Executive Functioning,Depressive Symptoms,and Intervention Engagement in a Sample of Adolescents Enrolled in a Weight Management Program

Abstract

Purpose:

This study examines how baseline demographics, psychosocial characteristics, and intervention delivery predict engagement among adolescents with overweight and obesity seeking treatment.

Methods:

Data originates from a multisite randomized control trial evaluating the efficacy of an app-based weight loss intervention, compared with standard in-clinic model in adolescents with overweight and obesity. Participants were randomized to one of the three arms: (1) AppCoach, (2) AppAlone, or (3) Control. Demographic, executive functioning (EF), and depression questionnaires were completed at baseline. Percent engagement was compared within and between groups defined by demographics and depressive symptoms. Quantile regression was used to evaluate the association between age and EF on percent engagement.

Results:

Baseline demographics were not associated with engagement within or between groups. Neither baseline self-reported depressive symptoms (p = 0.244) nor deficits in EF (p = 0.34) were predictors of engagement. Univariate analysis found that the control arm had the highest engagement (83%) compared with AppCoach (63.5%) and AppAlone (22.5%, p = 0.02). Hispanic ethnicity was predictive of higher engagement in the control arm (p = 0.02). On multivariate quartile regression no other baseline characteristics were significant predictors of engagement.

Conclusion:

Baseline demographics and individual psychosocial characteristics were not related to engagement in this cohort. The intervention arm that required parental involvement resulted in the greatest engagement suggesting that family involvement may overshadow individual behavioral phenotype and thus promote better engagement. Further investigation is needed to understand how program delivery can be leveraged to optimize treatment engagement and outcomes in adolescence.

Clinical Trial Registration number: NCT03500835.

Introduction

One in five adolescents has obesity in the United States.^1–3 Adolescents with obesity report higher rates of social isolation, poorer quality of life, and depressive symptoms compared with their nonobese peers.^4,5 In recent years, there has been an increased interest in understanding factors that may help explain treatment engagement and adherence among adolescents seeking weight-management treatment.^2–4 This is important because the effectiveness of weight loss interventions depends on adherence to treatment recommendations.^2,5

Studies in adults have shown that demographic characteristics, such as race, gender, age, household income, and insurance type are predictors of poor participation in prescribed treatment and attendance to visits, specifically in the management of chronic conditions such as obesity.^6,7 In children with obesity, who rely heavily on family involvement and participation, low household income, ethnicity, and race have been linked to lower engagement and attendance.⁸ Many obesity programs have attempted to address these disparities by developing culturally sensitive curricula and providing additional incentives to account for the financial burden of participating in specialized, outpatient obesity interventions.^7,8 However, it remains unclear how the developmental changes occurring in adolescents may influence engagement as they transition to greater individual autonomy and less reliance on family involvement.^9,10

During the transition from childhood to adolescence, individual characteristics may start to have a more significant impact on treatment engagement and outcomes. In a recent meta-analysis, adolescents with obesity were 1.3 times more likely to report depressive symptoms than their nonobese peers.¹¹ Depressive symptoms are associated with impairments in psychosocial and cognitive functioning,^9,12 likely to impede adolescents' ability to engage and adhere to weight loss and healthy lifestyle recommendations.^9,11,12 Previous studies have shown that in both pediatric and adult patients with obesity, baseline depressive symptoms can interfere with engagement in weight loss interventions.¹⁰ Interestingly, these relationships are more equivocal in adolescents, with some groups showing no association between baseline depression and intervention adherence.⁸ Further investigation is warranted to better understand how depressive symptoms are related to obesity treatment participation and attendance in the pivotal period of adolescent development.¹⁰

There is also growing evidence linking deficits in executive functioning (EF) and obesity in adults,^13–18 and impairments in EF have been linked to poor outcomes in weight loss interventions.^14,19 Less is known about the role of EF on treatment engagement and outcomes in adolescents. EF encompasses a range of higher cognitive capacities enabling goal-directed behavior, including: inhibition, cognitive flexibility, planning, and decision making.¹⁵ EF develops in childhood and manifests in adolescence; therefore, deficits in EF may lead to difficulties with self-regulation, impulse control, and problem-solving skills, which can all contribute to maladaptive eating behaviors and resulting in weight gain. Considering the role of these cognitive processes in behavior change, and in sustaining lifestyle modifications, EF may potentially be one of the predictors linked to adolescents' ability to consistently adhere to intervention recommendations.^20,21

Known associations exist between depression and deficits in EF.^22–24 It has been reported in the literature that deficits in EF are observed in adolescents and adults with depressive symptoms.^22–24 In a recent study, results showed that deficits in EF could worsen and increase depressive symptoms in adolescents who have a high BMI.²² In a systematic review and meta-analysis, impairment in EF and depressive symptoms is associated with low engagement and poor response to treatment interventions.²³ Considering that the role of depression and EF in adolescents affect treatment and lead to poor outcomes, it is imperative to assess these characteristics at baseline to better understand how future interventions can help adolescents be successful in treatment outcomes.

In addition to individual characteristics, the intervention delivery model may also contribute to adolescents' engagement in obesity treatment. The current clinical practice guidelines for pediatric obesity management recommend moderate-to-high-intensity (26 to >52 contact hours), multidisciplinary lifestyle modification programs for the treatment of obesity in youth. Unfortunately, these programs are labor intensive, costly, and difficult to implement on a large scale. Therefore, more simplified intervention curriculums and delivery modalities are needed. Modest-intensity weight loss interventions that recommend more than 26 contact hours demonstrated clinical meaningful BMI reduction.²⁵ For many families this is logistically difficult to implement given work and school schedules, transportation barriers, and financial burden of missed work.⁷ Multiple studies have shown that mobile health (mHealth) interventions and health coaching models may increase adherence to recommendations^26–28 and ameliorate some of the barriers interfering with treatment attendance.^27,29,30 Given ubiquitous adolescent use of technology, mHealth interventions may be a particularly well-suited modality for delivering weight loss interventions and promoting engagement.^28,31

This study examines associations between baseline individual characteristics and intervention delivery modality with treatment engagement in adolescents with overweight and obesity enrolled in a multisite randomized control trial that tests the efficacy of a behavioral app-based weight loss intervention compared with a multidisciplinary in-clinic model.³² We were particularly interested in understanding how demographic characteristics, depressive symptoms, EF, and intervention delivery (mHealth with or without coaching vs. in-person) were related to treatment engagement over the course of a 6-month weight loss intervention. Based on previous findings, we hypothesized that participants who report depressive symptoms or deficits in EF at baseline would have lower engagement, specifically in this sample of treatment-seeking adolescents with overweight and obesity.

Materials and Methods

Participants

Eligible participants were between the ages of 14–18 years with BMI ≥85th percentile for age and sex. Adolescents were excluded if they had: (1) previous diagnosis of developmental delay or hypertension; (2) inability to read English; or (3) current participation in other interventional studies. Participants were recruited from: (1) pediatric clinics at Children's Hospital Los Angeles (CHLA); (2) through direct mailing campaign; and (3) community outreach. Study procedures were approved by the CHLA's Institutional Review Board. The study was reported according to the Consolidated Standards of Reporting Trials (CONSORT) statement and is registered with ClinicalTrials.gov. Written informed consent was obtained from both the adolescents and a parent or guardian.

Intervention Arms

Participants were equally randomized to receive one of three intervention arms: (1) interactive mHealth with personalized health coaching (AppCoach), (2) interactive mHealth weight loss intervention alone (AppAlone), or (3) multidisciplinary in-clinic weight management program (Control). The full protocol was reported by Vidmar et al.³² Briefly, the app intervention was designed as a three-part intervention strategy that incorporates evidence-based nutrition guidance such as portion control and eliminating excessive snacking. The app intervention was intended to be implemented over a 6-month period equating to ∼15 contact hours per 6-month period. The arm that includes personalized coaching offered ∼21 contact hours per 6-month period. The control intervention was designed to replicate a low-intensity, multidisciplinary weight management program focused on evidenced-based nutrition recommendations for adolescents with overweight and obesity as recommended in the 2017 pediatric obesity clinical practice guidelines with ∼12 contact hours per 6-month period. The control arm required parental involvement as one parent was required to accompany the adolescent to each session.

Health Coaching

Two health coaches implemented the interventions. The coaches had completed undergraduate degrees in health-related fields and undertook structured training for the implementation of the coaching sessions. Each coach received 20 hours of curriculum training over a 4-week period with the principal investigator and app creator. The curriculum covered nonjudgmental communications, active listening, motivational interviewing, creating self-management goals, and simulated role play. In addition to curriculum training, the coaches were required to demonstrate understanding of the app intervention and mastery of coaching skills through role-play scenarios.

Measurements

Demographics and medical history

Participant and family member's demographics (i.e., race, ethnicity, household income, education) and medical history were collected at baseline. Participants received compensation in the form of gift cards to complete study assessments.

Anthropometrics

Participants' height and weight were assessed at baseline by a research coordinator. Height was measured using a Quick Medical stadiometer, accurate to 0.1 cm (Quick Medical, Issaquah, WA). Weight was measured on a self-calibrating Mobile Stand Digital Scale, accurate to 0.1 kg. Adolescents wore minimal clothing during the height and weight measurements. BMI was calculated as kilograms per meter squared and BMI z-score (zBMI) was determined utilizing the CDC growth charts.

Executive functioning

The Behavior Rating Inventory of Executive Function-2 (BRIEF-2) was used to evaluate baseline EF.³³ The BRIEF-2 provides theoretically and empirically derived clinical scales that measure aspects of EF in adolescents 11–19 years of age.³⁴ Items are scored on a 3-point Likert scale ranging from one (never) to three (often). Theoretically and statistically derived scales measure the adolescent's ability to regulate and monitor behavior (Behavior Regulation Index, BRI), emotional responses (Emotion Regulation Index, ERI), and cognitive processes (Cognitive Regulation Index, CRI). These scales can also be combined into a summary measure, global executive functioning composite (GEC) score.^33,35 T-scores are used to help interpret the participant's self-reported profile of EF on the BRIEF-2 self-reported form. Higher T-scores indicate higher level of impairment. For all BRIEF-2 indexes and clinical scales, T-scores between 60 and 64 are considered mildly elevated; T-scores from 65 to 69 are considered potentially clinically elevated; and T-scores at or above 70 are considered clinically elevated.³⁵

Depression

The Center for Epidemiological Studies Depression Scale for Children (CES-DC) was utilized to measure baseline self-reported depressive symptoms. The CES-DC has been developed and validated in children younger than 18 years of age. The CES-DC is a 10-item self-report scale designed to measure depressive symptoms in children and adolescents. Elevated scores on the CES-DC at or above 10 suggest depressive symptoms. The CES-DC has been reported to be a highly reliable and valid tool for assessing depressive symptoms across ethnic, gender, and age groups. Cronbach's Alpha coefficient for the scale was reported to be 0.75 for the general population. Test-retest reliability of the scale was reported to be 0.79 in 15-year-old Guatemalan adolescents. Concurrent, divergent, and convergent validity of the CES-DC has also been reported in numerous studies.^36–39

Engagement

Percent engagement was defined as the ratio of actual intervention dosage received to total dosage intended to be delivered. The total dosage intended to be delivered varied across the intervention arms: (1) AppCoach participants were required to attend 24 weekly coach calls and use the app daily for 180 days (6 months); (2) AppAlone participants were required to use the app daily for 180 days (6 months); (3) Control participants were required to attend 6-monthly multidisciplinary in-clinic visits. The engagement data for the AppCoach and AppAlone arms were extracted from the study database, which included date stamps of when the app was used and the number of coach calls completed. Engagement data for the control group was operationalized as the total sessions attended over the course of the 6-month intervention. Participants who completed the 6-month intervention period were considered “completers” (n = 55) and those who withdrew before intervention completion were considered “noncompleters” (n = 21), these categories were considered when calculating percent completion.

Data Analyses

Descriptive statistics were used to summarize the demographic data presented in Table 1. Continuous variables were summarized using mean with standard deviation and median with interquartile range. Categorical variables were summarized using frequency and percentage. We were interested in assessing the relationships between demographics, depression scores, EF, and intervention arm characteristics and participation in prescribed tasks and attendance (percent engagement) across intervention arms. Mathematical model specification was guided by our conceptual model, which treated age, sex, ethnicity, and household income, as well as CES-DC and BRIEF-2 scores (composite and subscale), as predictors of attendance and participation in intervention arm activities. Kruskal–Wallis and Wilcoxon Rank Sum tests were used to compare percent engagement distribution within and between the three groups defined by demographic variables with univariate and multivariate analyses. A quantile regression model was used to assess the relationships among BRIEF-2 T-scores (GEC and subscale scores), CES-DC scores, and percent engagement. This method was chosen to avoid the need of making assumptions about the distribution of residuals. Lastly, bivariate correlations were performed between CES-DC scores and BRIEF scores and subscales. The results were expressed as coefficient estimates, β with their associated 95th% confidence intervals (CIs). The significance level was set at 5% and was two sided. All statistical computations were performed in Stata/SE 15.1 (StataCorp, College Station, TX).

Table 1.

Demographic Characteristics among Study Completers and Noncompleters

	Total (n = 76)	Noncompleters (n = 21)	Completers (n = 55)	p
Age^a	15.39 ± 1.26	15.62 ± 1.20	15.31 ± 1.27	0.33^*
Gender^b
Male	31 (40.79)	9 (42.86)	22 (40.00)	0.82^**
Female	45 (59.21)	12 (57.14)	33 (60.00)
Ethnicity^b
Hispanic	50 (65.79)	8 (38.10)	42 (76.36)	0.002^**
Non-Hispanic	26 (34.21)	13 (61.90)	13 (23.64)
Insurance^b
Government	55 (72.37)	15 (71.43)	40 (72.73)	0.53^***
Private	16 (21.05)	6 (28.57)	10 (18.18)
Not reported	5 (6.58)	0 (0)	5 (9.09)
Parent education^b
High school or less	44 (57.89)	14 (66.67)	30 (54.55)	0.14^***
Trade/vocational	7 (9.21)	2 (9.52)	5 (9.09)
College	22 (28.95)	3 (14.29)	19 (34.55)
Graduate degree	3 (3.95)	2 (9.52)	1 (1.82)
Household income^b
<$50,000	50 (65.79)	13 (61.90)	37 (67.27)	0.17^***
$50,000–$149,999	16 (21.05)	3 (14.29)	13 (23.64)
≥$150,000	5 (6.58)	3 (14.29)	2 (3.64)
Not reported	5 (6.58)	2 (9.52)	3 (5.45)

All 76 study participants and their demographic characteristics between “noncompleters” and “completers.” The participants labeled as “completers” completed the 6-month intervention. The ones labeled “noncompleters” mean that they withdrew.

Two-sample t-test.

Chi-square test.

***

Fisher's exact test.

Mean ± standard deviation.

Frequency (percentage).

Results

Participant Characteristics

Table 1 summarizes participants' baseline characteristics and descriptive statistics both for participants who completed the 6-month intervention and associated study measurements (completers), and those who did not complete the study (noncompleters). A total of 76 adolescents with overweight and obesity (male/female: 31/45) completed the 6-month intervention and study measurements (Fig. 1). They had a mean age of 15.59 ± 1.3 years and baseline BMI z-score (zBMI) of 2.17 SD. Consistent with the demographics of patients served by CHLA, most participants (66%) were Hispanic, 72% were publicly insured, and 66% reported an annual household income of $50,000 or less (Table 1). There was no difference in baseline demographic, psychosocial characteristics, or BMI status between intervention arms. At baseline, one quarter (25%) of participants had high CES-DC scores suggestive of depressive symptomatology. There was no statistically significant difference between the participants with and without depressive symptoms regarding their baseline zBMI, obesity-related comorbidities, ethnicity, or age. More than half of participants (54.3%) had scores indicative of clinical impairment in EF (GEC-T score >60, Table 2). The bivariate correlation analysis shows that there is a positive correlation between CES-DC scores and EF subscale scores (Table 3) demonstrating that adolescents with higher depressive symptomatology were more likely to have executive dysfunction as demonstrated by higher composite and scale T-scores on the BRIEF-2 (Table 4).

Figure 1.

CONSORT Diagram. The CONSORT Diagram depicts all study participants that were enrolled in the Randomized Control Trial (citation of RCT protocol). Shaded boxes represent the subset of participants included in the analysis. Participants were included in the analysis if they were still in the intervention phase of the study. The participants labeled as “completers” completed the 6-month intervention. The ones labeled “noncompleters” mean that they withdrew. CONSORT, Consolidated Standards of Reporting Trials.

Table 2.

BRIEF T-Scores and CES-DC Score Distribution across Intervention Arms

	Total (n = 76)	Control (n = 22)	App (n = 28)	App+coach (n = 26)	p^a
BRIEF T-scores
BRI T-scores	57.42 ± 9.90	57.73 ± 10.07	57.5 ± 10.05	57.08 ± 10.00	0.97
ERI T-scores	59.30 ± 10.57	60.18 ± 11.34	58.86 ± 10.18	59.04 ± 10.69	0.89
CRI T-scores	60.01 ± 10.97	57.86 ± 11.61	60.39 ± 10.35	61.42 ± 11.22	0.52
GEC T-scores	59.93 ± 10.27	59.18 ± 10.98	60 ± 10.08	60.5 ± 10.23	0.90
CES-DC total scores	8.18 ± 5.15	7.91 ± 5.41	8.21 ± 5.26	8.38 ± 5.22	0.95

A breakdown of baseline T-scores and subscale scores for the BRIEF-2 and total CES-DC scores for all study participants by intervention arm.

Kruskal–Wallis test.

CES-DC, Center for Epidemiological Studies Depression Scale of Children; CRI, Cognitive Regulation Index; BRIEF, Behavior Rating Inventory of Executive Function; ERI, Emotion Regulation Index; GEC, global executive functioning composite.

Table 3.

Pearson Bivariate Correlation between CES-DC and BRIEF Scores

CES-DC scores	ρ	p
BRIEF
BRI T-score	0.44	<0.0001
ERI T-score	0.50	<0.0001
CRI T-score	0.47	<0.0001
GEC T-score	0.53	<0.0001

The bivariate correlation analysis between CES-DC scores and EF subscores.

EF, executive functioning.

Table 4.

Difference in Engagement as a Function of Demographic Characteristics within Each Intervention Arm

	Control (n = 22)			AppAlone (n = 28)			AppCoach (n = 26)
	Median (IQR)	CV (%)	P	Median (IQR)	CV (%)	p	Median (IQR)	CV (%)	p
Gender^a
Male	50 (33–100)	72	0.66	33.5 (17–78)	81	0.27	54.5 (26–71)	57	0.72
Female	83 (50–100)	54		16.5 (8.5–58)	97		64.5 (35.5–73)	47
Ethnicity^a
Hispanic	100 (50–100)	32	0.02	37.5 (9.5–87)	84	0.32	63 (34–73)	46	0.61
Non-Hispanic	17 (0–100)	138		18.5 (9–38.5)	83		64 (25–70)	58
Insurance^a
Government	83 (33–100)	64	0.58	34 (10–63)	85	0.437^a	63 (26–71)	52	0.50
Private	100 (50–100)	52		10 (8–53)	110		68 (46–91)	57
Parent education^b
≤High school	50 (33–83)	71	0.09	12 (8–63)	106	0.11	60.5 (31.5–73)	49	0.41
Trade/vocational	58.5 (17–100)	100		65.5 (34–97)	68		72 (11–74)	68
College	100 (100–100)	0		53 (21–89)	65		64.5 (46–71)	37
Graduate	58.5 (17–100)	100		—	—		—	—
Household income^b
<$50,000	50 (50–100)	67	0.47	20 (9–63)	96	0.76	65 (34–73)	49	0.44
$50,000–$149,999	100 (100–100)	45		41 (10–58)	87		58.5 (46–91)	44
≥$150,000	58.5 (17–100)	100		53 (53–53)	0		32 (0–64)	141
Self-report depression^a
Not depressed	100 (50–100)	55	0.2	32.5 (10–63)	85	0.49	64.5 (37–74)	52	0.54
Depressed	50 (17–83)	85		15.5 (9–34)	110		48.5 (29.5–72.5)	42
	β (95% CI)	P		β (95% CI)			β (95% CI)	p
BRIEF-2^b
BRI T-score	2.31 (−1.73 to 6.35)	0.24		0.20 (−2.24 to 2.64)	0.867		−0.55 (−2.09 to 1.00)	0.474
ERI T-score	−2.17 (−5.33 to 0.98)	0.166		−0.17 (−2.79 to 2.45)	0.897		−1.00 (−2.53 to 0.53)	0.191
CRI T-score	−2.24 (−5.18 to 0.70)	0.127		−1.35 (−3.30 to 0.60)	0.167		−0.29 (−2.01 to 1.44)	0.736
GEC T-score	−2.37 (−5.78 to 1.04)	0.162		−1.14 (−3.31 to 1.03)	0.292		−0.64 (−2.39 to 1.12)	0.462

The differences in engagement as a function of demographics, self-reported depressive symptoms, and EF scores within each intervention arm.

Wilcoxon Rank Sum test.

Kruskal–Wallis test.

CI, confidence interval; CV, coefficient of variation; IQR, interquartile range.

Percent Engagement

Percent engagement differed significantly across intervention arms (p = 0.02, Table 5). Specifically, participants assigned to the control group had the highest level of engagement (83%) compared with participants in the AppCoach (63.5%) and AppAlone (22.5%). There was no significant difference in total contact hours between the intervention arms. The AppAlone arm had 3.50 less contact hours compared with the control arm (95% CI = −8.74 to 1.73, p = 0.186) and AppCoach arm had 3.41 more contact hours compared with the control arm (95% CI = −2.06 to 8.87, p = 0.217). On univariate analysis between groups, Hispanic ethnicity was a positive predictor of engagement (p = 0.01, Table 5). With the multivariate regression analysis, this association between ethnicity as a positive predictor of engagement no longer remained (Table 6). The within-arm analysis showed that the association between Hispanic ethnicity and engagement was only present within the control arm (p = 0.02, Table 4). On univariate and multivariate analyses, age, gender, insurance status and household income were not significantly associated with percent engagement (Tables 5 and 6). There was no significant association between baseline BRIEF T-score (GEC T-score; β −1.00, 95% CI = −2.78 to 0.78, p-value 0.266) or self-reported depressive symptoms and percent engagement.

Table 5.

Engagement as a Function of Individual Characteristics and Intervention Arms

Categories	n (%)	% engagement		p
Categories	n (%)	Median (IQR)	Coefficient variation (%)	p
Intervention arm
Control	22 (29)	83 (33–100)	60	0.02^a
AppAlone	28 (37)	22.5 (9.5–63)	89
AppCoach	26 (34)	63.5 (34–73)	50
Gender
Male	31 (41)	50 (19–93)	69	0.99^b
Female	45 (59)	53 (17–83)	67
Ethnicity
Hispanic	50 (66)	63 (33–93)	57	0.01^b
Non-Hispanic	26 (34)	24.5 (10–65)	88
Insurance
Public	55 (72)	50 (17–86)	68	0.72^a
Private	16 (21)	58 (13.5–91.5)	68
Parent
High school or less	44 (58)	46.5 (9.5–73)	77	0.08^a
Education
Trade/vocational	7 (9)	72 (17–97)	64
College	22 (29)	66 (46–99)	45
Graduate degree	3 (4)	17 (0–100)	137
Household
<$50,000	50 (66)	50 (16, 74)	72	0.39^a
Income
$50,000–$149,999	16 (21)	64.5 (33, 99.5)	57
≥$150,000	5 (7)	53 (17, 64)	84
Self-reported
Not depressed	57 (75)	58 (18–92)	66	0.24^b
Depression
Depressed	19 (25)	34 (17–73)	71

A univariate analysis between groups and demographic and psychosocial characteristics.

Kruskal–Wallis test

Wilcoxon Rank Sum test

Table 6.

Multivariate Quantile Regression Model on Percent Engagement

	β	95% CI	P
Patient group
Control	Ref	—	—
AppCoach	−14.27	−51.32 to 22.78	0.444
AppAlone	−33.60	−74.09 to 6.89	0.102
Ethnicity
Non-Hispanic	Ref	—	—
Hispanic	30.25	−5.07 to 65.56	0.092
Gender
Male	Ref	—	—
Female	−15.47	−45.62 to 14.67	0.309
GEC T-score	−0.67	−2.55 to 1.22	0.480
CES-DC score	−0.94	−4.39 to 2.50	0.586
Age	1.99	−9.67 to 13.65	0.734
Insurance
Government	Ref	—	—
Private	30.02	−7.89 to 67.92	0.119

A multivariate analysis between intervention arms, demographic characteristics, and psychosocial characteristics.

Discussion

Treatment engagement and attendance are essential to effective weight loss interventions. Understanding the predictors of engagement can inform strategies to improve treatment outcomes. Predictors of engagement likely shift over individuals' life course as they transition through developmental milestones. The present study examined how baseline demographic factors, psychosocial characteristics, and intervention delivery were associated with engagement among adolescents enrolled in a weight loss intervention.

In contrast to findings in adult populations, EF and depressive symptoms were not detrimental to intervention engagement in this specific cohort of adolescents with overweight and obesity. Half of participants had BRIEF-2 composite scores suggestive of clinical impairments in EF, and one-fourth had CES-DC scores indicative of self-reported depressive symptoms, yet those adolescents were not less likely to complete the intervention requirements than their peers. It is unclear whether emotional and cognitive functioning simply represent greater barriers to engagement in adults than in adolescents, or whether the lack of association is unique to our sample. One possible explanation centers on the role of caregivers in weight management interventions. Adolescence may be characterized by increased autonomy, but parents are still heavily involved in their children's access to health services, including transportation to clinic appointments. Thus, caregivers and family members may provide additional contingencies and structure, independent of the adolescents' mood of cognitive function, thereby enhancing attendance to intervention requirements.

Consistent with our family structure hypothesis, engagement was the highest in the control group (in-person clinic model), which was designed to deliver a multidisciplinary clinical intervention meeting all current clinical practice guidelines for management of pediatric obesity. Participants in this arm spent prolonged face-to-face time, once a month, with 3–4 health care provides (physician, health educator, registered dietitian, and psychologist). These encounters established rapport, individualized goal setting, and harnessed a motivational interviewing approach. Notably, within-arm analysis revealed that Hispanic adolescents in the control arm demonstrated greater engagement than non-Hispanic participants. Specifically, parental involvement was required in this arm, as parents or guardians were required to attend the intervention visits regardless of the adolescent's mood or underlying cognitive functioning and this removes that factor as a barrier to engagement. Parental tangible (transportation) and emotional support may provide contingencies and structure, thereby increasing attendance to intervention visits. Although not directly assessed in this study, the value of familism central in many Hispanic cultures may further help explain differences in attendance between Hispanic and non-Hispanic participants. This difference is not sustained in a model considering multiple predictors, including age, gender, insurance type, EF, and CES-DC scores between groups and thus the sample size may be too low of power to detect multivariate differences. Further investigation is needed to elucidate the relationship between ethnicity and engagement in this type of intervention.

Between the two App-based intervention arms, engagement was highest among participants assigned to the AppCoach arm, which included frequent contacts with a health coach. Participants received weekly 15–20-minute calls from their health coach, who helped trouble-shoot barriers and promote adherence to intervention recommendations. This prescribed frequency of interaction in the AppCoach group was significantly more than that in the Clinic group; however, the percent engagement proved similar, further reinforcing the importance of social support in promoting engagement, regardless of the overall treatment dosage prescribed. Our findings are similar to those of other studies supporting the benefits of individualized interactions on engagement and app usage.^26–28 These findings suggest that family involvement and social support may be an essential component of effective interventions, regardless of whether they are delivered in-person or virtually using a mobile platform. This important finding is especially timely given the COVID-19 pandemic, which continues to drive new care delivery approaches, including working to optimize use of mobile platforms to promote sustained engagement and successful outcomes for chronic disease management.^26,40

Limitations

Specific study limitations must also be addressed. First, our design remains subject to omitted variable biases, such as unmeasured or uncontrolled factors that may actually drive the observed differences between groups. Second, the intervention engagement comparability is not exactly equivalent between intervention arms. Although this was defined post hoc, the percent engagement formulation with equal contact hours per arm provided a consistent way to discuss treatment arm dosage across groups. Third, our sample size may have precluded detection of other meaningful differences within each treatment arm. Fourth, further investigation on deeper aspects of engagement need to be considered beyond looking at app activity and frequency of use. In this cohort, we focused objectively defined metrics of engagement, including in-person attendance and recorded interactions with the app and health coach. However, we acknowledge that our operationalization of engagement fails to capture less tangible parameters of engagement. Finally, our study focused on treatment-seeking adolescents, and the results may not be generalizable to those adolescents with overweight/obesity not accessing treatment as they may have different levels of motivation and barriers to engagement than those who actively seek treatment. Nevertheless, it remains unproven whether the observed relationships among ethnicity, engagement, EF, and depressive symptoms further generalize to a community sample of nontreatment-seeking adolescents with overweight and obesity.

Conclusion

Overall, these results suggest that predictors of engagement may be different in adolescents with overweight and obesity compared with otherwise comparable adults seeking similar treatment. Family involvement and direct human contact, whether in-person or through mobile platform, appears to be the greatest positive predictor of engagement in this cohort of adolescents. Demographic characteristics previously associated with adults' treatment participation and attendance, such as household income and insurance status, were not associated with engagement in this sample of adolescents. This study adds to the growing literature analyzing how to improve engagement and adherence to specific weight management treatment strategies for adolescents. Those findings should help inform the design of innovative strategies that improve sustained behavior change and obesity treatment outcomes in this high-risk population.

Footnotes

Funding Information

This research was supported by a grant from eHealth International, Inc. and by NIH/NCRR SC-CTSI Grant Number UL1 TR000130 awarded to Dr. Vidmar.

Author Disclosure Statement

No competing financial interests exist.

References

Brewis

, Bruening

. Weight shame, social connection, and depressive symptoms in late adolescence. Int J Environ Res Public Health, 2018; 15:891.

Williamson

, Dilip

, Dillard

, et al. The influence of socioeconomic status on snacking and weight among adolescents: A scoping review. Nutrients, 2020; 12:167.

Burton

, Smith

, Thurston

, et al. Interdisciplinary management of pediatric obesity: Lessons learned in the Midsouth. Clin Pediatr, 2018; 57:509–518.

Small

, Aplasca

. Child obesity and mental health: A complex interaction. Child Adolesc Psychiatr Clin, 2016; 25:269–282.

Morrison

, Shin

, Tarnopolsky

, et al. Association of depression & health related quality of life with body composition in children and youth with obesity. J Affect Disord, 2015; 172:18–23.

Ponzo

, Scumaci

, Goitre

, et al. Predictors of attrition from a weight loss program. A study of adult patients with obesity in a community setting. Eat Weight Disord, 2020:1–8.

Demeule-Hayes

, Winters

, Getzoff

, et al. Pediatric weight management program outcomes in a largely minority, low socioeconomic status population. Clin Med Insights, 2016; 10:CMPed-S38457.

Williams

, Coday

, Somes

, et al. Risk factors for poor attendance in a family-based pediatric obesity intervention program for young children. J Dev Behav Pediatr, 2010; 31:705.

Chao

, Wadden

, Berkowitz

. Obesity in adolescents with psychiatric disorders. Curr Psychiatry Rep, 2019; 21:3.

10.

Jensen

, Aylward

, Steele

. Predictors of attendance in a practical clinical trial of two pediatric weight management interventions. Obesity, 2012; 20:2250–2256.

11.

Quek

, Tam

WWS

, Zhang

MWB

, et al. Exploring the association between childhood and adolescent obesity and depression: A meta-analysis. Obes Rev, 2017; 18:742–754.

12.

Diamantopoulou

, Verhulst

, Van Der Ende

. Gender differences in the development and adult outcome of co-occurring depression and delinquency in adolescence. J Abnorm Psychol, 2011; 120:644.

13.

Eichen

, Matheson

, Liang

, et al. The relationship between executive functioning and weight loss and maintenance in children and parents participating in family-based treatment for childhood obesity. Behav Res Ther, 2018; 105:10–16.

14.

Wyckoff

, Evans

, Manasse

, et al. Executive functioning and dietary intake: Neurocognitive correlates of fruit, vegetable, and saturated fat intake in adults with obesity. Appetite, 2017; 111:79–85.

15.

Pearce

, Leonhardt

, Vaidya

. Executive and reward-related function in pediatric obesity: A meta-analysis. Child Obes, 2018; 14:265–279.

16.

Yang

, Shields

, Guo

, et al. Executive function performance in obesity and overweight individuals: A meta-analysis and review. Neurosci Biobehav Rev, 2018; 84:225–244.

17.

Rodrigue

, Ouellette

A-S

, Lemieux

, et al. Executive functioning and psychological symptoms in food addiction: A study among individuals with severe obesity. Eat Weight Disord, 2018; 23:469–478.

18.

Mora-Gonzalez

, Esteban-Cornejo

, Cadenas-Sanchez

, et al. Physical fitness, physical activity, and the executive function in children with overweight and obesity. J Pediatr, 2019; 208:50–56.

19.

Brockmeyer

, Sinno

, Skunde

, et al. Inhibitory control and hedonic response toward food interactively predict success in a weight loss programme for adults with obesity. Obes Facts, 2016; 9:299–309.

20.

Hayes

, Eichen

, Barch

, et al. Executive function in childhood obesity: Promising intervention strategies to optimize treatment outcomes. Appetite, 2018; 124:10–23.

21.

Gettens

, Gorin

. Executive function in weight loss and weight loss maintenance: A conceptual review and novel neuropsychological model of weight control. J Behav Med, 2017; 40:687–701.

22.

Mac Giollabhui

, Swistun

, Murray

, et al. Executive dysfunction in depression in adolescence: The role of inflammation and higher body mass. Psychol Med, 2019; 50:683–691.

23.

Rock

, Roiser

, Riedel

, et al. Cognitive impairment in depression: A systematic review and meta-analysis. Psychol Med, 2014; 44:2029.

24.

Connolly

, Wu

, Ho

, et al. Resting-state functional connectivity of subgenual anterior cingulate cortex in depressed adolescents. Biol Psychiatry, 2013; 74:898–907.

25.

Styne

, Arslanian

, Connor

, et al. Pediatric obesity—Assessment, treatment, and prevention: An Endocrine Society Clinical Practice guideline. J Clin Endocrinol Metab, 2017; 102:709–757.

26.

Rice

, Jumamil

, Jabour

, et al. Role of health coaches in pediatric weight management: Patient and parent perspectives. Clin Pediatr, 2017; 56:162–170.

27.

Pellegrini

, Duncan

, Moller

, et al. A smartphone-supported weight loss program: Design of the ENGAGED randomized controlled trial. BMC Public Health, 2012; 12:1041.

28.

Alencar

, Johnson

, Gray

, et al. Telehealth-based health coaching increases m-health device adherence and rate of weight loss in obese participants. Telemed J E Health, 2020; 26:365–368.

29.

Dhaliwal

, Nosworthy

NMI

, Holt

, et al. Attrition and the management of pediatric obesity: An integrative review. Childhood Obes, 2014; 10:461–473.

30.

Skelton

, Goff Jr

, Ip

, et al. Attrition in a multidisciplinary pediatric weight management clinic. Child Obes, 2011; 7:185–193.

31.

Lin

P-H

, Grambow

, Intille

, et al. The association between engagement and weight loss through personal coaching and cell phone interventions in young adults: Randomized controlled trial. JMIR Mhealth Uhealth, 2018; 6:e10471.

32.

Vidmar

, Salvy

, Pretlow

, et al. An addiction-based mobile health weight loss intervention: Protocol of a randomized controlled trial. Contemp Clin Trials, 2019; 78:11–19.

33.

Baron

. Behavior rating inventory of executive function. Child Neuropsychol, 2000; 6:235–238.

34.

Roth

, Erdodi

, McCulloch

, et al. Much ado about norming: The Behavior Rating Inventory of Executive Function. Child Neuropsychol, 2015; 21:225–233.

35.

Rouel

, Raman

, Hay

, et al. Validation of the Behaviour Rating Inventory of Executive Function-Adult Version (BRIEF-A) in the obese with and without binge eating disorder. Eat Behav, 2016; 23:58–65.

36.

Faulstich

, Carey

, Ruggiero

, Enyart

, Gresham

. Assessment of depression in childhood and adolescence: An evaluation of the Center for Epidemiological Studies Depression Scale for Children (CES-DC). Am J Psychiatry, 1986; 143:1024–1027.

37.

Cosco

, Prina

, Stubbs

, et al. Reliability and validity of the Center for Epidemiologic Studies Depression Scale in a population-based cohort of middle-aged US adults. J Nurs Meas, 2017; 25:476–485.

38.

Berganza

, Aguilar

. Depression in Guatemalan adolescents. Adolescence, 1992; 27:771.

39.

Essau

, Olaya

, Pasha

, et al. Depressive symptoms among children and adolescents in Iran: A confirmatory factor analytic study of the Centre for Epidemiological Studies Depression Scale for Children. Child Psychiatry Hum Dev, 2013; 44:123–136.

40.

Pellegrini

, Pfammatter

, Conroy

, et al. Smartphone applications to support weight loss: Current perspectives. Adv Health Care Technol, 2015; 1:13.