Is symptom connectivity really the most important issue in depression? Depression as a dynamic system of...

According to the network theory strong associations between symptoms drive the disease process. We compared those with and without diagnosed depressive disorders (DD+/DD-) and analysed the effects of differences in (a) network connectivity, (b) symptom thresholds, and (c) autoregressive loops (i.e. how strongly specific symptoms predict themselves) on the potential activation of symptoms over time using simulations developed by Cramer and others (2016). The parameters for the simulation (symptom connectivity and symptom threshold) were obtained from Ising models and cross-lagged panel network analyses. Data were from the nationally representative samples (Health 2000-2011 Study) of 4190 participants measured in 2011 (cross-sectional analyses) and 3201 participants measured in 2000 and 2011 (longitudinal analyses). DD diagnosis was based on the Composite International Diagnostic Interview and depressive symptoms were self-reported using the 13-item version of the Beck Depression Inventory (BDI). Differences in symptom connectivity between participants with and without DD were not observed, but the mean probability (threshold) of symptom existence in the DD + group was higher than in the DD-group (0.41 vs. 0.12). Simulation showed that there are more active symptoms in the DD + group after 10 000 time points (means 1.2 vs. 4.6) than in the DD-group. This difference largely disappeared when we used longitudinal networks, including autoregressive loops, in the connectivity matrix. Our results suggest that the differences in symptom thresholds and autoregressive loops may be more important features than symptom connectivity in differentiating people with and without DD.