Associations between overweight and mental health problems among adolescents, and the mediating role of victimization

All secondary schools in Amsterdam require routine health assessments of their students by the Amsterdam Public Health Service (GGD). This assessment includes a clinical interview and completion of a self-reported electronic health questionnaire, before the interview. The data in this study were obtained from this questionnaire and from the Digital Child Health Care Registry (DCHCR), which administers the records of these physical examinations. Both data sources are part of the Youth Health Monitor (YHM) of the GGD and are described in more detail below. We combined three cohorts from the YHM. The first data wave was collected in school years 2010–2011 (cohort 1), 2011–2012 (cohort 2), and 2012–2013 (cohort 3). At that time the students were in the eighth grade, the second year of Dutch secondary education when students are on average 14 years old. Two years later, in the tenth grade, the second wave was collected for each cohort.

In addition, we used objectively assessed body height and weight, and sociodemographic variables from the DCHCR. The DCHCR is a standard government data registry system that includes information on ≥95% of Amsterdam’s youth. At their child’s first routine health assessment, parents are asked for their permission to use data from the DCHCR for research purposes by the GGD.

We aggregated and anonymised the data from the questionnaires and the DCHCR. To answer our research questions, data from the questionnaires and DCHCR were merged (Fig. 1).

We analyzed the association between overweight and obesity, and adolescents’ mental health, and whether this was mediated by victimization. For some students we had data available for either T0 or T1, and for others we had data available for both T0 and T1. We performed a multi-level mediation analysis with a two-level structure to incorporate the clustering of the measurements within individuals. As our study was observational, there was already a relationship between the variables established that we measured at time point one. Furthermore, our data was collected with two years between the first and second wave. As it was more likely that overweight, victimization and mental health influenced each other in a relatively short period of time [10, 20], the contemporaneous effects were a better representation of the relationships between the variables than the lagged relationships. By analyzing the data with a multilevel model we used all available information while taking into account the correlation among the repeated measures for part of the group. By using two waves of data for some of the pupils we get insight in both the within- and between-person effect, i.e. the effect estimates in our models were based on a combined within- and between-subject interpretation [47, 48]. We checked whether schools should be added as a third level in our analyses, but because the intra-class correlation was very small (.02) this level was disregarded. Class was not added as a cluster level, because students were in different classes in T0 and T1. Moreover, in the Dutch education system, education is given on the basis of subject clusters, with a general component that is the same for all pupils and an elective component. Furthermore, we tested for significant differences in the indirect effects between boys and girls. Sex was not a moderator (p < 0.10).

We used a two-step approach to analyze mediation. First, we examined the original total effect of overweight (independent variable) on psychosocial problems and suicidal thoughts (outcome variables). Second, we fitted a model in which overweight was related to victimization (a coefficient) and a model in which overweight and victimization were related to psychosocial problems or suicidal thoughts (direct effect and b coefficient respectively), adjusted for sex and ethnicity. Mediation was calculated as the product of the a and b coefficients. Confidence intervals (CI) for the indirect effect estimates (a*b) were based on Monte Carlo simulations [49] to account for the skewed distribution of the indirect effect.

Based on previous studies [36, 37] and on a sensitivity analysis, we hypothesized that adolescents with a more norm-deviating physical appearance, referring to adolescents with obesity, would be especially vulnerable to victimization and thereby to mental health problems. To test this hypothesis, we repeated all analyses for adolescents with obesity in comparison to their normal weight peers. All models were fitted using Mplus version 7 [50] using the Full Information Maximum Likelihood for handling missing data.

In total, 13,740 unique students were included in our study. From 3943 students we gathered information in grade 8 and grade 10, and from 9797 students we had one measurement moment from either grade 8 or grade 10 (Fig. 1). At T0 the mean age of the participants was 14 years, 47% were male and 40% of the students were of Dutch ethnic origin. Also at T0, 18% of the students had overweight and 6% obesity, 7% of the students had been bullied at school in the past three months, 11% had psychosocial problems and 17% had suicidal thoughts during the past 12 months. More details on population characteristics are presented in Table 1.

Compared to their normal-weight peers, adolescents with overweight reported psychosocial problems more often (original total effect OR: 1.5; 95% CI 1.3, 1.8) and victimization more often (a coefficient OR: 1.4; 95% CI 1.2, 1.7). Adolescents who reported being bullied more often also reported psychosocial problems (b coefficient OR: 11.0; 95% CI 8.2, 14.7). Victimization was a significant mediator in the relationship between having overweight and psychosocial problems (indirect OR: 2.3; 95% CI 1.5, 3.7). However, victimization only partly mediated the original association between having overweight and psychosocial problems; for psychosocial problems the direct effect also remained significant (direct effect OR: 1.4; 95% CI 1.2, 1.7) after the mediation effect of victimization was added to the model. This is illustrated in Fig. 2a and Additional file 1.

When we compared adolescents with obesity to their normal weight peers, the studied associations were even stronger. Although the direct effect remained comparable (OR: 1.4; 95% CI 1.0, 2.0), the indirect effect was larger (OR: 6.2; 95% CI 2.8, 14.7). This is illustrated in Fig. 3a and Additional file 1.

Adolescents with overweight reported suicidal thoughts more often than their normal weight peers (original total effect OR: 1.4; 95% CI 1.2, 1.6). Adolescents who reported victimization also reported suicidal thoughts more often (b coefficient OR: 8.0; 95% CI 6.2, 10.3). Victimization was a significant mediator in the relationship between having overweight and suicidal thoughts (indirect effect OR: 2.1; 95% CI 1.4, 3.2). The association between overweight and suicidal thoughts, i.e. the direct effect, remained significant after the mediation effect of victimization was added to the model (OR: 1.3; 95% CI 1.1, 1.5). This is illustrated in Fig. 2b and Additional file 1.

The associations between obesity and suicidal thoughts were stronger (indirect effect OR: 4.5; 95% CI 2.3, 9.1 and direct effect OR: 1.5; 95% CI 1.1, 2.0). This is illustrated in Fig. 3b and Additional file 1.

Compared to their normal weight peers, adolescents with overweight or obesity reported psychosocial problems and suicidal thoughts more often. These associations were mediated by whether or not the adolescents were victimized. This indirect effect was stronger for adolescents with obesity than for those with overweight. One explanation may be that the more an individual’s body type diverges from the norm, the more they are victimized and the stronger the association with mental health problems becomes. This explanation is in line with the findings from earlier studies on this topic [36, 37].

Studies investigating the relationship between victimization and relatively uncommon physical characteristics such as having red hair, wearing glasses, disabilities and gender dysphoria demonstrated inconclusive findings [51‐53]. However, our study, along with other research [27, 28, 54] shows that overweight is a physical characteristic that is related to being victimized. This may indicate that overweight is a specific trait that makes adolescents stand out negatively and thus leads to victimization. Perhaps overweight is seen as not just as a random physical trait, but as a trait with implicit social elements that may lead to stigmatization. Previous research on why overweight may lead to being bullied offered several explanations, which are all in line with the idea that overweight is perceived as a deviation from the social norm [23, 24]. As the large majority of the adolescents living in Amsterdam do not have overweight or obesity [5‐7], it is plausible that normal weight is the norm among adolescents. Other studies have confirmed this by showing that overweight negatively influences social status and that pupils with overweight received fewer friendship nominations and were more disliked and more often excluded by their peers [55‐57]. Such social damage can make adolescents with overweight more vulnerable to being victimized due to a lack of friends to defend them, lower self-esteem, and a lower social status among their peers [58, 59].

A recent review showed that the resulting social isolation is related to additional unhealthy behaviors: excessive food intake and decreased participation in sports and other physical activities because of increased stress and not enjoying or daring to participate in sports activities. Furthermore, overweight can lead to a low self-esteem. This can, in turn, lead to further weight gain and a vicious cycle of poor physical and social outcomes, which as a consequence could enhance the probability to become a victim [60, 61].

Besides being a victim of bullying, other potential mediators may be dissatisfaction with one’s own body [37] or reduced participation in sports and physical activity [60]. Therefore, further research using multiple mediation models is needed to better understand the relationship between overweight and negative mental health outcomes in adolescents.

Previous research demonstrated that the SDQ total difficulties score makes an excellent distinction between pupils who probably have no psychosocial problems and pupils who do, but also showed that the internal consistency of the SDQ subscales is low [39, 40]. A fruitful extension of our research would be to explore the association between overweight and different types of psychosocial problems.

This study is the first large-scale study examining being a victim of bullying as a potential mediator in the association between body weight and mental health problems. Our sample is representative for adolescents in Amsterdam, a multi-ethnic urban area, and likely representative for adolescents in other Dutch urban areas. Another strength is the combination of an objective report of body weight and validated self-reports of victimization and mental health problems. Being a victim of bullying and internalizing problems are less apparent to parents and teachers [62‐64]. Finally, the multilevel analyses in our study allowed us to use all available information from all participants, regardless of whether a participant had information on one or two time points. A limitation is that we explored only one potential mediator in the relationship of overweight or obesity and mental health, whereas other potential mediators could also play a role. We did not distinguish between types of victimization or intensity of victimization, which may help to explain the stronger association between obesity and mental problems. Given the cross-sectional nature of our study, we were not able to infer causality. We could not infer whether overweight caused victimization and, subsequently, mental health conditions or whether adolescents with mental health conditions were more vulnerable to becoming victimized or more inclined to develop overweight. Further research is needed to examine the causal relation between overweight and mental health problems and the role that victimization and other factors may play in this relationship. Finally, we did not use a specific weight based victimization question, so it is possible that the measured victimization could be caused by other factors such as personality factors underlining victimization and mental health problems. For future research it would be of added value to use the construct of ‘weight based victimization’ as developed and studied by Puhl et al. [30, 65].