Developmental Trajectories of Internalizing and Externalizing Symptoms in Youth and Associated Gender Differences: A Longitudinal Network Perspective
Rates of psychopathology in youth are high, with an estimated U.S. lifetime prevalence > 20% of disorders with severe impairment and/or distress (Merikangas et al.,
2010). A substantial proportion of adult psychopathology diagnoses have their origins in childhood and adolescence. For example, approximately two-thirds of lifetime depression cases among adults emerged in adolescence (Kessler et al.,
2005), and nearly all anxiety disorders begin in childhood (Kessler et al.,
2009). Additionally, antisocial personality disorder in adult men is associated with a diagnostic trajectory of attention-deficit/hyperactivity disorder (ADHD), oppositional defiant disorder, and other diagnoses in childhood (Beauchaine et al.,
2017). Variability in psychopathology (dis)continuity between childhood and adulthood (Rutter et al.,
2006; Shevlin et al.,
2017) highlight a critical need to understand psychopathology emergence and progression across development. Given the heterogeneity of disorder presentation and remarkable disorder co-occurrence, including between internalizing and externalizing disorders (e.g., nearly 50% of children with ADHD experience co-occurring depression; Wilens et al.,
2002), the application of categorical approaches (e.g.,
Diagnostic and Statistical Manual of Mental Disorders) may obscure developmental pathways of symptom trajectories. Understanding how symptoms may influence one another independent of diagnostic categories may provide insights that resolve disorder heterogeneity and co-occurrence. Here, we used a directed symptom-level network approach to characterize how internalizing and externalizing symptoms at baseline predict future symptoms within a homogenous age group of children.
Various methods have been used to study the development of psychopathology in youth, including epidemiological studies. For example, studies have examined the rates of different psychiatric disorders across childhood and adolescence, as well as proposed a general psychopathology factor that measures commonality among various disorders during youth (Costello et al.,
2011; Patalay et al.,
2015). These studies have yielded intriguing findings, including an adolescence-limited increase in generalized anxiety disorder prevalence among girls. In contrast, other anxiety disorders, such as social anxiety disorder and specific phobia, demonstrate a consistent increase in prevalence across adolescence and into adulthood (Costello et al.,
2011). Additionally, a general psychopathology latent factor explains shared variance across internalizing and externalizing disorders during childhood and adolescence, suggesting similar developmental pathways or risk factors between internalizing and externalizing domains (Patalay et al.,
2015). Research investigating gender differences in the development of youth psychopathology has demonstrated that boys experience higher rates of conduct disorders and girls experience higher rates of depression and anxiety (Lahey et al.,
2000; Zahn-Waxler et al.,
2008). However, previous research examining the broad spectrum of psychiatric disorders in youth have largely relied on cross-sectional data, preventing firm conclusions about the temporal dynamics or developmental pathways of psychopathology. As conflicting symptom and disorder progression have been reported in prior literature (e.g., Costello et al.,
2011; Shevlin et al.,
2017), comprehensive, longitudinal approaches are critical for delineating the nature of internalizing and externalizing symptom trajectories.
The application of a directed network approach to understanding psychopathology has broad appeal as it tests causal relations among symptoms, delineating their developmental trajectories (Borsboom,
2008; Cramer et al.,
2010). A network approach emphasizes mutually reinforcing relationships between symptoms across disorders, distinguishing itself from the latent factor approach to psychopathology in which disorders are thought to be the common cause through which presenting symptoms can be explained (Kendler,
2016). Though a detailed discussion comparing network and latent factor approaches is beyond the scope of this paper, it is notable that these approaches can be combined, such as with latent network modeling and residual network modeling (Epskamp et al.,
2017). Still, a network perspective can yield unique insights into the development and treatment of psychopathology. For example, the centrality hypothesis argues that symptoms which demonstrate high centrality (i.e., more numerous and stronger inter-symptom causal connections) are the most influential symptoms in both the development and remission of disorders (Borsboom & Cramer,
2013; Cramer et al.,
2010). Additionally, symptoms that have causal relationships with symptoms from different clusters, such as depression and anxiety, can be seen as “bridge symptoms” that provide a potential causal mechanism and explain disorder comorbidity (Cramer et al.,
2010). For example, Robinaugh et al. (
2014) found loneliness to be a bridge symptom between persistent complex bereavement disorder and depression symptoms. However, these conclusions were drawn from cross-sectional networks rather than directed networks derived from longitudinal data, limiting inferences about the causal direction and temporal dynamics of the observed relationships.
Longitudinal studies of youth psychopathology have used a developmental cascade perspective to investigate the mutual influence of internalizing problems (e.g., anxiety, depression, social withdrawal) and externalizing problems (e.g., peer aggression, rule-breaking) over time (Dearing et al.,
2006; Masten et al.,
2009). Studies have shown reciprocal associations across time between internalizing and externalizing problems and how co-morbid symptoms can arise in youth (Achenbach & Rescorla,
2000; Mesman et al.,
2001). Proposed causal mechanisms of these reciprocal associations include externalizing problems pre-disposing children to social rejection or academic failure thus leading to depression, particularly in boys (Patterson & Capaldi,
1990). However, conflicting evidence exists in the dynamics of these developmental cascades, with some studies finding that earlier externalizing problems predicted fewer internalizing problems later on (Panayiotou & Humphrey,
2018) and other studies finding that prior externalizing problems positively predicted later internalizing problems (Masten et al.,
2005; Moilanen et al.,
2010). Importantly, gender differences have also been found in these temporal dynamics with findings that internalizing problems predicted future externalizing problems consistently for girls but not boys (D’urso & Symonds,
2022), though the cascade effect of externalizing problems positively predicting later internalizing problems appeared consistent for both male and female participants (D’urso & Symonds,
2022; van Lier & Koot,
2010). More recently, Speyer et al. (
2022) used network approaches to investigate developmental cascades and found that ADHD symptoms were highly central to development of later socioemotional symptoms and that prosocial behaviors served as a potential bridge symptom between externalizing and internalizing difficulties. Black et al. (
2022) also utilized a network approach and found complex within-person effects between internalizing symptoms and indicators of well-being with indicators such as thinking clearly, unhappiness, dealing with stress, and worry being most central in the network.
The present study seeks to broaden understanding of developmental cascades by testing longitudinal relationships among sum scores of eight transdiagnostic symptom clusters––anxious/depressed, withdrawn/depressed, somatic complaints, social problems, thought problems, attention problems, rule-breaking behavior, and aggressive behavior––in pre-adolescents and adolescents over three timepoints (i.e., baseline, 1-year follow-up, and 2-year follow-up) using graphical vector autoregression (GVAR). GVAR models identify temporal relationships between symptoms by estimating edges, which represent the unique causal effects of one symptom cluster on another (Epskamp,
2020). By examining centrality indices, GVAR models can identify central symptoms that are most predictive of, or predicted by, other symptoms. The present study also tested gender differences in the longitudinal relationships among symptom clusters to highlight possible differences in developmental trajectories of psychopathology. Based on past studies that have identified depressed mood, attention difficulties, and anxiety as being the most central to psychopathology development in youth (Funkhouser et al.,
2021; McElroy et al.,
2018a,
b), we hypothesized that anxious/depressed, withdrawn/depressed, and attention problems would be the most central symptoms clusters, influencing changes in other symptom clusters at later timepoints. Additionally, we hypothesized that symptom clusters would group together such that internalizing and externalizing domains would exhibit higher within-group symptom associations than between-group symptom associations, though we anticipated depressive symptoms to be a bridge between internalizing and externalizing disorders based on previous research (McElroy et al.,
2018a,
b).
Method
This study used data collected from pre-adolescents and adolescents at baseline, 1-year follow-up, and 2-year follow-up assessments from the Adolescent Brain Cognitive Development (ABCD) study (data release 3.0; NDAR-
https://doi.org/10.15154/1520926). The ABCD study is an ongoing, longitudinal study within the United States that follows a nationally representative sample of 11,878 children aged 9–10 at baseline (see Garavan et al.,
2018 for information on sampling strategies across 21 data collection study sites, school and participant recruitment procedures, and informed consent; see Auchter et al.,
2018 for details on ABCD study Institutional Review Boards, Bioethics and Medical Oversight advisory group, and other advisory boards). To test temporal associations of directed symptom network structures in the development of psychopathology, we examined a subsample (
n = 6,414) who completed the assessment procedure at all three timepoints (timepoint 1 mean age = 10.0 years [
SD = 0.6]; timepoint 3 mean age = 12.0 years [
SD = 0.6]; 78.6% White; 82.4% 4
th or 5
th Grade at timepoint 3). Demographic information of the subsample is presented in Table
S1. Gender identity of participants for the purposes of the study was defined by parent-report at 2-year follow-up. There were no significant differences in racial identity (χ
2(180) = 192,
p = 0.256), ethnicity (χ
2(4) = 6,
p = 0.199), and combined family income (χ
2(180) = 192,
p = 0.256) between participants identifying as male and female.
Measures
Dimensional assessment of anxious/depressed, withdrawn/depressed, somatic complaints, social problems, thought problems, attention problems, rule-breaking behavior, and aggressive behaviors syndrome scales, representing symptom clusters, were assessed with the Child Behavior Checklist Parent’s Report Form (CBCL; Achenbach,
2001). On the CBCL, anxious/depressed, withdrawn/depressed, and somatic complaints are grouped as internalizing problems, while rule-breaking and aggressive behaviors are grouped as externalizing problems.
T-scores normed by sex, age, and ethnicity were used for the present study analyses.
T-scores had a lower bound of 50, representing 50th percentile or below, and an upper bound of 100, representing above 99th percentile. CBCL syndrome scales demonstrated one week test–retest reliability ranging from 0.80 to 0.94 (Achenbach,
2001). These syndrome scales have demonstrated concurrent validity with clinical diagnoses of anxiety disorders, mood disorders, attention-deficit/hyperactivity disorder, oppositional-defiant disorder, and conduct disorder (Ebesutani et al.,
2010; Eiraldi et al.,
2000; Kasius et al.,
1997; Seligman et al.,
2004).
Statistical Analysis
GVAR models for panel data (Panel GVAR) were used to examine longitudinal relationships between scores on CBCL syndrome scales, representing symptom clusters, across baseline, 1-year follow-up, and 2-year follow-up time points. Panel GVAR models illustrate how CBCL syndrome scale scores influence each other and themselves across time at the within-person level while controlling for between-person differences in these scales. Additionally, Panel GVAR constrains the effects of the syndrome scale scores so that they are stable across the three timepoints in order to assess for stable effects across timepoints rather than deviations between timepoints.
Panel GVAR models were computed using full information maximum likelihood estimation. First, the full model was estimated, then a model search procedure was used to maximize the Bayesian information criterion (BIC) by pruning edges that were not statistically significant at the
p < 0.05 level and adding edges that were significant at the
p < 0.05 level. The comparative fit index (CFI; Bentler,
1990), Tucker-Lewis index (TLI; Tucker & Lewis,
1973), and root mean square error of approximation (RMSEA; Steiger & Lind,
1980) were calculated to assess model fit. Centrality indices were then calculated for each syndrome scale score in the final model. Instrength centrality represents the degree to which syndrome scales are predicted by scores on other scales at the previous timepoint, while outstrength centrality represents to what degree syndrome scales predict scores on other scales at the next timepoint.
Similarly, bridge in-degree centrality represents the degree to which each syndrome scale in one community (i.e., internalizing or externalizing) is predicted by syndrome scales in the other community at the previous timepoint, while bridge out-degree centrality represents to what degree syndrome scales in one community predict scores on scales in the other community at the next timepoint. Gender differences in network structures were tested by estimating the Panel GVAR model as a multi-group model, constraining parameters to be equal across gender groups, and assessing significance of change in model fit. Of note, the 13 participants who identified as transgender/other or for whom gender identity was not known were not included in the multi-group model separating male and female groups, though they were included in the full Panel GVAR model consisting of all 6,414 participants. More details regarding specifics of Panel GVAR models can be found in Epskamp (
2020). Given concerns regarding the stability of network models (Forbes et al.,
2019), the robustness of the estimated networks was tested by applying the same model search procedure for Panel GVAR models of 1,000 non-parametrically bootstrapped samples and calculating how often each edge was included in the optimal model, with 50% inclusion probability being considered robust (see Betz et al.,
2020). Stability of centrality estimates was determined using case-dropping subset bootstrap, in which 20% of the sample was randomly dropped and the model was re-estimated across 1,000 iterations, and calculating 95% bootstrapped confidence intervals (see Epskamp et al.,
2018a,
b). All analyses involving Panel GVAR models were conducted using the
R package “psychonetrics,” version 0.8.1. R code for study analyses presented here is available on the Open Science Framework (
https://osf.io/fcuhm/?view_only=54cc8a191eed4da9a5a576bc35ef8c5b).
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.