Abstract
Adolescents with autism spectrum disorder are at increased risk of unhealthy eating behaviors and obesity. This study examined feasibility of a virtual implementation of Bringing Adolescent Learners with Autism Nutrition and Culinary Education, an 8-week nutrition intervention based on social cognitive theory that addresses autism spectrum disorder–specific eating behaviors and food environment challenges. The implementation process was measured with fidelity checklists, engagement records, and field notes. Feasibility of virtually assessing adolescents’ outcomes (food frequency questionnaire, psychosocial survey, height, and weight) was also evaluated. Adolescents with autism spectrum disorder aged 12–21 years were recruited through a local community partner. Six groups (n = 27; group size ranged 2–7) participated. Univariate data analyses were performed. Mean lesson attendance was 88%, participation was 3.5/4 (4 = Frequently), homework completion was 51.9%, fidelity was 98.9%, and technical difficulty prevalence was 0.4/2 (2 = Major technical difficulties). Assessment completion rate was 100% (98.9%–100%) at baseline and 92.6%–96.3% (99.5%–100%) at post-intervention. Data quality was high for 88% of food frequency questionnaires and 100% of psychosocial surveys. Findings suggest that a virtual implementation and evaluation of Bringing Adolescent Learners with Autism Nutrition and Culinary Education was feasible. Bringing Adolescent Learners with Autism Nutrition and Culinary Education may be implemented virtually to reach diverse populations of adolescents with autism spectrum disorder. Future research should examine the impact of the intervention on dietary behavior and obesity outcomes.
Lay abstract
Adolescents with autism spectrum disorder are at an increased risk of unhealthy eating behaviors and obesity compared to their typically developing peers. Many nutrition interventions for this population focus on improving autism spectrum disorder symptoms or managing weight rather than addressing participants’ healthy eating self-efficacy. The purpose of this study was to examine a virtual implementation of a new intervention for adolescents with autism spectrum disorder, Bringing Adolescent Learners with Autism Nutrition and Culinary Education. We used fidelity checklists, engagement records, and field notes to measure implementation. We also examined the feasibility of assessing outcome measures, including a food frequency questionnaire (FFQ), psychosocial survey, height, and weight. We recruited adolescents with autism spectrum disorder aged 12–21 years. Six groups of 2–7 adolescents (27 total) participated in the intervention and pre-/post-intervention measurements. Bringing Adolescent Learners with Autism Nutrition and Culinary Education consisted of eight weekly lessons: exploring taste, flavor, and texture; mealtimes and rules; food groups and nutrients; moderation; beverages; cooking; well-being; sustaining healthy eating habits. The virtual implementation was feasible based on lesson attendance, participation, homework completion, fidelity, and prevalence of technical difficulties. Evaluation was also feasible based on response rate, completion, and data quality for the food frequency questionnaire, psychosocial survey, and height and weight measurements. Bringing Adolescent Learners with Autism Nutrition and Culinary Education may be used in virtual settings to reach diverse populations of adolescents with autism spectrum disorder. Future research is needed to evaluate the impact of Bringing Adolescent Learners with Autism Nutrition and Culinary Education on dietary behavior and obesity outcomes.
Autism spectrum disorder (ASD), characterized by deficits in social communication and the presence of restricted, repetitive behaviors (American Psychiatric Association, 2013), is a pressing public health concern as one of the fastest growing developmental disabilities, with an estimated prevalence of 18.5 per 1000 (1 in 54; Maenner et al., 2020). Adolescents with ASD are at an increased risk of being overweight or developing obesity compared to typically developing adolescents (Must et al., 2017). In fact, by 2–5 years of age, children with ASD already have an increased prevalence of overweight and obesity (Hill, Zuckerman, & Fombonne, 2015), and there is evidence that odds of obesity increase with age in adolescents with ASD aged 10–17 years (Must et al., 2017). A recent meta-analysis found that children with ASD have 41.1% greater risk of developing obesity, with age as a positive moderator (Kahathuduwa et al., 2019). Obesity is associated with an increased risk of several poor health outcomes, including type 2 diabetes (Goran, Ball, & Cruz, 2003), hypertension (Friedemann et al., 2012), social marginalization (Strauss & Pollack, 2003), and family economic burden (Wang & Dietz, 2002) in typically developing children and adolescents. In youth with ASD, obesity and obesity-related complications pose a threat to independent living, self-care, and quality of life (Curtin, Jojic, & Bandini, 2014).
As children and adolescents with ASD exhibit an increased prevalence of problematic eating behaviors, such as food selectivity, or consuming a narrow range of foods (Marí-Bauset, Zazpe, Mari-Sanchis, Llopis-González, & Morales-Suárez-Varela, 2014), and consuming more energy-dense foods and fewer fruits and vegetables than typically developing children (Sharp et al., 2013), nutrition represents a critical modifiable risk factor for unhealthy weight gain in this population (Dhaliwal, Orsso, Richard, Haqq, & Zwaigenbaum, 2019). Existing nutrition interventions for children with ASD tend to focus on alleviating symptoms of ASD (Sathe, Andrews, McPheeters, & Warren, 2017) or targeting weight loss with inconsistent study quality and effectiveness (Healy, Pacanowski, & Williams, 2019) without addressing participants’ healthy eating self-efficacy. Furthermore, many nutrition interventions for youth with ASD aim to improve severe feeding difficulties, specifically for those who are diagnosed with a feeding disorder or a very narrow food repertoire (Sharp, Burrell, & Jaquess, 2014; Tanner & Andreone, 2015). In addition, interventions with samples that include adolescents with ASD often include adolescents with diverse disabilities rather than ASD-only samples, and few have addressed healthy eating habits and psychosocial factors such as self-efficacy. To promote healthy eating habits and prevent obesity as well as other diet-related health conditions such as type 2 diabetes or hypertension among adolescents with ASD, nutrition interventions should address both autism-specific feeding issues and long-term eating habits.
It is established that nutrition interventions are more likely to be effective in developing long-term healthy eating habits if they are theory-driven and behaviorally focused (Contento, 2012; Contento et al., 1995). Current nutrition interventions for adolescents with ASD do not examine psychosocial determinants of dietary intake, such as self-efficacy, behavioral skills, and social support. Social cognitive theory (SCT; Bandura, 1989) is commonly used to target psychosocial factors related to healthy eating in individuals without ASD (Vilaro, Staub, Xu, & Mathews, 2016), yet there is a lack of similar interventions for individuals with ASD.
To address these unmet needs in nutrition interventions for adolescents with ASD, a theory-driven nutrition intervention that focuses on ASD-specific eating challenges and healthy eating behaviors was created. The BALANCE (Bringing Adolescent Learners with Autism Nutrition and Culinary Education) intervention was developed with input from adolescents with ASD and their parents and piloted in a school-based setting in 2019 (Buro & Gray, 2020), which demonstrated good feasibility and high acceptability among adolescents with ASD. Due to the coronavirus disease of 2019 (COVID-19) pandemic, the intervention was adapted for implementation in a virtual setting. The purpose of this study was to examine the feasibility of a virtual implementation of the BALANCE program, an 8-week, SCT-based nutrition intervention for adolescents with ASD. The aims of the study were to: (1) assess feasibility of a virtual version of the BALANCE intervention based on fidelity checklists and engagement records and (2) examine feasibility of virtually administering instruments to assess outcome measures, including dietary intake and its psychosocial determinants, physical activity, and anthropometric measures.
Methods
Study design
Quantitative methods were used to measure feasibility of virtually implementing the intervention and assessing dietary intake, psychosocial determinants of dietary intake, physical activity, and anthropometric measures. Field notes were used to triangulate findings from quantitative data. To assess psychosocial determinants of dietary intake, a survey with measures developed and evaluated by Dewar, Lubans, Plotnikoff, & Morgan (2012) was administered online to BALANCE participants pre- and post-intervention. The Block Kids 2004 Food Frequency Questionnaire (FFQ) and Physical Activity Screener (PAS; Cullen, Watson, & Zakeri, 2008) were administered online to participants pre- and post-intervention to measure dietary intake, physical activity, and screen time. One parent of each adolescent was recruited to participate in virtual height and weight assessment appointments and fill out a demographic questionnaire and the Autism Behavior Inventory—Short Form (ABI-S; Bangerter et al., 2017). The ABI-S has the following scales: language level (one question scored 0–4), social communication quality (three questions scored 0–3), social communication frequency (three questions scored 0–3), restrictive behavior frequency (seven questions scored 0–3), mood and anxiety frequency (five questions scored 0–3), self-regulation frequency (three questions scored 0–3), and challenging behavior frequency (three questions scored 0–3). Responses were coded so that higher scores represent having more ASD-related features and 0 indicated an absence of ASD-related behaviors or challenges.
The 8-week curriculum was implemented via Microsoft Teams. Microsoft Teams was selected as the virtual platform because it was officially supported by the University of South Florida. A virtual setting was appropriate given the risk of contracting or transmitting COVID-19 in group gatherings during the timeframe for data collection (August–December 2020; Centers for Disease Control and Prevention, 2020).
Sample and recruitment
Adolescents with ASD aged 12–21 years and their parents were recruited with a target sample size of 30 adolescent–parent dyads. Participants were recruited through partnership with the Center for Autism and Related Disabilities at the University of South Florida (CARD-USF). The recruitment flyer was emailed through the CARD-USF listserv, posted on the CARD-USF Facebook page, and shared with other CARD centers throughout Florida. The study was approved by the University of South Florida Institutional Review Board in July 2020. Eligible adolescents were those clinically diagnosed with ASD and aged 12–21 years. Exclusion criteria included concurrent participation in another nutrition-related intervention, being non-English speaking, having below third grade reading level per parent report, or having eating disorder or feeding disorder diagnosis per parent report, as a severe eating or feeding disorder should be treated with an extensive feeding therapy and may require medical attention. Parents of adolescents participating in the intervention were eligible to participate. An exclusion criterion for parents was being non-English speaking. Informed consent/assent was obtained from all participants.
Intervention components
The conceptual framework for BALANCE incorporates SCT constructs and ASD-specific challenges, including sensory differences (Hazen, Stornelli, O’Rourke, Koesterer, & McDougle, 2014) and cognitive rigidity during mealtimes (Polfuss et al., 2016) to elicit dietary behavior change (Figure 1). The telehealth version of BALANCE consists of eight 45-min lessons to be delivered via Microsoft Teams once per week for 8 weeks. A lesson manual was created to guide the intervention, including aims, objectives, overview, preparation, procedure, and a teacher’s note for each lesson. A lesson booklet was created for participants with an overview, preparation instructions, handouts, and take-home activity for each lesson. Lesson activities were aligned with SCT constructs, as summarized in Table 1. Each lesson included a tasting session or an optional snack. The food suggestions were flexible so that participants could use food that was readily available in the home. Lessons 1–7 had brief homework assignments that were to be completed and returned the following week. If participants were unable to attend any of the lessons, a 15-min make-up video was sent to them to review. The make-up videos followed the same content and format as designed but did not include any interaction from participants.
Conceptual framework for the BALANCE intervention.
Application of social cognitive theory constructs to lesson activities.
USDA: US Department of Agriculture.
Lesson outlines were adapted from an early childhood nutrition intervention, Autism Eats, which was created by the research team (Van Arsdale, Gray, & Buro, 2020). Lesson content and activities were further modified based on participant feedback from a pilot study and discussion among the research team to tailor the intervention to the specific needs of adolescents with ASD. Lesson content was also designed based on evidence-based strategies and findings from formative research. The curriculum incorporates data-driven strategies for adults with ASD, such as social engagement, emphasis on the individual, sensory/motor enhancement, emphasis on choice (Goldschmidt & Song, 2017), and visual supports (Kluth & Darmody-Latham, 2003). Primary formative research for the study, including focus groups of adolescents with ASD and interviews with parents of adolescents with ASD, also indicated that social engagement, visual components, and teen-led initiatives should be incorporated in the intervention. Ideas for theory-based activities came from previous research (Perry et al., 1997), and one activity (in Lesson 4) was adapted from the Food Day Curriculum developed by the Laurie M. Tisch Center for Food, Education & Policy (Koch & Contento, 2011).
Parents were asked to participate in webinars via Microsoft Teams at baseline, after Lesson 4, and after Lesson 8. The webinars covered material from the lessons and showed parents how to provide social support and opportunities for their children to maintain healthy eating habits. In addition, a handout summarizing the lesson’s content and purpose was emailed to parents after each lesson. If parents were unable to attend any of the three parent webinars, then webinar slides and notes were provided to parents via email.
Planning and evaluation
The RE-AIM (Reach, Effectiveness, Adoption, Implementation, Maintenance) framework (Glasgow, Vogt, & Boles, 1999) was used to guide planning and evaluation for the study. Table 2, adapted from the RE-AIM Checklist for Inclusion of RE-AIM Issues by RE-AIM Dimension (RE-AIM, 2021), summarizes the components of the framework and demonstrates how each RE-AIM dimension was applied. The Maintenance dimension of the framework was not applicable to this stage of the research.
Application of RE-AIM.
RE-AIM: reach, effectiveness, adoption, implementation, maintenance; BMI: body mass index; N/A: not available.
Measures
Implementation measures
Fidelity and engagement records were completed by research assistants based on review of video-recordings for each of the eight lessons. Fidelity was monitored by a checklist for each lesson. Each checklist included 9–11 lesson-specific components and checkboxes for completion and modification, as well as room for notes if components were marked incomplete or modified (e.g. not enough time, instructor skipped it, participants did not bring food). Components were marked as modified if they were completed in a way that was modified from the lesson manual (e.g. none of the students brought recipe ingredients, so the instructor completed a demonstration and discussion instead of leading the students to make the recipe).
Engagement measures included attendance at the start and end of each lesson; minimum, maximum, and average minutes attended per student; verbal and nonverbal participation (frequently, occasionally, rarely, never); proportion of students who actively participated (all students, most students, few/some students, none); technical difficulties (major difficulties, minor difficulties, none); and number of students who completed the homework. Major technical difficulties were defined as those that interfered with the instructor’s ability to complete the lesson (e.g. instructor is disconnected, or students are unable to see the instructor). Minor technical difficulties were defined as those that did not interfere with the instructor’s ability to complete the lesson but may affect the lesson quality (e.g. student audio or video stops working). The engagement measures were the same for all eight lessons. Scales for engagement records were informed by a process evaluation study of a middle school nutrition curriculum intervention (Lee, Contento, & Koch, 2013).
Field notes were taken during and after each session based on a guide by Phillippi and Lauderdale (2018). Field notes included contextual information about participants, virtual setting, and overall process, as well as reflexive description of the researcher’s positionality, values, experiences, and relationships with the participants (Dodgson, 2019).
Outcome measures
Outcome data were collected at two time points: pre-intervention (baseline) and post-intervention (9 weeks from baseline). At both time points, a survey was administered to examine adolescents’ psychosocial determinants of dietary intake (Dewar et al., 2012); the Block Kids FFQ (Cullen et al., 2008) was administered to measure dietary intake; the Block Kids PAS was administered to adolescent participants to measure physical activity and screen time (Drahovzal, Bennett, Campagne, Vallis, & Block, 2003); height and weight of adolescents was measured via ruler and scale; and the ABI-S (Bangerter et al., 2017) was administered to parents to measure ASD symptoms and behaviors. Prior to this study, the FFQ and psychosocial survey were pilot tested in a sample of 12 adolescents with ASD aged 8–19 years in a school-based setting, and the results indicated that the instruments were feasible for adolescents with ASD who had teacher-reported high social communication skills and were 15 years or older. A full-size ruler with adhesive and a bathroom scale were delivered to each participant, and parents completed height and weight measurements for their children at baseline and post-intervention, as guided and observed by research staff via Microsoft Teams. Adolescents were asked to complete the psychosocial survey and FFQ + PAS. Parents were told that they could assist or complete surveys and questionnaires on behalf of the adolescents if assistance was required.
Data analysis
Univariate procedures including frequency distributions and descriptive statistics were performed for feasibility measures, including attendance, participation, homework completion, fidelity, and technical difficulties for the intervention lessons and response rate, completion, and quality for the FFQ + PAS and psychosocial survey. Fidelity checklists were used to calculate percentage fidelity for each lesson, and engagement records were used to calculate attendance, participation, homework completion, and technical difficulties. Wilcoxon signed-ranked tests were conducted to explore whether there were any changes in ASD symptoms from pre- to post-intervention.
Prior to analysis, quantitative data were reviewed, and unreliable records were flagged through a three-stage process of screening (e.g. detecting outliers or inconsistencies), diagnosing (e.g. errors, missing data), and editing (i.e. correction, deletion, or leaving unchanged; Broeck et al., 2005). Surveys were analyzed for response patterns, such as straightlining (choosing the same option for every item), diagonal lines, or a combination of both (Leiner, 2019). All survey data were also screened for inconsistent or unrealistic answers. Missing data were handled with pairwise deletion. No missing data analysis was performed because the amount of missing data was so low (4% of administered surveys and 0.4% of completed surveys) that it was assumed to be random rather than systematic. No data were missing from the completed FFQs due to the NutritionQuest forced-choice format. FFQ data were excluded if total energy intake was less than 500 kcal per day or greater than 5000 kcal per day based on previously defined cutoffs for outliers or implausible responses in children and adolescents (Rockett et al., 1997).
Community involvement statement
Adolescents with ASD and their parents and teachers were involved in the design of the BALANCE intervention. Input from adolescents with ASD and their parents was sought before the intervention was developed, and a school-based pilot study was conducted with a rapid-cycle evaluation approach (Shrank, 2013) to further refine the intervention based on participant feedback.
Results
Figure 2 depicts the overall flow of the study. A total of 34 parents expressed interest in the study, and 31 completed the eligibility screening and informed consent. All participants who completed the eligibility screening for the study were deemed eligible. Two participants did not respond to follow up after completing eligibility screening and one or more baseline measures. Two other participants dropped out of the intervention after Lesson 1. Both parents reported that their child’s challenging behaviors during the lesson contributed to their decision to drop out. One of the parents also reported that work- and school-related stress was a contributing factor. Results are presented for the 27 adolescents who completed the 8-week intervention.
Overall flow of the study.
Participant characteristics
Of those who completed the intervention, 74% were male, 26% were female, and the average age was 15 years (range 12–20 years). The race/ethnicity breakdown of participants was 63% White, 15% Hispanic, 7% Black or African American, 4% Asian, and 11% Other. Participants who selected “Other” for the race/ethnicity option identified as “Asian and White” (7%) and “Latino and White” (4%). Nearly half of participants (48%) came from households with reported income of US$75,000 or greater. There were two participants (7%) with a reported household income of less than US$20,000. Three participants (11%) responded “Strongly agree” or “Somewhat agree” to the food insecurity question.
The majority of participants were either homeschooled (44%) or attended public school (26%), with others attending private school (11%), or other school (15%). One participant had graduated from high school and was not attending any form of school at the time of study enrollment (4%). Description for “Other” school responses included virtual school (7%) and being in the process of transitioning from one type of school to another (7%; one transitioning from public to private and one transitioning from private virtual school to homeschool). Table 3 shows all demographic characteristics for adolescents who participated in this study and their families as reported by parents.
Demographic characteristics of study participants.
GED: general educational development.
Results represent mean and standard deviation.
Responses included: anxiety, auditory processing disorder, learning disabilities (dysgraphia, dyslexia, and non-verbal learning disability), cerebral palsy, hydrocephalus, fetal alcohol spectrum disorder, executive function disorder, epilepsy, periventricular leukomalacia, microcephaly, sleep apnea, progressive infantile idiopathic scoliosis cardiac, premature ventricular contractions, migraines, thyroid issues, apraxia, and failure to thrive.
Symptoms of ASD
Participants’ social communication scores ranged from 0 to 2 with a mean of 0.7 for quality and 1.2 for frequency (Table 4). According to Bangerter et al. (2020), social communication ABI-S scores for typically developing children were close to zero (0–0.2), indicating almost no problematic social communication challenges, while the scores for individuals with ASD were between 1.0 and 1.8, indicating significantly more problematic social communication challenges. Using these values as references, participants in this study had relatively high-quality social communication skills and low problematic social communication challenges compared to other individuals with ASD (Bangerter et al., 2020). There were no differences in pre- and post-intervention mean scores for any of the ASD symptom domains based on the ABI-S.
Pre- and post-intervention means for ASD symptoms.
ASD: autism spectrum disorder; SD: standard deviation; N/A: not available.
Response options: no language, signs, single words or 2–3-word utterances, simple sentences, full sentences.
Response options: not at all, with support, with some reminders, without help.
Response options: never, sometimes, often, very often.
Implementation measures
Table 5 summarizes the results for implementation of the intervention, including attendance, participation, homework, fidelity, and technical difficulties. There were six groups of adolescents who participated in the intervention. Group size ranged from two to seven participants. Four groups met on weekday afternoons or evenings (5:00 p.m. or 6:30 p.m.), one group met on weekday mornings (10:00 a.m.), and one group met on weekend afternoons (12:00 p.m.). Group meeting time and group size were determined based on the number of interested participants who were available at the same day and time of the week. Results for implementation are presented as group means. All lessons took place on their scheduled day/time by the scheduled instructor. Lessons lasted 30–45 min, with smaller groups (two to three participants) consistently having shorter lessons.
Intervention implementation: attendance, participation, homework, fidelity, and technical difficulties.
N/A: not available.
Participation consisted of: verbal participation and nonverbal participation (Response options: never, rarely, occasionally, frequently) and proportion of students who actively participated (none, few/some, most, all).
0 indicates no technical difficulties, 1 indicates minor technical difficulties, and 2 indicates major technical difficulties.
Mean lesson attendance was 88% and ranged 50%–100%. Participation was calculated from verbal participation (never, rarely, occasionally, frequently), nonverbal participation (never, rarely, occasionally, frequently), and proportion of students who actively participated (none, few/some, most, all). Mean participation was 3.5 of 4 (4 being frequent verbal or nonverbal participation or all students actively participating) and ranged 2–4 (2 being rare verbal or nonverbal participation or few/some students actively participating). Mean homework completion was 51.9% and ranged 0%–100%. Mean lesson fidelity was 98.9% with a range of 88.9%–100%. Mean prevalence of technical difficulties was 0.4 of 2 (2 indicating major technical difficulties) with a range of 0–1, indicating no technical difficulties or minor difficulties for all lessons. Mean parent webinar attendance decreased from 73% in Webinar 1 to 37% in Webinar 3, with attendance ranging 20%–91%.
Table 6 summarizes the mean, minimum, and maximum number of BALANCE lessons attended per student for each of the six groups. The total mean was 7.1 of 8 lessons. The minimum number of lessons attended was four, and the maximum was eight.
BALANCE lessons attended per student.
BALANCE: Bringing Adolescent Learners with Autism Nutrition and Culinary Education; N/A: not available.
Field notes
Emergent themes from field notes included engagement, modifications, prompts, distractions, and technical difficulties.
Engagement
Many adolescents were actively engaged and attentive throughout the lessons. Most adolescents followed each lesson’s preparation instructions and had food to share in front of the camera when instructed to do so. Occasionally, adolescents forgot to prepare, or, in Lesson 6, many adolescents did not have the ingredients for the guacamole-making activity.
Modifications
Modifications were made in four lessons overall. For three groups, there were no students who brought ingredients to make guacamole in Lesson 6, so the activity was modified to a demonstration by the instructor instead of a hands-on activity. For one group, the sharing snack activity in Lesson 3 was modified to the instructor showing and talking about snacks, as no participants brought a snack to share.
Prompts
Prompts successfully encouraged participation in all lessons. Sometimes adolescents only participated when supplied with visual or verbal prompts (e.g. instructor showing or reading the booklet) or when they were directly asked a question (e.g. “[Participant name], what do you think?”).
Distractions
Some adolescents were distracted by cell phones or other devices during lessons. Sometimes there was background noise that distracted participants. The participant with background noise was muted to alleviate the disturbance. Some participants had more verbal and nonverbal participation when there was no background noise or distraction.
Technical difficulties
Technical difficulties included connection issues causing lag or a frozen screen and audio or video not working. Two participants regularly had difficulty logging into Microsoft Teams; both mentioned that they were using Chromebooks to participate in the lessons. Participants who mentioned that they used desktop computers, laptops, or iPads did not report regular difficulties logging in.
Feasibility of outcome measures
The Block Kids PAS was included at the end of the FFQ. Of the 27 participants who completed the 8-week intervention, 27 (100%) completed the FFQ + PAS at baseline, and 25 (92.6%) completed the FFQ + PAS at post-intervention. Completion rate was 100% for those who filled out the FFQ + PAS. Parents were told that they could assist their children in completing the FFQ + PAS if clarification or other assistance was needed. Eight parents reported that they helped their children clarify questions or recall food items consumed (e.g. “I helped him remember milk and bread”). Data quality was high for 88% of the matched FFQs and 84% of the matched PASs. Two participants’ responses were excluded due to reported energy intake less than 500 kcals per day: one at baseline and post-intervention, and one at post-intervention only. Another participant’s responses were excluded due to a straightlining response pattern at post-intervention. Energy intake ranged 875–3121 kcals at baseline and 731–2469 kcals at post-intervention.
Of those who completed the intervention, 27 (100%) completed the psychosocial survey at baseline, and 26 (96.3%) completed the survey at post-intervention. The completion rate at baseline was 98.9% (ranged 86%–100%), and the completion rate at post-intervention was 99.5% (ranged 97%–100%). Data quality was high for 100% of the psychosocial surveys. None of the surveys had inconsistencies or unrealistic responses.
Height and weight measures were taken for all 27 participants (100%) at baseline and 26 participants (96.3%) at post-intervention.
Discussion
This study aimed to examine feasibility of a virtual implementation of the BALANCE intervention based on fidelity checklists and engagement records, as well as feasibility of virtually administering instruments to assess outcome measures, including dietary intake, psychosocial determinants of dietary intake, and anthropometric measures. The virtual implementation of BALANCE was determined to be feasible, with 88% attendance, high participation (3.5 out of 4), 52% homework completion, 99% fidelity, and no major technical difficulties. Of the 29 participants who completed Lesson 1 of the intervention, 27 (93%) completed all eight lessons. Adolescents participated verbally and nonverbally, and field notes indicated that verbal and visual prompts successfully increased participant engagement. However, field notes also indicated that some adolescents were distracted by other devices during the lessons. Future implementations should enforce rules about no devices via communication with both parents and adolescents to maximize participation and intervention effectiveness. Most absences on the fidelity checklists were due to children not having food for the guacamole-making activity in Lesson 6, suggesting that fidelity could be improved by making the food available for students via delivery or pickup or through more effective parent reminders. Two of the three participants who responded “Strongly agree” or “Somewhat agree” to the food insecurity question did not have guacamole for the guacamole-making activity, but these participants had food available for other lesson activities. As cost may be a barrier for the guacamole-making activity, the ingredients should be made available to students, or the lesson may be modified to include cheaper ingredients. There is a lack of virtual nutrition interventions for youth with ASD to compare findings on implementation. In-person nutrition interventions for youth with ASD report high fidelity, ranging 94%–100% (Cassey, Washio, & Hantula, 2016; Cosbey & Muldoon, 2017; Marshall, Hill, Ware, Ziviani, & Dodrill, 2015). Parent webinar attendance ranged 20%–91% and decreased from the first to last webinar, suggesting that an alternate format for the parent video component, such as short, asynchronous videos, may be more engaging and convenient for parents.
Homework completion rates were not as high as other implementation variables. As parents of children with ASD may be more likely than parents of children without ASD to help their children with homework (Zablotsky, Boswell, & Smith, 2012), the homework assignment may contribute to parent burden, particularly among younger adolescents. In contrast, a weight management intervention for younger children with ASD (aged 5–12 years) reported high parent adherence, including homework completion (Burrell et al., 2020). Since parents are not the participants of the BALANCE intervention, the homework component may be simplified, reduced, or eliminated for younger adolescents.
Response rate, completion, and data quality were high for the FFQ + PAS, psychosocial survey, and height and weight measurements. Baseline response rate was 100% with 98.9%–100% completion, and post-intervention response rate was 92.6%–96.3% with 99.5%–100% completion. These findings were similar to virtual obesity prevention interventions for typically developing youth. For example, 93% of participants completed baseline and follow-up measures for a web-based obesity prevention intervention for typically developing adolescents aged 12–15 years (Chen, Weiss, Heyman, Cooper, & Lustig, 2011). Data quality was high for 88% of matched FFQs, 84% of matched PASs, and 100% of the psychosocial surveys. FFQ and PAS data quality may be improved through research staff assisting adolescents in completion. Reasons for exclusion of the FFQs—daily energy intake less than 500 kcals and a straightlining response pattern—may suggest survey fatigue or lack of interest in completing the survey. It was also noted that some participants experienced technical difficulties using the NutritionQuest online system because their devices did not support Adobe Flash.
The high response rate, completion, and data quality for the psychosocial survey and the 100% response rate for height and weight measurements indicate that virtually implementing these measures is feasible for adolescents with ASD. Previous research has used electronic scales to send weight data to research or clinical centers, but the financial cost of research-grade scales ranges US$80–US$130 (Krukowski & Ross, 2020). The findings of this study suggest that conducting virtual height and weight measurements as instructed by research staff (e.g. through Microsoft Teams) may be a feasible and reliable option with low cost and response burden. Limitations of the virtual data collection include the potential for random error, as a parent of each adolescent conducted height and weight measurements. Furthermore, there may be challenges with research staff viewing the height and weight measurement process depending on the device used (i.e. desktop computer, laptop computer, or tablet). Due to the potential limitations and challenges, there is a need to validate this virtual method against in-person height and weight data collection.
The use of a novel, theory-based nutrition intervention developed specifically for adolescents with ASD was a strength of the study. The BALANCE intervention was developed based on formative research with adolescents with ASD and their parents, as well as evidence-based strategies for individuals with ASD (Goldschmidt & Song, 2017; Kluth & Darmody-Latham, 2003), theory-based activities (Perry et al., 1997), and nutrition education activities for children (Koch & Contento, 2011). The BALANCE intervention was designed and adapted based on 2 years of preliminary research, aided by perspectives and feedback from adolescents with ASD and their parents and teachers. Application of health behavior theory has been reported as a contributing factor to successful online nutrition education interventions (Ajie & Chapman-Novakofski, 2014; Murimi, Nguyen, Moyeda-Carabaza, Lee, & Park, 2019). The use of SCT to guide the intervention contributed to high transferability, and the use of the RE-AIM framework allowed for a multidimensional evaluation of the intervention implementation to guide future implementations of the BALANCE intervention.
Study limitations include that due to the small sample size and the fact that the study only reached adolescents with ASD who have high social communication skills as indicated by the ABI-S, the findings of this study cannot be generalized to all adolescents with ASD. However, the social communication scores of participants in this study were still consistent with ASD social communication symptoms rather than those of typically developing youth (Bangerter et al., 2020). In addition, due to the methods of data collection for the study, there is potential for self-report bias, recall bias, and social desirability bias. The FFQ + PAS asks participants to recall behaviors in the past week, and the psychosocial survey has questions about the past 3 months. Although the FFQ and psychosocial survey were pilot tested in a sample of adolescents with ASD as part of the formative research for this study, validity and reliability of the FFQ + PAS and psychosocial survey should be further examined in individuals with ASD. Despite these shortcomings, adolescents’ high response rates and data quality indicate that the outcomes from this study can be used to estimate sample sizes and statistical power for future studies.
Conclusion
This research examining the feasibility of a virtual group nutrition intervention for adolescents with ASD suggests that a small group virtual setting may be appropriate for many adolescents with ASD. Of the 29 adolescents who participated in Lesson 1, 27 adolescents completed the 8-week intervention. Many adolescents were engaged and attentive throughout the lessons, and visual and verbal prompts were effective at encouraging participation. There were no major technical difficulties, but minor technical difficulties were likely inevitable due to variations in Internet connection speeds and the number of participants in each Microsoft Teams meeting for the lessons. This study included a broad age range, with adolescent participants aged 12–20 years. While the live implementation allowed for individualized feedback, further iterations should tailor the intervention activities for specific age groups, for example, including a stronger focus on food preparation for adolescents aged 15 years and older. Further research is also warranted to examine the impact of BALANCE on dietary behavior and obesity outcomes.
Footnotes
Acknowledgements
We would like to thank all adolescents and parents who participated in this research.
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was funded by the University of South Florida College of Public Health Internal Grant and the University of South Florida College of Public Health Student Research Scholarship.
