Evaluating representativeness: recruitment for a virtual family-based intervention focused on the transition from pediatric to adult diabetes care^†

Abstract

Objective

Recruiting representative samples of youth for behavioral health interventions is challenging yet necessary to translate research into practice and eliminate health disparities. Transition-aged youth with type 1 diabetes (T1D) represent a vulnerable population; not enough attention is given to their inclusion in behavioral health interventions. Behavioral Family Systems Therapy for Diabetes Transition (BFST-DT) is an intervention aimed at improving transition readiness and is currently being pilot tested. The objectives of this study are to (1) evaluate the representativeness of the enrolled sample based on demographic and medical characteristics and (2) evaluate recruitment communication preferences.

Methods

Thirty adolescents (Mage = 16.57 years) with T1D and their caregiver(s) were recruited from a large urban hospital. Demographic and medical variables were collected via electronic medical record. Research staff recorded recruitment details about communication attempts and methods and reasons for participation decline.

Results

Those who enrolled in the intervention had more insulin pump usage than the recruitment population. Those who enrolled communicated primarily over email, while those who declined preferred phone. The length of time before a participation decision was communicated was similar between the enrolled and declined groups at about 6 weeks. The main reason for declining to participate was lack of interest.

Conclusions

Recruitment strategies were mostly effective in recruiting a representative sample of adolescents with T1D. Findings have implications for recruiting populations that are challenging to engage in intervention research. Future research should prioritize the stratification of historically underrepresented groups during recruitment.

Enhancing the representativeness of pediatric behavioral health intervention research is an area deserving of greater attention (Modi, 2023). In this paper, representativeness refers to having a study sample that is similar to the larger local, regional, national, or international population under study (Martínez-Mesa et al., 2016; Rooney & Evans, 2019; Shah & Peters, 2019). It may include demographic characteristics like race, ethnicity, gender, and socioeconomic status (SES). In pediatric behavioral health research, it may also include condition-specific characteristics (e.g., whether or not the sample uses diabetes technology at rates similar to the larger population). Obtaining a representative sample is necessary to effectively translate research into practice, and to better understand and eliminate health disparities (Ralston et al., 2019). However, many behavioral health studies often fail to recruit and retain representative samples despite growing acknowledgment of the importance in doing so (Winter et al., 2018).

This study will describe the recruitment efforts behind a pilot trial of a virtual family-based intervention focused on promoting the transition to adult healthcare for adolescents with type 1 diabetes (T1D). The importance of and challenges to recruiting representative samples in behavioral health intervention research is evident in research on youth with T1D, as racial and ethnic minority youth and youth from low SES backgrounds are often not represented. Racial and ethnic minority youth with T1D have higher hemoglobin A1c (HbA1c) levels, mortality rates, and lower diabetes technology use compared to their White counterparts (Agarwal et al., 2021; Keenan et al., 2021; Liese et al., 2022; Miller et al., 2020; Redondo et al., 2018). Few psychosocial interventions for T1D adequately recruit youth from racial or ethnic minority or low SES backgrounds (Ellis & Naar, 2023; Morone, 2019), and many do not report on recruitment priorities or strategies (Bindiganavle & Manion, 2022; Feeney et al., 2023; Le Roux, 2021). An exception to this is a study reporting on recruitment methods for a clinical trial investigating the effectiveness of an eHealth intervention to promote optimal glycemic control in young Black adolescents with T1D (Ellis et al., 2021). Researchers recruited Black individuals and implemented a prolonged recruitment for families that were logistically difficult to contact to address the disproportionate health disparities in T1D. Such recruitment efforts and reporting will result in more inclusive behavioral health intervention research for youth living with T1D.

Behavioral health intervention studies may fall short of recruiting a representative sample for several reasons. For one, researchers must first decide the sampling frame, which is the level (e.g., local, regional, national) at which they want their sample to be representative (Rooney & Evans, 2019; Shah & Peters, 2019). Sampling procedures (e.g., convenience, quota, stratification, or other sampling methods) can also affect representativeness (Rooney & Evans, 2019). Often, studies do not recruit representative samples because doing so requires recruiting individuals who face barriers to participating in research (Herbert et al., 2016). Families who do not have adequate access to medical care or those who face socioeconomic adversity are often difficult to contact or stay in contact with, and they may face more logistical barriers (e.g., work, childcare, transportation demands) that decrease the ability or willingness to participate (George et al., 2014; Winter et al., 2018). Additionally, legal issues such as immigration status, mistrust of researchers, perceived stigma, or limitations in the cultural considerations involved in the research may impact the willingness or ability to participate (George et al., 2014; Winter et al., 2018). Studies may have exclusion criteria to ensure consistent study conditions and improve the reliability of results, but this narrows the sampling frame and may be confounded with other demographic characteristics (Shah & Peters, 2019). For example, it is not uncommon for psychosocial studies of youth with T1D to exclude participants with high HbA1cs, and youth of color are more likely to have higher HbA1cs (Zaremba et al., 2024).

Strategies for addressing challenges that interfere with the recruitment of representative samples include infusing research protocols with greater flexibility, which is often needed when working with families from underserved backgrounds (Catlin & Van Hecke, 2021; Herbert et al., 2016) yet difficult to implement, especially in clinical trials (Paskett et al., 2008). Using a variety of recruitment methods is also necessary, such as in-clinic visits, phone calls, letters, and social media advertisements (Herbert et al., 2016; Miyamoto et al., 2013; Pyatak et al., 2023), though there is a lack of research on the communication preferences generally. The omnipresence of social media has resulted in increasing internet-based recruitment, and while some research has found this yields representative sample populations (Wisk et al., 2019), there is concern that it may exclude certain families. Using community-based participatory research methods to tailor the content, design, and delivery of interventions is another avenue for increasing representativeness (Byrne, 2019; Goodman & Sanders Thompson, 2017; Joo & Liu, 2020; Julian McFarlane et al., 2022; Morone, 2019). For example, research has found that Black youth with T1D and their families prefer interventions that address the family unit rather than the individual, therefore, experts have recommended for interventions to incorporate a family-lens to increase cultural responsiveness (Morone, 2019).

The current study

Behavioral health interventions tested with a representative sample of the intended clinical population facilitates successful future integration into the health care system (Barry-Menkhaus et al., 2020). Further, a greater focus on the representativeness of samples and the recruitment methods utilized in behavioral health research will promote health equity (Lescano et al., 2016). Inspired by this, the current study's objective is to describe the recruitment efforts of a virtual family-based intervention, Behavioral Family Systems Therapy for Diabetes Transition (BFST-DT), which is aimed at improving transition readiness in adolescents with T1D. BFST-DT is currently being tested via a pilot pre–post intervention design with an iterative mixed-methods assessment of its acceptability, feasibility, and impact on transition readiness factors. The first aim of the present study was to evaluate the representativeness of the enrolled sample based on demographic and medical characteristics by comparing it with the larger recruitment population (i.e., all eligible patients with T1D that were cared for at the medical center), those who declined to participate, and those who could not be reached. We hypothesized the following: there would be no significant differences in demographic and medical characteristics between the recruited sample and the larger eligible patient population, those who declined to participate, and those who could not be reached. The second aim was to evaluate recruitment communication preferences by comparing the method of contact and number of communications between those that enrolled and those who declined. Given the lack of previous research on recruitment communication preferences, this aim was considered exploratory, and no specific hypotheses were made.

Methods

Intervention

Data for this study were drawn from a trial of BFST-DT, a virtual family-based transition readiness intervention for adolescents with T1D. It was adapted from an existing empirically supported family-based intervention for youth with T1D (Wysocki et al., 2008) and informed by focus groups with adolescents and young adults with T1D, their caregivers, and pediatric and adult diabetes providers (Papadakis et al., 2024). The intervention targets individual (e.g., diabetes self-management behaviors, diabetes-related self-efficacy) and family-based (e.g., family communication, family problem-solving) modifiable factors known to impact the transition from pediatric to adult healthcare. The intervention is delivered over Zoom over a 6-month period and consisted of four 2-hr multifamily group meetings and six 1-hr individual family meetings facilitated by a masters-level clinical psychology graduate student. The multifamily group meetings involve learning content covering four areas specific to the transition period: adolescent and family development, problem-solving, communication, and cognitive restructuring. The individual family meetings focus on each family’s chosen transition readiness goals by applying concepts learned in the multifamily meetings.

Participants

The study was approved by the institutional review board of the academic medical center (2023-6210). This manuscript only describes data related to recruitment efforts (see Recruitment Protocol). Participants who were enrolled into the intervention included N = 30 adolescents with T1D and their caregiver(s), recruited from a pediatric academic medical center in the Midwest. Eligibility criteria included a junior or senior in high school, between the ages of 15 and 18 years old, plan to be living with a caregiver during the 6-month long intervention, have been diagnosed with T1D for at least 1 year, be fluent in English, and have a caregiver willing to participate that was also fluent in English. Adolescents with another significant chronic health condition that requires intensive daily management (e.g., cystic fibrosis, undergoing current cancer treatment) and adolescents with an intellectual disability were excluded given their transition to adult healthcare may be distinctly different. This study also reports data from potential participants who met inclusion criteria and were contacted by study staff, but did not enroll either because they declined, could not be contacted, or the study already met its enrollment goal.

Recruitment protocol

Recruitment occurred between October 2023 and May 2024 on a rolling basis with the goal of recruiting 30 patient families into one of four cohorts receiving the intervention, with each cohort spaced approximately 1 and a half months apart. A targeted sample size of 30 is considered appropriate for a pilot study aimed at evaluating initial efficacy, acceptability, and feasibility through qualitative and quantitative methods (Teresi et al., 2022). The goal of recruitment was to use stratified sampling methods to enroll a sample that was representative of the population of all patients with T1D that was cared for at the medical center, with respect to gender, race, ethnicity, insurance status, HbA1c, and diabetes technology use. All recruitment efforts for BFST-DT were conducted by a single behavioral research coordinator with the support of the larger research team. The principal investigator trained the coordinator in all research recruitment strategies, which included mock recruitment calls and informed consent visits. All recruitment scripts and communications (e.g., emails, flyers) were created by the research team and approved by the institutional review board.

Eligible patients were identified via an electronic medical record (EMR; Epic) database that contained all patients with T1D receiving care at the medical center. This database was created and maintained by the clinical coordinators of the medical center’s T1D program using Epic’s SlicerDicer tool. The research coordinator reviewed the adolescent’s EMR to obtain contact information including patient name, caregiver name(s), address, telephone numbers, and email addresses. All eligible patient families were sent a letter to the home address listed in the patient’s EMR. The letter described the study in detail, invited the family to participate, and provided contact information if interested. If eligible patient families had an activated digital patient portal, they also received a message with information describing the study and were asked to select either “I’m Interested” or “Decline/Not Interested.” The families that indicated “Decline/Not Interested” were not contacted and were removed from the active recruitment contact list. Those that selected “I’m Interested” were contacted via email to discuss the study more or to schedule a phone call to discuss it. The majority (n = 183, 81.3%) of potential participants had their patient portal activated. However, recognizing that there may be certain families who are less likely to use patient portals to communicate, and wanting to eliminate a potential bias, it was decided to focus on phone and email for further recruitment communication. Approximately 1 week from the expected date of the mailed letter’s arrival, preliminary recruitment phone calls were conducted to the patient’s primary caregiver. The start of the recruitment calls was defined as the start of active recruitment. During the phone call, study information was provided, and all questions and concerns were answered appropriately. Phone calls were made weekly in the evening after common working hours. If requested, the research coordinator called families at their preferred date/time. All details of the contact attempts were noted in a secure database, including the date, time, whether contact was made, who contact was made with, topics discussed, and decisions made. Once families decided whether to participate, they were categorized into either the “Enrolled” or “Declined” groups. If the caregiver did not respond after 10 unsuccessful contact attempts, or all phone numbers associated with the patient chart were inactive, they were categorized as “Unable to Contact.” Recruitment calls were made to a random group of 50 families at a time. As families enrolled, declined, or were unable to be contacted, more families were added to the weekly call list so that the research coordinator was consistently calling 50 families once per week.

We used a stratified sampling method in an effort to enroll a sample that was representative of the pool of eligible participants. To do this, we determined the characteristics of all the eligible participants with respect to gender, race, ethnicity, insurance status, HbA1c, and diabetes technology use. On a weekly basis, the characteristics of enrolled participants were compared to those of the eligible pool. If the enrolled sample did not reflect that of the eligible pool, efforts were made to prioritize potential participants based on specific characteristics, by contacting those participants first within each batch of 50 families.

After speaking on the phone about the study, the research coordinator sent an email to potential participants summarizing the details and next steps for enrolling in the study. Once a caregiver indicated interest, a remote informed consent visit was scheduled and conducted. Occasionally, a caregiver requested the research coordinator speak directly to the adolescent before the informed consent visit to confirm the adolescent’s interest in participating. The informed consent visit occurred over HIPPA-compliant video conferencing software (Zoom), with a link to the meeting emailed to the family immediately after scheduling and 1 day before the visit.

Measures

Demographic and medical information

The EMR was also reviewed for patient demographic and medical information including age, gender, race, ethnicity, insurance, city of residence, number of medical and psychiatric comorbidities, date of T1D diagnosis, diabetes technology use (continuous glucose monitor, insulin pump), and most recent HbA1c as indicated by lab value or documented physician note. As a broad indicator of SES, the median income of the family’s zip code was determined through a census search (U.S. Census Bureau, n.d.).

Recruitment communication

“Communications” included communications that occurred directly with the family (e.g., responsive phone calls and emails) and communication attempts that did not result in direct contact with the family (e.g., voicemails or unread sent email messages). The number and format of each communication attempted was recorded in a routinely managed secure database.

Statistical analyses

All data were entered and analyzed in SPSS version 29.0. Means, SDs, and frequencies were evaluated for all study variables. The first aim was to evaluate the representativeness of the enrolled sample based on demographic and medical characteristics by comparing it with the larger recruitment population, those who declined to participate, and those who could not be reached. First, each group—enrolled, declined, and could not be reached—was compared with the larger recruitment population (i.e., all other potential participants not included within the group being analyzed). Shapiro–Wilk tests indicated data were not normally distributed; therefore, nonparametric Mann–Whitney U tests were conducted to make comparisons on the following continuous variables: age, median income, and HbA1c. Chi-square goodness-of-fit tests were used to make comparisons on the following dichotomized categorical variables: gender (male vs. female), race (White vs. Black, Asian, American Indian/Alaska Native, Biracial, or Other, which were dichotomized due to limited sample sizes), ethnicity (non-Hispanic vs. Hispanic), location (urban vs. non-urban), glucose monitoring method (CGM vs. glucometer), insulin delivery method (insulin pump vs. multiple daily injections), insurance (private vs. public), presence of medical comorbidities (yes vs. no), and presence of psychiatric comorbidities (yes vs. no). Second, the groups—enrolled, declined, and could not be reached—were compared to each other. Nonparametric Kruskal–Wallis tests with Dunn–Bonferroni post hoc tests were used to compare groups on the continuous variables stated previously, and Chi-square tests were used to compare groups on the categorical variables stated previously.

The second aim was to evaluate recruitment communication preferences by comparing the number of communications and method of contact between those who enrolled and those who declined. Nonparametric Mann–Whitney U tests were conducted to compare the enrolled and declined groups on the following continuous variables: the number of overall communications that occurred before the participation decision was communicated, and the number of days between the initial contact and when the decision was communicated. Chi-square tests were conducted to compare the enrolled and declined groups on the following categorical variable: The method of communication used to communicate participation decision (phone call vs. mailing/email/digital patient portal). Lastly, reasons for participation decline (which were reported directly to the research coordinator and systematically documented) were examined with descriptive statistics.

Results

The total number of patients of any age (11 months–24 years) with T1D who received care at the medical center during the past 2 years was 1433. The total number that was potentially eligible based on inclusion (age, year in school, time since diagnosis, English-speaking) and exclusion (major chronic illness comorbidity, intellectual disability) criteria was 225 (15.70% of the total population receiving care). Those 225 patients were contacted for recruitment. Of those 225, 30 families enrolled in the study and began the intervention (30 was the targeted enrollment number), 88 families declined to participate (including 8 that initially consented but then shortly thereafter changed their mind and withdrew from the study before engaging in the intervention), and 41 families could not be contacted after 10 attempts. There were an additional 66 families who were contacted, but contact attempts were stopped as the study met its enrollment goal; these families were not included as a subgroup in analyses but were included in the overall recruitment population group (i.e., N = 225). See Figure 1 for the CONSORT diagram. As all data were pulled from EMR, there was very limited missing data and included only three participants for whom there were no race or ethnicity data reported in the EMR.

Aim 1: comparisons between groups based on enrollment status

Table 1 displays the means/SDs or frequencies/percentages for all demographic and medical characteristics for all groups. Our hypothesis was mostly confirmed as there was only one significant difference between the enrolled group and the recruitment population; the enrolled group had a higher percentage of pump usage (80% [n = 24]) compared to the recruitment population (55.4% [n = 108], χ² [1, N = 30] = 7.36, p < .01). In addition, the could not be reached group was older (M [SD] = 17.07 [0.72] years) than the recruitment population (M [SD] = 16.74 [0.98] years), and the enrolled group (M [SD] = 16.57 [0.82], χ² [2] = 6.75, p < .05).

Aim 2: comparison of communication methods

Table 2 displays the means/SDs for communication variables between the enrolled and declined groups and results from nonparametric Mann–Whitney U tests. There were three significant findings. There were more communications with those who enrolled (M [SD] = 9.27 [3.24]) compared to those who declined (M [SD] = 5.03 [3.22], z = −5.63, p < .001). Related, there were more singular emails exchanged (e.g., either sent or received) with those who enrolled (M [SD] = 5.20 [1.9]) compared to those who declined (M [SD] = 1.11 [1.4], z = −7.58, p < .001). There were no significant differences in the number of calls or digital patient portal messages sent to each group, no significant differences in the percentages of patient portal activation between groups (83.3% of the enrolled group and 82.5% of the declined group had their patient portal activated), and no differences in the number of days before the participation decision was communicated (ps > .05). When examining which mode of communication was used to communicate a decision, a larger majority of those who declined communicated their decision over the phone (n = 68 [85%]), compared to those who enrolled (n = 18 [60%], χ² [1, N = 110] = 7.99, p < .05).

See Figure 1 for the reasons families provided when declining to participate. The majority indicated a lack of interest in the intervention as the reason for declining. Another common reason was generally not having the time or being “too busy.” The remaining reasons included situational barriers (e.g., planning to be out of the country during the intervention period, conflicting schedule with extra-curricular activities), recently moved/no longer receiving care at the hospital, expecting there may be a language barrier, disliking that the intervention was being delivered virtually, feeling diabetes was too challenging at the time to participate, or no answer was given.

Discussion

The present study aimed to recruit a sample of adolescents with T1D and their caregiver(s) to participate in a virtual family-based intervention (i.e., BFST-DT) to promote the skills necessary to promote a smooth and successful transition from pediatric to adult diabetes care. The goal was to recruit a sample that represented the larger T1D patient population of the study’s medical center located within a large Midwestern urban city. This paper’s objective is to describe the study’s recruitment approach and methods in detail and to evaluate their effectiveness. Motivation for this paper comes from the acknowledgment that more transparent reporting about recruitment will aid in collective efforts to increase representativeness in behavioral health research and will assist in efforts to demonstrate the effectiveness of evidence-based interventions with “real-world” populations.

The first aim was to evaluate the representativeness of the enrolled sample based on demographic and medical characteristics by comparing it with the larger recruitment population, those who declined to participate, and those who could not be reached. Using a stratified sampling method was mostly effective in enrolling a representative sample, in that there were no differences between the enrolled sample and the recruitment population in terms of age, gender, race, ethnicity, residence, estimated median income, insurance, HbA1c, and comorbidities. However, compared to the recruitment population, those enrolled were more likely to use insulin pumps (80% compared to 58.7%). Although diabetes technology use is commonly associated with demographic variables (e.g., race, income), this finding highlights the importance of stratifying these characteristics separately to increase representativeness. In other words, a recruited sample that is stratified based on race, ethnicity, income, or insurance status does not necessarily ensure that the sample is representative based on diabetes technology use. In addition, this study had targeted enrollment deadlines because of needing to enroll a certain number of participants into each cohort before the official start date for that cohort. This hard stop to recruitment sometimes challenged efforts to meet stratification goals (i.e., as in with insulin pump usage), and often required a balancing of research priorities (e.g., enrolling the targeted n into each cohort vs. ensuring each cohort was representative as planned). Thinking through these priorities ahead of time may help research teams be intentional about study design and methods and avoid potential inconsistencies.

The second aim was to evaluate recruitment communication preferences by comparing the number of communications and method of contact between those who enrolled and those who declined. We found that, on average, there were about nine communications with families who eventually enrolled compared to five communications with those who declined. The main driver of that difference was that families who enrolled communicated over email much more than families who declined. The increased number of email communications made with the enrolled group reflects those participants’ preference for email (e.g., they emailed the research coordinator back instead of calling them back) and interest, as more emails were sent if participants emailed questions or inquiries. Regardless, it is likely that potential participants will have different communication preferences, so the more research teams can increase the ways they communicate with families, the more likely it is they will recruit the targeted population. In addition, EMR systems often allow for tracking of communication preferences such as for appointment reminders or communication from providers. There may be opportunities to document preferences for research communication that can benefit studies that recruit clinical populations.

A potential missed opportunity could have been that this study only sent one digital patient portal message to potential participants. As previously stated, the patient portal was not utilized for ongoing recruitment communication to eliminate a potential bias against those families who had not activated their patient portal. With the growing use of patient portals to communicate with medical teams (Shenson et al., 2016), more focus should be placed on this method of communication to recruit participants, as it signals to potential participants that the research is supported by the medical system they are familiar with.

Interestingly, there was not a difference in the length of time it took to communicate a participation decision between families who enrolled and those who declined; it took each group about 42 and 44 days, respectively, from the day they received a mailed invitation letter to communicate their preference. There is reason to hypothesize that those who enroll do so more quickly than those who decline (e.g., they are sure of their interest or ability), or that they may require more time (e.g., they need time to consider it). Yet, these data suggest that all families communicate their decision after about 6 weeks, regardless of if they choose to enroll or decline. This finding may reflect the cadence of recruitment outreach (e.g., placing a call every week for up to 10 weeks). Regardless, this is useful data as it informs planning of future, related projects, where knowing how long it may take to recruit participants can influence overall project timelines. Indeed, in the current study, knowing the average length of time until a participation decision would be communicated would have helped with recruitment planning in the present study; as mentioned previously, there were conflicting priorities between enrolling the targeted n into each cohort in time versus ensuring each cohort was representative.

As reviewed previously, there are many barriers to recruitment for behavioral health intervention studies, including for T1D (Ellis & Naar, 2023; Morone, 2019). In examining why potential participants declined to enroll in the study, we found that about half (53.8%) declined for lack of interest. The current study did not inquire further as to exactly why these families were not interested, but it may be helpful for future research to do just that. Given the challenges that families can face when an adolescent has T1D, and the difficulties faced during the transition to adulthood, it could be considered surprising families do not find an intervention focused on this to be of interest to them. Also, it is assumed that most caregivers expressed a lack of interest, but the lack of interest may have been from the adolescents. The fact that family-based interventions often require the caregiver and child/adolescent to buy in is a potential barrier that should not be overlooked. This study was developed and designed with significant family engagement through focus groups held with adolescents and young adults with T1D, their caregivers, and pediatric and adult diabetes providers. In theory, it should reflect the needs and interests of these groups. However, this aspect may not have been effectively communicated to potential participants. Receiving immediate feedback about interest could help recruiters better tailor their communication about the study to potential participants. In addition, as expected, many families (22.6% of those who declined) also cited lack of time or scheduling/time conflicts as barriers. Future studies will benefit from continued focus on how behavioral health interventions can be delivered with flexibility to increase access to more families. Indeed, this was the precise motivation behind offering BFST-DT virtually.

Limitations

This study intentionally used methods to address barriers to recruiting representative samples, such as involving patients and families in the development of the intervention, not excluding participants based on HbA1c, reducing communication biases (e.g., recognizing certain families are less likely to use patient portals to communicate and contacting families at times convenient to them), and using a stratification approach. Still, limitations of this study include that it did not use in-person recruitment methods. Given this study was virtual, and many behavioral health interventions are being delivered through virtual means (e.g., Kelber et al., 2024) alongside the delivery of diabetes telemedicine care (Stallings et al., 2023), perhaps the lack of in-person recruitment was not a hindrance. This is an area deserving of additional attention in future research. In addition, adolescents or caregivers who were not English-speaking were excluded from enrollment. Such language exclusions can decrease representativeness and potentially increase health disparities. This study could have utilized simultaneous translation to deliver the intervention to non-English-speaking families. Future research, including evaluations of this intervention, should strive to ensure interventions are available to families of other languages. Similarly, adolescents were excluded if they had another significant chronic health condition that required intensive daily management or if they had an intellectual disability. Additionally, race was dichotomized into White and non-White groups for statistical analysis because of the limited sample size. This is not ideal and diminishes potential differences among groups. Further, this study sought to recruit a sample of youth representative of the T1D patient population at the larger medical institution; however, the study did not evaluate how representative the sample is compared to the larger geographical area, or national samples, given participants were not being recruited from those locations.

Conclusion

Recruitment approaches and methods are integral to all scientific research, including pediatric behavioral health intervention research, yet attention to and prioritization of recruitment approaches and reporting is not always observed. Aligning recruitment efforts with the goal of obtaining samples that are representative of the populations that these interventions are meant to serve is critical for improving implementation and increasing equity in medical and behavioral health outcomes. Studies evaluating behavioral health interventions focused on pediatric T1D or the transition to adult healthcare will certainly benefit from attention to increasing representativeness, and through that, hopefully lead to improved outcomes for many youth and young adults living with T1D.

Author contributions

Madeleine Claire Suhs (Conceptualization [lead], Data curation [lead], Formal analysis [equal], Investigation [equal], Methodology [equal]), Julia Ellis (Conceptualization [supporting], Resources [equal]), Jill Weissberg-Benchell (Project administration [supporting], Supervision [supporting]), Michael A. Harris (Project administration [supporting], Supervision [supporting]), and Jaclyn Lennon Papadakis (Conceptualization [equal], Data curation [equal], Formal analysis [lead], Methodology [equal], Project administration [lead], Supervision [lead])

Funding

Data collection and preliminary analysis were sponsored by Breakthrough T1D (GR001892-01), formerly known as Juvenile Diabetes Research Foundation.

Conflicts of interest: None declared.

Data availability

The data underlying this article cannot be shared publicly for the privacy of individuals who participated in the study. The data will be shared on reasonable request to the corresponding author.

References

Author notes