Medires Publishers - Article Full Text

Abstract

This study applied the Hierarchical Taxonomy of Psychopathology (HiTOP) framework and exploratory graph analysis (EGA) to examine the dimensional structure of 17 common psychological disorders in adolescents. Network analysis was then used to assess centrality and connectivity among these disorders. Data were drawn from 874 adolescents (aged 12–17) who completed the Child and Adolescent PsychProfiler (CAPP-SRF), with responses recoded to indicate risk status for each disorder. EGA identified five dimensions corresponding to HiTOP’s subfactors or superspectra: distress and fear, neurodevelopmental difficulties, cognitive problems, eating pathology, and a depression/conduct disorder (CD) spectrum. The most central disorders in the network were learning disorder (LD), generalized anxiety disorder (GAD), and CD. While several dimensions aligned with the HiTOP model (e.g., fear, distress, eating pathology), others—such as neurodevelopmental and cognitive difficulties, and the depression/CD spectrum-suggest potential extensions or refinements to the original model. The modest connectivity between disorders challenges HiTOP’s assumption of strong hierarchical relationships and suggests that its current structure may be incomplete for adolescent populations. Our findings support the inclusion of distinct spectra for neurodevelopmental and cognitive difficulties, and propose a refined HiTOP model that better reflects adolescent psychopathology, including reclassifications for ODD and CD.

Introduction

Psychological disorders are defined as patterns of thoughts, feelings, or behaviours that lead to significant distress or functional impairment [1]. Various theoretical frameworks have been developed to conceptualize and classify these disorders, with the Hierarchical Taxonomy of Psychopathology (HiTOP) emerging as one of the most prominent dimensional models [2]. Rather than viewing disorders as discrete categories, the HiTOP model conceptualizes them along continua of symptom severity [3]. Building on this dimensional perspective, statistical methods have been employed to group symptoms into clusters that reflect underlying psychopathological constructs. More recently, the network approach has been proposed as an alternative framework, which conceptualizes psychological disorders as complex systems of interacting symptoms [4]. In the network approach, symptoms are not merely passive indicators of latent disorders but are seen as mutually reinforcing, giving rise to broader dimensions of psychopathology. Despite growing interest in this approach, there remains a lack of network-based research focusing on common psychological disorders in adolescents. Addressing this gap, the current study applies a network analysis to examine the structure and dimensionality of 17 frequently observed psychological disorders in an adolescent sample.

Network Theory, Network Analysis and Exploratory Graph Analysis

Much of our current understanding of psychological disorders dimensionality stems from latent variable models, which conceptualize dimensions as unobservable constructs that give rise to observable symptoms or disorders. In contrast, the network model offers an alternative perspective, proposing that disorders themselves directly interact with one another, and that these interactions give rise to broader dimensions of psychopathology [4]. This conceptualization aligns with the HiTOP framework’s emphasis on dimensionality but diverges by focusing on observable relationships between disorders themselves, rather than latent causes.

In network analysis, the variables of interest, such as symptoms or disorders, are represented as nodes, and the relationships between them are known as edges. These edges typically reflect partial correlations, meaning the relationship between two nodes is estimated while controlling for all others in the network [4,5]. When computed using Markov Random Fields with regularization techniques, the resulting network suppresses spurious correlations, revealing a conditional independence structure that highlights the most meaningful associations [6-8]. Network models offer both visual and quantitative insights into the structure of psychopathology. Centrality metrics (e.g., expected influence) quantify how connected a node is within the network, with highly central nodes considered more influential and potentially more clinically relevant [4,9]. Although some debate this assumption evidence suggests that targeting central nodes in treatment may enhance therapeutic outcomes [10-13]. Edge weights, meanwhile, indicate the strength of associations between disorders, offering nuanced information about comorbidity and symptom clustering beyond simple co-occurrence [14].

Closely related to network analysis, Exploratory Graph Analysis (EGA) uses these network structures to identify dimensions, or densely interconnected clusters of nodes that may represent underlying factors [15,16]. Unlike traditional dimensionality reduction methods (e.g., factor analysis), EGA is grounded in network theory and has been shown to outperform standard methods in large samples or when items are highly correlated [17,18]. It also allows for automatic dimension detection and intuitive visualisation, making it particularly accessible for interpreting complex data structures.

Existing Network Analysis of Common Psychological Disorders/Syndromes

Several studies have used network analysis to explore how common psychological disorders or syndromes are related [5,19-21]. Boschloo et al. (2015) examined 12 DSM-IV disorders in adults and found that symptoms tended to cluster within their respective disorders. In a sample of pre-adolescents, Boschloo et al. (2015) observed general clustering of internalizing, externalizing, attention, thought, and social problems, although the structure was less clearly defined than in standard models. McElroy et al. (2018) investigated eight disorders in adolescents using the Child Behavior Checklist items and found that symptoms clustered into two broad spectra: internalizing (e.g., depression, anxiety, PTSD) and externalizing (e.g., ADHD, ODD, CD), with generalized anxiety disorder and ODD emerging as the most central disorders. Similarly, Elovainio et al. (2018) found distinct clustering of internalizing, externalizing, and other problems (e.g., attention, thought, social) among preadolescent adopted children, with anxiety/depression, social problems, and aggression being the most central domains. More recently, Watkeys et al. (2024) used EGA and identified two broad clusters in children: one comprising internalizing and externalizing disorders (e.g., anxiety, conduct, affective disorders), and another comprising neurodevelopmental and sleep-related disorders, with affective/emotional and developmental disorders being most central.

While the specific disorders examined vary across studies, a few consistent patterns emerge. Most notably, internalizing and externalizing dimensions are commonly identified, with some evidence for a separate neurodevelopmental dimension. Emotional disorders, particularly generalized anxiety disorder, often appear as central nodes in these networks. However, past studies differ in their target populations, ranging from children to adults, making it difficult to draw uniform conclusions. It has been proposed that the structure of psychopathology may differ across developmental stages, contributing to variability in past findings [22].

To date, only one network analysis study has focused exclusively on adolescents, despite adolescence being a key developmental period in which many psychological disorders emerge, including depression, anxiety, behavioural issues, eating disorders, and psychosis [20, 23-25]. Adolescence is also a period marked by high rates of comorbidity, underscoring the need for more targeted network-based investigations in this age group [26]. Gaining a clearer understanding of how psychological disorders interact during this critical period could offer valuable insights for both clinical assessment and intervention.

Aims of the Study

Given existing limitations and omissions, the current study had two major aims. Firstly, this study aimed to use use network analysis to examine the network properties (centrality and edge weights) of 17 common psychological disorders (nodes) in a group of adolescents, recruited from a psychology clinic in Australia. Following the disorders captured by the Self-Report Form of Child and Adolescent PsychProfiler (CAPP-SRF; Langsford et al., 2014), the disorders included in the current study were (N = 17) [2]: Attention Deficit/Hyperactivity Disorder Predominantly Inattentive Presentation (ADHDI) and ADHD Predominantly Hyperactive-Impulsive Presentation (ADHDHI), Oppositional Defiant Disorder (ODD), Conduct Disorder (CD), Specific Learning Disorder (SLD) with impairment in Reading (SLDR), Written Expression (SLDW) and Mathematics (SLDM), Autism Spectrum Disorder (ASD), Language Disorder (LD), Speech Sound Disorder (SSD), Generalised Anxiety Disorder (GAD), Persistent Depressive Disorder (PDD), Separation Anxiety Disorder (SAD), Obsessive-Compulsive Disorder (OCD), Posttraumatic Stress Disorder (PTSD), Anorexia Nervosa (AN), and Bulimia Nervosa (BN). Secondly, this study aimed to use EGA to examine the dimensionality the of psychological disorders in the network. Based on the network findings report in previous study in this area, we expected that GAD, PDD and CD would have higher centrality values [19,21]. Additionally, we expect to find at least dimensions relevant to internalizing, externalizing, and neurodevelopmental disorders.

Method

Participants

The initial sample comprised 951 adolescents with self-ratings for the CAPP (CAPP-SRF ratings, who completed either an online or paper version of the CAPP-SRF. Of these, the responses of 71 adolescents contained missing values and could not be used to establish if these were at risk or not for the different disorders. An additional six individuals were removed from the analysis due to not providing their gender status. The final sample resulted in 874 adolescents. Participants’ average age was 14.52 years (M_age = 1.66 years; range = 12-18). There were 523 (59.84%) boys, (M_age=14.53 years, SD = 1.70) and 351 (40.16%) girls (M_age = 14.52 years, SD = 1.62). There was no significant difference for age across boys and girls, t (df = 872) = 0.140, p = 0.889. Supplementary Table S1 provides missing values, mean, and standard deviation scores, and skewness and kurtosis values for the different disorders for this group.

Measures

Adolescents completed the Self-Report Form of the Child and Adolescent PsychProfiler (CAPP-SRF; Langsford et al., 2014) that is part of the collection of measures referred to as the PsychProfiler [2]. The CAPP-SRF is used for the screening of 17 DSM-5 disorders in individuals aged 2 to 18 years: ADHDI, ADHDHI, ODD, CD, SLDR, SLDW, SLDM, ASD, LD, SSD, GAD, PDD, SAD, OCD, PTSD, AN, and BN. Considering that the skewness scores ranged from −0.64 to 5.22, and the kurtosis values ranged from −1.03 to 6.21 (see Supplementary Table S1), the data can be considered to have relatively normal univariate distribution (Brown, 2015).

Supplementary Table S1: Missing values, mean and standard deviation scores, and skewness and kurtosis values for the different disorders in the study

Note. GAD = Generalized Anxiety Disorder; SAD = Separation Anxiety Disorder, ADHDHI = Attention-Deficit/Hyperactivity Disorder: Hyperactive/Impulsive; ADHDI = Attention Deficit/Hyperactivity Disorder: Inattentive; ASD = Autism Spectrum Disorder; LD = Language Disorder; SSD =Speech Sound Disorder; PDD = Persistent Depressive Disorder; CD = Conduct Disorder; ODD = Oppositional Defiant Disorder; AN = Anorexia Nervosa; BN = Bulimia Nervosa; OCD = Obsessive-Compulsive Disorder; SLDR = Specific Learning Disorder: Reading; SLDW = Specific Learning Disorder: Written Expression; SLDM = Specific Learning Disorder: Mathematics; PTSD = Posttraumatic Stress Disorder.

Procedure

The PsychProfiler measures (including the CAPP) have a designated website (www.psychprofiler.com) that can be used by those interested in the online screening of DSM-5 disorders using any of the PsychProfiler forms. The primary users are psychologists, psychiatrists, paediatricians, and the general public. The participants involved in the study provided data through the website. On completion of the PsychProfiler, individuals and parents were requested, if they so wished, to click a statement consenting to their data being used for future research and instrument validation purposes. Only adolescents who provided consent (including their parents’ consent) were included in the study.

Statistical Analysis

Network analyses were conducted using the network module in Jeffreys' Amazing Statistics Program [27]. This module incorporates the bootnet package for network estimation and qgraph for visualisation [6]. We applied the Least Absolute Shrinkage and Selection Operator (LASSO) with a hyperparameter set at 0.05, following recommendations by Foygel and Drton (2010) and Epskamp and Fried (2018) [7,28]. This approach estimates a regularized partial correlation network by computing associations between pairs of nodes while controlling for all others, thereby identifying only the most robust relationships. Missing data were handled using listwise deletion. Network visualisation was based on the Fruchterman–Reingold algorithm, which positions nodes with stronger or more frequent connections closer together, with more central nodes placed near the centre of the layout. Positive associations were depicted as blue lines, and negative associations as pink lines, with the strength of associations reflected in the thickness and brightness of the edges.

Centrality was assessed to identify the most influential nodes in the network. In accordance with network theory, we used expected influence as the centrality index, as it accounts for the direction of edge weights and avoids some of the interpretive issues associated with other indices such as strength, betweenness, or closeness [29]. Nodes with high expected influence are understood to exert greater impact within the network, either through activating or inhibiting connected nodes. Edge weights were interpreted using effect size guidelines proposed by Christensen and Golino (2021), where values of 0.15 to 0.24 were considered small, 0.25 to 0.34 moderate, and 0.35 or greater large. Moderate and large effect sizes were prioritised for interpretation due to their relative strength.

To evaluate the network’s reliability, we assessed the stability of centrality estimates and the accuracy of edge weights using bootstrapping methods. The stability of centrality indices was examined using case-dropping bootstrapping, which involves iteratively re-estimating the network with subsets of data and calculating a correlation stability (CS) coefficient [6]. This coefficient reflects the consistency of centrality rankings when data is reduced. Although the JASP module does not report CS coefficients for expected influence, we used strength centrality as a proxy. A CS coefficient above 0.25 is considered acceptable, and values above 0.50 indicate high robustness. Edge weight accuracy was evaluated using non-parametric 95% bootstrap confidence intervals, where narrower intervals indicate more precise estimation. For both network models, these procedures were repeated using 1,000 bootstrap samples.

To explore the dimensional structure of ODD and CD symptoms within the network, we conducted an EGA. This was implemented using the Walktrap community detection algorithm within an EBIC-glasso framework [15,30]. The resulting dimensions were presented graphically, with each community of symptoms represented in a different colour.

Data availability statement

The data for this study can be accessed by contacting the corresponding author upon reasonable request.

Results

Sample Size Requirements for the Network Analysis

The mean scores for the 17 CAPP-SRF scales (see Supplementary Table S1) were used in the network analysis. Thus, a total of 171 [(18) + (18 x17/2)] parameters were estimated in network analysis [31]. For a network analysis, sufficient power can be assumed if the sample size in the network is more than the number of estimated parameters [7]. Given the sample size in the present study was 874, it can be considered adequate for the network analysis.

Network Analysis

Visualization

The maximum number of edges in the network was 136. The elasso estimation reduced the number of edges estimated to 105 (sparsity = 0.23). Figure 1 shows a visualization of the network structure for the 17 CAPP-SRF nodes in this model. As can be seen, the reflecting the EGS findings, the neurodevelopmental disorders (ADHDHI, ADHDIA, ASD, LD, SSD, SLDR, SLDW, and SLDM), anxiety disorders (GAD, SAD, OCD and PTSD), and eating disorders (AH and BN) were grouped together, but in different positions.  Also, CD had links with ODD but was associated closer with depression.

Figure1: Network of the Disorders in the Network Analysis

Note. GAD = Generalized Anxiety Disorder; SAD = Separation Anxiety Disorder, ADHDHI = Attention-Deficit/Hyperactivity Disorder: Hyperactive/Impulsive; ADHDI = Attention Deficit/Hyperactivity Disorder: Inattentive; ASD = Autism Spectrum Disorder; LD = Language Disorder; SSD =Speech Sound Disorder; PDD = Persistent Depressive Disorder; CD = Conduct Disorder; ODD = Oppositional Defiant Disorder; AN = Anorexia Nervosa; BN = Bulimia Nervosa; OCD = Obsessive-Compulsive Disorder; SLDR = Specific Learning Disorder: Reading; SLDW = Specific Learning Disorder: Written Expression; SLDM = Specific Learning Disorder: Mathematics; PTSD = Posttraumatic Stress Disorder.

Blue lines represent positive associations, and red lines negative associations. The thickness and brightness of an edge indicate the association strength. The layout is based on the Fruchterman–Reingold algorithm that places the nodes with stronger and/or more connections closer.

Centrality

The standardized estimates of the centrality indices for betweenness, closeness, strength, and expected influence are presented in Table 1.  Supplementary Figure S1 presents the plot for these. As shown in the table and the supplementary figure, in sequence, the three nodes with the highest expected influence values (used as our index of centrality) were LD, GAD, and CD.

Table1: Centrality Indices of the CD Symptoms from Network Analysis of CD Symptoms

Note. GAD = Generalized Anxiety Disorder; SAD = Separation Anxiety Disorder, ADHDHI = Attention-Deficit/Hyperactivity Disorder: Hyperactive/Impulsive; ADHDI = Attention Deficit/Hyperactivity Disorder: Inattentive; ASD = Autism Spectrum Disorder; LD = Language Disorder; SSD =Speech Sound Disorder; PDD = Persistent Depressive Disorder; CD = Conduct Disorder; ODD = Oppositional Defiant Disorder; AN = Anorexia Nervosa; BN = Bulimia Nervosa; OCD = Obsessive-Compulsive Disorder; SLDR = Specific Learning Disorder: Reading; SLDW = Specific Learning Disorder: Written Expression; SLDM = Specific Learning Disorder: Mathematics; PTSD = Posttraumatic Stress Disorder.

Note. Higher numbers indicate that the variable is more central to the network.

Supplementary FigureS1: Centrality Plots (Betweenness, Closeness, Degree, and Expected Influence) in the Disorders in the Network

Note. GAD = Generalized Anxiety Disorder; SAD = Separation Anxiety Disorder, ADHDHI = Attention-Deficit/Hyperactivity Disorder: Hyperactive/Impulsive; ADHDI = Attention Deficit/Hyperactivity Disorder: Inattentive; ASD = Autism Spectrum Disorder; LD = Language Disorder; SSD =Speech Sound Disorder; PDD = Persistent Depressive Disorder; CD = Conduct Disorder; ODD = Oppositional Defiant Disorder; AN = Anorexia Nervosa; BN = Bulimia Nervosa; OCD = Obsessive-Compulsive Disorder; SLDR = Specific Learning Disorder: Reading; SLDW = Specific Learning Disorder: Written Expression; SLDM = Specific Learning Disorder: Mathematics; PTSD = Posttraumatic Stress Disorder.

Edges

The edge weights for the nodes in network analysis are shown in Table 2. There was a total of 136 nodes across all the CAPP-SRF nodes. Of these, 30 (22.06%) were negative, and 47 (35.38%) were not connected. The remaining 58 (42.65%) were positive. Negative associations involved mainly GAD, OCD and ODD with SLDR and SLDW; and lack of connection involved mainly CD and PDD with the neurodevelopmental disorders; and CD, PDD and BN with the anxiety disorders. There were large effect size connections between ADHDHI and ADHDIA, LD and SSD, SLDR and SRDW, CD and PDD, and AN and BN; and medium effect size connections between ASD and LD, LD and SLDR, ADHDHA and ODD, and GAD and OCD. All other connections were either negligible (≤ .14) or small (≥ 0.15 to < .25).

Table2: Weights Matrix Between the Disorder in the Network Analysis

Reliability

The reliability or stability of the centrality indices, examined using case-dropping bootstrapping, is shown in Supplementary Figure S2. The figure shows that although there was a drop in the correlation stability coefficient for strength (the specific index used in the study for evaluating stability of centrality from the original sample and the subsample reduced to around 25% of the original sample, it well above .5. This indicates robustness in the stability for the strength centrality index [6]. Supplementary Figure S3 shows the accuracy of the edge weights estimated using bootstrap 95% non-parametric CIs. As shown, the 95% CI for most of the estimated edges were relatively narrow, thereby indicating robust stability of the edges in the network.

Supplementary FigureS2: Stability of Central Indices for the Disorders in Network Analysis

Note. The graph shows the average correlation between bootstrap centrality indices of networks sampled with node-dropping. A strong correlation after dropping a high percentage of participants indicate that centrality measures in the original network can be considered robust.

Supplementary FigureS3: Edge Stability Estimate for the Disorders in Network Analysis

Note. The x-axis represents the edges, while every line on the y-axis represents a specific edge. The red line shows the estimate of the edge weights, and the gray bars the 95% confidence intervals for the estimates

Exploratory Graph Analysis

Figure 2 shows the results of the EGA. As shown, the EGA identified 5 dimensions, represented by different colors in Figure 2. Dimension 1 included GAD, OCD, SAD and PTSD; dimension 2 included ADHDHI, ADHIA, ASD and ODD; and dimension 3 included LD, SSD, SLDR, SLDW, and SLSM. Dimension 4 included CD and PDD; and dimension 5 included AN and BN. Seen in the context of the original HiTOP model and Forbes’ (2025) extension of the original HiTOP model, dimension 1 included subfactors from the distress and fear spectra [32]. Except for ODD, dimension 2 was reflective of neurodevelopmental subspectra, and dimension 3 was reflective of the cognitive subspectra, respectively. Dimension 5 was reflective of the eating pathology spectrum.  Dimensions 4 was unique in that it included a distress subfactor (depression) and an antagonistic externalizing problems subfactor. Thus, our findings included spectra for internalizing, eating pathology, neurodevelopmental (with neurodevelopmental and cognitive difficulties subspectra), and a mixed depression/CD spectrum.

Figure2: Exploratory Graph Analysis (EGA) for Detecting Communities of the Disorders in the CAPP-P D

Note. GAD = Generalized Anxiety Disorder; SAD = Separation Anxiety Disorder, ADHDHI = Attention-Deficit/Hyperactivity Disorder: Hyperactive/Impulsive; ADHDI = Attention Deficit/Hyperactivity Disorder: Inattentive; ASD = Autism Spectrum Disorder; LD = Language Disorder; SSD =Speech Sound Disorder; PDD = Persistent Depressive Disorder; CD = Conduct Disorder; ODD = Oppositional Defiant Disorder; AN = Anorexia Nervosa; BN = Bulimia Nervosa; OCD = Obsessive-Compulsive Disorder; SLDR = Specific Learning Disorder: Reading; SLDW = Specific Learning Disorder: Written Expression; SLDM = Specific Learning Disorder: Mathematics; PTSD = Posttraumatic Stress Disorder.

Discussion

The first aim of this study was to examine the network properties of 17 common psychological disorders in a sample of adolescents. Network analysis revealed that the most central disorders were learning disorder (LD), generalized anxiety disorder (GAD), and conduct disorder (CD). Regarding edge weights, 42.65% were positive, 35.38% showed no connection, and 22.06% were negative. Strong associations (large effect sizes) were found between ADHD-HI and ADHD-IA, LD and SSD, SLDR and SLDW, CD and PDD, and AN and BN. Medium-strength connections were observed between ASD and LD, LD and SLDR, ADHD-HI and ODD, and GAD and OCD; all other connections were negligible. The second aim was to use exploratory graph analysis (EGA) to identify the dimensional structure of these disorders. Five dimensions emerged: anxiety disorders (GAD, OCD, SAD, PTSD), neurodevelopmental disorders (ADHD-HI, ADHD-IA, ASD, ODD), cognitive difficulties (LD, SSD, SLDR, SLDW, SLSM), eating pathology (AN, BN), and a CD/PDD cluster. These findings have important clinical and theoretical implications, including for the refinement of the Hierarchical Taxonomy of Psychopathology [33,34].

Clinical and Theoretical Implications

Given the centrality values for the nodes in network analysis, the nodes for LD, GAD, and CD can be considered as having relatively stronger influence on the other nodes in the network. Therefore, clinicians may wish to pay particular attention to the presence and severity of those disorders with high centrality values as they could be relatively more important for understanding, assessing and managing individuals with many cooccurring disorders.  Relatedly, the network centrality hypothesis suggests that the more central nodes are the most influential in a network, and that intervening on the central nodes can potentially maximize the impact of an intervention, especially on those with which they are closely connected [4,9,12,13,35,36]. Therefore, given the more central nodes in the network analysis, it could be argued that intervening on LD, GAD, and CD could have the potential (more than the other disorders) to reduce overall psychopathology when treating individuals with many cooccurring disorders. At the practical level, a network-based based treatment approach could require clinicians to view the disorders present in an individual in terms of their centrality (based on the findings reported here), and choose disorders with higher centrality values to target [35]. 

Our findings showed that only 42.65% of the edges were associated were positively, with 35.38% not connected, and 22.06% connected negative. Additionally, only 6.62% of edge weights were of either large or medium effect sizes, and all others were of small or negligible. Considering these findings, it can be speculated that there were only modest associations between the nodes.

Considering that HiTOP assumes that the building blocks of psychopathology are related to one another hierarchically [37], we expected more and stronger associations between the nodes. As we did not find this, it could be argued that the HiTOP is an incomplete overarching model for psychopathology, as lest for adolescents (who were examine in the current study). Indeed, when considered together, some features of psychopathology may be better viewed as entities outside the HiTOP model. In particular GAD, OCD and ODD with SLDR and SLDW; CD and PDD with the neurodevelopmental disorders; and CD, PDD and BN with the anxiety disorders.

Implications for the Hierarchical Taxonomy of Psychopathology (HiTOP).

The HiTOP is a is a fast-developing dimensional model of psychopathology [33,34]. In brief, the original HiTOP model proposed by Kotov et al. (2017; Ruggero et al., 2019) organizes psychopathology at five different hierarchical levels, moving upwards from narrow to broader constructs of psychopathology (i.e., signs, symptoms, and maladaptive traits of psychopathology → syndromes→ subfactors→ spectra→ superspectra). Each spectra include constellations of highly-related syndromes comprise, and they represent the most basic factors of psychopathology just below a general predisposition (g-factor) to mental disorder. The original HiTOP model identified six spectra: internalizing, somatoform, disinhibited externalizing, antagonistic externalizing, thought disorder, and detachment. More recently, Michelini et al. (2019) has provided evidence supporting a neurodevelopmental spectrum [38]. In general, the spectra can be viewed as broad dimensions reflecting factor analysis of a group of psychological disorders/syndromes.

The syndromes and disorders for the internalizing spectrum includes distress (e.g., depression, anxiety and PTSD), mania (e.g., bipolar I disorder, and bipolar II disorder)], fear (e.g., panic disorder, social phobia, and specific phobia, obsessive-compulsive disorder, separation anxiety disorder), eating pathology (e.g., anorexia nervosa and bulimia nervosa), and sexual problems (e.g., hyperactive sexual desire disorder, and delayed ejaculation) [39]. The syndromes for the externalizing spectrum include disinhibition (substance use, hyperactivity, inattention) antisocial behaviour (aggression, destruction of property, fraud, and theft), and antagonism, reflecting primarily antisocial behaviour (e.g., antisocial personality, oppositional defiant disorder, conduct disorder, narcissistic personality, and histrionic personality) [40]. The somatoform spectrum includes somatic symptom disorder and illness anxiety disorder; the detachment spectrum includes anhedonia, suspiciousness, social withdrawal, intimacy avoidance, unassertiveness, risk aversion, and restricted affectivity; and the thought disorder spectrum include psychoticism, hallucinations, and delusions [41,42]. Recently, Forbes (2025) has proposed an extension for the initial HiTOP model to include what they referred to as “neurodevelopmental and cognitive difficulties” specturm. In their model, this spectrum comprises neurodevelopmental (subfactors for social communication difficulties, altered sensation and attentional control, and ritualistic behavior), and cognitive difficulties subspectra (subfactors for neurocognitive impairment, difficulties with organization, and forgetfulness). According to them these subfactors of the neurodevelopmental and cognitive difficulties spectrum covers the features of the six neurodevelopmental disorders specified in DSM-5 (intellectual, communication, autism spectrum, attention-deficit/hyperactivity, motor, and specific learning disorders).

Seen against the original HiTOP model and Forbes’s (2025) extension of the original HiTOP model, one of our dimensions that included subfactors from the distress and fear spectra could be viewed as an internalizing spectrum.  With the exception of ODD, there was a dimension reflective of a neurodevelopmental subspectra, and another reflective of the cognitive subspectra, that together can be viewed as a neurodevelopmental spectrum. There was also a dimension reflective of the eating pathology spectrum.  Additionally, there was a dimension that included a distress subfactor (PDD) and an antagonistic externalizing problems subfactor (CD). Thus, it can be viewed as a distress/externalizing spectrum. Absent in our network were the dimensions for thought disorder, detachment, somatoform, and disinhibited externalizing, and antagonistic externalizing problems (spectra in the HiTOP model). However, this was expected as these features were not entered in the network.

The locations of ODD and CD in our network were not as predicted from the HiTOP model. In the initially published HiTOP model, ODD and CD were in the externalizing spectrum. However, our findings located ODD as was part of the neurodevelopmental spectra. Strictly speaking that is not unexpected. A comprehensive review by Wakschlag et al. (2018) concluded that the weight of evidence from developmental, clinical, and neuroscience areas indicate that ODD (and CD) are more closely related to neurodevelopmental disorders and problems [43]. For example, Kerekes et al. (2014) found strong associations for ODD and CD with autism spectrum disorder (ASD), and ADHD (HI in boys, and IA in girls) [44]. Also, genetic and environmental effects linked to ADHD and ASD also influenced ODD, especially in boys. Taking such findings into consideration, it can be argued that ODD and CD could potentially be better located in the neurodevelopmental spectrum rather than the externalizing spectrum of the HiTOP.

Despite this possibility, we found that CD was not located with the neurodevelopment disorders, but with PDD. This is also not unexpected. It has been proposed that in children and adolescents, the externalizing dimension includes behaviors that have the propensity to express distress outwards, such as CD [45,46,47]. Also, existing studies have shown high comorbidity between depression disorder and CD, and genetic studies suggest that these disorders share common genetic vulnerability factors (for a review) [48-51].  

Overall, out findings suggest that at the level of the spectra, an appropriate HiTOP model for adolescents would include spectra for internalizing (anxiety disorders), neurodevelopmental (ASD, ADHD, and ODD), cognitive difficulties (LS, SSD and SLS), eating pathology (AN and BN), and externalizing/distress (PDD and CD). In addition, while this was nor examined in the current study, other spectra could include thought disorder, detachment, somatoform, and disinhibited externalizing.

Study Strengths, and Limitations

This is the first study to use EGA to examine the properties and dimensionality of 17 common psychological disorders concurrent in a large group of adolescents, recruited from a psychology clinic. Consequently, the study provides novel, and theoretically and clinically meaningful findings regarding the psychometric properties of common psychological disorders that are relevant to researchers and clinicians.

Despite the strengths, the results of the study must be interpreted in light of a number of limitations. Although the network centrality hypothesis suggests that the more central nodes are the most influential in a network, we used cross-sectional data and therefore we cannot infer causality [4,9]. However, it is possible that the findings could be interpreted as eliminating spurious candidates for causal relations. Although participants in the study were from a psychology clinic, they were not formally diagnosed. They were identified as have the different disorders in terms of meeting DSM-5 threshold number of symptoms for the relevant disorders, based on recoding the symptoms in terms of symptoms absent (ratings of 0, 1 and 2) or present (ratings of 3, 4 and 5). Thus, there are limitations regarding the generalization of the findings to adolescents with clinical diagnoses. Similarly, as data were obtained using the CAPP-SRF, they cannot be generalized to other measures. Only one sample was examined and therefore replication studies are needed. Moreover, as the data collected were self-report they may have been confounded from common method variance.

In concluding, although our findings showed only marginal support for the HiTOP model. It supported neurodevelopmental and cognitive difficulties spectra in the Hi-TOP framework, with ADHDIA, ADHDHI, ODD, ASD, LD and SSD being part of the neurodevelopmental spectrum; and SLDR, SLDW and SLDM being part of the cognitive difficulties spectrum. At a more general level, we demonstrated the applicability of EGA and network analysis for exploring the HiTOP model.

References

1. American Psychiatric Association [APA]. (2013). Diagnostic and statistical manual of mental disorders, Fifth Edition. https://doi.org/10.1176/appi.books.9780890425596
   View at Publisher | View at Google Scholar
2. Langsford, S., Houghton, S., & Douglas, G. (2014). Child and Adolescent PsychProfiler (CAPP) v5 (oriented to DSM-5). Langsford, Houghton, Douglas and Strategy West Consulting.
   View at Publisher | View at Google Scholar
3. Kotov, Roman, Robert F. Krueger, David Watson, David C. Cicero, Christopher C. Conway, Colin G. DeYoung, Nicholas R. Eaton et al. "The Hierarchical Taxonomy of Psychopathology (HiTOP): A quantitative nosology based on consensus of evidence." Annual review of clinical psychology 17, no. 1 (2021): 83-108.
   View at Publisher | View at Google Scholar
4. Borsboom, Denny, and Angélique OJ Cramer. "Network analysis: an integrative approach to the structure of psychopathology." Annual review of clinical psychology 9, no. 1 (2013): 91-121.
   View at Publisher | View at Google Scholar
5. Boschloo, Lynn, Claudia D. Van Borkulo, Mijke Rhemtulla, Katherine M. Keyes, Denny Borsboom, and Robert A. Schoevers. "The network structure of symptoms of the diagnostic and statistical manual of mental disorders." PloS one 10, no. 9 (2015): e0137621.
   View at Publisher | View at Google Scholar
6. Epskamp, Sacha, Denny Borsboom, and Eiko I. Fried. "Estimating psychological networks and their accuracy: A tutorial paper." Behavior research methods 50 (2018): 195-212.
   View at Publisher | View at Google Scholar
7. Epskamp, Sacha, and Eiko I. Fried. "A tutorial on regularized partial correlation networks." Psychological methods 23, no. 4 (2018): 617.
   View at Publisher | View at Google Scholar
8. von Klipstein, L., Cramer, A. O. J., van Borkulo, C. D., & Vreeker, A. (2021). Network structure of symptoms of the depressive episode in bipolar disorder. Bipolar Disorders, 23(1), 40–51. https://doi.org/10.1111/bdi.12934
   View at Publisher | View at Google Scholar
9. Robinaugh, Donald J., Ria HA Hoekstra, Emma R. Toner, and Denny Borsboom. "The network approach to psychopathology: A review of the literature 2008–2018 and an agenda for future research." Psychological medicine 50, no. 3 (2020): 353-366.
   View at Publisher | View at Google Scholar
10. Bringmann, Laura F., Timon Elmer, Sacha Epskamp, Robert W. Krause, David Schoch, Marieke Wichers, Johanna TW Wigman, and Evelien Snippe. "What do centrality measures measure in psychological networks?." Journal of abnormal psychology 128, no. 8 (2019): 892.
    View at Publisher | View at Google Scholar
11. Spiller, T. R., Levi, O., Borsboom, D., & Epskamp, S. (2020). Investigating causal relationships in psychopathology networks. Psychological Methods, 25(6), 747–761. https://doi.org/10.1037/met0000250
    View at Publisher | View at Google Scholar
12. McNally, Richard J., Donald J. Robinaugh, Gwyneth WY Wu, Li Wang, Marie K. Deserno, and Denny Borsboom. "Mental disorders as causal systems: A network approach to posttraumatic stress disorder." Clinical psychological science 3, no. 6 (2015): 836-849.
    View at Publisher | View at Google Scholar
13. Rodebaugh, Thomas L., Natasha A. Tonge, Marilyn L. Piccirillo, Eiko Fried, Arielle Horenstein, Amanda S. Morrison, Philippe Goldin et al. "Does centrality in a cross-sectional network suggest intervention targets for social anxiety disorder?." Journal of consulting and clinical psychology 86, no. 10 (2018): 831.
    View at Publisher | View at Google Scholar
14. Meng, Y., Wang, X., Liu, Y., & Li, M. (2022). Comorbidity structure of psychological disorders in Chinese adolescents: A network analysis. Child Psychiatry & Human Development, 53, 540–552. https://doi.org/10.1007/s10578-021-01230-w
    View at Publisher | View at Google Scholar
15. Golino, Hudson F., and Sacha Epskamp. "Exploratory graph analysis: A new approach for estimating the number of dimensions in psychological research." PloS one 12, no. 6 (2017): e0174035.
    View at Publisher | View at Google Scholar
16. Golino, Hudson, Dingjing Shi, Alexander P. Christensen, Luis Eduardo Garrido, Maria Dolores Nieto, Ritu Sadana, Jotheeswaran Amuthavalli Thiyagarajan, and Agustin Martinez-Molina. "Investigating the performance of exploratory graph analysis and traditional techniques to identify the number of latent factors: A simulation and tutorial." Psychological methods 25, no. 3 (2020): 292.
    View at Publisher | View at Google Scholar
17. Christensen, Alexander P., and Hudson Golino. "On the equivalency of factor and network loadings." Behavior research methods 53, no. 4 (2021): 1563-1580.
    View at Publisher | View at Google Scholar
18. Santiago, P. H., Moreira, H., & Nunes, C. (2021). Is Exploratory Graph Analysis a reliable method for identifying dimensionality of mental health scales? A comparison with traditional factor analytic procedures. International Journal of Methods in Psychiatric Research, 30(3), e1884. https://doi.org/10.1002/mpr.1884
    View at Publisher | View at Google Scholar
19. Elovainio, M., Hakulinen, C., Pulkki-Råback, L., Virtanen, M., Josefsson, K., Jokela, M., & Hintsanen, M. (2018). Children's behavioral and emotional problems and early risk of adult psychiatric disorders: A population-based study of adopted children. Journal of Child Psychology and Psychiatry, 59(11), 1225–1233. https://doi.org/10.1111/jcpp.12901
    View at Publisher | View at Google Scholar
20. McElroy, Eoin, Jay Belsky, Natacha Carragher, Pasco Fearon, and Praveetha Patalay. "Developmental stability of general and specific factors of psychopathology from early childhood to adolescence: dynamic mutualism or p‐differentiation?." Journal of Child Psychology and Psychiatry 59, no. 6 (2018): 667-675.
    View at Publisher | View at Google Scholar
21. Watkeys, O. J., Carver, T., Simkiss, N., Khan, C. T., Danese, A., Goodyer, I. M., & Davey Smith, G. (2024). Mapping the structure of childhood psychopathology using exploratory graph analysis: Evidence from a UK population cohort. Journal of Child Psychology and Psychiatry. Advance online publication. https://doi.org/10.1111/jcpp.13894
    View at Publisher | View at Google Scholar
22. Stevanovic, D., Urban, R., Knez, R., Mihajlovic, G., & Epskamp, S. (2024). Developmental changes in the structure of psychopathology: A network perspective. Journal of Child Psychology and Psychiatry. Advance online publication. https://doi.org/10.1111/jcpp.13925
    View at Publisher | View at Google Scholar
23. Polanczyk, Guilherme V., Giovanni A. Salum, Luisa S. Sugaya, Arthur Caye, and Luis A. Rohde. "Annual research review: A meta‐analysis of the worldwide prevalence of mental disorders in children and adolescents." Journal of child psychology and psychiatry 56, no. 3 (2015): 345-365.
    View at Publisher | View at Google Scholar
24. Scheiner, N., Kranzler, A., Zago, A., Barzilay, R., Reardon, T., & Calkins, M. E. (2022). Psychopathology in adolescence: A network analysis of comorbidity patterns and longitudinal outcomes. Journal of the American Academy of Child and Adolescent Psychiatry, 61(2), 204–214. https://doi.org/10.1016/j.jaac.2021.02.021
    View at Publisher | View at Google Scholar
25. Solmi, Marco, Joaquim Radua, Miriam Olivola, Enrico Croce, Livia Soardo, Gonzalo Salazar de Pablo, Jae Il Shin et al. "Age at onset of mental disorders worldwide: large-scale meta-analysis of 192 epidemiological studies." Molecular psychiatry 27, no. 1 (2022): 281-295.
    View at Publisher | View at Google Scholar
26. Merikangas, Kathleen Ries, Jian-ping He, Marcy Burstein, Sonja A. Swanson, Shelli Avenevoli, Lihong Cui, Corina Benjet, Katholiki Georgiades, and Joel Swendsen. "Lifetime prevalence of mental disorders in US adolescents: results from the National Comorbidity Survey Replication–Adolescent Supplement (NCS-A)." Journal of the American Academy of Child & Adolescent Psychiatry 49, no. 10 (2010): 980-989.
    View at Publisher | View at Google Scholar
27. JASP Team. JASP. Version 0.14.1. 2020. Computer Software
    View at Publisher | View at Google Scholar
28. Foygel, Rina, and Mathias Drton. "Extended Bayesian information criteria for Gaussian graphical models." Advances in neural information processing systems 23 (2010).
    View at Publisher | View at Google Scholar
29. Robinaugh, Donald J., Alexander J. Millner, and Richard J. McNally. "Identifying highly influential nodes in the complicated grief network." Journal of abnormal psychology 125, no. 6 (2016): 747.
    View at Publisher | View at Google Scholar
30. Pons, Pascal, and Matthieu Latapy. "Computing communities in large networks using random walks." In Computer and Information Sciences-ISCIS 2005: 20th International Symposium, Istanbul, Turkey, October 26-28, 2005. Proceedings 20, pp. 284-293. Springer Berlin Heidelberg, 2005.
    View at Publisher | View at Google Scholar
31. Leme, Daniel Eduardo da Cunha, Erika Valeska da Costa Alves, Vinícius do Carmo Oliveira Lemos, and André Fattori. "Network analysis: a multivariate statistical approach for health science research." Geriatrics, Gerontology and Aging 14, no. 1 (2020): 43-51.
    View at Publisher | View at Google Scholar
32. Forbes, M. K. (2025). Expanding the HiTOP model: The neurodevelopmental and cognitive difficulties spectrum. Journal of Abnormal Psychology. Advance online publication. https://doi.org/10.1037/abn0000782
    View at Publisher | View at Google Scholar
33. Kotov, Roman, Robert F. Krueger, David Watson, Thomas M. Achenbach, Robert R. Althoff, R. Michael Bagby, Timothy A. Brown et al. "The Hierarchical Taxonomy of Psychopathology (HiTOP): A dimensional alternative to traditional nosologies." Journal of abnormal psychology 126, no. 4 (2017): 454.
    View at Publisher | View at Google Scholar
34. Ruggero, Camilo J., Roman Kotov, Christopher J. Hopwood, Michael First, Lee Anna Clark, Andrew E. Skodol, Stephanie N. Mullins-Sweatt et al. "Integrating the Hierarchical Taxonomy of Psychopathology (HiTOP) into clinical practice." Journal of consulting and clinical psychology 87, no. 12 (2019): 1069.
    View at Publisher | View at Google Scholar
35. Jones, Payton J., and Donald R. Robinaugh. "An answer to “so what?” Implications of network theory for research and practice." Focus 19, no. 2 (2021): 204-210.
    View at Publisher | View at Google Scholar
36. Ruzzano, Laura, Denny Borsboom, and Hilde M. Geurts. "Repetitive behaviors in autism and obsessive–compulsive disorder: New perspectives from a network analysis." Journal of autism and developmental disorders 45 (2015): 192-202.
    View at Publisher | View at Google Scholar
37. Conway, Christopher C., Robert F. Krueger, and HiTOP Consortium Executive Board. "Rethinking the diagnosis of mental disorders: data-driven psychological dimensions, not categories, as a framework for mental-health research, treatment, and training." Current Directions in Psychological Science 30, no. 2 (2021): 151-158.
    View at Publisher | View at Google Scholar
38. Michelini, Giorgia, Deanna M. Barch, Yuan Tian, David Watson, Daniel N. Klein, and Roman Kotov. "Delineating and validating higher-order dimensions of psychopathology in the Adolescent Brain Cognitive Development (ABCD) study." Translational psychiatry 9, no. 1 (2019): 261.
    View at Publisher | View at Google Scholar
39. Watson, David, Miriam K. Forbes, Holly F. Levin-Aspenson, Camilo J. Ruggero, Yuliya Kotelnikova, Shereen Khoo, R. Michael Bagby, Matthew Sunderland, Praveetha Patalay, and Roman Kotov. "The development of preliminary HiTOP internalizing spectrum scales." Assessment 29, no. 1 (2022): 17-33.
    View at Publisher | View at Google Scholar
40. Krueger, Robert F., Kelsey A. Hobbs, Christopher C. Conway, Danielle M. Dick, Michael N. Dretsch, Nicholas R. Eaton, Miriam K. Forbes et al. "Validity and utility of hierarchical taxonomy of psychopathology (HiTOP): II. Externalizing superspectrum." World psychiatry 20, no. 2 (2021): 171-193.
    View at Publisher | View at Google Scholar
41. Kotov, Roman, Katherine G. Jonas, William T. Carpenter, Michael N. Dretsch, Nicholas R. Eaton, Miriam K. Forbes, Kelsie T. Forbush et al. "Validity and utility of hierarchical taxonomy of psychopathology (HiTOP): I. Psychosis superspectrum." World Psychiatry 19, no. 2 (2020): 151-172.
    View at Publisher | View at Google Scholar
42. Zimmermann, Johannes, Thomas A. Widiger, Lara Oeltjen, Christopher C. Conway, and Leslie C. Morey. "Developing preliminary scales for assessing the HiTOP detachment spectrum." Assessment 29, no. 1 (2022): 75-87.
    View at Publisher | View at Google Scholar
43. Wakschlag, Lauren S., Susan B. Perlman, R. James Blair, Ellen Leibenluft, Margaret J. Briggs-Gowan, and Daniel S. Pine. "The neurodevelopmental basis of early childhood disruptive behavior: Irritable and callous phenotypes as exemplars." American journal of psychiatry 175, no. 2 (2018): 114-130.
    View at Publisher | View at Google Scholar
44. Kerekes, Nora, Sebastian Lundström, Zheng Chang, Armin Tajnia, Patrick Jern, Paul Lichtenstein, Thomas Nilsson, and Henrik Anckarsäter. "Oppositional defiant-and conduct disorder-like problems: neurodevelopmental predictors and genetic background in boys and girls, in a nationwide twin study." PeerJ 2 (2014): e359.
    View at Publisher | View at Google Scholar
45. Krueger, Robert F. "The structure of common mental disorders." Archives of general psychiatry 56, no. 10 (1999): 921-926.
    View at Publisher | View at Google Scholar
46. Krueger, R. F., & Finger, M. S. (2001). Using item response theory to understand comorbidity among internalizing disorders. Psychological Assessment, 13(2), 140–151. https://doi.org/10.1037/1040-3590.13.2.140
    View at Publisher | View at Google Scholar
47. Gomez, R., Stavropoulos, V., Vance, A., & Griffiths, M. D. (2019). A network analysis of the relationships between impulsivity and psychopathology in youth. European Child & Adolescent Psychiatry, 28(11), 1567–1575. https://doi.org/10.1007/s00787-019-01296-9
    View at Publisher | View at Google Scholar
48. Ingoldsby, Erin M., Gwynne O. Kohl, Robert J. McMahon, Liliana Lengua, and Conduct Problems Prevention Research Group. "Conduct problems, depressive symptomatology and their co-occurring presentation in childhood as predictors of adjustment in early adolescence." Journal of abnormal child psychology 34 (2006): 602-620.
    View at Publisher | View at Google Scholar
49. Olsson, G., & Hansson, K. (2009). Depressive symptoms in Swedish adolescents: A prospective study on the significance of parental gender and mood disorder. Scandinavian Journal of Psychology, 50(1), 37–44. https://doi.org/10.1111/j.1467-9450.2008.00693.x
    View at Publisher | View at Google Scholar
50. Wolff, Jennifer C., and Thomas H. Ollendick. "The comorbidity of conduct problems and depression in childhood and adolescence." Clinical child and family psychology review 9 (2006): 201-220.
    View at Publisher | View at Google Scholar
51. Caserini, A., Forresi, B., & Frigerio, A. (2023). Comorbidity between conduct and depressive disorders in youth: A systematic review of genetic, biological and psychosocial risk factors. Neuroscience & Biobehavioral Reviews, 147, 105079. https://doi.org/10.1016/j.neubiorev.2023.105079
    View at Publisher | View at Google Scholar