tracking algorithms and provide messages that prompt device uploads and provide automated feedback. MobileCoach sends messages written by the research team in the style of Facebook Messenger or WhatsApp from a digital coach who communicates with the app user to provide scripted feedback about goals as shown in Figure 4. Participants select from a random sample of 4 coaches offered from a bank of 12 coaches of diverse gender and race/ethnicity. The participant (Briana in Figure 4) can answer the digital coach (Taylor in Figure 4) using predefined responses.

The SweetGoals app will automatically download the participant’s device data each Sunday at midnight using the Glooko application programming interface. The app will use the device data to provide feedback each Monday about goals that target glucose monitoring and mealtime behaviors tailored by device as shown in Table 2. The glucose monitoring goal for participants who use a glucometer only (no CGM) is ≥5 checks, each ≥2 hours from another check. Additional checks—regardless of spacing—are encouraged based on clinical needs (eg, hypoglycemia). This same criterion was used successfully in prior studies [21,26,27]. For participants who use CGM, we conservatively define adequate daily CGM wear time as 80% of expected values each day. We selected this daily glucose monitoring goal to provide clinically meaningful data to the participant but also allow for legitimate disconnect time (sports, leisure activities) and sensor changes.

The mealtime goal for participants who use a glucometer only (no CGM) is to check their glucose before they enter their carbs. Those who use MDI will enter their carb counts in the Glooko app. Participants who use CSII will enter their carb value in their pump. A glucose check must be documented within 30 minutes of the carb entry, an evidence-based criterion [56] that should occur at least 3 times per day according to clinical practice guidelines [57]. An algorithm based on the type of devices used will evaluate whether the mealtime goal was met. The daily mealtime target is 3 properly timed paired glucose levels and carb count values.

The weekly goal for all participants is to meet the daily goal on 6 or more days per week. Each week, participants will also be asked to set personal goals for the next week for the number of days they think they can meet each daily goal, to help participants build motivation to gradually increase their self-management behaviors. Each week, participants will receive feedback messages and graphical reports about goals met (see Figure 5 for an example of a participant who uses a CGM). The app will send reminders to upload device data and set personal goals.

Participants assigned to the incentive group will earn $1 per day for meeting each daily goal (glucose monitoring, mealtime behaviors) and will receive a $5 bonus for meeting the weekly target of 6 or more days per week meeting each goal, for a total of $24 maximum per week (approximately $3.50 per day). The maximum earnings for all goals are $600 across 25 weeks. The app will provide messages and graphical reports about incentives (see Figure 6). Incentives accrue weekly throughout the 25 weeks but are paid weekly from weeks 1 to 11, with payments fading in frequency from weeks 12 to 25 (ie, paid at weeks 13, 16, 20, and 25). Fading of reinforcement delivery has been shown in human and animal research to engender resistance to extinction after reinforcers end, increasing the maintenance of behavior change beyond the intervention period [58-62]. For this reason, we have incorporated a lengthy (3-month) fading period to promote maintenance of improvements in daily self-management.

This fading procedure is used to promote habit automaticity and long-lasting improvements in adherence and HbA_1c. We have successfully used this fading paradigm in our prior studies, which resulted in no weakening of intervention effects on HbA_1c up to 6 months after the end of the intervention.

Participants assigned to the coaching group will receive 16 web health coaching sessions by a bachelor’s level health coach. Weekly coaching sessions occur from weeks 1 to 11, and then fade over the second half of the intervention with 4 sessions held during weeks 12 to 25 (weeks 13, 16, 20, and 25). Coaches will use their own chat interface in SweetGoals and text messages to communicate with participants. Coaches will also remind participants of meetings, offer encouragement, and respond to any messages the participant sends. In each meeting, coaches will review weekly glucose monitoring and mealtime goals and incentives earned in the context of diabetes device data reviewed using the Glooko website. Motivational exercises are completed in the first few meetings to develop rapport and allow the young adult to explore and reflect on their self-management strengths and challenges [63]. The young adult will work together with the coach to identify barriers to adherence. Coaches will also teach a structured problem-solving method to the young adult [64].

During the second half of the intervention, as coaching sessions are spaced at increasing intervals, participants are encouraged to review their diabetes data weekly on the Glooko website or app and complete the problem-solving steps on their own to address self-management concerns. Success and challenges with independent problem solving are reviewed at coaching sessions during the fading period. Throughout the intervention, if the young adults experience challenges meeting glucose monitoring or mealtime goals, coaches encourage them to use the problem-solving steps to develop a plan to improve adherence. As they gain more success meeting those goals, they are encouraged to select new goals regarding glucose levels and problem solve potential barriers to achieving those goals. Participants are always encouraged to reach out to their providers with any concerns about the insulin regimen, hypoglycemia, or hyperglycemia as coaching does not address medical management of diabetes. The structure of the problem-solving method is constant across participants; however, the content is highly personalized based on the unique challenges faced by each young adult.

Coaches receive training in communication skills consistent with a motivational interviewing approach [65]. Importantly, motivational interviewing has been identified as an effective approach for improving diabetes outcomes clinical trials [35,66,67]. Coaches are also trained in providing instruction and support in problem-solving [64,68]. Training involves didactic material and extensive role play practice.

Participants in all conditions will receive ongoing diabetes treatment from their current medical provider. Study staff will not provide medical care or intervention. The project includes an endocrinologist who will monitor device data regularly in Glooko, and research staff will encourage participants to follow-up with their medical provider as necessary.

Measures will be collected at baseline, 6 months (end of intervention), and 12 months unless otherwise noted. Assessments will not be blinded. All measures will be completed by participants online via the app or LimeSurvey or are objective measures (eg, device downloads, HbA_1c tests, or recorded directly via the app).

All subjects, once randomized, will be included in the intent-to-treat sample. We will strive to collect all primary and secondary outcomes even if a participant does not engage in assigned interventions. Data will be stored in the Research Electronic Data Capture, MobileCoach, and a secure Dartmouth server.

This study focuses on investigating the utility of two intervention components, seeking to optimize an intervention for improving glycemic outcomes among young adults with T1D. Innovations include recruiting participants with T1D into an intervention via social media and intervention delivery via an app plus coaching and incentives both delivered remotely, strategies that may be vital to reaching the targeted, understudied, and underserved population of young adults. In addition, this is the first intervention designed to target adherence via incentives among patients using diverse glucose measurement methods (glucometer, CGM) and methods of insulin delivery (MDI, CSII). This trial also makes innovative use of existing diabetes device technology by integrating a diverse array of devices with an app to automate tracking of and feedback on daily adherence habits. This digital and web-based model is designed to be broadly applicable across the range of health conditions in which patients struggle with self-management.

The study design was informed by research evidence across multiple domains. For example, the focus on young adults was based on evidence suggesting they are at unique risk for above target HbA_1c levels and that periods of above target HbA_1c levels during these years have a long-lasting negative impact on health [96]. We chose to focus on early young adulthood (ages 19 to 25 years) due to their higher risk for above target HbA_1c [97] and possible developmental differences across the later 20s and early 30s. Further, we chose to target young adults outside the traditional clinic setting directly via social media, empowering them to address their diabetes self-management. Many young adults are less compliant with obtaining regular medical care [98], which suggests it may be important to offer services to young adults with T1D outside the endocrinology setting to problem solve their barriers to clinical care and encourage them to schedule regular visits with their provider. This proposed model focuses on directly engaging participants rather than targeting changes in provider behavior. As such, it offers a practical outreach approach that could be deployed nationally and adopted by medical practices, health systems, or insurers outside of office visits to engage young adults with T1D in working together with their provider and managing their own health.

This study is designed to test the role of two evidence-based behavioral intervention components (incentives and health coaching) in supplementing a mobile health approach to improve glycemic outcomes among high-risk young adults with T1D who have above target HbA_1c levels. We will also test the effect of these intervention components on key self-regulatory mechanisms hypothesized to be impacted by these components, including two aspects of adherence (glucose monitoring and mealtime behaviors), plus self-efficacy and problem solving. Overall, results will provide critical data on enhancements to a digital intervention for T1D that are highly disseminable. They will also advance the field of theoretically driven interventions aimed at improving self-management and glycemic outcomes among high-risk young adults with elevated HbA_1c levels.