Abstract
Background
Based in centuries of historical contingencies, the assumption that mental disorders are latent “disease entities” has been codified in diagnostic manuals since 1980. In scientific research, this has led to a proliferation of the variable model of mental disorder. Recently, it has been argued that there is a lack of sound theory underpinning the latent variable model. In response to this, the network theory of mental disorders has emerged as an alternative theoretical framework for studying mental disorders. Under network theory, mental disorders are complex systems which emerge from causal relations between symptoms. It is argued that network models may be theoretically more plausible than the latent variable model.
Objectives
There are three overarching objectives to the present thesis. First, to review the network theory of mental disorders against its historical context. Second, to report three empirical network studies. These three stand-alone network studies each have a primary exploratory aim: (1) Identify childhood anxious/withdrawn behaviours (7–9 years) which predict later anxiety disorder (14–21 years and 21–40 years); (2) Estimate a psychosocial complex system within which internalising disorders (major depression and anxiety disorder) are embedded during late adolescence (16–18 years); and (3) Identify childhood anxious/withdrawn behaviours (7–9 years) that predict adolescent neuroticism items (14 years). The third overarching aim of this thesis is to critically review the network theory and network psychometric literature.
Methods
A narrative literature review contextualises network theory against relevant scientific history. Each empirical network analysis uses data gathered from the Christchurch Health and Development Study database (original cohort N = 1,265). All network models are bootstrapped to assess accuracy and robustness, and various supplementary analyses are performed to appropriately address research questions.
Results
In Study 1, a preponderance of social and emotional childhood anxious/withdrawn behaviours predicted later anxiety disorder. Conversely, situational fears and anxiety around adults did not tend to predict later anxiety disorder. In Study 2, a well-connected psychosocial system was estimated (33/105 model parameters were non-zero). All but three psychosocial risk factors related to major depression or anxiety disorder either directly or indirectly. In Study 3, 18 childhood anxious/withdrawn behaviours predicted adolescent neuroticism items. Of these pairwise associations, only two linked items with explicit content overlap.
Conclusions
Network psychometrics appears both a valid and promising framework for studying mental disorders. The empirical studies presented in Chapters 5–7 demonstrate the application of exploratory network analysis to three areas of research interest. From Study 1, it is suggested that precocious exhibition of social and emotional anxious/withdrawn behaviours may be more indicative of later pathology than exhibition of developmentally typical behaviours, such as situational fears. In Study 2, a well-connected network suggests that the psychosocial environment of internalising disorders is complex. One identified conditional relationship was between childhood anxiety/withdrawal and adolescent neuroticism. Study 3 investigated this association by modelling it at the item level. Results suggest that the continuity between childhood anxiety/withdrawal and adolescent neuroticism may be explained by both content overlap and true causal relations. The strengths, limitations, and future directions of network psychometrics are discussed, and the contributions to knowledge made in the present thesis are summarised.