Longitudinal Associations Between Teasing and Health-related Quality of Life Among Treatment-seeking Overweight and Obese Youth

Chad D. Jensen; PhD ; Ric G. Steele; PhD; ABPP


J Pediatr Psychol. 2012;37(4):438-447. 

In This Article



Ninety-three overweight or obese children and adolescents (ages 7–17 years; mean BMI percentile = 98.2) who participated in a randomized clinical trial of a behavioral/educational pediatric weight management program and their participating parent/guardian (i.e., parent–child dyad) comprised the study sample. Participants were recruited through pediatric medical clinics, school nurses in public primary and secondary schools, flyers posted in community centers, and advertisements in newspapers. All participants attended either public or private Midwestern elementary or secondary schools. Eligibility criteria for participation in the study included: (a) the participating child or adolescent was between the ages of 7–17, (b) the participant's BMI percentile was categorized as overweight (i.e., BMI ≥ 85th percentile) or obese (i.e., BMI ≥ 95th percentile), (c) one parent/guardian participated in the intervention, (d) the participant had no serious mental illness or developmental delay, (e) the parent and child spoke English, (f) the parent provided written informed consent, and (g) the child verbally assented to participation. Participants who met enrollment eligibility criteria were stratified by age (i.e., ages 7–12 years; ages 13–17 years) and randomly assigned into intervention (Positively Fit) and control groups (Brief Family Intervention [BFI]) in blocks of 4–7 families using a random number generator.

Positively Fit This manualized intervention (Steele et al., n.d.) was comprised of 10 90-min weekly group treatment sessions held over conducted separately for parents and children. Groups ranged in size from 4 to 8 families per group and separate groups were held for children (7–12 years) and adolescents (13–17 years). Of particular interest to the present study, one session directly addressed bullying and teasing from peers. Children were instructed about appropriate actions to deter future victimization and cognitive and behavioral strategies intended to reduce the effects of victimization were presented. Steele et al. (2011) reported that this intervention produced favorable reductions in BMI percentile and increases in HRQOL among children.

Brief Family Intervention Participants randomly assigned to the BFI intervention participated in the Trim Kids manualized treatment program (Southern, von Almen, & Schumacher, 2002). Consistent with recommendations made by the authors of this program, participants received 36-min individual face-to-face visits with a licensed nutritionist within a 10-week period. Families in this condition received the Trim Kids manual at initial (pretreatment) assessment and were instructed to read the first four book chapters prior to their first meeting with the nutritionist. Subsequently, participating families attended three meetings with a nutritionist over the course of 10 weeks.


Teasing This construct was measured using the Perceptions of Teasing Scale (POTS; Thompson et al., 1995), an 11-item measure of teasing that assesses two distinct teasing constructs: WRT (six items; e.g., "People made jokes about you being too heavy") and Teasing about Abilities/Competency (Competency Teasing; CT; five items; e.g., "People laughed at you because you didn't understand something"), from kindergarten until the present. Consistent with Thompson et al., WRT scale scores consist of numerical ratings of frequency for six weight-related items plus associated ratings of distress, and CT scores consisted of five competency teasing ratings plus attendant distress ratings. Higher scores indicate more self-reported teasing. Good test–retest reliability was reported by Thompson and colleagues in the measure's original development among young adults (r = .88). Test–retest reliability in the current study was acceptable between Times 1 and 2 (i.e., 10–12 weeks; r = .79) and between Times 2 and 3 (i.e., 1 year; r = .57). Internal consistency in the present sample was very good at all three time points (α = .91, α = .92, α = .90, respectively).

Quality of Life

Self-reported HRQOL was measured using the Pediatric Quality of Life Inventory (PedsQL) 4.0 Generic Core Scales. This 23-item self-report measure of HRQOL yields scores on four subscales: physical functioning (eight items); emotional functioning (five items); social functioning (five items); and school functioning (five items). The PedsQL has demonstrated good reliability and validity with internal consistency statistics consistently > .70 (Varni et al., 2001). Test–retest reliability for child-reported HRQOL in the present sample was acceptable between Times 1 and 2 (r = .61) and between Times 2 and 3 (r = .45). Internal consistency was acceptable at all three time points in the present sample (α = .86, α = .87, α = .81, respectively).

Parent-report of HRQOL was measured using the PedsQL 4.0 Parent Proxy Report. Internal consistency statistics for this measure have also consistently been >.70. Construct validity has been established using the known-groups method (Varni, Limbers, & Burwinkle, 2007). In the present study, internal consistency was good at all three measurement occasions (α = .91, α = .85, α = .87, respectively) and test–retest reliability was comparable to self-report measures between Times 1 and 2 (r = .57), and between Times 2 and 3 (r = .48). Consistent with Varni and colleagues (2001) scoring instructions, items were reverse scored and converted to a 0–100 scale such that higher scores indicated better HRQOL and the mean scale score was used in study analyses.

Procedures Participants completed study measures at three time points over the course of the study. After completing informed consent and assent procedures, the participating child/adolescent and one parent/guardian completed study measures prior to beginning treatment (Time 1), following treatment completion (approximately 10 weeks after commencing treatment; Time 2) and approximately 1 year following treatment completion (Time 3). The institutional review board of the authors' institution approved these procedures.

A priori power analyses were conducted to determine the likelihood of detecting good and not-good global model fit using a SAS program created by MacCallum, Brown, and Sugawara (1996). Generally, power estimates of .80 or above are considered sufficient (Muthén & Muthén, 2002). Results of this analysis, conduced with α set to .05, 123 degrees of freedom, and a sample size of 93, indicated a 98% chance of detecting close model fit. Further tests revealed a 47% chance of detecting not-good model fit.

Missing data resulting from participant attrition were best judged to be missing at random. Therefore, a maximum likelihood multiple imputation procedure (Buhi, Goodson, & Neilands, 2008) was performed prior to conducting study analyses using SPSS version 19. Study variables associated with missing data were included as predictors in the imputation procedure and m = 100 imputations were conducted (Graham, Olchowski, & Gilreath, 2006). Overall, 24.6% of the raw data necessary for analyses in the present study was imputed. Missing data resulted primarily from participant attrition and unwillingness (or inability to make contact with participants) to complete assessments after completion of the weight control intervention.

Analytic Plan

Measurement and predictive analyses were conducted using structural equation modeling (SEM) techniques in LISREL 8.80 (Jöreskog & Sörbom, 2007). An advantage of SEM that is particularly salient to this investigation is the ability to test bidirectional associations within the same structural model. That is, a variable can be analyzed as both a cause and an effect of other variables simultaneously (Farrell, 1994). Because the χ2 statistic (routinely used to evaluate model fit in SEM) is highly sensitive to sample size (Kline, 2005), alternative fit statistics such as RMSEA, CFI, and NNFI were used to evaluate model fit for all CFA and SEM analyses.

Because the study sample was comprised of participants receiving two different treatments (i.e., Positively Fit and BFI), a two-group confirmatory factor analysis (CFA) and structural equation model were constructed to establish measurement comparability. This analysis supported measurement equivalence across both groups and no group-specific differences in predictive relationships were observed. These findings indicate that the participants receiving Positively Fit did not experience greater improvements in HRQOL attributable to teasing compared to those receiving the BFI. Thus, the two intervention groups were combined for all subsequent study analyses.

Consistent with guidelines for conducting statistical analyses using SEM (Brown, 2006), the present investigation began with a CFA including both measures of interest (POTS and PedsQL). Consistent with previous validation studies (Thompson et al., 1995; Varni et al., 2001), two parcels were created for the POTS (WRT and CT) while four parcels were specified for the PedsQL (Physical, Emotional, Social, and School functioning). These parcels represent the subscales of each measure and allowed for evaluation of specific subscale loadings on latent constructs (i.e., teasing and HRQOL) at each of the three time points.

To evaluate the aforementioned hypotheses, subsequent directional regressions were performed using latent variables within the SEM framework. Using a longitudinal panel design (Little, Preacher, Selig, & Card, 2007), the associations between latent variables at all three time points were tested simultaneously (Figure 1). Moreover, the structural model incorporated autoregressive paths designed to determine the stability of the latent constructs over time. Advantages of using a fully cross-lagged, autoregressive longitudinal panel design within an SEM framework include the ability to control for measurement error by modeling constructs using latent representations, the ability to assess stability of constructs over time, the ability to test complex models with multiple interacting dependent variables, and the ability to test bidirectional associations (i.e., a variable can be analyzed as both a cause and an effect of other variables simultaneously) within the same structural model (Farrell, 1994).


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as: