Depression and anxiety are highly common in adolescents, with a 12-month prevalence of 10% of any mood disorders and 24.9% of any anxiety disorders in adolescents in the US (Kessler et al. 2012
), and are often comorbid (Garber and Weersing 2010
). This comorbidity causes further debilitation than depression or anxiety alone; it results in higher suicidality, poorer prognosis, worse treatment outcomes, lower life satisfaction, more physical health problems, less likelihood to attend college, greater overall impairment, and academic difficulties (Newman et al. 1998
; Cummings et al. 2014
; Schoevers et al. 2005
). As such, prevention and reduction of comorbidity is important, yet at the same time, necessitates a deep understanding of the inter-relationship between depression and anxiety (Jones et al. 2019
). Importantly, the mechanisms that underlie this co-occurrence, to date, are poorly understood (Karlsson et al. 2006
). Delineating symptom to symptom relationships and the relevance of risk and protective factors might constitute a sound step toward understanding these mechanisms.
The network approach is considered a particularly useful method to investigate comorbidity; it conceptualizes comorbidity as resulting from mutual interactions among symptoms (Cramer et al. 2010
). Symptoms that are highly associated with both depression and anxiety can be thought of as bridge symptoms, as they are considered active and reinforcing between the two disorders (Fried et al. 2017
). Bridge symptoms are thus thought to contribute to the development and maintenance of comorbidity between disorders (Cramer et al. 2010
). As a result, network modelling allows for identifying specific pathways, at the level of symptoms, through which disorders interact with each other (Cramer et al. 2010
). This implies that interventions targeting bridge symptoms may help reduce comorbidity (Borsboom and Cramer 2013
A handful of network studies examined the associations specifically between depression and generalized anxiety symptoms. These studies showed that several bridge edges (i.e. associations between two symptoms belonging to different disorders) existed between the disorders. The results from two adult samples (Cramer et al. 2010
; Beard et al. 2016
) and one adolescent sample (McElroy et al., 2018
) were comparable in that the depression symptom of “sad mood” and anxiety symptom of “excessive worrying” were among the most influential bridge symptoms. Depression symptom of “guilt” was also a bridge symptom in two of these studies (McElroy et al., 2018
; Beard et al. 2016
). However, a common limitation across these studies was that some of the items in the measures were highly similar either within the same measure (e.g. anxiety symptoms of “too much worry” and “unable to control worry”, or the depression symptoms of “trouble sleeping” and “sleeps less than most children”) or between measures (e.g. symptom of “feeling restless” being measured in both anxiety and depression measures). An overlap in symptoms can artificially inflate edge weights and centrality indices (Beard et al. 2016
Assessing how relevant risk or protective factors can affect symptom networks and bridge symptoms is also important (Fried and Cramer 2017
). Cramer et al. (2010
) suggested that “etiological factors” such as stressful life events can provide insight into the complex mechanisms underlying bridge symptoms between anxiety and depression. However, only a few studies included risk factors in network models to date (Pereira-Morales et al. 2019
; Schellekens et al. 2020
). For example, Schellekens et al. (2020
) investigated how risk and protective factors were interconnected with symptoms of depression, a fatigue sum score and one anxiety symptom (“I felt fearful”) in cancer patients. Their results showed that risk factors of helplessness and physical symptoms were positively associated with depression symptoms and fatigue. They also found that the protective factor of illness acceptance was negatively associated with the anxiety symptom and depression symptoms. These results shed light on the pathway to depression in cancer patients. Similarly, Pereira-Morales et al. (2019
) examined the associations between different types of childhood maltreatment, personality traits, and a few selected depression and anxiety symptoms. Their results indicated that sexual abuse in childhood was associated with anxiety symptoms of being unable to relax and feelings of tingle. Emotional neglect in childhood was positively associated with anxiety symptoms of worrying and feelings of tingle, and negatively associated with the personality trait of openness. While these studies examined some depression and anxiety symptoms together with other psychological distress items, no study to date investigated how relevant risk factors and protective factors interacted with depression and anxiety symptoms specifically, or the role these factors played in relation with comorbidity of these disorders.
In the present research, we included risk factors that affect both anxiety and depression symptoms. We examined these risks factors at 13 years of age, as early adolescence is a common age of onset for anxiety and depression symptoms (Kessler et al. 2007
) and adolescent-onset is associated with poorer prognosis and stronger impairment (Copeland et al. 2014
; Dunn and Goodyer 2006
). We examined peer victimization, bullying peers, low quality of peer relationships, and experiencing stressful life events (Stapinski et al. 2015
; Ivarsson et al. 2005
; Oppenheimer and Hankin 2011
; Young and Dietrich 2015
). We also included protective factors for the onset of anxiety and depression. These factors included parental knowledge of child whereabouts, adolescent disclosure/discourse with caregivers about their whereabout, and prosocial behaviors towards peers and adults (Garthe et al. 2015
; Hamza and Willoughby 2011
; La Greca and Harrison 2005
Therefore, we aimed to explore, by network analytic methods, which bridge symptoms and risk/protective factors play an important role in the co-occurrence of depression and anxiety in a large community sample of adolescents. We then explored how the initial network changed in terms of structure and bridge centrality statistics after introducing several risk/protective factors. We thus aimed to examine how risk/protective factors contribute to high levels of comorbidity between depression and anxiety in adolescents (Garber and Weersing 2010
). Based on previous research (McElroy et al., 2018
; Beard et al. 2016
; Cramer et al. 2010
), we expected depression symptoms related to “guilt” and “sad mood” to be among the most influential bridge symptoms. Since GAD worry symptoms have not previously been examined in this context, we did not have a priori hypotheses regarding the GAD worry symptoms with the highest bridge properties. In addition, we expected all of the risk factors included (e.g. peer victimization/bullying, peer relational problems, and SLEs) would have several associations with symptoms of both disorders since these are well-established risk factors for depression and anxiety in youth (Stapinski et al. 2015
; Oppenheimer and Hankin 2011
; Young and Dietrich 2015
). Finally, in line with prior studies (La Greca and Harrison 2005
; Garthe et al. 2015
) we hypothesized the protective factors (e.g. prosocial behavior and parental monitoring) would have negative associations with the symptoms of depression and anxiety.
In this study, we aimed to examine bridge symptoms between depression and non-overlapping anxiety symptoms (i.e. GAD worry symptoms) as these bridges can say a lot about how one disorder is actively associated with – or can innervate –another disorder. In addition, we explored the risk (and protective) factors that may have facilitated (and impeded) these associations. While this study is the first to specifically examine the bridge symptoms of depression and anxiety in adolescents, we based our expectations on existing network research on adults and adolescents (Beard et al. 2016
; Cramer et al. 2010
, McElroy et al., 2018
). This study is also one of the first in this line of research to show that certain risk factors played an important role in the comorbidity of depression and anxiety as well as to identify an influential protective factor. These results contribute to the existing knowledge of comorbidity between depression and anxiety in three main ways.
Firstly, in Step-1, we identified the most influential bridge symptoms for depression (“feeling unhappy”, “feeling lonely”) and GAD worries (“worrying about past”, “worrying about future”). It is important to note that even though “feeling lonely” is not a depression symptom listed in the versions of the DSM, it has been included in several other depression questionnaires as a theoretically relevant item (e.g. Child Behavior Check List, Birleson’s Depression Self Rating Scale, and Children’s Depression Scale). Previously, research has found other non-DSM symptoms to be more influential than some of the DSM symptoms (Fried et al. 2016
) and assessing the importance of non-DSM depression symptoms were advised (Fried and Nesse 2015
). Our results support the idea that examining relevant non-DSM depression symptoms might reveal more insights into the nature of depression. Overall, the said most influential bridge symptoms remained essentially unchanged in Step-2, suggesting these symptoms may be maintaining comorbidity between depression and anxiety by innervating inter-disorder associations (Cramer et al. 2010
), even after controlling for the effects of risk factors and protective factors included in Step-2.
The cross-sectional nature of this study precludes interpretating the directionality of the bridge pathways (Boschloo et al. 2016
). Accordingly, the edges present in this study could be reflecting unidirectional relationships of either direction or bidirectional relationships which were previously shown to exist between depression and anxiety (Jacobson and Newman 2017
). Hence, future research will want to explore directional inferences in network models. To this end, the edges between depression symptoms of “hating self” and “feeling lonely” and GAD worry symptom of “worrying about past” may be interpreted as worrying about past events leading one to have negative feelings toward themselves, as well as to feel lonely, which might represent an important pathway to comorbidity. Equally, feeling lonely may result in negative feelings about the self and prompt worries about past behavior that have led to this. Interestingly, rumination, a process involving perseveratively thinking about one’s problems and their causes, is also found to strongly associate with hating self (Flett et al. 2020
) and can even be involved in a loss of social support (Nolen-Hoeksema et al. 2008
). In addition, as rumination increases, the association between depressed and anxious mood gets stronger (Starr and Davila 2011
). Accordingly, this finding is in line with the existing literature given the close relationship between rumination and worries about the past.
Secondly, we examined the role of several risk factors and protective factors with regard to these bridge symptoms. Peer relational problems and SLEs were the risk factors that exhibited the highest bridge properties. For example, GAD worry symptom of “worrying about future” and depression symptom of “feeling tired” were directly associated in Step-1, but they became associated via peer relational problems in Step-2. Accordingly, it is possible that peer relational problems can act as a common cause for depression and GAD worry symptoms; or that the associations between these symptoms were facilitated by this risk factor (Epskamp and Fried 2018
In addition to facilitating relationships between symptoms, a risk factor can influence the comorbidity between the disorders due to having several associations with multiple symptoms of both mental health problems. Here, peer relational problems was strongly associated to depression symptoms related to feelings of loneliness and low self-esteem; and the GAD worry symptom of “worrying about future.” This may be interpreted as having problems with peers leading to both feelings of loneliness/worthlessness and worries about the future, hence adding to the direct association between these symptoms and contributing to comorbidity. Conversely, having peer problems due to having more severe depression/anxiety symptoms is equally plausible. Indeed, previous research indicated depression and anxiety symptoms may lead to peer relational problems (Oppenheimer and Hankin 2011
) and vice versa (La Greca and Harrison 2005
) without being able to provide insight into which specific symptoms might play the most important role in this phenomenon.
SLEs showed similar results as peer relational problems. Anxiety symptom of “worrying about bad things happening to others” and depression symptom of “feeling unhappy” became associated via SLEs in Step-2. This may be interpreted as experiencing SLEs may lead one to both experience low mood and worry about the safety of the loved ones. That is plausible given SLEs include loved ones getting harmed and one might worry about experiencing a similar event in the future and feel unhappy. SLEs also had edges with multiple depression symptoms of “feeling unhappy”, “feeling restless”, “feeling tired”; and GAD worry symptoms of “worrying about past” and “worrying about bad things happening to others.” Thus, SLEs may lead to exacerbations in all of these symptoms, some of which also inter-related with each other. Accordingly, in line with previous research (Young and Dietrich 2015
; Van Veen et al. 2013
), it may be hypothesized that experiencing SLEs exacerbates symptoms of both disorders and contributes to comorbidity by functioning like a common cause.
Thirdly, our results showed that prosocial behavior acted as a protective factor; it was negatively associated with a large number of depression and GAD worry symptoms. This indicates engaging in prosocial behavior may function as an important protective factor by decreasing symptom severity (Kramer et al. 2014
), or that individuals less frequently engage in prosocial behavior as symptom severity increases (Broeren et al. 2013
). Prosocial behavior was also positively associated with other protective factors of parental knowledge and child disclosure, which may be enhancing its protective effect indirectly. Interestingly, prosocial behavior was positively associated with GAD worry symptom of “worrying about bad things happening to others.” That suggests while engaging in prosocial behavior and caring about others is generally a protective factor, this characteristic might also predispose one to worry about safety of others.
Our results have several clinical implications. Depression symptoms of “feeling lonely” and “feeling unhappy”, and GAD worry symptoms of “worrying about past” and “worrying about future” emerged as the most important symptoms in enabling and maintaining the comorbidity between depression and anxiety. Accordingly, targeting these bridge symptoms, thus preventing the interactions between the disorders, may be a good strategy to treat comorbid depression and anxiety (Fried et al. 2017
). Similarly, our results suggest that peer relational problems and SLEs were highly influential in contributing to the co-occurrence of depression and anxiety. Hence, experiencing these may put adolescents in higher risk of developing comorbid depression and anxiety. Thus, prevention of peer relational problems and providing early intervention to adolescents who experienced peer relational problems and SLEs may lead to decreases in comorbidity.
Finally, our findings should be considered in light of a number of limitations. Firstly, network analysis has been criticized on a number of grounds, including failure to replicate within and between samples, being non-causal, and that edges can result by chance (Forbes et al. 2017a
; Steinley et al. 2017
; Forbes et al. 2021
). However, these critics themselves, have been criticized, on a number of grounds, including estimating unsuitable network models for the given data, using different network models that result in different network structures, and ignoring the impact of sampling variability on the results of network models (Borsboom et al. 2017
; Epskamp et al. 2018a
; Fried et al. 2020
; Jones et al. 2020
). Secondly, the data analyzed here is cross-sectional as these risk and protective factors, and symptoms of anxiety and depression were present together only at age 13. Additional research should test if the associations identified here replicate in longitudinal network analysis that spans early- to late-adolescence. Thirdly, the variability of the items was low due to the low symptom and risk factor endorsements of the non-clinical sample, which was argued to possibly influence edge strengths (Terluin et al. 2016
). Finally, most of the measures used were parent-reported and that might have limited the capacity to capture the true states of the psychological constructs.
These limitations notwithstanding, our study is first to specifically examine bridge symptoms/edges between depression and anxiety in a large community sample of adolescents and to investigate how risk and protective factors are relevant to these associations. Examining these associations around 13 years of age may be particularly informative given age of onset of depression and anxiety is in early adolescence (Kessler et al. 2007
) and the associations around this time point may be setting in place the patterns for long term comorbidity. Thus, our findings provide first steps for understanding symptom to symptom associations between depression and anxiety, while also considering prominent risk/protective factors.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.