This study was conducted at Hasbro Children's Hospital (Brown University Health) in Providence, RI from July 2022 to May 2023. Participants were recruited from the Pediatric Endocrinology Clinic and the Primary Care Clinic during their regular clinic appointments, by members of the research staff. A convergent mixed methods design was used, with survey and interview data collected concurrently. Data were analyzed independently, and then integrated to identify areas where qualitative themes explained, or expanded on the quantitative results.

Participants were recruited from the Pediatric Endocrinology and Primary Care Clinics at Hasbro Children's Hospital. Participants were required to be 13–17 years old, have overweight or obesity (BMI for age/sex ≥ 85th percentile), speak English, and have daily access to a smartphone. Exclusion criteria included medical conditions that would preclude adherence to dietary recommendations (such as inflammatory bowel disease), genetic or endocrine causes of obesity, a diagnosis of type 1 or type 2 diabetes mellitus, severe developmental delay and/or serious cognitive impairment, current use of psychotropic medications which could affect weight, or being non-English-speaking.

Eligibility was determined via chart review by research staff and confirmed by discussion with the participant and their parent. All participants received compensation for their participation in the study.

Informed consent and assent were obtained from all participants prior to the study visit. Each participant met with a member of the research team once over a HIPAA-compliant video platform. Demographic data, including age, school grade, sex, gender identity, race, ethnicity, type of health insurance, BMI (extracted from the Electronic Health Record), and technology use data, was collected in REDCap and de-identified.

Participants reported their current weight-related stage of change by selecting the statement that best described their current weight situation, based on the transtheoretical model for weight management (64): (1) At the moment, I am not doing anything to lose weight, and I have no intention about doing anything over the next 6 months; (2) At the moment, I am not doing anything to lose weight, but I am thinking about doing something over the next 6 months; (3) I have been making an effort to lose weight (by dieting and/or exercising) for less than 6 months; (4) I have been making an effort to maintain my weight (by dieting and/or exercising) for 6 months or more.

Based on subject enrollment order, participants were assigned to review (via videoconference) one of three different subsets of video lessons, video animations, and mindfulness exercises from a commercially available mindful eating application (Eat Right Now®), which was designed for and tested with adult participants (39). The goal of this grouping was to keep each interview time-limited and engaging. After playing the clip of each video lesson, video animation, or mindfulness exercise to the participant, the researcher conducted a short semi-structured interview to elicit participant opinions about the app, including likes and dislikes about app features, suggestions for change, and how to adapt the app for adolescents (Table 1). Interviews were audio-recorded, transcribed, and then reviewed for accuracy and appropriate deidentification.

The study visit concluded with a survey that was developed by our group based on the Unified Theory of Acceptance and Use of Technology (UTAUT) constructs. Participants completed the survey online using REDCap electronic data capture tools hosted at Brown University Health (65). The survey included 13 statements, organized in eight constructs assessing Performance Expectancy, Effort Expectancy, Social Influence, Facilitating Conditions, Attitudes toward Technology, Self-Efficacy, Anxiety, and Behavioral Intention. Participants responded to each statement via a 5-point Likert scale (66). We analyzed average scores for each construct to determine participant perceptions of the app. We assessed the percent of participants that provided favorable ratings for each statement, a 4 or 5/5 (agree or strongly agree), or 1 or 2/5 (strongly disagree or disagree) for the Anxiety construct, which is reverse-coded. An a-priori benchmark of 80% of participants reporting favorable ratings was used to gauge acceptability.

Reliability and validity were also estimated for the UTAUT-based survey, using principal components analysis. Scale validation focuses on correlations. If the questions are all valid measures of the same underlying trait, in this case willingness to use technology for managing mindful eating, then participants who endorse high scores on one question should also endorse high scores on other questions as well– a pattern of positive correlations. Principal components analysis summarizes these correlations in a loading, denoted λ, which can be interpreted like a correlation between an individual question and all other questions on the questionnaire. Values close to 1 represent positive correlations with most other questions, values close to −1 represent generally negative correlations with other questions, and values close to 0 represent no association. This can be used to evaluate whether or not all questions are equally relevant to participant attitudes, which reflects internal structure validity. This provides information on how to interpret what scores represent, as well as which questions are most insightful to individual attitudes.

The values of the loadings were used to estimate reliability, interpreted as representing the percentage of total variation in scores that is meaningful. Reliability of 0.70, or 70% meaningful variation, is often considered acceptable. Lastly, the correlation between summed scores on the UTAUT and the two questions about Behavioral Intention was estimated to represent concurrent validity. If the UTAUT validly represents willingness to use a mindful eating app, then participants who endorse higher scores on the questionnaire will also endorse greater willingness to use the app when directly asked, a positive correlation. Using this approach allowed us to investigate this empirically. A Bayesian estimation procedure was used because the sample size is modest (N = 25 participants).

All data was de-identified before analysis.

Framework matrix analysis (FMA) and applied thematic analysis (ATA) were used to analyze qualitative data from the interview transcripts. Considering the time-consuming process of ATA in the context of app development (67), we conducted FMA first, as it provided real-time feedback for the design and initial development of our mindful eating web-based app (SAMBA). Together with the results from the UTAUT survey, which gauged adolescent interest in a mindful eating app, the results from the FMA of the interview transcripts informed the development of the initial version of the SAMBA app. Once app development was underway, we again analyzed the interview transcripts, via comprehensive ATA, to gain a more nuanced perspective on the app's successes and potential areas of improvement for future iterations of the app.

First, all 25 audio-recorded participant interviews were transcribed by members of the research team. Full review of the transcripts yielded inductive (based on the researcher's interests) and deductive (based on interview responses) FMA codes. A research assistant entered summaries of participants' comments, relevant quotes, and key decisions regarding app development into the framework matrix. After the framework matrix was built, matrix contents were reviewed by the principal investigator and deemed accurate and complete. Data helped inform app design and development (67).

After FMA was complete, we used the framework matrix to create a second codebook for ATA. Two coders reviewed the ATA codebook and independently coded the first transcripts, then met to review their coding, finalize the codebook and add additional inductive codes. With the finalized codebook, the two coders then independently coded all 25 transcripts. After the initial coding was completed, the two coders met to review 5 (20%) randomly selected transcripts to review codes and resolve discrepancies. The master coded transcripts were entered into NVivo software (version 14.0). Concordance was calculated at 90%, demonstrating a high consistency between the coders.

In total, 25 participants completed the research study, including providing demographic and technology use information, and completing the semi-structured interview and UTAUT questionnaire. Fifteen (60%) identified as male. In terms of race and ethnicity, 5 identified as African-American (20%), 4 identified as white (16%), and 15 (60%) identified as Hispanic or Latino. Seventeen (68%) had severe obesity, defined as a BMI percentile at ≥120% of the 95th percentile. All 25 participants reported owning a smartphone and using apps. Detailed demographic information is summarized in Table 2. When asked about their weight-related stage of change, 2 (8%) were contemplating changing their behaviors, 3 (12%) were preparing to change their behaviors, 16 (64%) were actively changing behaviors, and 4 (16%) were in the maintenance stage of behavior change.

Findings from the UTAUT survey are summarized in Table 3. The Performance Expectancy, Social Influence, Facilitating Conditions, and Attitudes toward Technology constructs received average scores between 4 and 5/5, indicating that most participants responded “Agree”, or “Strongly Agree” to the questionnaire statements for these constructs. However, the Effort Expectancy and Self-Efficacy constructs had lower average scores, between 3 and 4. Over 80% of participants selected either “Agree” (4/5) or “Strongly Agree” (5/5) with at least one statement from each construct, including Behavioral Intention, with the exception of the Self-Efficacy construct. The Anxiety construct, which was reverse coded, received an average score of 2.27 [1.62, 3.07], indicating that most participants responded “Disagree” (2/5) or “Neither Agree nor Disagree” (3/5) to these statements. These results suggest that adolescents are motivated and think a mindful eating app would be helpful, but they may lack confidence in their ability to use it. Overall, the high levels of agreement for these constructs support the development of a mindful eating app for adolescents.

In terms of internal validity of the survey, the loadings (which range from −1 to 1) representing common association between an individual question and all other questions, were large (λ > 0.5) except for facilitating conditions (λ = 0.01). This is because all teens endorsed this question, which indicates it provides no meaningful information. The association with behavioral intention was positive [correlation > 0.15, 95% CI (0.07, 0.27)], which supports the validity of the survey, meaning that those who endorse the constructs say they would be more likely to use the app. Reliability was appropriate [0.70, 95% CI (0.50, 0.85)], suggesting that similar youth would get similar scores using this questionnaire.

FMA supported the findings from the survey in that participants overall were receptive to a mindful eating mobile app. It also pointed to additional app features that adolescents consider important, which were not present in the app that was presented to them during this study, as it was developed for an adult audience (39).

Analysis of inductive codes “attitude and tone” and “user flexibility” found that youth overall found the app relatable and encouraging; however, they disliked content containing moralizing language about food and weight. They also preferred the self-directed exercises that provided many choices.

Analysis of the deductive codes “easy to understand” and “educational” revealed that adolescents found the app educational and easy to understand. Analysis of the deductive codes “engaging” and “visually appealing,” contained mixed responses. While some youth found the app engaging, others wanted shorter videos, captioning, and more visual engagement. Similarly, while some youth found the app visually engaging, others provided suggestions to increase engagement, including text message reminders, introduction videos, gamification features, and teen-specific examples.

Given that adolescents disliked moralizing language about food and weight, we avoided the use of such language in the SAMBA app. The SAMBA app was designed to include directions for following certain exercises depending on each person's daily experience, and at their own pace, since adolescents preferred self-directed exercises that provided choices,

Youth preferences highlighted the need for shorter videos, captioning, and more visual engagement; SAMBA's lesson videos include captions, have an abundance of visual aids, and are shorter and more colorful than those in apps designed for adults, such as Eat Right Now®. We followed participants' suggestions to add customizable text reminders, and thus the SAMBA app sends five text reminders per week. We included two introduction videos that must be watched before the video lessons. We also included personalization features, including the opportunity to choose between two different characters who are wearing different clothes, have different gender and different body weight. In terms of gamification, the SAMBA app also allows participants to accrue experience points, XP, for their app use (i.e., completion of audio meditations, video lessons, mindful eating exercises). As their XP increase, youth progress through different experience levels based on their XP accrual: beginner, intermediate, advanced, and pro-mindful eater, and these levels are displayed on the main screen (Figure 1). For the pilot study to test the app, our plan is to offer prizes if participants accrue enough XP reflecting that they have used the app as intended. The app also includes teen-specific examples to increase engagement (i.e., the lesson videos include examples pertinent to school experiences, as opposed to work experiences which would match an adult audience).

During review of the master coded transcripts, three themes emerged. These themes embodied the participants' opinions about the app's visual engagement and interactivity, ease of use, and the overall tone of the app.

Results from the UTAUT-based survey showed that a mindful eating app is acceptable to adolescents overall, with high scores for Performance Expectancy, Social Influence, Facilitating Conditions, and Attitudes Toward Technology. However, the survey results also identified potential challenges in terms of Effort Expectancy and Self-Efficacy. Qualitative analysis supported and contextualized the survey findings. Advanced thematic analysis showed adolescents to be receptive to a mindful eating app, and also identified themes pointing to essential features for successful youth engagement.

Similarly to prior work, given that the FMA methodology has a faster turn-around time, it was used in real time to inform app design and development, whereas final insights from ATA were available after app development (67). Importantly, analysis of the semi-structured interviews using FMA and ATA yielded consistent results, an approach for analyzing qualitative data that has been described in past research (67). Specifically, both FMA and ATA codebooks elicited similar themes regarding the importance of an easy-to-use, engaging, and visually appealing app for adolescents.

While the FMA codebook summarized the participants' individual interview responses about the app and shaped the development of the SAMBA app, the ATA codebook identified broader themes. In addition to the major themes elicited by analysis of the semi-structured interviews, participants provided constructive suggestions regarding visual engagement, gamification, and the addition of teen-specific examples.

Participants' answers to the UTAUT survey constructs demonstrated that they believe that a mindful eating app may improve their relationship with food and help them be healthier. Overall, participants agreed that they have the resources and time to use this type of app, and that they (and their parents) would have a positive perception of this technology. However, participants cited lower agreement with the Effort Expectancy and Self-Efficacy constructs, indicating that they may find a mindful eating app confusing or intimidating, or they anticipate potential challenges with learning to use a mindful eating app. Qualitative themes that emerged from thematic analysis suggested several potential features that may mitigate youths' lack of confidence in successfully using such a digital tool. Specifically, youth endorsed that a mindful eating app for adolescents needs to have a specific design, allowing it to be engaging, interactive, and visually appealing; it should also have a specific structure and flow, allowing it to be flexible, adaptable, and easy to use, and finally, it should have content that is relatable, non-judgmental, and helpful.

Pilot testing of health apps, such as the SAMBA app, should include exit interviews to assess for specific barriers to Effort Expectancy and Self-Efficacy that may affect youth engagement with apps and their content. This would inform user feedback-driven iterations of the SAMBA app to optimize adherence and clinical outcomes.

The SAMBA web-based app, a mindful eating app for adolescents, was developed with the incorporation of adolescent feedback from this study. As a next step, we had planned to test the SAMBA app in an NIH-funded pilot randomized controlled trial (RCT), to assess the feasibility of app use by adolescents and its effect on eating behavior. Unfortunately, the web-based app was hacked and due to the high cost of launching it again, plans were halted. Instead, we plan to proceed stepwise, and to first test the core curriculum (didactic video lessons, audio meditations) for acceptability and feasibility. For this, we plan to use either text messaging or e-mail to deliver links to our didactic videos and audios, which are hosted on a streaming platform, or to use one of the existing robust online platforms that are designed to centralize the delivery of educational content. This approach of utilizing text messaging, e-mail, or an existing platform as opposed to a brand-new web-based app would allow for a much lower cost, shorter timeline, and increased robustness against real-world challenges such as hacking.

With the rising clinical use of glucagon-like peptide receptor (GLP-1s) agonist medications for weight management in adolescents, the practice of mindful eating remains highly relevant (68, 69). Recent literature has highlighted the need for a focus on nutritional interventions to optimize GLP-1 efficacy (70–72). GLP-1s slow down gastric emptying and intestinal motility and thus can cause early fullness, and they are frequently associated with side effects including nausea, vomiting, bloating, and abdominal pain (73). Interventions focused on eating practices, such as mindful eating, may be able to promote medication compliance and decrease side effects by encouraging youth to pay attention to their bodily cues of hunger and satiety. Following these body cues can lead to eating smaller, more frequent meals, at a slower pace, and stopping before feeling full, which may be less likely to be associated with gastrointestinal side effects (70, 71). The practice of mindful eating remains highly important, and may be beneficial, when used in conjunction with obesity medications such as GLP-1s. Based on our clinical experience and the literature, the proportion of adolescents with obesity seen in weight management clinics who are eligible and interested in taking GLP-1s as an adjuvant to intensive health behavior and lifestyle treatment has been rapidly increasing (74). Thus, we plan to test the SAMBA app on adolescents who are preparing to start GLP-1s. Other future directions include adapting our mindful eating curriculum for youth with type 1 diabetes mellitus, as this population may also benefit from mindful eating practices given their increased risk for weight gain (75–77) and food-centric management.

The limitations of this preliminary study include a relatively small sample size of 25 participants from a single urban institution, which does not allow us to draw conclusions about app perceptions by gender or severity of obesity, and which may limit generalizability of findings to the general population. We did not include non-English-speaking participants (as the app and its curriculum are only available in English at this time) or older adolescents (greater than 17 years old), as we wanted to focus on middle adolescence, which is different developmentally and socially from late adolescence (18–21 years) (78).

All participants attended a clinical weight management program, and thus may be more motivated than the general adolescent population, as reflected in their answer to the weight-related stages of change questionnaire. Participants viewed excerpts from a mindful eating app and answered a questionnaire endorsing the behavioral intention to use the app, which is different from actually using the app in their daily lives. A future pilot randomized controlled trial (RCT) will seek to address these limitations by assessing the acceptability and feasibility of the core SAMBA curriculum vs. a control, followed by a larger RCT.

Tailored mindful eating apps should continue to be studied as a potential intervention for adolescents seeking to improve their relationship with food, as an adjunct to multicomponent weight management interventions. Mindful eating apps designed and developed for adolescents, such as SAMBA, should prioritize non-judgmental and relatable features that encourage interactivity with the app.

As technology changes and app use increases among adolescents, prospective studies should continue to assess feasibility of app use within this population. Research measures such as the UTAUT model can help evaluate what influences youth to adopt and use technology, and qualitative analysis can provide depth to those findings. Competing timelines and high costs for app development, unexpected challenges which are becoming more commonplace, such as hacking, and research implementation present challenges to creating and studying health apps including mindful eating apps. Creative solutions are needed to address these challenges to developing and studying digital MBIs for adolescents.