This study was approved by the institutional review board of the first author’s university (IRB# 015117B3E) using a single institutional review board agreement covering all participating institutions. The primary caregiver and adolescent provided informed consent and assent to participate. Participants were provided with US $50 at each study visit to compensate them for their participation. The trial was registered at ClinicalTrials.gov (NCT03168867).

Adolescent participants and their primary caregivers were recruited from 3 pediatric diabetes clinics located in the greater metropolitan Detroit area and 4 in the Chicago area. The study took place from 2017 to 2021. Eligible adolescents had to be aged between 10 years, 0 months, and 14 years, 11 months, diagnosed with T1D for at least 6 months, self-identify as Black, and be residing with a caregiver who was willing to participate in the study. Study exclusion criteria were psychiatric diagnoses, such as suicidal ideation or psychosis, cognitive impairments that limited the ability to complete study measures, not being able to speak in English, or having an additional medical diagnosis leading to atypical diabetes management.

The data regarding study eligibility (based on adolescents’ age, race, and medical diagnosis) were obtained from the electronic medical records of the participating diabetes clinics, along with their contact information. Families were first sent an introductory letter describing the study. Subsequently, study research staff contacted the adolescent’s primary caregiver by phone or at a clinic visit to provide more information and screen interested families for additional eligibility criteria.

Of the 1916 families screened for participation, 1569 were ineligible, and 23 could not be contacted. Of the remaining 324, a total of 89 (27.5%) declined to participate, citing lack of interest or time. An additional 86 (26.5%) expressed an interest in the study but did not enroll before the closure of recruitment. A total of 149 families were enrolled (89 from Detroit clinics and 60 from Chicago clinics), of whom 75 were assigned to The 3Ms and 74 to standard care. A total of 5 of The 3Ms families and 1 of the standard care families dropped out of the study and did not complete follow-up data collection. The overall study retention rate was 96% (143/149). Enrollment and flow through the study are shown in Figure 1.

The study was a multicenter controlled trial using a randomized, controlled, parallel arm design. Participants were allocated to either The 3Ms plus standard medical care or standard medical care control in a 1:1 ratio using block randomization within 14 strata defined by the 7 sites and hemoglobin A_1c (HbA_1c) level (most recent HbA_1c <9.5% vs ≥9.5%) after baseline data collection. The allocation sequence was generated by the study statistician using randomization software. Assignment to condition was completed by study research staff immediately after baseline by opening a sequentially numbered, sealed envelope with the allocation.

This study was designed as an effectiveness trial to test the effects of The 3Ms under “real-world” conditions. Caregivers who were randomized to receive The 3Ms completed between 1 and 3 intervention sessions, depending on the number of diabetes clinic visits attended by the family during the 12-month intervention window. A maximum dose of 3 sessions was chosen based upon routine practice in the care of youth with diabetes [15], as standards of care include quarterly visits to a diabetes specialty care center. The first intervention session was delivered after the baseline data collection to ensure all caregivers received at least 1 intervention session. The subsequent 2 sessions were completed during any clinic visit that occurred during the 12 months after baseline.

The planned study design called for follow-up data collection visits to be completed in the family home to minimize study attrition. T2 data collection visits were completed 6 months after baseline, T3 data collection visits were completed 13 months after baseline (1 month after the 12-month intervention window was complete), and T4 data collection visits were completed 18 months after baseline (6 months after intervention completion). However, the COVID-19 pandemic caused the study’s institutional review board to place restrictions on all face-to-face contact with trial participants in March of 2020, which precluded any subsequent in-home data collection. For follow-up data collections completed after this date, participants either had HbA_1c test kits dropped at their home or were mailed the test kit to complete and return. In both cases, study staff watched the adolescents complete the test during a videoconference call to ensure reliable collection of the specimen. Due to the difficulties associated with the completion of study follow-up visits during the pandemic, the planned study design, in which follow-up data were only collected within narrow study windows (±2 weeks from the planned visit date, or 30 days in total), were modified to obtain data whenever possible within 18 months after baseline. About 87.3% (124/142), 86.7% (118/136), and 87.4% (106/121) of data collection visits were within a 45-day window of the planned visit dates at T2, T3, and T4, respectively. Study staff were not blinded to the treatment conditions; however, the objective nature of the HbA_1c measure mitigated the risk of bias.

The 3Ms intervention was delivered using Computer Intervention Authoring Software, an internet-based, interactional software [31]. Session content was delivered by an interactional and emotive 3-dimensional narrator that reads, speaks aloud, reflects participant responses, and functions as an engaging guide throughout the intervention. This approach is particularly useful in populations such as those for whom the present intervention was designed, where challenges with health literacy could affect engagement with the eHealth intervention [32]. Caregivers used a tablet computer provided to them at the diabetes clinic visit by research staff to complete The 3Ms.

The early development process for The 3Ms intervention has been reported elsewhere [17], as have the results of pilot testing [16]. In brief, The 3Ms was based on the “Information-motivation-behavioral Skills” model of behavior change [33], which posits that health behavior change is driven by 3 critical components: “accurate information” about both risk behaviors and their replacement health behaviors (eg, benefits of daily parental monitoring), “motivation” to change behavior, and “behavioral skills and confidence” (eg, self-efficacy) necessary to perform the behavior. As The 3Ms was designed to be delivered during regular diabetes clinic visits, each session lasted approximately 15-20 minutes. To ensure the cultural relevance of The 3Ms for Black caregivers, the early intervention development process included input and review of intervention content and language from Black pediatric researchers and beta-testing by caregivers of Black adolescents with T1D.

The intervention’s informational content encouraged parents to use 3 strategies for increasing parental supervision and monitoring of adolescent diabetes management. Called “The 3Ms,” the strategies were (1) watch your child give as many doses of insulin each day as possible (medicine), (2) check your child’s glucose monitor at least once a day (monitor), and (3) eat at least 1 meal each day with your child so carbohydrate counting can be assessed (meals). This informational content was delivered through psychoeducational video clips where a Black endocrinologist and a Black caregiver provided advice regarding these parenting behaviors and encouragement to use them. To increase caregivers’ motivation and self-efficacy to engage in daily supervision of adolescent diabetes management, the intervention used multiple strategies consistent with motivational interviewing [34], including evoking change talk and commitment language (ie, statements regarding desires, reasons, needs, and abilities to make behavior change) and eliciting the pros and cons of behavior change. Intervention content was tailored based on caregivers’ ratings of the importance of engaging in daily parental supervision and their ratings of self-efficacy for parental supervision. Tailoring also included the completion of different content in follow-up sessions based on caregiver appraisals of their success in completing daily parental supervision, as well as the completion of optional goal-setting activities at the end of each session (Figure 2 provides sample intervention content).

HbA_1c level was used to evaluate glycemic control. Values were obtained during data collection visits using the Food and Drug Administration (FDA)–approved Accubase fingerstick capillary blood collection test kit. Due to the COVID-19 pandemic and higher than expected missed data collection visits, these data were also obtained from the clinic medical record for follow-up points if a clinic visit fell within ±30 days of data collection and data were otherwise missing. A total of 88.5% (485/548) of follow-up HbA_1c measurements were obtained using the Accubase test kit, and 11.5% (63/548) were obtained from the medical record. Previous studies have shown high comparability between samples collected using methods similar to those of the Accubase kit compared to venous samples [35].

A self-report questionnaire was used to obtain information from the adolescent’s primary caregiver on demographic variables. The adolescent’s medical chart was reviewed to obtain clinical information such as the duration of diabetes and insulin delivery method.

Adolescent depressive symptoms were measured at baseline using an adapted version of the 8-item Patient-Reported Outcome Measurement Information System Pediatric Short Form Depressive Symptoms (PROMIS-D; version 1.0) [36]. The self-report scale assesses mood, positive or negative affect, and views of self. Items were rated from 1 to 4, with higher scores reflecting more depression. The internal consistency of the measure in this study was high (α=.94). For the analyses, PROMIS-D was dichotomized at a score of 1 SD above the sample mean (<23 vs ≥23). This approach is similar to using a T-score of 60 or higher; PROMIS-D T-scores in this range indicate mood-related difficulties [37].

Analyses were conducted using a repeated-measures linear mixed effects (LME) regression model. The LME model included 3 fixed factors and 4 fixed covariates. The fixed factors were treatment group (The 3Ms vs control), data collection point (at baseline, 6 months, 13 months, and 18 months), and treatment moderator (PROMIS-D ≥23 vs <23). The 4 covariates were age, income, and 2 dummy codes for insulin delivery method. These covariates were selected from medical and demographic characteristics (Table 1). A correlation below the threshold value of P=.10 with either the treatment group variable or HbA_1c determined selection. The intercept and study site were random factors. The treatment effects were evaluated with change-from-baseline–planned comparisons in HbA_1c levels at 6-, 13-, and 18-month follow-up. Planned comparisons were statistically evaluated with a 2-sided P<.05 for significance. Moderation effects were investigated with post hoc simple effect tests.

Analyses were intent-to-treat, and all randomized cases were included. Of the 149 enrolled cases, 122 cases provided complete HbA_1c data across 18 months. Under the assumption that data are missing at random, the LME model used all available data to estimate model parameters. Explicit data imputation was not required.