Introduction
Psychiatric comorbidities occur with greater-than-chance frequency [
1] and are associated with increased symptom severity and a poorer overall prognosis [
2‐
5]. The mechanisms by which psychiatric symptoms and disorders are related, however, remain poorly understood. Hierarchical dimensional models, which have long been favoured in child and adolescent psychiatric research, account for comorbidity by framing psychopathology as a small number of broad transdiagnostic dimensions (e.g. internalizing, externalizing). However, due to strong associations between the dimensions themselves, a general psychopathological factor,
p, has been proposed to explain the co-occurrence of virtually all psychiatric symptoms and disorders [
6,
7]. Although this general factor has been supported in both adult [
6,
7] and child samples [
8,
9], a consistent interpretation of the
p-factor has so far proven elusive [
6‐
11]. The most popular interpretation posits that
p reflects a shared aetiological factor, or set of factors (e.g. genetic vulnerability, personality, environmental factors) that predispose individuals to any and all forms of psychopathology [
11].
In recent years, an alternative school of thought, the network approach, has gained considerable momentum. This perspective posits that psychiatric disorders reflect complex networks of locally associated symptoms [
12]. In such models, variables are presented graphically as nodes (points in space) and the associations between nodes are presented as edges (lines, with thickness denoting strength). This approach assumes that the effects of causal factors spread throughout networks via direct, symptom-level interactions and reinforcement (as opposed to causal factors influencing all symptoms simultaneously) [
12]. Borsboom and Cramer [
12] provide the following chain as an example; chronic stress → depressed mood → self-reproach → insomnia → fatigue → concentration. As such, the network approach accounts for comorbidity through ‘bridging edges’; i.e. direct associations that serve to link reasonably distinct clusters of symptoms/disorders [
12,
13]. Under such an interpretation,
p represents the statistical reduction of a plethora of lower-level interactions between different components of psychopathology. The main advantage of the network approach is that, by focussing on local interactions, we can determine not only how important a symptom/disorder variable is in terms of its overall connectivity (aka centrality), but also where a symptom/disorder is important within the network (i.e. the strongest edges).
Although an abundance of network studies has been published recently, the majority have focussed their enquiry on the structure of single disorders [
14,
15], or a narrow range of comorbid disorders [
16]. To our knowledge, only two studies have used network methods to model broader psychiatric comorbidity; Boschloo and colleagues [
17] examined the network structure of 120 symptoms from 12 supposedly distinct DSM-IV disorders in the second wave of the National Epidemiologic Survey on Alcohol and Related Conditions (NESARC,
N = 34,653). They found that, although symptoms clustered in a manner broadly consistent with the DSM structure, all 120 symptoms were connected, either directly or indirectly, and each cluster was connected to at least three others. Boschloo et al. [
18] then examined the network structure of emotional and behavioural problems in a large sample of pre-adolescents (mean age 11.1 years;
N = 2175). Ninety-five symptoms were assessed using the Youth Self-Report [
19], and in the resultant network symptoms clustered together in patterns broadly reflective of the YSR domains. Connections were generally stronger and more common within, rather than between, these domains; however, numerous symptom pairs served to link domains, suggesting that the domain boundaries were not as defined as had previously been assumed in factor analytic studies.
Further studies of broad comorbidity networks may help unpack the overlap between higher order dimensions of psychopathology and provide a more complete view of psychiatric comorbidity. The present study expands on previous work in two ways. This is the first study to examine the network structure of comorbid internalizing and externalizing at the disorder level (i.e. where nodes in the network reflect distress/impairment aggregates from internalizing and externalizing symptom groups). To date, network analysis has mainly been used to model the association between symptoms, assuming that symptoms are the base level of psychopathological expression. However, as Borsboom et al. [
12] note, symptoms themselves may be broken down into networks of emotional, behavioural and external factors; e.g. difficulty sleeping may be understood in terms of the interplay between mood, melatonin production, routine, caffeine intake and/or screen time. As such, network analysis is flexible and may be used to study psychopathology at various levels of complexity [
20]. Indeed, there has been an increase in studies that have examined psychopathology at the construct level [
20‐
23], and this approach has the advantage of reducing the amount of nodes and edges in a network, in turn simplifying interpretation [
24]. As the aim of the present study was to provide a network analogue to
p (which has primarily been modelled at the disorder-level), we chose to focus our enquiry at the disorder level (e.g. depression, general anxiety, hyperactivity).
Second, this study expands our knowledge of the network structure of psychopathology in childhood and adolescence, and explores whether this structure changes over this key developmental period. The transition from childhood through adolescence is a period marked by significant biological, cognitive and socio-environmental change. It is during this period that psychopathology commonly emerges [
1,
25]. Furthermore, disorders that emerge during this period typically serve as precursors to similar problems later in life [
26]. As such childhood/adolescence is an ideal period to focus on when examining direct associations between symptoms and/or disorders and may offer key insights into the development of psychiatric sequelae. To date, only one study has focussed on this age group [
18]; however, the network structure was examined at only one time point (age 11). The present study examined the network structure of psychopathology (internalizing and externalizing) within a single cohort across three time points (7.5, 10.5, and 14 years) and tested whether this structure remained stable over this period. Although this study was exploratory in nature, it was predicted that disorders would form two distinct clusters of nodes analogous to the internalizing and externalizing dimensions of psychopathology. No a priori hypotheses were made regarding potential ‘bridging edges’.
Discussion
The present study sought to examine the network structure of internalizing and externalizing disorders assessed between middle childhood and early adolescence. To our knowledge, this is the first study to take a network approach to the modelling of internalizing and externalizing at the disorder level. Furthermore, this is the first study to apply network techniques to a cohort of children assessed from middle childhood through adolescence. A coherent network structure was identified at three time points (age 7.5, 10.5 and 14 years), and tests of structural invariance indicated that the networks remained generally stable, despite the many biological, cognitive and social changes that typically occur over this developmental period. As predicted, two regions of clustered nodes emerged, reflecting strong associations between internalizing disorders, and externalizing disorders. These clusters were most strongly bridged via the edges GAD-ADHD, and DEP-ODD. The centrality indices indicated that GAD and ODD were most important to the networks as a whole. An examination of the robustness of the centrality indices suggested that they could be interpreted with a degree of confidence.
Internalizing and externalizing as networks
Borsboom and colleagues [
41] suggested that, if modelled using network techniques, traditional hierarchical measurement models of psychopathology would be reflected in clusters of highly associated nodes, analogous to the higher order dimensions of internalizing and externalizing. The findings of the present study support this claim. While this finding may not be surprising (both network and latent variable models are derived from covariance), it offers a different interpretation of psychiatric comorbidity. This perspective suggests that, rather than a single causal factor (or amalgam of causal factors) driving the association between symptoms/disorders, causal factors spread their effects throughout psychopathological networks via local interactions and reinforcement. In other words, higher order dimensions (e.g.
p) may be capturing a plethora of local-level interactions. Under this interpretation, the dimensions of internalizing and externalizing are correlated due to certain disorders (which themselves are comprised of networks of associated symptoms) acting as ‘bridges’ between these two broad spectra.
The nature of these interactions (i.e. edges), however, is far from clear and serves to further emphasise the complexity of psychiatric comorbidity. Given that the analysed data were cross-sectional (within-time points), a significant edge could represent a multitude of possible relationships. First, it is possible that disorders directly influence each other. For example, there is a long history of research looking at the comorbid anxiety and depression, with evidence suggesting that anxiety tends to precede and lead to subsequent depression [
42]. One proposed mechanism for such a relationship is that cognitive/neurophysiological processes (e.g. sustained heightened physiological arousal) may lead to an exhaustion of the body which manifests as depression [
43]. Other edges may reflect more indirect associations, e.g. the edges ODD-DEP and ADHD-GAD in the current networks. Developmental cascade models have long suggested that externalizing behaviour may indirectly lead to internalizing problems through mediating variables [
44‐
46]. To illustrate, frequent disruptive behaviour in childhood/adolescence may lead to negative reactions from parents, teachers and/or peers, e.g. shouting, punishment, ostracisation from peer group, and/or academic failure. Such negative outcomes may in turn foster feelings of irritability, distress and worthlessness within the child, and if left unchecked, these experiences may eventually progress to levels of clinical significance [
45,
46]. Edges in any given network may also represent spurious associations due to an unmeasured common cause. For example, shared biological (e.g. genes) or environmental (e.g. trauma) risk factors may influence the development of multiple symptom domains simultaneously [
42]. If these risk factors are not represented in a given network, this may give rise to non-causal associations between disorders. Edges are further complicated by the possibility of bidirectional feedback loops, equifinality (multiple risk factors, pathways and processes leading to similar outcomes), and multifinality (specific risk factors leading to multiple outcomes) [
47].
Given that the networks in the present study were undirected and cross-sectional, any causal interpretations, such as those described above, are purely speculative. The aim of the present study, however, was not to infer strict causal relationships but to demonstrate how network analysis can be used to quantify the importance of disorders and identify key associations between disorder pairs and as such generate hypotheses regarding the complex mechanisms that drive psychiatric comorbidity. For example, in the present study, GAD and ODD were identified as the two most influential nodes within wider comorbidity networks, suggesting that symptoms within these disorders may be influential in the initiation and/or maintenance of comorbid psychopathology. Furthermore, a number of significant edges bridged the spectra of internalizing and externalizing, the strongest of which was the edge ODD-DEP. This suggests that local-level interactions between these two disorders may go some way to explaining the correlations between internalizing and externalizing when modelled as continuous dimensions [
24].
Strengths and limitations of the present study
The main strength of the present study was the large sample size relative to the number of parameters estimated [
49]. With regard to limitations, diagnostic data were not available for the participants at all three of the time points; therefore, a comprehensive recoding strategy was adopted. It must be noted, however, that the diagnostic algorithms used to create the quasi-diagnostic variables were based on skip patterns (i.e. those who did not endorse ‘core’ symptoms were scored as having no disorder, and subsequent distress and burden were not assessed). This likely introduces an element of bias to the data in favour of DSM scoring conventions. Similar bias likely affects all analyses that employ DSM skip structures.
Skip patterns are particularly problematic when using symptom-level data, as they introduce deterministic dependencies (i.e. secondary symptoms can only be endorsed if primary symptoms are first endorsed, thus artificially inflating correlations between these symptom pairs). However, this was not the case in the present study, as a diagnosis of one disorder was not dependent on the diagnosis of another disorder. Finally, any study that employs network analysis comes with the caveat that this approach is highly divisive. Indeed, this has been evidenced by a series of back-and-forth papers focussed on the replicability of networks [
50‐
52]. A methodological critique of network analysis and/or traditional latent variable models is beyond the scope of this article (for recent reviews, see [
13,
48,
50‐
52]); however, our findings do appear to support the replicability of psychopathological networks within a cohort of children and adolescents assessed repeatedly over a period of significant development.
Acknowledgements
We are extremely grateful to all the families who took part in this study, the midwives for their help in recruiting them, and the whole ALSPAC team, which includes interviewers, computer and laboratory technicians, clerical workers, research scientists, volunteers, managers, receptionists and nurses. The UK Medical Research Council and the Wellcome Trust (Grant ref: 092731) and the University of Bristol provide core support for ALSPAC. This publication is the work of the authors and the corresponding author will serve as guarantor for the contents of this paper.