Recruitment will take place in the Sparrow Hospital’s Pediatric Subspecialty Clinic in Lansing, Michigan, United States. The Sparrow Endocrinology Clinic team consists of 1 registered pediatric dietitian, 1 full-time nurse practitioner, 2 certified diabetes educators, and 2 endocrinologists, all of whom provide diabetes care 4 days per week within the larger subspecialty clinic. All patients are scheduled to come to the clinic every 3 months for care.

The clinic nurse diabetes educator will act as the site coordinator. Participants for the pilot testing will be recruited from the hospital’s Pediatric Endocrinology Clinic. Invitations will be mailed three times over 2 months to all of the potential participant dyads, as identified by hospital providers. Additionally, the health care providers in the clinics will be given a presentation on the study by the researchers so they can offer it to the families who meet the inclusion criteria during any typical interactions they have with the adolescent and parent.

If interested, the parent will contact the researchers to learn more about the study. A researcher will be in the waiting room of the pediatric endocrinologists, providing study information to possible participants. The researcher will explain the study and determine if the adolescent and parent are eligible through a brief screening questionnaire. We will recruit a total of 70 adolescent and parent dyads, which would correspond to similar past research studying T1DM and mobile phone interventions [12]. Inclusion and exclusion criteria have been presented in Textboxes 4 and 5, respectively.

During the study, there will be 4 contact points with the participants: the enrollment or baseline data collection visit (Visit 1), a midpoint review (Visit 2, 3 months after Visit 1), an end-of-study visit (Visit 3, 6 months after Visit 1), and a follow-up visit 3 months after app usage has stopped (Visit 4). Visit 1 (month 0) will be scheduled during the screening process. At Visit 1, a member of the research team will review and obtain informed consent (parent) and assent (adolescent) from the participants. The participants will then complete the baseline surveys to measure self-efficacy, quality of life, and social support and conflict; the laboratory will test the teens’ HbA_1c, and we will download data from their glucometer or pump. If a participant does not have a mobile device that can download and operate the app, the research team will provide the participant with a device and data plan for the duration of the study. The researcher will demonstrate how to use the technology and have the participants use each of the functions. At Visit 2 (month 3), a research team member will test the adolescents’ HbA_1c and download data from their glucometer or pump. The participants will also complete the survey at the midpoint. The midpoint data will be used to demonstrate whether any change at 6 months is part of an increasing trend, a plateau, or a waning trend. At Visit 3 (month 6), the participants will complete the survey and participate in interviews. Additionally, the adolescents’ HbA_1c will be tested, and the meter or pump readings will be collected. Finally, any participant that has borrowed any equipment will need to return it at this time. Qualitative methods will also be used to better examine perceptions and intentions, whether participants were light users, or why participants dropped out of the study [39]. After Visit 3, the participants will not use the app for 3 months. At Visit 4 (month 9), all the participants will complete the survey. The researchers will test the adolescents’ HbA_1c and download data from their glucometer or pump.

Specifically, the aim 2 seeks to determine the feasibility and impact on diabetes of the self-management mobile app among the adolescents and their parents. Demographic information will also be collected including current age, age at diagnosis, race, gender, parental status (ie, mom, dad, stepdad, etc), and method of diabetes treatment (eg, pump).

Hypothesis 1 seeks to demonstrate that adherence to self-management correlates with mobile app usage and a decrease in HbA_1c. HbA_1c data will be collected through the hospital laboratory system. The patient will be provided with a laboratory slip to obtain the baseline HbA_1c value. The patient will be instructed to obtain the lab results within 1 week. Once the results are obtained, they will be faxed to the principal investigator via a secure fax line. Adherence will be measured in two ways. First, the participants will complete the Diabetes Behavior Rating Scale [40]. Second, we will use data from the server to determine if they are monitoring blood sugar on schedule and what mechanisms of the app helped the adolescents follow the medical recommendations. For example, we will use time-stamped log-ins, number of sections the user visited, number of individual logs, communication through the app, educational information accessed, and duration of use per log-on.

Hypothesis 2 states that through use of the app, the participants’ self-efficacy [29,41,42], quality of life [43-45], and perceptions of social support [46-48] will improve, and conflict [49] between the child and parent will be reduced from baseline to 3 months. These secondary outcomes are key in helping the adolescents maintain adherence and a normal HbA_1c [50-52] and will be measured through surveys. The survey instruments, the participants who will complete them, and reliability for each measure can be found in Table 2.

This is a feasibility study, so all analyses will be exploratory, focusing on identifying temporal trends in the self-management mobile app usage and changes in study outcomes and identifying associations between outcomes. Preceding the analyses, we will calculate summary statistics for all measures as well as exploratory plots (histograms and box-plots) for all measures at each time point to identify outliers and potentially spurious values.

Trends over the 4 visits will be examined graphically as a function of time using nonparametrically smoothed plots to help identify when potential effects reach their maximum, plateau, and begin to decrease, which may indicate a time to target for additional interventions. At each time point, we will calculate pairwise correlations between the self-management mobile app usage measures and all study outcomes and examine bivariate scatter-plots with estimated trends for each pair. Distribution-appropriate correlations (Pearson or Spearman) will be calculated as appropriate for each measure and formally tested for statistical significance. These analyses will also be repeated using change from baseline as the outcome rather than the score itself, both as a simple bivariate analysis and using linear regression to model the change as a function of the app usage measures while controlling for baseline values. Analyses will also be repeated using paired t tests for differences in HbA_1c values from baseline to posttest to measure self-management adherence.

Statistical power will be established based on a small pilot study. A 2010 report on a small randomized controlled trial (RCT) of an mHealth intervention for T1DM in teenagers showed a significant change in self-management adherence from a mean of 3.7 (SD 0.4) at baseline to 3.9 (SD 0.4) at the 12-week follow-up, at an effect size of d=0.64 [53]. Given our longer follow-up period (6 months), we expect to observe a larger effect size in our feasibility study because HbA_1c has low variability in 3 months [54]. While using a small pilot study to calculate power may not be ideal, it provides a foundation for us to approximate this study’s sample size in order to estimate the critical parameters necessary for this study [55]. The 2010 study also reported no change in HbA_1c for the intervention group, but there was an increase in the control group, with SDs ranging from 1.2 to 1.9. Considering that their groups were imbalanced with respect to HbA_1c at baseline, it is difficult to project a difference for our study except to note that observing no change in HbA_1c may be an improvement over a possible increase in adolescents with usual care. Despite our limited sample size, assuming the use of two-sided tests with a type 1 error rate of 0.05, with 70 participants we will have 80% power to detect relatively modest correlations of .29.

In the event of unanticipated difficulties with recruitment or of attrition of up to 20%, leaving a final sample of 56, the correlation detectable with 80% power increases only slightly to .32. For t tests, N=70 provides 80% power to detect an effect size of 0.30, which is smaller than that reported in the RCT, while the effect size detectable with N=56 increases slightly to 0.34 [53].

Additionally, the primary foci of this study include assessment of the following: (1) the effect size of the intervention to power a larger study; (2) feasibility of the app; (3) practicality of data collection procedures; and (4) the ability to implement. We will also perform a post hoc analysis on personal characteristics and satisfaction related to use of the system. At the completion of the study (month 9), we will ask both those who completed the whole intervention and those who withdrew early about their experience with the intervention, including what worked well and what did not. Additionally, we will use the IBM computer usability satisfaction questionnaire to assess usability [56].

The interview data will be analyzed by developing broad code categories based upon perception themes and will serve as a preliminary sorting tool. The researchers will then use thematic analysis and create a list of common perceptions. Once a coding scheme is developed, two coders will perform a pretest by coding randomly selected interview transcripts in order to measure reliability. Coders will work together to reconcile any disagreements; this will allow us to achieve a high level of reliability.