The network structure of paranoia in the general population

Bebbington and colleagues' influential study on 'the structure of paranoia in the general population' used data from the British National Psychiatric Morbidity Survey and latent variable analysis methods. Network analysis is a relatively new approach in psychopathology research that considers mental disorders to be emergent phenomena from causal interactions among symptoms. This study re-analysed the British National Psychiatric Morbidity Survey data using network analysis to examine the network structure of paranoia in the general population.

We used a Graphical Least Absolute Shrinkage and Selection Operator (glasso) method that estimated an optimal network structure based on the Extended Bayesian Information Criterion. Network sub-communities were identified by spinglass and EGA algorithms and centrality metrics were calculated per item and per sub-community.

We replicated Bebbington's four component structure of paranoia, identifying 'interpersonal sensitivities', 'mistrust', 'ideas of reference' and 'ideas of persecution' as sub-communities in the network. In line with previous experimental findings, worry was the most central item in the network. However, 'mistrust' and 'ideas of reference' were the most central sub-communities.

Rather than a strict hierarchy, we argue that the structure of paranoia is best thought of as a heterarchy, where the activation of high-centrality nodes and communities is most likely to lead to steady state paranoia. We also highlight the novel methodological approach used by this study: namely, using network analysis to re-examine a population structure of psychopathology previously identified by latent variable approaches.