Introduction
Bullying is a form of aggressive behavior that is repeated over time against one or more individuals who are relatively powerless (Monks et al.
2009; Salmivalli
2010). The aggressive behavior can take many forms such as name calling, hitting, spreading rumors, and social exclusion. Bullying is considered a significant public health problem world-wide, with prevalence rates ranging from approximately 10 to 25% (Analitis et al.
2009; Nansel et al.
2001; Thomas et al.
2017). There is mounting evidence that the experience of being bullied associates with overall mental health problems, including both internalizing and externalizing symptoms (Klomek et al.
2015; Moore et al.
2017; Reijntjes et al.
2010; Schoeler et al.
2018; Singham et al.
2017; van Lier et al.
2012).
However, it is unclear at present whether these identified associations with exposure to bullying are specific to particular psychiatric outcomes, or shared between them. Psychiatric disorders and their symptoms co-occur substantially, even across the broadly defined internalizing and externalizing domains (Achenbach et al.
2016; Angold et al.
1999; Kessler et al.
2005; Krueger
1999). Additionally, psychiatric disorders often have a multifactorial etiology that includes (i) shared risk factors, such as bullying, childhood maltreatment, maternal depression, and stressful life events (Caron and Rutter
1991; Kessler et al.
1997; Kim et al.
2003; Vachon et al.
2015), and (ii) a shared genetic vulnerability associated with, for example, the risk of exposure to bullying (Schoeler et al.
2019). In recent years, studies have suggested that this general vulnerability to psychopathology may be usefully represented by a bifactor model that captures (i) shared variance among a broad range of mental health problems (i.e., general factor), as well as (ii) specific influences beyond those explained by shared variance (i.e., specific factors) (for studies in children and early adolescents, see Lahey et al.
2015; Neumann et al.
2016; Olino et al.
2014; Patalay et al.
2015). Although in children the focus has often been on internalizing and externalizing dimensions of psychopathology, thought disorders (e.g., obsessive–compulsive disorders, schizophrenia) have been identified as a distinct third dimension in adolescents and adults (Caspi et al.
2014).
The general factor of psychopathology in childhood has been found to predict various psychiatric outcomes in adolescence, such as anxiety disorders, mood disorders, and substance abuse (Pettersson et al.
2018). The specific factors, however, predicted only a subset of psychiatric outcomes. For example, the specific internalizing factor associated with anxiety and mood disorders but not with, for example, substance abuse. Furthermore, previous research found evidence for both genetic and environmental influences on the general factor of psychopathology (Brikell et al.
2020; Brodbeck et al.
2018; Caspi et al.
2014; Neumann et al.
2016; Riglin et al.
2019). Of particular interest is the finding that childhood maltreatment (e.g., harsh discipline, physical abuse, sexual abuse) associated most strongly to a general factor of psychopathology as opposed to symptom specific factors (e.g., internalizing, externalizing) (Brodbeck et al.
2018; Caspi et al.
2014). However, to the best of our knowledge, no study to date has examined the association of bullying exposure in childhood with the general factor of psychopathology in early adolescence.
The primary focus of the current study is to assess the prospective associations of bullying victimization (8, 10 years) with both general and specific psychopathology factors (7, 13 years) while accounting for other risk exposures known to associate with overall mental health problems (e.g., childhood maltreatment, maternal depression, stressful life events). To this end, the aim of the study was threefold; (i) testing general psychopathology bifactor models at different time points across development, specifying a general factor in addition to specific internalizing and externalizing factors, (ii) systematically characterizing the chronicity and type (overt/relational) of bullying victimization in relation to both general and specific psychopathology factors, and (iii) incorporating the repeated assessment of bullying exposure and general psychopathology into a path model, together with the other risk exposures, to test for prospective interrelations.
Discussion
The primary aim of this study was to investigate prospective associations of exposure to bullying with general psychopathology factors across childhood, accounting for a wide range of pre- and postnatal factors. Using prospective data from a large birth cohort study, we found that exposure to bullying in primary school associated with the general psychopathology factor in early adolescence. Our integrative longitudinal model showed that pre-existing psychopathology, together with cumulative risk exposure, is also a vulnerability factor for bullying exposure.
Although it is well established that those whose are bullied are at increased risk for a wide range of mental health disorders and symptoms, less is known about the extent to which these findings are specific to these outcomes or reflect a general vulnerability to psychopathology. We found that bullying exposure was associated with both internalizing and externalizing factors from the correlated-factors model that has been commonly used in research on the structure of psychopathology. However, after extracting the shared variance that belongs to the general factor of psychopathology, effect estimates for the internalizing factor decreased considerably in size and dropped to insignificant. In the correlated two-factors model, the observed link between bullying exposure and the internalizing factor was likely inflated because the internalizing factor contained variance shared with the externalizing factor, which associated with bullying exposure. The bifactor model, however, has the advantage of directly assessing the shared variance between internalizing and externalizing domains (general factor), while simultaneously modeling the unique variance in each domain (specific factors).
Concerns have been raised, however, about the stability of these general and specific factors across samples or indicators (Bornovalova et al.
2020). Although we found support for a general factor of psychopathology with borderline acceptable construct replicability across three different time points (ages 7, 10, 13), the pattern of the factor loadings for the specific factors did not replicate well. This finding supports previous research by Gluschkoff et al. (
2019), showing that despite an equal degree of strong longitudinal invariance for the correlated factors model and the bifactor model, the specific internalizing factor demonstrated unacceptable construct replicability estimates. For example, some of the factor loadings dropped to insignificant once the general factor was included in the model. Although these irregular loading patterns can challenge the interpretation of the specific factors, our longitudinal data demonstrated consistent, moderate autocorrelations over time, as well as consistent correlations with bullying exposure.
In particular, youth chronically exposed to multiple forms of victimization (i.e., both overt and relational) displayed greater levels of general psychopathology in early adolescence. This finding is in line with previous research indicating a dose–response relationship between chronic (8 and 10 years) or combined (overt and relational) exposure to bullying and borderline personality symptoms at age 11 years (Wolke et al.
2012). All of these associations considerably attenuated but remained significant after correcting for pre-existing general psychopathology. This finding supports prior evidence that the association between exposure to bullying and mental health outcomes can partially be accounted for by pre-existing vulnerabilities of bullied individuals (Hodges and Perry
1999; Reijntjes et al.
2010; Singham et al.
2017). Children exposed to bullying may differ from children not exposed in individual characteristics, such as having fewer friends, withdrawal or aggressiveness, which in turn increase their risks for being bullied (Arseneault et al.
2010; Monks et al.
2009). It has been previously shown that children who displayed aggressive behaviors in early childhood were more likely than nonaggressive children to experience chronic or high levels of bullying in preschool (Barker et al.
2008).
Integrating repeated assessments of general psychopathology (7, 13 years), exposure to bullying (8, 10 years) and environmental risk exposure (prenatal, early-childhood, late-childhood) in a longitudinal risk model, we also found that general psychopathology prospectively associated with bullying exposure, with higher levels of general psychopathology at age 7 rendering youth more susceptible to exposure to bullying at age 8. Bullying exposure at age 8, in turn, associated with general psychopathology at age 13 through its two-year continuity. Although statistically significant, the observed association of exposure to bullying with general psychopathology was small in magnitude when controlling for the various variables included in the model. This is in line with multifactorial influences on overall mental health problems (Lereya et al.
2015). For example, we found that the effect size of bullying exposure was similar to that observed for the cumulative risk exposure score, including a wide range of adversities known to be associated with mental health.
Self-reports involve children’s subjective perception of being victimized and likely tap into children’s feelings and well-being. Indeed, it has been demonstrated that self-reports of bullying victimization are stronger predictors of internalizing problems than peer reports (Bouman et al.
2012). However, this also implies that parents may have under-reported their children’s internalizing symptoms. Nevertheless, we found that the internalizing and externalizing factors were largely equally indicative of the general factor and were similarly associated with bullying exposure in the correlated-factors model. By using different reporters for bullying exposure and child outcomes, we were able to control for common method variance attributable to the informants.
The current findings should be interpreted in the context of several limitations. First, as mentioned previously, the replicability of the specific factors was sub optimal, owing in part to their small number of indicators. Therefore, in the future, our findings need to be replicated in longitudinal data that allows a higher number of psychiatric domains and disorders. Second, as with most longitudinal studies, considerable attrition occurred. Attrition might result in a loss of power to detect effects and may also bias the findings to those individuals who continued participating in the study. However, bullied children with missing data on covariates were as likely as children with complete data to have higher levels of psychopathology. This supports previous simulations using ALSPAC data, demonstrating that associations between predictors and outcomes are unlikely to be substantially altered by selective dropout (Wolke et al.
2009). Of note, the full information likelihood (FIML) approach enabled our integrative path model to be conducted on the full sample of children with available data on general psychopathology at age 13. Third, the analyses are correlational in nature and, hence, causality cannot be inferred. Fourth, in line with previous child studies, the current study examined the structure of psychopathology using the broadly defined internalizing and externalizing domains. Other disorders not studied here (e.g., obsessive–compulsive disorders, schizophrenia) might also contribute to a general factor of psychopathology (Caspi et al.
2014). Finally, although we controlled for a wide range of covariates, spanning the prenatal period up to early adolescence, the possibility of residual confounding cannot be fully excluded.
These limitations notwithstanding, this is the first population-based longitudinal study to show that exposure to bullying in primary school is a risk factor for a more general vulnerability to psychopathology in early adolescence. Although small in magnitude, the findings of the current study highlight the potential value of a transdiagnostic approach to understanding psychopathology.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.