University of Leicester
In search of disorders: internalizing symptom networks in a large clinical sample.pdf (342.37 kB)

In search of disorders: internalizing symptom networks in a large clinical sample.

Download (342.37 kB)
journal contribution
posted on 2019-10-21, 09:17 authored by Eoin McElroy, Praveetha Patalay
BACKGROUND: The co-occurrence of internalizing disorders is a common form of psychiatric comorbidity, raising questions about the boundaries between these diagnostic categories. We employ network psychometrics in order to: (a) determine whether internalizing symptoms cluster in a manner reflecting DSM diagnostic criteria, (b) gauge how distinct these diagnostic clusters are and (c) examine whether this network structure changes from childhood to early and then late adolescence. METHOD: Symptom-level data were obtained for service users in publicly funded mental health services in England between 2011 and 2015 (N = 37,162). A symptom network (i.e. Gaussian graphical model) was estimated, and a community detection algorithm was used to explore the clustering of symptoms. RESULTS: The estimated network was densely connected and characterized by a multitude of weak associations between symptoms. Six communities of symptoms were identified; however, they were weakly demarcated. Two of these communities corresponded to social phobia and panic disorder, and four did not clearly correspond with DSM diagnostic categories. The network structure was largely consistent by sex and across three age groups (8-11, 12-14 and 15-18 years). Symptom connectivity in the two older age groups was significantly greater compared to the youngest group and there were differences in centrality across the age groups, highlighting the age-specific relevance of certain symptoms. CONCLUSIONS: These findings clearly demonstrate the interconnected nature of internalizing symptoms, challenging the view that such pathology takes the form of distinct disorders.


This research was supported by the Wellcome Trust grant 204366/Z/16/Z. The authors would like to thank all partnerships that took part in the Children and Young People's Improving Access to Psychological Therapies (CYP IAPT) service transformation programme between 2011 and 2015 for providing the data presented here. The authors would also like to thank members of the Child Outcomes Research Consortium (CORC), its committee at the time of writing: Miranda Wolpert, Ashley Wyatt, Mick Atkinson, Kate Martin, Ann York, Alan Ovendon, Duncan Law, Julie Elliot, Isobel Fleming – and the CORC team at the time of writing: Julian Edbrooke‐Childs, Benjamin Richie, Kate Dalzell, Jenna Jacob, Jenny Bloxham, Elisa Napoleone, Victoria Zamperoni, Carin Eisenstein, Meera Patel, Andy Whale, Alison Ford, Sally Marriott, Lee Atkins, Danielle Antha, Rebecca Neale. The authors would like to especially thank Elisa Napoleone for her assistance in aiding access to and preparing this dataset. The authors have declared that they have no competing or potential conflict of interest.



Journal of Child Psychology and Psychiatry, 2019, 60 (8), pp. 897-906

Author affiliation

/Organisation/COLLEGE OF LIFE SCIENCES/Biological Sciences/Neuroscience, Psychology and Behaviour


  • VoR (Version of Record)

Published in

Journal of Child Psychology and Psychiatry


Wiley for Association for Child and Adolescent Mental Health (ACAMH)



Acceptance date


Copyright date


Available date


Publisher version


Additional supporting information may be found online in the Supporting Information section at the end of the article: Appendix S1. Methods. Table S1. Item-level means, standard deviations and 95% confidence intervals. Table S2. Mean RCADS scores pre and post propensity score matching. Figure S1. Results from tests of edge weight accuracy. Figure S2. Results from tests of centrality stability. Figure S3. Networks estimated using alternative methods. Figure S4. Regularized partial correlation networks for the three age groups. Figure S5. Bootstrapped difference tests of strength values by propensity score matched groupings. Figure S6. Networks estimated separately by gender (propensity score matched). Figure S7. Centrality values by gender



Usage metrics

    University of Leicester Publications




    Ref. manager