background In psychometric and behaviour genetic analyses, psychiatric symptoms are often modelled as being dependent on a latent variable. This latent variable receives genetic and environmental effects, and then acts as a common cause of the symptoms. We present an alternative way of looking at symptom covariation as arising from reciprocal relations between symptoms, where we conceptualize disorders as networks of causally coupled variables.
aim To provide a novel conceptualization of mental disorders that may be used to furnish improved methodological approaches to psychiatric and behaviour genetic research.
methods We used network analyses of symptom overlap to shed light on the structure of the symptom space, as defined in the Diagnostic and Statistical Manual of Mental Disorders (dsmiv). Subsequently, we used network properties derived from this analysis to explain empirical comorbidity patterns. Finally, we carry out simulation studies to analyse whether running a dynamic processes on the network topology yields reasonable results in view of empirical findings.
results The network topology implicitly defined in the structure of the dsm shows that a) the network features a giant component, as about 50% of the symptoms in the dsm is connected through symptom overlap, b) the network topology has small world properties, c) the distance between disorders in the network topology is inversely related to empirical comorbidity rates, and d) simulations on the network give empirically plausible results.
conclusion The application of network methodology to mental disorders appears to be promising. However, if mental disorders are indeed networks of causally coupled variables, the consequences for psychiatric research are farreaching. We discuss these consequences with special attention to the role of genetics in symptom development.