JMIR Diabetes Help with

Background: Physical activity is a major component of treatment for adolescents with obesity and prediabetes or type 2 diabetes; however, sedentary behavior remains pervasive. An SMS text message–based intervention paired with financial incentives may be an effective way to promote physical activity in this population. Objective: This study aims to obtain end-user feedback on SMS text message content and assess the acceptability of a planned to be doing when messages were sent (eg, homework or watching television) and messages that were upbeat or informative. Specific physical activity suggestions were rated favorably. Attitudes toward financial incentives varied, with differing opinions about whether loss-framed incentives would be motivating or discouraging. Many participants highlighted the role of intrinsic, rather than extrinsic, motivation in achieving and sustaining behavior change. Conclusions: The engagement of adolescents with obesity and diabetes or prediabetes allowed for the refinement of SMS text messages for our planned intervention, with an emphasis on short, upbeat, relatable, and informative messages. Although an SMS text messaging intervention using financial incentives to motivate youth with prediabetes or type 2 diabetes to be more physically active is theoretically acceptable, the impact on actual activity levels in this population requires prospective evaluation in a clinical trial. Abstract Background: People with gestational diabetes have enhanced learning requirements during pregnancy, and management of their disease often requires the translation of health information into new health behavior changes. Seeking information from the internet to augment learning from health professionals is becoming more common during pregnancy. YouTube is a popular free and accessible web-based resource, which may be particularly useful for individuals with low health literacy or other barriers to receiving high-quality health care; however, the quality and content of YouTube videos varies, and little is known about those covering gestational diabetes. Objective: We aimed to systematically evaluate the quality, content, and reliability of YouTube videos about gestational diabetes. Methods: A systematic search of YouTube videos was conducted over the course of 1 week in April 2020 using the following keywords: “gestational diabetes,” “gestational diabetes management,” “gestational diabetes treatment,” and “pregnancy and diabetes.” The search results were displayed by relevance, replicating a default YouTube search attempt. The first 60 results from each keyword were reviewed (n=240). Exclusion criteria were videos unrelated to gestational diabetes, videos not in English, and those for which the full video was not available at the time of review. For each unique video, a gestational diabetes content score was used to rate video comprehensiveness and accuracy, and the DISCERN instrument, a validated metric to assess consumer health information, was used to evaluate the


Background
Once a disease of adulthood, type 2 diabetes is now becoming increasingly common among youth [1]. Among individuals with youth-onset type 2 diabetes, severe diabetes-related complications such as retinopathy, neuropathy, and nephropathy begin to emerge within 10 to 20 years of diagnosis and contribute to a 15-year reduction in life expectancy [2,3]. The increasing incidence is driven largely by obesity, which is present in 21% of persons aged 12 to 19 years in the United States [4] and quadruples the risk of youth-onset type 2 diabetes [5]. Globally, nearly one-fifth of children and adolescents aged 5 to 19 years old were overweight or obese in 2016, representing a more than 4-fold increase over 4 decades [6]. Prediabetes, which increases the risk of developing type 2 diabetes [7,8], is already present in approximately 20% to 30% of adolescents with obesity [9,10]. A healthy lifestyle, including adequate physical activity, plays a central role in the prevention and treatment of diabetes. Unfortunately, adolescents with type 2 diabetes typically do not reach the recommended duration of 60 minutes per day of moderate to vigorous physical activity [11], averaging only 8 (girls aged 15-18 years) to 26 (boys aged 10-14 years) minutes daily [12]. To reduce the risk of poor outcomes, effective methods to increase physical activity in youth with prediabetes and type 2 diabetes are needed.
Approximately 95% of teenagers in the United States now report ownership or access to a smartphone [13]. Smartphone ownership is also common throughout the world, although it is more prevalent in advanced economies (median 76% ownership of a smartphone; 17% of mobile phones that are not smartphones) than in emerging economies (median 45% ownership of a smartphone; 33% of mobile phones that are not smartphones). Notably, smartphone ownership is more common among younger individuals [14]. With such a high penetrance of smartphone and mobile phone ownership, one strategy to engage and motivate adolescents to be more physically active is via a text message-based intervention. Text message-based physical activity interventions have been studied in adolescents with overweight and obesity [15]; however, explicit focus on youth with prediabetes and type 2 diabetes has been limited [16]. To promote engagement and retention in a text message-based lifestyle intervention, however, it is critical to elicit adolescent viewpoints as part of intervention development [17,18]. This engagement is particularly needed for youth with obesity-related complications such as type 2 diabetes or prediabetes, in whom the perceived risk of poor health may impact responses to message tone and content.
Another potentially effective strategy to motivate youth to be more physically active is the provision of financial incentives. In adults, incentives framed as losses (upfront endowment with losses applied for failure to meet a goal) rather than gains (money earned upon completion of a goal) are more effective at inducing physical activity-related behavior change [19]. Financial incentives may also help counteract the rapid habituation of text message-based interventions that have been demonstrated in adolescents and young adults [20]. However, to our knowledge, the effectiveness of, or perspectives on, different financial incentive strategies to motivate adolescents to engage in physical activity have not been explored. In the Behavioral Economics for Activity Motivation (BEAM) trial (NCT04874415), a mobile health (mHealth)-based optimization trial, we are studying text messaging, loss-framed financial incentives, and gain-framed financial incentives as candidate components in a factorial experiment with youth with overweight or obesity and prediabetes or type 2 diabetes, with the aim of increasing time spent in moderate to vigorous physical activity.

Objectives
To prepare for the trial, we conducted group interviews with adolescents from the target population. We sought insights into message content and financial incentive strategies, including financial incentive preferences and the anticipated impact on one's motivation to be physically active.

Participants
A convenience sample of individuals from the target population of the BEAM trial, adolescents aged 13 to 18 years with overweight or obesity (BMI ≥85th percentile for age and sex) and prediabetes (hemoglobin A 1c 5.7%-6.4%, fasting glucose 100-125 mg/dL, and 2-hour oral glucose tolerance test glucose 140-199 mg/dL) or type 2 diabetes (hemoglobin A 1c ≥6.5%, fasting glucose ≥125 mg/dL, and 2-hour oral glucose tolerance test glucose ≥200 mg/dL and negative diabetes autoantibodies) were recruited from the pediatric endocrinology clinic at the Children's Hospital of Philadelphia, a large academic children's hospital in Philadelphia, Pennsylvania, United States, from December 2020 to April 2021. Potentially eligible participants were approached about the study at clinic visits or by phone within 2 weeks after a clinic visit by the study research coordinator or principal investigator. Individuals with limited English proficiency were excluded from this study. Eligible individuals were invited to participate in a 1-hour group interview. Electronic consent and assent (for participants aged <18 years) were obtained via REDCap (Research Electronic Data Capture; Vanderbilt University). Group interviews were roughly segmented by age when possible (13 to 15 or 16 to 18 years) to promote more open dialogue among similarly aged peers but not by other demographic or clinical characteristics. Participants were compensated with US $30 for their time. Limited medical record reviews to include demographic information and pertinent medical history were performed by the study team with the permission of the participants.

Ethics Approval
Ethics approval was obtained from the Children's Hospital of Philadelphia Institutional Review Board (IRB 20-017554).

Message Development
The text messages were developed by the investigators with the assistance of one high-school student (female, aged 17 years) and one college student (female, aged 20 years). Both students became involved after approaching the primary investigator regarding temporary research opportunities in pediatrics, independent of any school-required research experience. Messages were designed with the target population in mind; limited financial resources and transportation barriers are common, so physical activity suggestions included only those that were free or low-cost, as well as many that could be performed at home. Initial message development was guided by common barriers to being more physically active, including feeling too tired, lack of interest, lack of peer or family support, lack of equipment, lack of space indoors or outdoors, lack of motivation, preference to avoid sweating, embarrassment, lack of confidence in skills, and lack of knowledge about how to do so [21]. Messages were crafted to have different tones, including informative, encouraging, or funny, to enhance variety and improve interest and engagement. Messages often included emojis, and some included graphic interchange formats (GIFs). The initial messages were reviewed and edited by the study team for clarity and their potential to engage end users. A total of 84 messages (1 per day for the 12-week intervention) were created.

Web-Based Group Interviews
Group interviews were chosen rather than individual interviews to promote an atmosphere of engagement and the sharing of ideas among peers. Smaller group interviews (goal of 2 to 6 participants) were chosen over larger group interviews or focus groups (the original goal of 7 to 10 participants) because of the virtual nature of the interviews, which was necessitated by the COVID-19 pandemic. Although the original study design included the use of focus groups, the richness of the data gathered from focus groups is heavily dependent on the ability to promote and maintain dynamic group discussions. It is possible to conduct focus groups in a virtual setting, but potential challenges include technological difficulties and the inability to control each individual's environment, which may lead to distractions and interruptions [22]. Therefore, we chose group interviews, in which the primary objective of the interviewer questions was to obtain individual responses rather than to stimulate group discussion as a method of information gathering [23]. Interviews were conducted by 2 primary interviewers (NM and AE), clinical research coordinators who identify as Black women in their 20s to 30s. NM had previous training and experience as a focus group facilitator and served as the primary interviewer. AE also participated in the recruitment of participants. MEV, the principal investigator, is a female pediatric endocrinologist who provides clinical care and conducts research with youth with prediabetes and type 2 diabetes; she assisted with interviews and identified the participants as White and Hispanic.
A semistructured interview guide was designed to elicit open-ended feedback on text messages for physical activity motivation (including acceptability, practical considerations, and preferred or nonpreferred content), experience with and motivation for behavior change, and experience with and attitudes toward the use of financial incentives for behavior change. The interview guides were piloted with nonmedically trained adults and revised as needed for clarity. The interviews were rehearsed by the interviewers with oversight by the study principal investigator (MEV), who had formal training in qualitative research methods. Interviewers practiced neutral, nondirective responses to participant answers and the use of open-ended probing questions to elicit more detailed responses. Before the start of the interview, interviewers informed participants that there were no correct answers and that all feedback, including critical feedback, was welcomed.
In addition to specific questions, 8 messages were reviewed during each interview. The number of messages was restricted to 8 to optimize the amount of time available for discussion, with the goal of understanding what aspects of the message were engaging or not rather than simply rating the specific message. Messages were selected to include a variety of barriers addressed and tones used (informative, encouraging, or funny) such that all combinations were tested at least once throughout the study. The messages were not repeated across groups; in total, 64 representative messages were reviewed. The remaining 20 messages were not substantially different in tone or content and thus were not reviewed by the study participants. During the interviews, message content was shared with participants via the videoconferencing platform's screenshare function and read aloud. Participants were asked to rate the messages as great, OK, or bad; these specific rating words were chosen to maximize the ability to identify message outliers-those that truly resonated (great) and those that were strongly disliked (bad)-so that themes relating to message success or failure could more easily be identified. Participants were invited to type their initial ratings in the platform's chat or respond aloud. After all initial ratings were shared, the participants were asked

XSL • FO
RenderX to explain their ratings, including why they liked or disliked the message and how it could be improved.
All interviews were conducted via the videoconferencing platform BlueJeans (Verizon), accessed by participants using their preferred personal internet-connected device (smartphone or computer). Participants were asked to keep their camera on if comfortable doing so and to remain unmuted in a quiet setting if possible. If participants wrote in the chat rather than speaking, the interviewers reread the typed content, stating which participants responded.

Analysis
Participant characteristics were presented using summary statistics, including means and SDs for continuous variables and proportions for categorical variables. All interviews were recorded securely on a videoconferencing platform, and audio was transcribed using a professional transcription service. Transcripts were reviewed by the study staff, corrected as needed, and supplemented by observer notes. Transcripts were deidentified before analysis. Content analysis was conducted to explore these themes. An a priori set of codes was created based on the semistructured interview guide; for example, Text messages acceptability and Text messages: preferred content. Next, two of the study team members with experience and training in qualitative study methods (TAH and MEV) reviewed the transcripts, identified emergent themes, and then refined the initial codebook. For example, an additional code for pandemic was considered owing to the frequent mention of how the COVID-19 pandemic impacted behavior and motivation but ultimately not included as an independent code but rather incorporated as a subcode to motivation for behavior change. After independently coding the 3 transcripts, the 2 reviewers compared the coding and clarified the codebook as needed. NVivo (version 12; QSR International) was used for qualitative analysis.
In all, 8 group interviews, each consisting of 2-4 adolescents, were conducted from December 2020 to April 2021. Each group interview lasted for approximately 30 to 45 minutes. The participants had variable degrees of engagement, with many leaving their cameras off. However, all (20/20, 100%) participants responded to the entire set of interview questions and rated all text messages per session. Many used the chat feature in the videoconferencing platform to respond but spoke aloud when asked to do so.

Overview
Thematic saturation was achieved with no new themes generated from later interview groups. Instead, previously identified themes were also present in the later groups. The interrater reliability of the 8 transcripts was high (κ=0.92). Several themes and subthemes emerged from focus group discussions that pertained to the development of text messaging content as well as the use of financial incentives in an intervention designed to motivate youth with overweight or obesity and prediabetes or type 2 diabetes to be more physically active.

Theme 1: Near-Continuous Use of Smartphones and Acceptability of Text Message Reminders
All but one (19/20, 95%) participant had smartphones and felt that text messaging was a good way to remind teens to be more active: We always on our phone, so we gonna see it [and] texting is a really good way to sort of spread words. Participants varied in the acceptable frequency of text message reminders, with a participant suggesting that messages as frequently as once per hour would be appropriate. However, most participants preferred 1-2 messages per day, either at noon or after school. They emphasized the need to know from whom the text was sent, so they did not mistake it for spam. Most felt that they used their phone more during the summer but had it with them at nearly all times, regardless of the time of year or day of the week.

Theme 2: Text Messages Should Be Short, Upbeat, Informative, and Relatable
Short and upbeat messages were strongly preferred. Table 2 highlights representative text messages and participant responses. Participants reported that they would most likely read and enjoy messages that were 1 sentence or phrase long and noted that they may not open the message to read beyond the message preview. Regardless of the originally intended tone (informative, encouraging, or funny), messages that had upbeat characteristics such as exclamation points, smiley faces, or encouraging words were felt to be most motivational. However, even if brief and encouraging, those stating a generally known fact (eg, Exercise helps your body and your brain) resulted in mixed responses. Importantly, a participant highlighted the need for caution when discussing self-esteem. Reflecting  Participants reported satisfaction with texts that reported facts or specific suggestions on how to overcome barriers: Feel like you're tight on time? Maybe it's in your head. A brisk walk actually changes the chemicals in your brain to make you feel more relaxed and less anxious was felt to explain a bit of the science and to be memorable.
Another informative message (Physical activity that gets your heart rate up releases endorphins that raise your energy level. Ready for your energy boost?) was appreciated for providing a specific reason for exercising. Messages that gave examples of how to overcome specific barriers were also well-rated. For example, in response to a message that identified the common barrier of feeling embarrassed to exercise in front of others and suggested a solution (It is completely normal to feel embarrassed to work out around others. Don't let that stop you. Exercise by yourself or with people you're comfortable with), a participant rated the message highly: There was agreement on the utility of sharing ideas to be physically active, but the preferred activities to suggest varied by participant. Several noted that they did not like sports, that "yoga sounds hard," exercise classes sounded like school ("Why must I have a class? Why can't we just go outside and just do whatever? Why does it have to be a class?" [Participant 13, group 6; male, aged 15 years]), and not everyone has room to or enjoys dancing at home. Suggestions about how to be physically active at home were appreciated, but participants differed in the perceived utility of including a link to a video or description of activities, with many stating that they would click the link, but others suggesting that clicking the link is "just kind of another thing to do." Messages that mentioned school or sitting on the couch and watching television were highly relatable. A message that emphasized that being active can help with school performance resonated with a participant:  • "I said great because think about like if we were all sitting on the couch or whatever being lazy, then someone got this text message, they could maybe motivate everybody else, like let's just go on a walk together" (participant 11, group 5; female, aged 16 years). • "...When you get this, let's say you're watching a movie. Everybody's on the couch and eating chips or whatever, and you see this emoji or you see this message. It's like a reminder to like go out, and it's like motivational" (participant 12, group 5; female aged 13 years).
Funny Sometimes large gyms are not for everyone; make use of smaller spaces such as a bedroom or a kitchen to fit in a quick work out. At least it will smell better !
• "[It's] ok-I would change it to make use of the space you are most comfortable in" (participant 4, group 2; female, aged 14 years).
Informative Exercise helps your body and your brain, so get to it ! • "I heard this before still, but I think it's like a short and simple nice quote to educate people on and to send a quick reminder because like that exercising can put you in a better mood and stuff like that" (participant 11, group 5; female, aged 16 years).

•
"And I like that it's short, but it doesn't give me motivation or anything. It's just like another text that would be sent, read, and not replied to, and wouldn't have any affect in my life" (participant 12, group 5; female, aged 13 years).
Informative Feel like you are tight on time? Maybe it's in your head . A brisk walk actually changes the chemicals in your brain to make you feel more relaxed and less anxious .
• "I feel like this one's great as well because if you're ever feeling anxious, you can go back to this in your brain and remember that one time you got a text message telling you how to fix it. So I think this is good for any situation where you're ever feeling anxiety" (participant 19, group 8; female, aged 14 years). • "I think this one's great because it explains it. It explains a bit of the science" (participant 20, group 8; male, aged 14 years).
Informative A for effort: being active can help you do better in school .

•
"I like how it ties in the school thing. I know that I would like to do better in school. And I think that it's the same for a lot of people my age. So, I think that this one could really capture a lot of people" (participant 8, group 4; female, aged 14 years).
Encouraging Improving your fitness is a slow process, but quitting will not speed it up! Keep up the great work!
• "I know a lot of people do better with motivation, so it'll help motivate you during that day. Whatever you're doing, you see this and you're like okay. Somebody believes in me..." (participant 12, group 5; female, aged 13 years).
Encouraging Rise and shine . Do something today that your future self will thank you for. Get yourself pumped up!

Theme 3: Extrinsic Versus Intrinsic Motivation to Change Behavior
Participants frequently reported that they had a goal to exercise more, but they differed in their sources of motivation or encouragement to do so. Parents and physicians were often cited as the person setting physical activity goals for participants, and frequent reminders were felt to be particularly frustrating. A participant noted that "a lot of adults tell kids to exercise and that they need to get out more" by quoting the following:

Theme 4: Effect of Financial Incentives May Depend on Intrinsic Motivation
Nearly all participants reported that their parents had offered rewards for behavior change, most commonly money or a desired item such as a video game or clothes. However, the perceived effectiveness and acceptability of this approach varied. Some reported that a monetary reward or desired object "kind of pushes you more and gives you a good boost of energy and motivates you a lot" (participant 11, group 5; female, aged 16 years) and that "I like the idea of getting something out of doing things" (participant 10, group 4; male, aged 18 years Financial rewards were also seen as a way to overcome barriers even for goals that were previously intrinsically motivated: .

Theme 5: Potential Benefits and Drawbacks of Loss-Framed Incentives
When asked about loss-framed incentives, using either money or physical objects, participants gave examples of times that their parents had threatened or actually removed objects for poor behavior. Most participants expressed frustration, stating that this approach would make them "sad," "mad," and "defeated or upset" and that it would be "unmotivating." However, some also acknowledged that this approach could be viewed as fair: .

Content Revision
On the basis of the participant feedback, messages were refined by study team members for use in the BEAM trial. Those rated as bad by even a participant were edited to omit unfavorable content or discarded entirely if the overall concept was disfavored. For example, the message Did you know that being physically active improve your self-esteem? Why not start right now? You got this was omitted, and messages that were supportive of self-confidence and self-efficacy but that did not unintentionally imply that the individual had poor self-esteem, were substituted instead (Ever feel like you're in a tough situation? Going for a walk can help you prepare to face that problem and overcome it). On the basis of the favorable response to messages referencing feeling lazy or describing how being physically active may help with school performance, more messages with these themes were created. Messages with mixed responses were edited based on specific negative feedback or included if responses were positive and neutral. Additional encouraging messages were created to address problem solving to overcome barriers to being more physically active and to remind the value of a support person to help achieve activity goals; for example, Does your step count goal seem like a stretch? Break it up! Every minute counts! and Who will be your workout buddy today? Bring a friend or family member with you on a walk or to the gym (blame us if you need to!). Messages were also shortened and adapted to fit the preferred timing in mid-day or afternoon (eg, You have the power to start your day off strong! Wake up, have your favorite breakfast and go for a walk, run, or bike ride. Get moving and you'll be ready to face the day was changed to Tomorrow, you have the power to start your day off strong! Plan to wake up, be active, and you'll be ready to face the day), and additional encouraging messages were created (eg, Is working toward your fitness goal a bit challenging? That's great! The more challenging a task, the more rewarding it will be in the end!) . Additional specific lessons learned are summarized in Textbox 1.

Principal Findings
Through virtual group interviews of adolescents with obesity and prediabetes or type 2 diabetes, we obtained end-user feedback to refine a bank of text messages for an mHealth-based physical activity intervention and identified several themes related to the use of text messaging and financial incentives to motivate youth to be more physically active. The participants' responses highlighted the challenge of developing message content that appeals to all end users [24], as well as the importance of keeping messages brief, upbeat, informative, and relatable. Adolescents' familiarity with rewards as a means to encourage behavior change allowed a rich discussion of their anticipated responses to financial incentives for physical activity motivation.
Our approach to intervention development is person-based, in which qualitative research is used to inform design by gaining insight into the perspective and psychosocial context of individuals who will use the intervention [25]. This qualitative assessment will be used not only in the intervention design but also in the evaluation of the intervention. Stakeholder involvement is an essential component of digital health interventions [26], but to date, text messaging interventions promoting healthy lifestyle changes for adolescents with obesity have been heterogeneous, with a limited description of whether, and how, the intervention was co-designed with end users [15,27]. One recent exception is the TEXTBITES intervention, in which Partridge et al [27] used an iterative mixed methods approach to develop a text message program for Australian adolescents with obesity. We took a similar approach to develop a text-message-based physical activity intervention for adolescents with overweight or obesity and prediabetes or type 2 diabetes (BEAM trial). Like TEXTBITES, our population will also include adolescents with obesity, but as reflected in this study, a large proportion of participants in our trial will be youth of minority race or ethnicity and from disadvantaged socioeconomic backgrounds towing to the higher prevalence of type 2 diabetes and prediabetes in these populations [28,29].
The likelihood of limited financial resources or safe spaces to engage in physical activity shaped our physical activity suggestions, which were intentionally free or low-cost and included in-home options. In addition, our population also likely differs from otherwise healthy adolescents in terms of overall stress and depressive symptoms; youth with type 2 diabetes report high levels of life stressors, which correlate with poor psychosocial functioning and impaired treatment adherence [30]. Thus, for our ultimate goal of helping adolescents with prediabetes and type 2 diabetes to become more physically active, we worked to develop messages that were both encouraging and contained activity suggestions that were perceived as realistic and achievable.
Our finding that adolescents preferred messages that were brief, upbeat, and informative was similar to previous reports [18,31]. In their TXT Me! intervention, Thompson et al [18] used pedometers and stand-alone text messages grounded in concepts from the Self-Determination Theory, including autonomy, competence, and relatedness, to promote increased physical activity among adolescents [18]. They used web-based surveys and telephone interviews with 30 persons, aged 14 to 17 years, to assess the acceptability of pedometer use and daily texts to help achieve a step-count goal. Similar to our findings, the adolescents in the study by Thompson et al [32] favored short messages that were positive, used exclamation points, did not nag, and focused on facts. In a pilot study of their intervention [32], they demonstrated feasibility, with an enrollment of 160 adolescents and complete data available in 86% of cases. Notably, they found that participants who were excluded owing to insufficient pedometer use or missing data collection were more likely to be older and African American. Postintervention feedback included the suggestion that other types of physical activity should be promoted in addition to walking and that a step count goal should be set for participants rather than having the goal be self-selected. In total, 80% (16/20) of the adolescents enjoyed receiving daily texts for 12 weeks, and 75% (15/20) chose noon or afternoon as the best time to receive texts. Notably, participation was not based on body weight or health conditions, including obesity, type 2 diabetes, or prediabetes. In contrast, Woolford et al [31] explored attitudes and preferences regarding automated SMS text messages among adolescents with obesity as part of a multidisciplinary weight management program. The messages included an explicit focus on topics central to weight management. As demonstrated in our study and in TXT Me!, adolescents reported that an automated SMS text messaging strategy would be acceptable and preferred positive upbeat messages that included exclamation points [31]. Again, the study by Woolford et al [31] did not explicitly focus on adolescents with obesity-related complications such as prediabetes or type 2 diabetes.
To our knowledge, ours is the first study to evaluate the acceptability of a text message-based intervention to increase physical activity in adolescents with obesity and prediabetes or type 2 diabetes. Because of the critical importance of lifestyle changes in improving health outcomes, youth with prediabetes or type 2 diabetes are likely to have already been instructed by multiple health care providers and their caregivers to be more physically active. It is notable that despite this commonly heard advice, adolescents still reported that daily or twice-daily text messages encouraging and reminding them about physical activity would be acceptable and potentially motivating.
In addition to specific feedback about the reviewed text messages, participants shared valuable insights about experiences when encouraged to exercise by caregivers and health care providers. Their responses highlighted the fine line required when showing concern about an adolescent's health and health behaviors. Although expressing genuine concern may be encouraging for some, repeated reminders may be perceived negatively by others. There is a danger that repeated text message reminders to exercise may also be perceived in this light. Our revised text message bank now places greater emphasis on being more physically active in any activity of their choice, rather than structured and repetitive exercise. Indeed, distinguishing between exercise (a planned, structured, purposeful, and repetitive behavior) and more general physical activity may help clarify expectations for patients, as increased nonexercise physical activity may be an acceptable goal for both patients and providers [33]. As part of the planned intervention, we will counsel participants' caregivers about the benefit of showing support for the adolescent's physical activity in ways that can be more naturally incorporated throughout their days, rather than insisting that the adolescent engages in narrowly defined exercise at specific times. The specific linkage of the health benefits of physical activity to diabetes-related outcomes may also prove motivational to youth already experiencing an obesity complication, driving intrinsic motivation rather than relying primarily on extrinsic motivation. As described by the health belief model, the perception of health vulnerability associated with prediabetes or type 2 diabetes may serve as an additional motivation to act [34] on healthy lifestyle changes. This concept was suggested by some youth who referred to their health as a primary motivator to increase physical activity, but again, health care providers should balance using health as a motivator while not inducing shame or stigma that is commonly associated with obesity and type 2 diabetes [35].
Because of the challenges of maintaining engagement in mHealth and physical activity interventions as well as motivating adolescents to be more physically active, we also tested the impact of different financial incentive strategies in our BEAM trials. Specifically, we evaluated loss-versus gain-framed financial incentives to achieve step-count goals.
Although the comparative efficacy of these financial incentive approaches has been evaluated in physical activity interventions in adults, it has not been assessed in youth [36], who may differ in their responses to the negativity of a loss-framed incentive. Among adolescents with type 1 diabetes, financial incentives appear to be an acceptable approach for promoting self-monitoring behaviors [37]. However, the acceptability of loss-framed incentives has not been explored. When evaluated in a clinical trial, loss-framed financial incentives led to more frequent blood glucose monitoring among adolescents with sub optimally controlled type 1 diabetes [38]. Participants in that trial endorsed the feasibility of daily financial incentives to motivate behavior change, with some reporting feeling motivated by the loss of money they believed was already their own. Notably, however, the increase in self-monitoring behaviors was transient, and the effect was extinguished after financial incentives were stopped. The possibility of financial incentives to crowd out intrinsic motivation has led to the concern that financial incentives may do more harm than good [39]; fortunately or unfortunately, this is less of a concern for health behaviors, many of which are associated with low baseline intrinsic motivation [40]. Our exploration of attitudes toward gain-and loss-framed incentives suggested that the acceptability of loss-framed incentives is mixed but that some adolescents may find them highly motivating. This study was not designed to quantify the differences in the anticipated motivation or acceptability of different financial incentive approaches. However, in the BEAM trial, we explore the heterogeneity of the treatment effect of the incentives in an effort to identify characteristics that predicted objective responsiveness to different financial incentive approaches.
Although not initially planned, owing to the COVID-19 pandemic, we conducted group interviews using a virtual format using videoconferencing technology. Virtual interviews had several benefits: (1) ease of recruitment, particularly for adolescents with financial or transportation barriers; (2) efficient and high-quality audio recording via the platform; and (3) availability of the chat feature, which allowed participants who may not otherwise have interrupted someone speaking to share their thoughts and engage in the discussion. The virtual format allowed participants to mute video or audio, which was done at times the strain discussion progressed. However, the ability to retain control of their privacy, which is not practical in most traditional focus groups [41], may have encouraged participants to share more freely than they would have in an in-person setting. Technical challenges occasionally arose, including poor audio quality related to internet connectivity. Overall, the virtual format appears to be a promising way to engage adolescents in qualitative research in a way that minimizes participant burden while facilitating recruitment. Our successful engagement of participants in a remote format also provides further evidence of the feasibility and acceptability of mHealth interventions targeting adolescents [42,43].

Limitations
Attitudes and perspectives about text message content and financial incentives may vary across populations; therefore, our findings may not be generalizable to other adolescent subgroups, such as those with obesity but without comorbidities, or to individuals of different age groups or at different neurodevelopmental stages [36]. However, our participants represent a diverse group whose perspectives are not often elicited or reflected in research settings and represent the epidemiology of youth with prediabetes and type 2 diabetes, who are disproportionately from minority backgrounds [29]. As with any group interview or focus group study, particularly those including adolescents, social desirability bias may have influenced the responses [44]. We did not use single-sex groups, as has been recommended when conducting qualitative research with groups of unfamiliar adolescents [44]. However, our small group sizes may have limited the potential discomfort or embarrassment of offering differing opinions. Because the BEAM trial uses text messages in English, we did not include participants with limited English proficiency. We acknowledge that this has the potential to limit the intervention's reach among Hispanic youth, who have a higher prevalence of type 2 diabetes than non-Hispanic White youth [29]. Importantly, to adapt text message interventions for individuals with limited English proficiency, it is critical to embark on a transcreation process, in which the text messages are not simply translated but adapted in a way that is linguistically and culturally appropriate [45]. Thus, additional qualitative research is required to evaluate the acceptability and appropriateness of the adapted intervention.
The focus of this study was on the perspectives and preferences regarding text messages and financial incentives to motivate behavior change, which may not perfectly correlate with objectively measured behavior change; objective measurement will occur as part of the BEAM trial. Because of our study's sample size, we were unable to reliably characterize differences in attitudes and perceptions across patient characteristics, such as age, gender, or prediabetes versus diabetes; however, we will do so when evaluating the acceptability of the intervention in the BEAM trial. The use of virtual interviews made engaging participants challenging at times; this limited engagement often contributed to more limited discussion than may have occurred in face-to-face interviews and thus to the shorter actual duration of interviews than originally planned (30-45 vs 60 minutes). Finally, although participants reported that they found even frequent text messaging acceptable, their actual behavior may differ, as additional interruptions from push notifications and text messages from friends also occur. The BEAM trial will also evaluate the impact of message frequency.

Conclusions
In summary, among the participants in our study, all adolescents with obesity and prediabetes or type 2 diabetes preferred text messages that were short, upbeat, informative, and relatable. Through virtual group interviews, participants' feedback allowed for the creation of an end-user-refined bank of text messages that will be used in an mHealth physical activity intervention targeting the same population (BEAM trial). The perceived acceptability of loss-framed incentives was mixed, but the effect of incentive framing will be objectively measured in the BEAM trial. Overall, the ability to elicit end-user feedback increased the likelihood of acceptability of our intervention among the target population, and we will continue to engage participants throughout the trial to further refine our approach to intervention development.
Objective: We aimed to systematically evaluate the quality, content, and reliability of YouTube videos about gestational diabetes. Methods: A systematic search of YouTube videos was conducted over the course of 1 week in April 2020 using the following keywords: "gestational diabetes," "gestational diabetes management," "gestational diabetes treatment," and "pregnancy and diabetes." The search results were displayed by relevance, replicating a default YouTube search attempt. The first 60 results from each keyword were reviewed (n=240). Exclusion criteria were videos unrelated to gestational diabetes, videos not in English, and those for which the full video was not available at the time of review. For each unique video, a gestational diabetes content score was used to rate video comprehensiveness and accuracy, and the DISCERN instrument, a validated metric to assess consumer health information, was used to evaluate the reliability of information presented. Videos were further categorized by quality: videos with DISCERN scores lower than 3 (out of 5) or a content score less than 4 (out of 7) were categorized as low quality, and all others were designated high quality. We performed descriptive analysis and compared video characteristics by source and quality rating.

Results:
For 115 unique videos, the mean content score (out of 7) was 3.5 (SD 2.0) , and the mean DISCERN score (out of 5) was 2.7 (SD 0.7), representing low to moderate information comprehensiveness and reliability respectively. Video sources were categorized as personal vlog (12/115, 10.4%), web-based education (37/115, 32.2%), medical (52/115, 45.2%), business or company (13/115, 11.3%), and media clip (1/115, 0.9%). DISCERN and content scores trended higher among medical and web-based education videos. The majority of videos (n=88) were categorized as low quality, while 27 videos were categorized as high quality. Video duration was longer for high-quality videos (P<.001); high-and low-quality videos otherwise had similar views and viewer interaction numbers.

Conclusions:
Although high-quality videos about gestational diabetes exist, reliability, accuracy, and comprehensiveness were low overall, and higher quality was not associated with increased viewer interaction. It is important to acknowledge the limitations of this platform and to assist patients in accessing high quality content and differentiating the quality of information sources.

Introduction
Gestational diabetes is one of the most common complications of pregnancy. Untreated or undertreated gestational diabetes is associated with perinatal and maternal complications, including preeclampsia, macrosomia, neonatal hypoglycemia, and maternal risk of developing type 2 diabetes [1]. Because management of gestational diabetes may include a combination of nutritional therapy, physical activity, blood glucose monitoring, or medication, affected individuals require health behavior changes beyond those required during routine pregnancy care in order to have best outcomes [2][3][4]. These increased requirements are particularly noteworthy because individuals with gestational diabetes are more likely to belong to racial, ethnic, and socioeconomic groups that face barriers to access to high-quality care [1,5,6]. Because an individual's access to and application of informative resources about gestational diabetes testing, treatment, and self-management may determine their risk of adverse pregnancy outcomes, improving access to high-quality health information and support for health behavior change is crucial [6][7][8][9][10].
The internet is a widely available, increasingly accessed resource for health information, and it is a common source of information for pregnancy-related health concerns [11,12]. Multiple studies [13][14][15][16][17] have found high prevalences of internet and other digital media use among pregnant patients in the United States-estimated to be greater than 70% among patients of different ages, races or ethnicities, and socioeconomic groups. Web-based information-seeking has been reported for many pregnancy-related topics including fetal development, pregnancy complications, prenatal care, medication safety, nutrition, and gestational weight gain [15,[17][18][19]. The rapidity of information access and availability of supplemental explanation are reported advantages of seeking information from the internet [17]. YouTube is a particularly important source of health information, both because of its popularity as one of the most used web-based platforms, and because its information is unregulated [20][21][22][23][24][25]. Evaluations on a variety of topics, including infertility treatment, medication safety during pregnancy, and diabetes outside of pregnancy have reported variability in the value of YouTube as a health information source and some found that videos provided incomplete or misleading information, while others have reported advantages, including the availability of personal experience-based information [26][27][28]. However, little is known about the content and quality of information from YouTube on gestational diabetes.
Given the emerging role of web-based platforms-including YouTube-as health information resources, and the importance of optimizing health care information and interventions for individuals with gestational diabetes, it is important to learn more about the content and quality of resources about gestational diabetes that are available on YouTube. Restrictions on access to care [8], limited health literacy [18], and lack of availability of culturally appropriate interventions [8,29,30] are well-known barriers in the provision of prenatal care for gestational diabetes, and understanding resources-such as videos available on YouTube-that may help combat those barriers is vital. Thus, our objective was to examine the quality and content of videos about gestational diabetes in order to ultimately develop opportunities for better, more accessible health care delivery.

Search Strategy
We systematically reviewed YouTube content on gestational diabetes on April 30, 2020. Search terms were identified in consultation with 2 perinatal diabetes care specialists (CN and LY), and Google Trends was used to identify top searches associated with gestational diabetes. The search was then conducted on a cache-cleared web browser using incognito mode with the following terms: "gestational diabetes," "gestational diabetes management," "gestational diabetes treatment," and "pregnancy and diabetes." The search was conducted via public access-a registered account was not used.
Search results were displayed by relevance, which is the default search setting. The first 60 results-representing the first 3 pages of results-from each keyword search were reviewed and analyzed. The sample size and sort strategy were selected based on research indicating that the majority of internet searchers view only 1 page of results, and 83% of searchers view no more than 3 pages [31]. Duplicate videos, videos unrelated to gestational diabetes, videos in a language other than English, and clips for which the full video was unavailable were excluded.

Data Extraction
Descriptive characteristics of each video were collected, including video length, number of comments, channel number of subscribers, number of views, number of likes, and number of dislikes. Collection of this information was completed by 1 reviewer (EB), over the course of 1 week (from April 26, 2020 to May 2, 2020) to minimize variability in number of views and other characteristics collected. Source characteristics were also gathered from each video. These included the video source and whether the video was a character video, which was defined as a video in which a specific, identifiable person was presenting. Video source was determined based on the affiliation of each video's author when available (eg, identified as an employee of a hospital system or pharmaceutical company), or the channel description and theme (eg, medical school test preparation videos or personal vlog in which the author presents primarily personal experiences).

Content and Quality Assessment
Criteria to judge video content were formulated from American College of Obstetricians and Gynecologists recommendations for management of gestational diabetes, and in consultation with perinatal diabetes care specialists (CN and LY). Video comprehensiveness and accuracy of information provided about (1) screening or testing guidelines; (2) treatment for gestational diabetes self-management; (3) nutrition information; (4) information about the purpose of gestational diabetes treatment and benefits; (5) blood glucose level monitoring guidelines; (6) gestational diabetes effects on long-term maternal health; and (7) gestational diabetes effects on the fetus. These topics were selected because they represent key foundational knowledge regarding gestational diabetes that is essential for participants to understand their condition, participate in self-monitoring and management, and engage in health care to prevent adverse perinatal outcomes. To evaluate videos, we used a content score-1 point for each content area addressed accurately in the video (ie, the video included relevant information in that topic area, though depth of explanation varied); thus, a maximum score of 7 was possible if a video covered the comprehensive range of topics about gestational diabetes accurately. Similar assessments have been performed in prior evaluations of YouTube videos as a health information source [32][33][34][35][36].
The DISCERN instrument [37] was used to assess the quality and reliability of the videos as information sources. The DISCERN instrument consists of 16 questions assessing the (1) reliability, (2) quality, and (3) overall rating of web-based publications. DISCERN criteria are written as questions and are rated on a scale from 1 (not satisfied) to 5 (fully satisfied).
The DISCERN tool has been validated for use in a variety of settings, and it has been used in similar studies that have evaluated YouTube videos as a source of health information [34,35,[38][39][40][41][42].
A combination of both scoring techniques was used to divide videos into high-and low-quality categories. Videos with a mean DISCERN score less than 3 or a content score less than 4 were defined as low quality; otherwise, videos were defined as high quality. These quality cutoffs were chosen because they correspond to a DISCERN rating representing potentially important shortcomings or an inclusion of fewer than half of the expert-developed content topics, respectively, both of which affect the overall quality of the video as a reliable comprehensive source of information.
After determining evaluation criteria, 2 videos were assessed collaboratively by 3 reviewers (EB, JJ, ED) to establish consensus on the application of these criteria and scoring. Subsequently, these 3 reviewers independently evaluated the same 15 randomly selected videos for content and reliability. The remaining videos were then divided evenly among the reviewers, and each was evaluated using the established standardized criteria. Data for each video were collected and stored using REDCap (Vanderbilt University) software.

Statistical Analysis
Statistical analyses were performed using Excel software (version 2020; Microsoft Inc). Interrater agreement was analyzed using intraclass correlation coefficients and single-factor analysis of variance. Descriptive statistics were used to evaluate study data. Frequencies were used to describe categorical variables; normally distributed continuous variables were presented as means, and nonnormally distributed variables as medians. The Kruskal-Wallis test for intergroup comparison, Spearman rank correlation, and Wilcoxon rank-sum test were used for analysis involving nonnormally distributed continuous variables. A P value<.05 was considered significant.

Ethics Statement
This study was deemed exempt from review by the Northwestern University institutional review board because it does not involve human participants.

Results
The mean scoring disparity for the 15 videos was 0.27 (for DISCERN), and correlation was 88% (intraclass correlation 0.81), indicating a high level of interrater agreement. Of the 240 videos initially identified, 109 were duplicates, 3 were in a language other than English (Urdu, Nepali, Bengali), 2 with full videos unavailable at the time of data recording, and 11 had a primary topic that was not gestational diabetes. After applying exclusion criteria, 115 videos remained for analysis ( Figure 1; Multimedia Appendix 1). Video characteristics (Table 1)  Of 115 videos reviewed, 27 videos were designated as high quality, and 88 videos were designated as low quality. The low-quality group included 12 videos that were rated as low only because they included fewer than 4 topics; their DISCERN scores ranged from 3.0 to 3.6. Therefore, the scoring system did not result in a large number of videos with reliable in-depth information about a small number of topics, but instead, a small number of videos with moderate reliability and low topic comprehensiveness, being found in the low-quality categorization. Longer video duration was associated with higher quality (P<.001), but the numbers of views, comments, likes, dislikes, and subscribers were not associated with quality (Table  4). Among high-quality videos, 24 out of 27 were from web-based education or medical sources, 2 were personal vlogs, and 1 was from a business or company.

Principal Findings
In this study, we identified no relationship between video quality and indicators of video popularity. These findings indicate that the information available to pregnant individuals with gestational diabetes is highly variable in quality, and may not deliver accurate or comprehensive information to users.

Comparison With Prior Work
YouTube has been evaluated as a source of health information on a variety of topics, including emerging infectious diseases and pregnancy [43,44], medication use in pregnancy [27], and diabetes outside of pregnancy [34], as well as colorectal cancer [36], ankylosing spondylitis [39], and rheumatoid arthritis [40]. The average DISCERN score in this study is concordant with that found in many previous studies of YouTube videos [34,35,[38][39][40]43]. Scores in previous studies have typically fallen in the low to moderate reliability range, indicating potentially important shortcomings. Although studies evaluating other health conditions used various topic-specific comprehensiveness scales, they found similar levels of topic coverage, accuracy, and comprehensiveness to those found by us [36,40,41]. We found that this level of topic inclusion represents notably incomplete coverage of the health condition. The majority of videos we evaluated covered only 3 to 4 key topics about gestational diabetes accurately, and only 1 topic-gestational diabetes management-was included by more than 50% of videos. The overall low comprehensiveness of individual videos evaluated has implications for their utility as health information resource. In further research, it could be helpful to investigate YouTube user viewing patterns to ascertain whether these shortcomings can be overcome through careful combination of videos and topics covered.
Interestingly, in our study, one of the lowest rated videos overall by both DISCERN criteria and topic coverage had some of the highest views, likes, comments, and channel subscribers. This video was a media clip related to a television personality; therefore, the high numbers of views, likes, comments, and subscribers are likely because viewers accessed the clip for entertainment rather than for information about gestational diabetes. However, this highlights an important aspect of finding health information on YouTube-the search algorithm may not take the quality of the video into account for any specific search term, and quality varies significantly between videos. As we found and as has been seen previously, there are rare videos available with both good reliability and comprehensive topic coverage; however, it is difficult to ensure that these videos are accessed, as they are not necessarily the most viewed, liked, or highly ranked videos available [38,45,46].

Clinical and Research Implications
Pregnant patients increasingly seek health education via web-based sources, including YouTube, and web-based media can be helpful adjuncts to clinical care [15,17,25,[47][48][49]. Trust in information found on the internet is often reported to be high, and many people who sought information on the internet did not discuss it with their providers [15,17]. Therefore, it is important for providers to be aware of the overall use and limitations of YouTube as a resource when counselling patients with gestational diabetes, and it may be useful for clinicians to query and understand the resources patients are using to augment their clinical care.
However, it is also important to note that YouTube has the potential to address the specific needs of individuals with gestational diabetes. Health information and supportive social networks-which are both easily accessed via the internet-have been found to influence the challenges experienced by people with gestational diabetes [49][50][51]. In addition, patients' health literacy levels and access to information are known to affect glycemic control and gestational diabetes-related outcomes [52][53][54][55]; interventions for pregnant individuals at an appropriately targeted health literacy level are particularly essential to address health inequities present in gestational diabetes care and outcomes [8,56]. Video-based information has frequently been used to combat care barriers related to health literacy [57][58][59]. Despite the limitations found in this study, YouTube videos remain a low-cost, easily accessible resource, and further interventions should investigate ways to address these limitations in order to harness YouTube's potential as a patient education tool. Clinicians also have the opportunity to assist patients who are using YouTube as resource by recommending specific videos, discussing ways to identify reliable resources, or even creating high-quality content. Health care providers involved in gestational diabetes care are likely to have both the knowledge base and the resources available to create accurate comprehensive videos with information that is helpful for patients with gestational diabetes.

Limitations
To the best of our knowledge, this study is the first to assess the reliability, content, and quality of YouTube videos about gestational diabetes, but it has several limitations. First, while the sample size was chosen to encompass the majority of videos likely to be encountered by individuals searching for information about gestational diabetes, we cannot ensure that all videos that a user may find were included, given the dynamic nature of video uploading. Search history or location may also affect ranking and accessibility of videos in ways that were not demonstrated in this paper. Second, we could not evaluate the target audience for each video, because most videos do not indicate their audience and default search setting will not necessarily preclude people from accessing videos aimed at different groups. As such, our assessment of quality and reliability may not reflect all videos with which pregnant individuals engage. Additionally, our search was limited to English-language videos, and future work is necessary to corroborate our findings with those for videos in other languages.
We also noted little gender, racial, or ethnic diversity among YouTube presenters in videos about gestational diabetes.
Especially because social media sites, and YouTube specifically, may be sought as a more personal or personal-experience based resource, this lack of diverse representation is a noteworthy limitation. Based on the representation found in the videos evaluated, many patients may not see themselves or their experiences reflected fully.

Conclusions
Despite the relatively low quality and comprehensiveness found among YouTube videos for individuals with gestational diabetes, YouTube is an easily accessible and increasingly important source of health information. Providers caring for individuals with gestational diabetes should remain aware of the limitations of this resource, especially when discussing information sources with patients. Further study may be helpful in elucidating ways to harness the potential strengths of YouTube for providing high-quality accessible health information.