Effectiveness of a Pediatric Weight Management Intervention for Rural Youth (iAmHealthy): Primary Outcomes of a Cluster Randomized Control Trial

Abstract

Background:

Youth in rural areas are disproportionally affected by obesity. Given the unique barriers rural populations face, tailoring and increasing access to obesity interventions is necessary.

Objective:

This paper evaluates the effectiveness of iAmHealthy, a family-based pediatric obesity intervention delivered to rural children, compared to a Newsletter Control.

Methods:

Participating schools (n=18) were randomly assigned to iAmHealthy or Newsletter Control. iAmHealthy consists of individual health coaching and group sessions delivered via televideo to a participant’s home. Child and parent body mass index (BMI), child physical activity, and child dietary intake were assessed at baseline, post-treatment (8 months) and follow-up (20 months). Multilevel modeling estimated the effect of treatment at both timepoints.

Results:

Parent and child dyads were recruited (n=148) and randomized to iAmHealthy (n=64) or the Control group (n=84). The Control group had significant increases in child BMIz from baseline to follow-up. iAmHealthy youth had no significant changes in BMIz from baseline to post or follow-up. Child dietary intake, physical activity and parent BMI results are also discussed.

Conclusions:

This trial extends previous pediatric obesity work by simultaneously increasing convenience and dose of treatment. Results suggest iAmHealthy resulted in a change in BMIz trajectories and long-term health behavior for youth.

1. Introduction

Pediatric obesity in the United States is a complex and widespread public health issue with current estimates indicating approximately 31.8% of youth have overweight or obesity (body mass index [BMI]≥ 85^th percentile).¹ Pediatric obesity poses significant health consequences such as higher incidence of cardiometabolic risk factors, physical comorbidities including obstructive sleep apnea, and psychological comorbidities (e.g., anxiety and depression).^2,3 Youth in rural areas have 26% greater odds of having obesity compared to youth in urban areas.⁴ These increased rates are likely caused by multiple factors including limited access to recreational facilities^5,6 and healthy food at grocery stores and restaurants,^7,8 low sidewalk availability,⁹ difficulties finding recreational alternatives to screen time⁸ which results in youth engaging in fewer health promoting behaviors.^10–12 Additionally, rural primary care providers face extensive barriers in relation to implementing treatment and preventative interventions for childhood obesity (e.g., limited specialists in community, time constraints, few community resources to refer families to) which results in families from rural communities having limited access to programs or medical advice that may support weight management.¹³

Given these unique barriers and the limited access to care that rural populations face, tailoring and increasing accessibility to obesity interventions is necessary. Historically, few interventions have been designed to target this population. In recent years, attempts have been made to reduce barriers to care by providing obesity services via televideo with results suggesting that this may be an effective way to deliver care to rural populations.^14–16 The majority of these prior interventions have depended on families gathering at the child’s school^15,16 or in pediatric offices to use televideo services.¹⁴ Given that the United States Preventative Services Task Force recommends that comprehensive behavioral obesity treatment consist of 26 hours of contact with the family,¹⁷ asking families to commute to appointments so frequently may impact recruitment and attendance given the barriers rural families face (e.g., geographic dispersion). Moving televideo treatment into rural family homes may allow for interventions to increase dose and convenience simultaneously.

The purpose of this paper is to evaluate the effectiveness of a family-based pediatric obesity intervention delivered to rural families via televideo to a participant’s home (iAmHealthy) compared to a newsletter control group (Control). Specifically, the iAmHealthy intervention was hypothesized to result in significant decreases in child BMIz and parent BMI compared to the Control intervention. We also hypothesized that child participants enrolled in the iAmHealthy intervention would make significantly better food choices (i.e., servings of sugar-sweetened beverages, “red” foods, and fruits and vegetables) and spend significantly greater time in moderate to vigorous intensity physical activity (MVPA) than the Control intervention.

2. Methods

All methods were approved by the Human Subjects Committee at the University of Kansas Medical Center. The methods for this study (ClinicalTrials.gov NCT03304249) have been previously published in detail.¹⁸

2.1. Recruitment

2.1.1. School Recruitment.

School recruitment included sending paper flyers to 395 rural public elementary schools in Kansas. Schools were eligible if they were in cities or counties with a population of <20,000 individuals and served students in 2^nd-4^th grade. Interested schools contacted study personnel and identified an onsite school representative (e.g., school nurse, physical education teacher) who underwent training on research ethics and study procedures (e.g., recruitment, study measurement).

2.1.2. Participant Recruitment.

Once a school was fully enrolled, school personnel utilized phone calls, emails, in-person conversations, and/or flyers to recruit families. Schools were asked to recruit 8 parent/child pairs but were included in the study if they recruited at least 5 parent/child pairs. Inclusion criteria included: family living in a rural area, child in 2nd-4th grade whose BMI percentile is ≥85th percentile, and both the child and caregiver speak English. Exclusion criteria included: children with a known medical issue or physical limitation or injury which significantly limited physical mobility, children and parents with significant developmental delay/cognitive impairment, and children who have a sibling already enrolled in the program. Figure 1 (CONSORT Diagram) provides detailed information on reasons youth were excluded. Prior to participation, parents and children provided informed consent and assent.

Figure 1.

CONSORT Diagram.

2.1.3. Randomization & Study Timeline.

This study utilized a cluster randomized design, with schools randomly assigned to either iAmHealthy or Control and parent/child pairs at each school assigned to the same group. The study statistician (KF) assigned schools to conditions by utilizing a random number generator (M = 0, SD =1). Recruited schools were stratified by school size and percent free/reduced lunch. Within each strata, the schools were sorted according to the random number (higher values were assigned to iAmHealthy). Random numbers were generated after participants consented, and the random number sequence and assignment was concealed until assignment occurred. Nine schools were randomized to the iAmHealthy group, and nine schools were randomized to the Control group. The 18 schools began their participation from December of 2017 to March of 2020. Post measurement occurred at the end of the intervention (8 months after baseline) and follow-up measurement occurred one year after the intervention was completed (20 months after baseline). Of the 18 participating schools, seven schools received at least part of the intervention during the COVID-19 pandemic, and for six additional schools, the COVID-19 pandemic occurred after the 8-month intervention but prior to their follow-up measurement.

2.2. Intervention Groups

2.2.1. iAmHealthy Intervention Group.

iAmHealthy is a family-based obesity intervention delivered fully via televideo consisting of group and individual sessions. The content was based on Cognitive Behavioral Theory,¹⁹ Davison and Birch’s Child Weight Theory,²⁰ Bandura’s Social Cognitive Theory,²¹ and tailored based on our previous work with families from rural communities.²² This tailoring included providing specific free home-based physical activity recommendations due to the lack of recreational activities^5,6 and sidewalk availability in rural communities,⁹ problem solving barriers to healthy food access, an increased focus on finding alternatives to engaging in screen time, additional time focused on promoting self-esteem due to previous findings that parents from rural areas are hesitant to address obesity due to concerns about child self-esteem,²² and an increased focus on meal planning and budgeting given the limited number of restaurants²³ and lower socioeconomic statuses of residents in rural communities.²⁴ The group sessions focused on behavioral, nutrition, and physical activity topics at the individual, family, and school/community level, allowing families to discuss barriers specific to their rural communities (see summary of topics in Davis et al., 2019).¹⁸ As a family-based program, strategies and encouragement to improve health behaviors were provided to the entire family. This included discussing the importance of parents modeling healthy habits and providing tailored nutrition and physical activity recommendations to both adults and children.

The intervention included eight weekly group sessions followed by six monthly group sessions (with all other participating families from their local elementary school), as well as 16 individual health coaching sessions lasting 30–45 min each with members of their household every other week. This resulted in 14 hours of group intervention and 11 hours of individual sessions planned over the eight-month intervention period. All group sessions were led by a Ph.D. level psychologist or doctoral trainee with experience in health behavior change intervention who followed the iAmHealthy treatment manual and received training and supervision from the PI (AD). Registered dietitians or trainees led the individual health coaching sessions which focused on goal setting, problem solving, reinforcement for tracking, and completion of homework assignments. Sessions were delivered via teleconferencing software over a provided iPad tablet (Apple, Inc.) with a data connection provided (Verizon, Inc.) at times that were convenient for participating families (evenings, weekends). Families participated from their homes, and both the target child and “primary” parent (as indicated by the family) were expected to attend each session, but the entire family was welcome. Make-up sessions were offered when needed and text message reminders were sent to promote attendance.

2.2.2. Newsletter Control Group.

Consistent with prior pediatric obesity intervention trials,^25,26 a newsletter control group was utilized in which families received a monthly newsletter with similar content to that delivered in the iAmHealthy groups. The Control group received eight newsletters consisting of fitness and nutrition tips, fun child activities, parenting tips and skill building information (e.g., goal setting).

2.3. Measures

All measures were collected at baseline, post-treatment (8 months) and follow-up (20 months). At baseline, a demographic questionnaire assessed the child’s age, gender, race, ethnicity, free/reduced lunch status, household income, and parent education.

2.3.1. Primary Outcome Measures

Child Body Mass Index z-score (BMIz) and Parent Body Mass Index (BMI):

Parent height was measured at baseline. Parent weight, child height, and child weight were initially assessed monthly by fully trained school personnel on standardized equipment provided by the study team. Standing height was assessed via a Harpenden Holtain stadiometer, Model 603 (Holtain, Crymych, UK), and weight was measured on a portable SECA digital scale (SECA, Hamburg, Germany). Once COVID-19 impacted school availability for measurement, families were asked to either use their own home scale or were provided with a digital scale (Etekcity High Precision Digital Body Weight Bathroom Scale) and tape measure. Measurement then occurred over a proctored virtual session with study personnel; the full protocol and differences in these measurement methods are published in detail.²⁷ These results suggested only small and clinically insignificant differences were found between the measurement methods.²⁷ Standard formulas/equations were utilized to calculate child BMIz and parent BMI.

2.3.2. Secondary Outcome Measures:

Dietary Intake:

Food intake was measured via the multiple-pass 24-hour food recall.^28,29 Two weekdays and one weekend day of dietary information were collected at each assessment point by highly trained staff. Trained staff used the Nutritional Data System for Research (NDSR; University of Minnesota, Minneapolis, MN) to analyze the recalls and calculated the per day number of servings of sugar-sweetened beverages, “red” food items (foods with ≥ 7 g of fat and/or ≥ 12 g of sugar), and servings of fruits and vegetables. To reflect the marketplace throughout the study, dietary intake data were collected using NDSR software versions 2017–2020. Final calculations were completed using NDSR version 2020.

Physical Activity:

Physical activity was assessed using the ActiGraph wGT3X-BT accelerometer (ActiGraph LLC, Pensacola, FL), which was worn on the child’s non-dominant hip for seven consecutive days at each assessment point with raw data sampled at 40 hertz. The data were reintegrated to 15s epoch via ActiLife software, and screened for non-wear using the Choi algorithm.³⁰ Participants who wore their ActiGraph for at least 10 hours per day on ≥4 of 7 days were included in analyses. Analyses only considered wear time between 6:30am – 9:30pm on weekdays and 6:30am – 10:30pm on weekend days. Total daily minutes spent in MVPA were calculated, and to account for total wear time, average (across valid wear days) percent of wear time spent per day in MVPA and the percent of wear time spent sedentary were calculated. Evenson cut-points were utilized to create intensity specific estimates.³¹

2.4. Analysis Plan

Both Intent-to-Treat (ITT) analyses and analyses of completers were planned a priori (Davis et al., 2019). All participants who consented to participate, completed baseline measures, and had at least one measurement after baseline were included in the ITT analyses. Completer analyses included participants who received at least 80% of iAmHealthy. For dyads in the Control group, they must have completed at least 80% of their data collection time points.

For each primary and secondary outcome, two analyses were conducted, one examining outcomes at 8-months (post) and one examining outcomes at 20-months (follow-up). To accommodate the cluster randomized design where schools are randomized to groups, multilevel modeling analyses were conducted using SAS PROC MIXED. In this design, observations (Level-1 units) were nested in children (Level-2 units) which in turn are nested in school treatment clusters (Level-3 units). Each school had exactly one treatment cluster.

A fixed effect indicating treatment group was estimated in each model as a Level-3 predictor to determine if there are significant differences between participants based on group. Gender was included at Level-2 to control for differences in all models. The hypotheses were directional and anticipated that the iAmHealthy group would have a better response than the Control group. Full information maximum likelihood estimation, which uses all available data to build a multidimensional likelihood function for each highest-level unit (schools), was used for the multilevel models. For variables brought into the likelihood function, any missing cases were assumed missing at random, which means random only after conditioning on model predictors and the observed outcomes.

2.4.1. Power analysis

A priori power analyses for treatment group differences were conducted using Optimal Design software for cluster-randomized trials with person-level outcomes and treatment at Level-2. These analyses assumed equal cluster sizes and indicated that 18 clusters with eight dyads per cluster allowing for attrition would address the primary aims of this study. See Davis et al. 2019¹⁸ for more information regarding power analyses.

3. Results

A total of 18 schools were randomized into groups. The schools had an average of 8.22 parent-child dyads each (range 5–14) for a total of 148 dyads consented and randomized. Thus, the target number of participants based on a priori power analyses were obtained. Of the 148 families, 64 dyads were in the nine schools assigned to the iAmHealthy group, and 84 dyads were in the nine schools assigned to the Control group. At baseline, 56.8% of children identified as female, ranged in age from 6–10 years (M = 8.93, SD = .86), and identified as predominantly White (87.1%), followed by biracial (10.1%); 13.5% of our sample identified as Hispanic. Approximately 50.0% of children were eligible for free/reduced lunch. 87.8% of youth participated with their mother, and mean caregiver BMI was in the obese range (M = 34.43, SD = 8.91). There were no significant differences at baseline among any of the demographic variables.

A participant CONSORT diagram is presented in Figure 1. For the iAmHealthy group, participants completed between 0–22 hours of treatment (M = 15.84, SD = 7.13). Participants on average obtained 10.62 hours (SD = 4.37) of group sessions and 5.22 hours (SD = 2.93) of individual health coaching. Of note, while participants were able to have 45-minute health coaching sessions, these sessions lasted on average 31 minutes. This resulted in a maximum of only 8 hours of health coaching delivered to families for a total of 22 total contact hours delivered. Of the 64 iAmHealthy dyads, 38 (59%) completed at least 80% of the 22 delivered contact hours, and 48 (75%) completed 50% or more. When considering all dyads, 91.22% completed post-measures, and 87.16% completed follow-up measures. For dyads in the iAmHealthy group and Control group, 52 and 75 dyads completed 80% of their data collection time points, respectively. A one-way ANOVA suggested no significant differences in the number of dyads who completed 80% of measurement timepoints between the two groups, (F(1,146)= 1.92, p = 0.167). For all analyses, ITT and completers analyses were conducted; results did not significantly differ, thus, ITT analyses are presented.

Fidelity coding was conducted on all group sessions, and 20% were selected for double coding at random.³² A content checklist of 8–10 topics was developed for each session, and independent trained coders scored the videotapes for the presence/absence of topics. Analyses indicate fidelity was high as 96.75% of topics from the manual were covered by group leaders. Interrater agreement was calculated using percent absolute agreement which indicated high agreement between raters³³ (percent agreement = 95.52).

The least squares means for all primary outcomes are provided in Table 1. To help streamline results, only outcomes with significant effects by group are presented in subsequent tables. Additional information regarding both ITT and completer analyses is available upon request from the authors. All analyses included each of the 18 clusters. Table 2 summarizes the results from baseline to post. ICCs suggested 7.2% of the residual variation in change in child BMIz and 19.8% of change in parent BMI from baseline to post was attributable to systematic differences between schools. Child BMIz means were not significantly different between groups at baseline. For child BMIz change from baseline to post, the change in the Control group from 1.91 to 1.96 represents a marginally significant increase, F(1,130)= 3.82, p = 0.053, while the iAmHealthy group had no significant changes F(1,130)= 2.20, p = 0.140. For parent BMI, there were no differences between groups at baseline and neither group changed significantly from baseline to post.

Table 1.

Primary Outcomes Over Time by Group.

Outcomes	Baseline iAmHealthy*	Post iAmHealthy	Follow-up iAmHealthy	Baseline Control*	Post Control	Follow-Up Control
	Mean (SD)
BMI Outcomes
Child BMIz	1.79 (.07)	1.83 (.08)	1.83 (.07)	1.91 (.07)	1.96 (.07)	2.03 (.07)
Parent BMI	35.76 (1.70)	36.18(1.16)	36.10 (1.69)	33.21 (1.65)	33.43 (1.65)	33.79 (1.63)
Dietary Outcomes
Sugar-Sweetened Beverages	1.23 (.11)	0.80 (.12)	0.96 (.14)	1.15 (.10)	1.08 (.11)	1.22 (.13)
Red Foods	6.72 (.48)	5.17 (.53)	5.87 (.59)	7.21 (.46)	5.95 (.48)	6.89 (.57)
Fruit and Vegetables	3.19 (.20)	3.13 (.23)	2.86 (.22)	2.69 (.17)	2.47 (.19)	2.57 (.19)
Physical Activity Outcomes
MVPA	52.10 (3.34)	49.36 (3.56)	47.03 (3.28)	53.45 (3.15)	49.58 (3.28)	42.34 (2.99)
% Time in MVPA1	6.72 (.43)	6.32 (.45)	6.07 (.42)	6.93 (.40)	6.44 (.42)	5.56 (.38)
% Sedentary Time1	59.07 (1.08)	60.57 (1.15)	63.27 (1.06)	56.60 (0.98)	59.69 (1.03)	63.36 (0.91)

Note.

^*Baseline means and standard deviations presented are from baseline to post comparisons. Significant differences in baseline measurements from baseline to follow-up comparisons are discussed in text.

¹Percent time in MVPA and percent sedentary time represent the total amount of time in that activity type divided by total wear time.

Table 2.

Outcomes from Baseline to Post (8 months).

Intent to Treat Models Pre/Post
Fixed Effects
Outcome	Group	Time	Time*Group
Child BMIz	F(1,130) = 1.60 P = 0.21	F(1,130) = 5.78 p = 0.02	F(1,130) = 0.03 p = 0.87
Sugar-Sweetened Beverages	F(1,104) = 0.53 p = 0.47	F(1,104) = 9.11 p = 0.003	F(1,104) = 4.38 p = 0.04
Red Foods	F(1,104) = 1.14 p = 0.29	F(1,104) = 16.57 p < 0.001	F(1,104) = 0.18 p = 0.67
Fruit and Vegetables	F(1,104) = 6.72 p = 0.01	F(1,104) = 0.71 p = 0.40	F(1,104) = 0.25 p = 0.62
% Sedentary Time 1	F(1,106) = 1.48 p = 0.23	F(1,106) = 15.10 p < 0.001	F(1,106) = 1.80 p = 0.18

Note. All models included the planned covariate of gender.

¹Percent time in MVPA and percent sedentary time represent the total amount of time in that activity type divided by total wear time.

There are three outcomes related to diet: number of sugar-sweetened beverages, number of “red” foods, and number of fruits and vegetables consumed per day. At baseline, youth in both groups reported consuming slightly more than one sugar-sweetened beverage per day. There was no difference between groups at baseline. While no significant decrease was observed for the Control group (F(1,104)= 0.50, p = 0.480), there was a significant decrease in consumption of sugar-sweetened beverages for the iAmHealthy group from baseline to post (F(1,104)= 11.34, p = 0.001). For the average number of servings of “red” foods per day, both groups saw significant decreases over the intervention period of approximately one serving of red food daily for each group (F(1,104)= 8.66, p = 0.004 for Control; F(1,104)= 9.16, p = 0.003 for iAmHealthy), but no difference was observed between groups at baseline, post, or follow-up measurement. For number of fruits and vegetables consumed daily, groups were not significantly different at baseline, and neither changed significantly over the course of the intervention. However, small differences in starting values combined with somewhat uneven changes led to the iAmHealthy group having significantly higher average numbers of fruits and vegetables per day at post (F(1,104)= 5.08, p = .026). The Control group ate approximately 2.5 fruits and vegetables daily at each time point, while the iAmHealthy group had slightly more than 3.

There are three outcomes related to physical activity, average daily minutes in MVPA, percent of time spent in MVPA, and percent of time spent in sedentary activity. For average daily minutes of MVPA and percent of time spent in MVPA, no differences related to time or condition were observed from baseline to post. Percent of wear time spent in sedentary activity increased from baseline to post intervention for both groups overall, and specifically for the Control group. When simple effects were examined, a significant increase in percent of sedentary time was observed in the Control group from baseline to post (F(1,106)= 16.07, p < 0.001) but not for the iAmHealthy group (F(1,106)= 2.81, p = 0.097).

Table 3 summarizes the results from baseline to follow-up for the primary outcomes with significant effects by group. For the outcome of child BMIz from baseline to follow-up, the change in the Control group from 1.91 to 2.03 was a significant increase in BMIz, F(1,121)= 12.88, p < 0.001, while the iAmHealthy group had no significant changes F(1,121)= 1.19, p = 0.279. At follow-up, the difference between groups was significant, F(1,121)= 3.99, p = 0.048, with the iAmHealthy group having an average BMIz of 1.83 and the Control group having an average BMIz of 2.03. Parent BMI increased significantly from baseline to follow-up for the Control group F(1,121)= 170.81, p < 0.001 and the iAmHealthy group, F(1,121)= 52.96, p < 0.001. The two groups were significantly different at follow-up F(1,121)= 4.36, p = 0.039 with the iAmHealthy group having a higher parent BMI (M = 36.10) than the Control group (M = 33.79).

Table 3.

Outcomes from Baseline to Follow-up (20 months).

Intent to Treat Models Pre/Follow-Up
Fixed Effects
Outcome	Group	Time	Time*Group
Child BMIz	F(1,121)=2.97, p = 0.09	F(1,121) = 9.71, p = 0.002	F(1,121) = 2.04, p = 0.16
Parent BMI	F(1,119) = 1.06, p = 0.30	F(1,119) = 2.81, p = 0.10	F(1,119) = 0.12, p = 0.73
Average Daily Minutes in MVPA	F(1,97) = 0.21 p = .65	F(1,97) = 19.10 p < .001	F(1,97) = 2.43 p = .12
% Time in MVPA 1	F(1,97) = 0.10 p = 0.76	F(1,97) = 18.21 p < 0.001	F(1,97) = 2.18 p = 0.14
% Sedentary Time 1	F(1,97) = 1.19 p = 0.28	F(1,97) = 64.42 p < 0.001	F(1,97) = 3.31 p = 0.07

Note. All models included the planned covariate of gender.

¹Percent time in MVPA and percent sedentary time represent the total amount of time in that activity type divided by total wear time.

Given the increases in parent BMI from baseline to follow-up, we evaluated if change in parent BMI predicted child BMIz. Results suggested parent BMI change from baseline to post, F(1,100)= .67, p = 0.414, B = −.001, did not impact child BMIz at post. However, parent BMI change from baseline to follow-up, F(1,95)= 4.50, p = 0.036, B = .03, was positively associated with child BMIz at follow-up. Additionally, we evaluated if the COVID-19 pandemic influenced treatment outcomes by separately evaluating child BMIz outcomes for participants who were either unaffected by COVID-19 and participants who experienced the COVID-19 pandemic either during the intervention or prior to their follow-up assessment. Regardless of whether participants were impacted by COVID-19, iAmHealthy did not outperform the Control group, and those impacted by COVID-19 had greater increases in BMIz at post and follow-up time points compared to those who completed their assessments prior to the COVID-19 pandemic.

There were no significant differences by groups observed for any of the dietary outcomes at the follow-up time point. For the average daily minutes in MVPA, no differences were observed between groups at baseline. When simple effects were examined, a significant decrease in MVPA was observed in the Control group from baseline to follow-up (F(1,97)= 20.59, p <.001) while the iAmHealthy group had no significant changes in MVPA from baseline to follow-up (F(1,97)= 3.45, p = .067). Similarly, for the percent of time spent in MVPA, no differences were observed between groups at baseline. When simple effects were examined, a significant decrease in percent of time in MVPA was observed in the Control group from baseline to follow-up (F(1,97)= 19.34, p < 0.001), while the iAmHealthy group had no significant changes in percent of time spent in MVPA from baseline to follow-up (F(1,97)= 3.39, p = 0.069). Similar results were also observed for percent of time sedentary. A significant increase in percent of time sedentary was observed in the Control group (F(1,97)= 56.98, p < 0.001) and iAmHealthy group (F(1,97)= 16.76, p < 0.001) from baseline to follow-up. There was also a difference at baseline (F(1,97)= 4.17, p = 0.044) between the iAmHealthy group (M = 58.90) and Control group (M = 56.42) which was not observed in the 8-month baseline comparison.

4. Discussion

The objective of this study was to examine the effectiveness of a family-based intervention delivered over interactive televideo to treat pediatric obesity among children from rural areas (iAmHealthy) compared to a Newsletter Control group. While no differences were seen between groups with respect to BMIz at post-intervention, results at follow-up suggested that iAmHealthy may have resulted in a change in BMIz trajectories and long-term health behavior for youth. More specifically, at the one-year follow-up, youth in the iAmHealthy group had a slight but non-significant increase in BMIz while those in the Control group had significant increases in BMIz. While these results suggest iAmHealthy may have helped to attenuate the increase in BMIz trajectory for youth, youth in the iAmHealthy condition ultimately did not have significant improvements in BMIz. This limited effectiveness highlights the importance of continuing to find strategies to treat youth with overweight and obesity in rural communities.

Regarding dietary intake, youth in the iAmHealthy group significantly decreased their sugar-sweetened beverage consumption at post-intervention while the Control group did not; both groups significantly decreased their “red” food consumption at post-intervention. Despite these improvements, youth were still consuming approximately six “red” foods per day at the end of the intervention. While this is consistent with previous literature,¹⁶ this number of “red” foods consumed is much higher than the recommended 4 “red” foods per week discussed in the stoplight diet.³⁴ Regarding physical activity, youth in both groups were initially engaging in approximately 52 minutes of MVPA per day. This is less than other children their age³⁵ and slightly less than the recommended 60 minutes of daily MVPA.³⁶ Despite initially low levels of MVPA, at follow-up, the Control group engaged in a significantly lower percentage of MVPA, while the iAmHealthy group had no significant changes. It is possible these minor changes in health behaviors contributed to the BMIz maintenance in the iAmHealthy group.

Twelve months after the intervention completed, the Control group had significant increases in BMIz while the iAmHealthy group did not. This outcome demonstrates the importance of long-term follow-ups for pediatric obesity treatment studies. Reviewing post-treatment results exclusively, there were minimal differences between groups; however, when evaluating treatment outcomes at follow-up (20 months), iAmHealthy youth were on a significantly better BMI trajectory compared to Controls. This improved trajectory demonstrates the importance of early treatment and prevention for pediatric obesity.

Despite not being a part of the inclusion criteria, parents in both groups had an average BMI in the obese range at baseline. Both groups also had significant increases in BMI from baseline to follow-up. Previous research indicates decreases in parent BMI are associated with improved health behaviors in youth.³⁷ In this study, parent changes in BMI from baseline to follow-up were associated with child BMIz at follow-up, such that parents whose BMI decreased had children with BMIz decreases and parents with increases in BMI were more likely to have youth with increased BMIz scores. Thus, our study suggests this lack of parent improvement has an influence on child outcomes. This lack of positive change in parent BMI in family-based treatment is consistent with past research^16,38 and illuminates a need for parent health behaviors to be more directly addressed in family-based behavioral treatment.

Past obesity intervention dropout rates have ranged from 1–42% at post-intervention and 12–52% for 12-month follow-ups.³⁹ In this study, less than 10% of participants did not complete post-measures, and less than 13% did not complete follow-up measures. This low attrition may be explained by the participation of school personnel who had established connections with the participating families. These personnel personally contacted families to obtain BMI measurements and encourage measurement completion. This low attrition is similar to our past research in which participants attended telehealth or telephone sessions at their local school¹⁶ suggesting this remote approach in partnership with rural elementary schools is an acceptable option for rural families that may protect against drop out. Despite this strong retention, only 59% of iAmHealthy families completed 80% of the delivered contact hours. These rates are low given recent research with families from urban areas finding contact hour rates range from 68–95%.¹⁷ Despite iAmHealthy delivering the intervention via telehealth directly to the participants homes and sending text message reminders to promote attendance, overall attendance was low. Anecdotally, our interventionists reported families often identified busy schedules being the reason for non-attendance. Future studies should evaluate predictors of attendance to telehealth obesity programs for families in rural communities so families at high risk for non-attendance can be identified early and reasons for potential non-attendance can be addressed (e.g., problem solve barriers, utilizing motivational interviewing techniques to increase engagement).

One important consideration for future research is whether to utilize the clustered design approach. We initially chose to randomize schools to treatment groups to decrease potential contamination among research participants and allow for community-level factors to be discussed amongst the group (e.g., ideas of places in the community to exercise). However, this design did not allow researchers to randomize other important factors that may influence study results. For example, some schools had unique demographic or school characteristics (i.e., personnel) that may have impacted study outcomes. Thus, future studies may want to consider randomization within schools to overcome these limitations.

This study has many strengths including the utilization of a randomized control design and both intent-to-treat and completer statistical models that accounted for the multilevel structure of the data. Further, this intervention’s design was able to increase convenience for families and achieve a very low attrition rate. Despite these strengths, this study has multiple limitations to consider. Child BMIz changes, while positive, were not as large as we had hoped. Similarly, despite this program being family-based, parents in both groups had significant weight gains at follow-up. Thus, iAmHealthy may insufficiently meet the needs of parents. Given the time intensive nature of iAmHealthy, the lack of significant improvement highlights a need to take a step back and ensure the recommended strategies are effective for families in rural communities. It is possible a multi-level approach (e.g., targeting changes in the school, community, and policy while working with the families) may be more effective given the community level barriers these families face. It is also possible this lack of improvement for parents and children may be partially explained by the COVID-19 pandemic. Our results suggested those impacted by COVID-19 consistently had greater increases in BMIz compared to those who participated prior to the pandemic. Thus, any potential intervention effects from either iAmHealthy or the Control group may have been overwhelmed by the body mass changes that occurred during the time of COVID-19.^40,41 Additional research⁴² has found that obesity intervention efficacy was diminished during COVID-19, and the parents enrolled in this iAmHealthy intervention⁴³ reported that the COVID-19 pandemic increased variability in family meals, unstructured time, caregiver stress, and child behavioral problems.

Additionally, despite planning to provide 25 hours of treatment to the iAmHealthy group, only 22 hours were delivered and families on average obtained only 15 hours of treatment. This low dose received may be a result of the COVID-19 pandemic, indicative of additional time constraints of families from rural areas, or indicative of a lack of interest/engagement of families with the iAmHealthy program. Future studies should identify what factors impact attendance in pediatric obesity treatment for families from rural communities. Further, although iAmHealthy and the Control group were matched in terms of length of time involved in the study, they were not matched in the total amount of time invested into the study (as only 8 newsletters were sent to the Control group). This may have impacted expectations and engagement of the Newsletter Control. Lastly, participants were predominately White (representing the majority demographics of the rural areas of the one state from which they were drawn), thus findings may not generalize to other rural populations.

4.1. Conclusions

In summary, the current study evaluates the effectiveness of a group and individual family-based intervention delivered over interactive televideo to treat pediatric obesity. While only minor changes were seen at post-treatment, results at follow-up indicated the Newsletter Control was not sufficient to prevent increases in obesity and iAmHealthy may have helped to mitigate the BMIz trajectory for youth. Child BMIz and parent BMI results ultimately suggest that more work is needed to effectively change BMI outcomes for families in rural areas. This trial extended previous pediatric obesity work in an innovative direction by moving the mHealth intervention into rural family homes while simultaneously increasing convenience and opportunity for contact hours. Results demonstrated very low dropout rates suggesting this home-based telemedicine intervention in partnership with rural elementary schools may be an effective option to increase access to care for families in rural communities.