Weight-related quality of life and temperament as predictors and moderators of outcomes among treatment-seeking, low-income, ethnically diverse children with obesity

Children with obesity who avoided social interaction due to their weight had significantly better weight-related outcomes in a comprehensive clinic-community treatment program compared to a clinic-only program. Practitioners may need to consider children's weight-related quality of life in their decisions about treatment recommendations and carefully counsel children experiencing weight negatively.

Abstract

Within any childhood obesity treatment program, some children have better outcomes than others. Little is known about predictors or moderators of more positive outcomes. We aimed to identify whether child temperament and weight-related quality of life predict or moderate childhood obesity treatment outcomes at 6 months. From 2015 to 2016, children (n = 97) ages 5–11 years old with obesity were randomized to a clinic–community (Bull City Fit) or a clinic-only treatment program. Linear regression was used to explore whether dimensions of child temperament and weight-related quality of life predicted or moderated 6-month anthropometric and physical activity outcomes. Children who had more social avoidance due to their weight at baseline had significantly better improvements in body fat percent in the clinic–community model compared with the clinic-only model at 6 months. Across programs, better baseline social quality of life predicted greater increases in waist circumference; conversely, better physical quality of life predicted a decrease in percent of the 95th percentile. Also, children with longer attention spans had greater increases in physical activity. Our findings suggest that children who have social avoidance due to their weight may benefit most from comprehensive clinic–community treatment. Weight-related quality of life may influence outcomes across all treatments, and practitioners need to carefully counsel children experiencing weight negatively.

Implications

Practice: Childhood obesity treatment providers may need to screen children for issues of focus impairment and weight-related social avoidance to inform counseling and treatment plans.

Policy: Policymakers should consider expanding childhood obesity treatment options in low-income, racially diverse communities that involve partnerships between clinical and community-based agencies.

Research: Future research should examine if matching children based on temperament and weight-related social avoidance to more comprehensive interventions improves efficiencies and optimization of outcomes.

INTRODUCTION

Nearly one in five children in the USA have obesity, and the burden is greatest among low-income, African American, and Hispanic children [1]. Unfortunately, the prevalence of childhood obesity does not show signs of abating [1], and advances in treatment are needed. Comprehensive nutrition and physical activity treatment programs that simultaneously engage families in clinical and community settings are considered the gold standard compared with clinical programs with lower frequency and shorter duration that have been shown to be less effective [2]. However, children within and across different types of treatment have varied outcomes, thereby creating gaps in knowledge about factors that predict (i.e., affect outcomes across all types of treatment) or moderate outcomes (i.e., affect how much a child benefits from receiving a specific treatment). Knowledge about predictors and moderators can elucidate strengths and gaps in treatment strategies, help practitioners match children to optimal treatment models, and potentially result in improved outcomes.

Prior research has identified several factors that predict outcomes. Higher parent body mass index (BMI), parent psychological conditions, and poor family functioning predict less improvement in child obesity-related outcomes [3–6], highlighting the importance of including family members in interventions. Other predictors of improved outcomes include entering treatment with a lower BMI, having greater early response to treatment, younger age, female sex, White race, and coming from a family with higher socioeconomic status [7–12]. These demographic factors are difficult to intervene on, and therefore, more understanding is needed of other individual predictors and moderators, especially among low-income and racial and ethnic minorities.

Our study draws on theory and research of temperament that include strong self-regulatory components and point to underlying behavioral mechanisms that could be simultaneously addressed within obesity treatment programs [13]. Temperament is a set of characteristics that reflect predispositions to behave in a certain way [14], and research has shown that these predispositions can drive self-regulation processes [15]. For example, laboratory studies have shown that children who exhibit higher levels of self-regulation when completing effortful tasks (e.g., block tower building) are less likely to be rated by their parents as having negative affectual and impulsive temperaments [16,17].

Research has also shown relationships exist between temperament and weight and dietary behaviors among children and indicate that certain dimensions of temperament can increase risk for the development of childhood obesity [16,18–20]. Extensive research has documented a relationship between temperament and physical activity [21] and highlighted the critical role that temperament-related constructs such as self-regulation may play in moderating the link between physical activity intentions and behavior [22]. Most of this research has been among adolescents and adults. Among children, one study found that activity-related temperament was a better predictor of weight status than accelerometer-measured activity and concluded that the temperament construct possibly captured important components of activity and energy expenditure [23]. These studies highlight the relevance of temperament constructs to childhood obesity, but little research has considered if dimensions of temperament predict more or less positive outcomes among children in weight-loss treatment programs [7,24]. In treatment, children and their parents are often counseled to change eating and physical activity behaviors. The response to such counseling could be directly altered by temperament characteristics such as the intensity of and ability to monitor and regulate emotional reactions. For example, a child who has higher self-regulation may be more able to resist sedentary activities and replace them with more physical activity than a child with lower self-regulation.

Issues of quality of life may also be important to examine among children affected by obesity since they often face stigmatization. Stigmatization places them at risk for psychosocial challenges such as low self-esteem, which has been found to predict poorer outcomes [7]. More specific to body weight than self-esteem, children’s perceptions about how their size influences their social and physical functioning, that is, weight-related quality of life [25], may influence their interest and ability to make lifestyle changes and participate in treatment. For example, children who are avoiding social interactions because of their weight may find it difficult to pursue physical activity opportunities that often occur in group and social formats.

Few studies have identified temperament and quality of life factors that predict treatment outcomes among low-income, racial, and ethnic minority children, and scholars have noted that knowledge about moderators of outcomes is a limitation of the field of obesity treatment [2,26]. To our knowledge, only one prior study exists that identified factors that moderated (as opposed to predicted) effects of treatment. The study examined family functioning and child self-esteem as moderators of outcomes in a family-based nutrition and physical activity program versus a clinic-only treatment model [7].

We conducted a randomized trial of a clinic–community treatment model that combines clinical care with community-based physical activity and nutrition programing for children with obesity and their families (Bull City Fit) versus a clinic-only tertiary care treatment model [27]. The results of the traditional trial evaluation found that participants in the clinic–community treatment arm had improved changes in physical activity and waist circumference compared with participants in the clinic-only arm. However, we were interested in learning more about how temperament and quality of life related to study outcomes (predictive analyses) and how temperament and quality of life were differentially related to study outcomes in each treatment group (moderation analyses). The trial data provide an ideal opportunity to explore predictors and moderators because both groups received active treatment. Using Bull City Fit trial data, the objective of this study is to conduct an exploratory analysis to identify child temperament and weight-related quality of life factors that predict or moderate anthropometric and physical activity outcomes. We hypothesized that more positive child temperament and weight-related quality of life (i.e., higher scores on measurement scales) would predict decreases in anthropometric and increases in physical activity outcomes and that participants with more negative temperament and weight-related quality of life (i.e., lower scores on measurement scales) would have better outcomes in the clinic–community treatment model compared with clinic-only treatment model.

METHODS

The Bull City Fit participants, study design, treatment groups, and outcomes have been detailed previously [27]. The institutional review boards from Duke University (Pro00066366) and the University of North Carolina at Chapel Hill (15–1867) approved the study protocol (ClinicalTrials.gov Identifier: NCT02573142).

Participants

Participants were a consecutive sample of patients aged 5–11 years with a BMI ≥ 95th percentile, referred by their primary care provider to a pediatric obesity treatment clinic, along with each child’s primary caregiver (“parent”) aged 18 and older. Of the 100 child and parent dyads enrolled in the study (n = 50 clinic-only, n = 50 clinic–community), 68% (n = 68) completed the study and are included in this analysis. Participant enrollment occurred between October 29, 2015 and August 15, 2016, and we collected 6-month follow-up data at the end of the intervention (through March 13, 2017).

Study design

The primary study used a two-group, randomized controlled trial to compare a clinic–community model (i.e., Bull City Fit) to a clinic-only treatment model.

Clinic-only model

Participants met once per month with a multidisciplinary team (medical, nutrition, physical therapy, and mental health) to set and monitor lifestyle behavioral goals and manage health conditions. Additional details have been previously published [28].

Clinic-community model (Bull City Fit)

Participants received standard clinical care (i.e., the clinic-only intervention) plus an invitation to a free community-based program at a local recreation center. The program, Bull City Fit, is open 6 days per week, from 6 pm to 8 pm on weekdays and 1 pm to 3 pm on weekends. Participants are recommended to attend at least two sessions per week and can attend as many of the 6 weekly sessions as desired. At least one adult household member must attend with the child, and all other household members are welcome to also attend and fully participate. The sessions follow recommended evidence-based practices that include a family focus and developmental stage-appropriate activities that are both suitable and enjoyable [29,30]. During the sessions, trained staff supervise and facilitate structured games, team-building sports, cooking and swimming lessons, and peer support activities. As detailed in a prior publication, participation in clinical appointments and Bull City Fit sessions was tracked electronically. The clinic-community arm achieved an increase in treatment hours over the 6-month period (11.7 vs. 4.4 hrs in the clinic-only group), which is consistent with adherence reported in other studies [27]. Additional details have been previously published [27,31].

Anthropometric and physical activity outcome measures

We assessed child outcomes of waist circumference, body fat percent, percent of the 95th percentile, and physical activity. Trained research staff collected measures at baseline and 6 months by administering surveys or abstracting data from the clinical record. A nurse collected anthropometric measures, including waist circumference, body fat percent, height, and weight using standard practices. Height and weight were measured with a stadiometer (Health o meter Professional CE#92977; Health o meter, McCook, IL) and digital scale (Seca CE#96990; Seca, Chino, CA) and used to calculate percent of the 95th percentile [32]. We also evaluated physical activity as an outcome because Bull City Fit has a strong emphasis on active play. Physical activity was measured using a modified version of a Physical Activity Questionnaire for Children (PAQ-C) [33], a 7-day recall instrument administered using a parent-assisted interview with the child. The PAQ-C has internal reliability (Cronbach’s alpha range of .72 to .88) [34] and convergent validity with alternate measures of moderate to vigorous physical activity (r = .63) [35]. The instrument was scored to provide a summary physical activity score derived from nine items.

Predictor and moderator variables

Child temperament

Using a parent survey (Colorado Child Temperament Inventory), we assessed temperament along six dimensions: (i) sociability—the tendency to connect with others and respond to social stimuli—children high on this scale are generally more comfortable interacting with others, (ii) emotionality—how quickly a child becomes agitated or reacts negatively to environmental stimuli—children with high emotionality tend to cry easily and be more fearful, (iii) activity—level of physical energy and output—children high on this scale tend to explore and enjoy physical activity and games, (iv) attention span-persistence—the extent that a child stays with an activity for a long time and can stay on task through frustrations—children high on this scale tend to be able to focus and attend to relevant information, (v) reaction to food—level of a child’s neophobia related to food, and (vi) soothability—extent that a child can be calmed after unexpected or frustrating events—children who are higher in soothability tend to be easier to console [36,37]. Each dimension consists of five items (e.g., “my child gives up easily when difficulties are encountered”) that parents rated on a 5-point Likert type scale (1 = “not at all like my child”; 5 = “a lot like my child”). Scores were calculated if an individual answered at least 75% of the items for each dimension. Higher scores in each temperament dimension indicate greater sociability, emotionality, activity, attention span-persistence, food reactivity, and soothability. The measures are reported to have strong internal consistency (α = .70–.88) [36] and convergent validity when examined against other commonly used temperament questionnaires (r = .50–.75) [38]. Our sample had moderate internal consistency for measures of sociability and activity (α = .62 and .59, respectively), and strong internal consistency for emotionality, reaction to food, and soothability (α = .84, .81, and .75, respectively). The attention span-persistence measure had lower internal consistency (α = .27). On further investigation, the three negatively phrased items that required reverse-coding for this measure were problematic. We removed these items from the attention span-persistence subscale, which improved internal consistency (α = .38), though it remained low.

Child weight-related quality of life

Sizing Me Up [25] is an obesity-specific instrument that measures a child’s perception of how his or her body size influences his or her personal, physical, and social functioning, and the instrument was administered using a child interview. The instrument assesses a child’s perception of how his or her body size affects five dimensions: (i) emotional functioning—feelings and emotions, (ii) teasing/marginalization—whether they were teased or left out due to their size, (iii) physical functioning—ability to keep up with physical activities and perceptions of “fitting in” while being physically active, (iv) social avoidance—comfort in and avoidance of social activities, and (v) positive attributes—positive qualities and strengths. Each dimension consists of two to six items (e.g., “during the past month, tell us how much you felt left out because of your size”) that children rated on a 4-point Likert type scale (1 = “none of the time” to 4 = “all the time”). Scores were calculated if an individual answered at least 75% of the items for each dimension. Higher scores in each dimension indicate better emotional functioning, less teasing/marginalization, better physical functioning, less social avoidance, and more positive attributes. The measures are reported to have good internal consistency (α = .68–.85) and convergent validity with the PedsQL (r = .35–.65). Our sample had moderate internal consistency for teasing/marginalization and positive attributes (α = .53 and .67, respectively) and strong internal consistency for emotional, physical, and social avoidance measures (α = .85, .70, and .73, respectively).

Analysis

To identify potential predictors, we used linear regression models and regressed each of the 6-month outcome variables on each of the baseline child temperament and quality of life dimensions. All models included baseline outcome measure and treatment arm. To identify potential moderators of our treatment models, we included an interaction term between the treatment arm and each potential moderator variable (i.e., baseline child weight-related quality of life and temperament dimensions). We adjusted models by including covariates of parent race, child gender, and parent age. To aid interpretation of predictors, we created figures that plotted the regression line for each model, that is, change in the outcome by each predictor variable. To aid interpretation of significant moderators, we created figures that plotted the change in the outcome by the moderating variable level for each treatment arm. We defined the familywise error rate based on different tests of the same hypothesis [39]. Thus, we considered independent hypotheses for each of the quality of life and temperament dimensions as potential predictors or moderators. We used a Bonferroni adjusted alpha level of .0167 for the three different tests regarding anthropometric outcomes (.05/3) and an alpha of .05 for physical activity. All analyses were completed in SAS version 9.4.

RESULTS

As reported previously, participant demographic characteristics did not significantly differ between the randomized groups [27]. Participants came from families with low socioeconomic status (68% had incomes below $50,000 per year and 44% of parents had less than a college degree). Approximately half of the children came from single-parent households and lived a mean of 8.6 (SD = 6.1) miles from the Bull City Fit location. The participants who completed the study (n = 68) were approximately equal numbers of girls (52.9%) and boys (47.1%). The majority of completers were African American (44.1%) and Hispanic (42.7%). Approximately one third (27.8%) had a household income of less than $5,000 per year. Study completers had lower baseline physical activity scores compared with individuals who did not complete the study. There were no other significant differences in demographics or baseline outcomes between study completers and noncompleters.

Child temperament

Figure 1 provides the anthropometric and physical activity outcome changes for all temperament predictors. We plot all combinations of temperament and outcomes to illustrate overall trends, and significant predictive relationships are indicated with a black line. Table 1 provides additional details of all coefficients, standard errors (SE), and p values from regression models that assessed dimensions of child temperament as a predictor and moderator. Having a higher attention span predicted a 0.19 (SE = 0.07) unit increase in physical activity (p < .05).

Fig 1.

| Plots of child temperament by obesity-related outcomes. Higher values of child temperament indicative of more positively viewed temperaments (e.g., increase on emotionality scale indicates less emotionally reactive; increase in reaction to food means less picky). Relationships that were significant in models adjusted for parent age, parent race, and child gender are shown in black, and relationships that were insignificant are shown in gray. Vertical axis units differ by outcome. Note: PAQ = physical activity questionnaire.

Table 1.

| Effect of dimensions of child temperament as a predictor of and as a moderator of treatment model on obesity treatment outcomes

Temperament dimensions	Waist circumference (cm)		Body fat percent		% of the 95th percentile		Physical activity score
	B (SE)	p	B (SE)	p	B (SE)	p	B (SE)	p
Predictor models
Sociability	−0.64 (1.21)	.6009	−0.27 (0.6)	.6519	1.52 (1.28)	.2414	0.09 (0.11)	.4033
Emotionality	−0.56 (0.79)	.4849	0.03 (0.43)	.9525	0.63 (0.91)	.4949	−0.1 (0.08)	.2019
Activity	0.14 (1.5)	.9261	−0.84 (0.72)	.2508	−0.14 (1.58)	.932	−0.06 (0.14)	.6866
Attention span	−0.05 (0.87)	.9536	−0.20 (0.45)	.6549	−0.46 (0.96)	.6349	0.19 (0.07)	.0134*
Soothability	−1.62 (1.16)	.1714	0.09 (0.6)	.8763	2.39 (1.26)	.0649	0.15 (0.11)	.2091
Reaction to food	0.74 (0.85)	.3884	0.07 (0.45)	.8707	−0.21 (0.97)	.8333	0.01 (0.09)	.8694
Moderator models
(Coefficients represent change in response based on moderating variable for clinic–community relative to clinic-only treatment.)
Sociability	1.56 (2.59)	.5512	1.4 (1.25)	.2667	0.67 (2.66)	.8021	−0.19 (0.24)	.4294
Emotionality	−0.79 (1.64)	.6349	−0.95 (0.87)	.2824	−1.71 (1.84)	.3572	−0.13 (0.15)	.393
Activity	−5.24 (2.58)	.0491	2.45 (1.34)	.073	−0.27 (2.97)	.928	0.2 (0.25)	.4144
Attention span	−1.45 (1.88)	.4468	−0.04 (1.01)	.9684	−2.07 (2.11)	.3317	−0.02 (0.18)	.9223
Soothability	−2.3 (2.43)	.3499	2.21 (1.23)	.0784	−0.92 (2.67)	.7322	−0.18 (0.24)	.443
Reaction to food	0.37 (1.69)	.8295	−1.35 (0.88)	.1319	−1.93 (1.9)	.3172	−0.13 (0.17)	.4426

Higher values of child temperament indicative of more positively viewed temperaments (e.g., increase on emotionality scale indicates less emotionally reactive; increase in reaction to food means less picky); models are adjusted for intervention arm, parent race, child gender, and parent age; B presented for moderator models is the coefficient of the interaction term.

*Significant, p < .0167 for anthropometric outcomes after Bonferroni adjustment, p < .05 for physical activity outcome.

Temperament dimensions were not found to be significant moderators of treatment on outcomes.

Child weight-related quality of life

Figure 2 provides the anthropometric and physical activity outcome changes for all weight-related quality of life predictors. We plot all combinations of quality of life and outcomes to illustrate overall trends, and significant predictive relationships are indicated with a black line. Table 2 provides the coefficients, SE, and p values from regression models that assessed dimensions of weight-related quality of life as a predictor and moderator. The baseline weight-related quality of life dimension of teasing/marginalization was a significant predictor of increased waist circumference (p = .0045). Each one-unit increase in a child’s teasing/marginalization score (i.e., less teasing/marginalization) at baseline predicted a 1.31 (SE = 0.44) centimeter increase in waist circumference at 6 months. Conversely, physical quality of life (i.e., ability to keep up with physical activities) was a significant predictor of decreased percent of the 95th percentile (p < .0167). Each one-unit increase in physical quality of life at baseline predicted a 0.65 (SE = 0.26) decrease in percent of the 95th percentile in the adjusted model.

Fig 2.

Plots of weight-related quality of life by change in obesity-related outcomes. Child quality of life dimensions were standardized to mean 0 for comparability in graphs. Higher values of weight-related quality of life indicate more positive quality of life (e.g., increase in emotional dimension means less negative emotional problems related to weight). Relationships that were significant in models adjusted for parent age, parent race, and child gender are shown in black, and relationships that were insignificant are shown in gray. Vertical axis units differ by outcome. Note: QoL = quality of life; PAQ = physical activity questionnaire.

Table 2.

| Effect of dimensions of child quality of life as a predictor of and as a moderator of treatment model on obesity treatment outcomes

Quality of life dimensions	Waist circumference (cm)		Body fat percent		% of the 95th percentile		Physical activity score
	B (SE)	p	B (SE)	p	B (SE)	p	B (SE)	p
Predictor models
Emotional	0.52 (0.24)	.0339	0 (0.13)	.9948	−0.27 (0.27)	.3161	−0.03 (0.03)	.3155
Physical	0.2 (0.25)	.4279	−0.04 (0.13)	.7858	−0.65 (0.26)	.0166*	−0.04 (0.03)	.0868
Teasing	1.31 (0.44)	.0045*	−0.07 (0.26)	.7959	−0.58 (0.57)	.3104	−0.06 (0.05)	.2212
Positive attributes	0.24 (0.23)	.3034	−0.17 (0.12)	.1799	−0.06 (0.27)	.8151	0.02 (0.02)	.4262
Social avoidance	0.19 (0.33)	.5625	−0.09 (0.19)	.6475	−0.15 (0.39)	.7017	−0.03 (0.05)	.5762
Moderator models
(Coefficients represent change in response based on moderating variable for clinic–community relative to clinic-only treatment—see Fig. 3 to aid interpretation of significant factor.)
Emotional	0.54 (0.5)	.2899	0.19 (0.26)	.4725	0.19 (0.56)	.731	0.05 (0.06)	.3685
Physical	0.73 (0.5)	.1522	0.16 (0.27)	.5656	0.18 (0.54)	.7445	0 (0.06)	.9829
Teasing	−0.14 (0.96)	.8826	0.47 (0.53)	.3794	0.8 (1.12)	.4798	0.07 (0.1)	.5138
Positive attributes	0.27 (0.48)	.5736	0.08 (0.26)	.7468	0.14 (0.56)	.8084	−0.01 (0.05)	.8956
Social avoidance	1.11 (0.71)	.1265	0.95 (0.37)	.0132*	1.34 (0.82)	.1104	0.1 (0.12)	.3877

Higher values of weight-related quality of life indicative of more positive quality of life (e.g., increase in emotional dimension means less negative emotional problems related to weight); models are adjusted for intervention arm, parent race, child gender, and parent age; B presented for moderator models is the coefficient of the interaction term.

*Significant, p < .0167 for anthropometric outcomes after Bonferroni adjustment, p < .05 for physical activity outcome.

One weight-related quality of life dimension, social avoidance, was a statistically significant moderator of body fat percent in the adjusted model (p < .0167, Fig. 3). Children who had poorer weight-related social avoidance had significantly smaller increases in body fat percent in the clinic–community model compared with children with similarly poor social avoidance scores in the clinic-only model. Conversely, children who were less socially avoidant (higher scores) had unimproved outcomes in both treatment models. More specifically, among clinic-only participants, a one-unit worsening in social avoidance led to an increase in body fat percent by 0.65, whereas among clinic–community participants, a one-unit worsening in social avoidance led to a decrease in body fat percent by 0.29.

Fig 3.

Plot of change in body fat percentage by baseline social avoidance for the clinic-community treatment model and clinic-only treatment model. Baseline social avoidance was a significant moderator of treatment type on body fat percent outcome. Circles indicate data points for participants in the clinic–community treatment model and crosses indicate data points for participants in the clinic-only treatment model.

DISCUSSION

In a largely low-income, racially diverse population, we found evidence that dimensions of temperament and weight-related quality of life influenced and altered treatment outcomes. Our findings suggest that clinic–community treatment models may be especially valuable for children who have been avoiding social interactions due to their size. Children with lower attention span-persistence scores had poorer outcomes across treatment compared to their peers with higher attention span-persistence, which suggests a potential mechanism that could be addressed to improve outcomes. Also, weight-related quality of life influenced waist circumference and percent of the 95th percentile across types of treatment, which suggests it may provide practitioners information that could guide nonstigmatizing counseling to children with obesity.

We found that increased attention span-persistence predicted increased physical activity outcomes across treatment models. No other studies to our knowledge have assessed this specific characteristic as a predictor, but relatedly, a few studies have found that children with less impulsive temperaments and a better ability to delay future gratification had better response to treatment [24,40]. Other research has suggested that attention span-persistence is one component of self-regulation, which has been connected to many health behaviors [41]. More specifically, research has shown that the attentional network in the brain is activated during tasks that require focus and attention and is important for regulation of behavior [42,43]. Although parents do clearly play a role in regulating child physical activity, evidence has also shown that children are rapidly developing self-regulatory abilities that can influence their physical activity patterns [44,45]. Higher attention span-persistence may allow children to better monitor their own behaviors and persist in making behavioral changes and may be an important factor for children to positively respond to currently available treatment options. We caution, however, that the measure of attention span-persistence had relatively low internal consistency among our sample, and further research is needed to confirm our exploratory findings.

We found that children who felt more negatively about how their weight influenced their emotional and social functioning had smaller increases in waist circumference compared with their peers who had higher weight-related quality of life, which was contrary to our hypothesis. It has been proposed that some degree of body dissatisfaction may motivate weight control behaviors [46], and it is possible that children who had more negative perceptions about their weight were more motivated to lose weight. However, a large body of evidence has accumulated that indicates that negative experiences and an internalization of weight-based stigmatization may lead to detrimental and nonsustainable lifestyle changes [47,48]. We also found that children who had better physical quality of life had a greater decrease in their percent of the 95th percentile. It is possible that children who have fewer barriers to physical movement were more able to participate.

Overall, our findings highlight a need for treatment providers to better understand how children’s weight-related quality of life influences their motivations and help children and their parents set healthy goals. Children who report lower physical quality of life may benefit from additional consultation with a physical therapist who can prescribe individualized activity plans that are feasible, comprehensive, and pleasant [49]. Children with obesity who have high emotional and social functioning related to their weight may need validation and support to maintain these positive feelings while also receiving counsel about how healthy changes to diet and physical activity could improve other aspects of their life (e.g., short-term emotional benefits, decreased risk for long-term health problems). In contrast, children with lower social and emotional quality of life may need specific counseling to improve positive body image, and a focus on healthy behaviors and removal of environmental barriers may be favorable compared with body size-related treatment goals.

Finally, we found that children who reported more weight-related social avoidance performed better in the clinic–community treatment model than in the clinic-only model, whereas children with less social avoidance had similar outcomes across both models (Fig. 3). One past study evaluated moderators between a family-based and standard clinical model and found that children with low self-esteem (a construct probably associated with social avoidance) had better outcomes in a comprehensive compared with clinical care treatment program, whereas children with higher self-esteem had similar outcomes [7]. Clinic–community and family-based models such as Bull City Fit are designed to provide children affected by obesity a supportive and nonjudgmental environment for engaging in healthy activities. This type of model may fulfill an important need for children with psychosocial challenges due to their weight.

This study has several limitations. We had a small sample size, which reduces our statistical power to detect significant effects. The internal consistency of several measures was lower in our sample than found in the measurement tools’ validation studies, possibly due to lower literacy among our sample as well as the small sample size. Participants dropped out of the study, and there were missing data due to incomplete answers, which may bias our findings. Specifically, we found that study completers had lower baseline physical activity scores than participants who dropped out, and thus, some of the significant effects related to physical activity may be overestimated. However, the analyses still provide a valid “treatment on the treated” effect. Finally, physical activity was self-reported, which is subject to social desirability bias.

Moving forward, dimensions of temperament and weight-related quality of life deserve further exploration to improve how we match children to treatment, which is especially important among low-income and underserved populations. Our finding that the previously validated scale of temperament dimensions performed less strongly on some and poor on one construct among our sample indicates the need to advance measurement research of temperament among low-literacy and non-English speaking populations. Further research is needed to assess and confirm factors of temperament that influence childhood obesity treatment outcomes, and we need quality instruments to measure these dimensions. More research should examine if matching children based on temperament and weight-related quality of life to more comprehensive interventions improves optimization of outcomes. Also, providers may need to screen for attention deficit issues and consider additional treatment strategies for children with lower attention span-persistence. Future research with adaptive trial designs [50] is a potentially efficient way to improve our understanding of treatment options. This type of design could assign children to increasing intensity of treatment based on response or allow for adaptions based on specific characteristics.

Research has demonstrated that quality of life is lower among children with obesity and reductions in quality of life adversely affect weight changes [47,48]. Treatment should be an opportunity to address emotional, social, and physical concerns children may have due to their weight. Motivational interviewing with children and their parents is a recommended strategy for obesity treatment programs [51] and is a patient-centered approach to help increase motivation and commitment for change. However, additional research is needed to better understand how to best guide individuals who may be motivated due to negative and stigmatizing experiences, as well as what types of goals (weight, diet, family meals, etc.) produce favorable and sustainable physical and emotional health outcomes.

We found new evidence about potential predictors and moderators of treatment outcomes. As the prevalence of childhood obesity remains high and demand for treatment will probably only increase, advances are critically needed in treatment. Our findings indicate factors of child temperament and weight-related quality of life are worth further research and may be useful to guide providers of obesity treatment programs.

Compliance with Ethical Standards

Conflict of Interest: The authors have indicated they have no potential conflicts of interest.

Ethical Approval: All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.

Welfare of Animals: This article does not contain any studies with animals performed by any of the authors.