A complex systems model of temporal fluctuations in depressive symptomatology

Symptom fluctuation is one of the hallmarks of major depression. However, a principled, quantitative framework explaining when and how symptoms change remains elusive. Following a complex systems perspective, we model clusters of depressive symptomatology across N = 2059 longitudinal biweekly measurements using a mobile phone app from N = 113 patients based on Beck Depression Inventory-I (BDI-I). We show that 1) the five largest clusters cover more than 60% of all symptom profiles, 2) in close to 90% of all cases, major temporal fluctuations in depression scores (i.e. sum of BDI-I items) coincides temporally with the change between the symptom clusters, 3) the probability of an approaching transition increases the further away the symptoms move from the center of the clusters (r = 0.78, p < 0.008), and 4) the approaching symptoms clusters are those that are closest to the BDI measurement at any given moment (r = 0.42, p < 0.0001). Taken together these results support the notion that symptom fluctuations can be predicted with high precision, potentially opening the possibility to augment interventions by predicting upcoming shifts in symptom structures.