Abstract
BACKGROUND: Major depressive disorder (MDD) and anxiety disorders (AD) have high degrees of comorbidity and show great overlap in symptoms. The analysis of covarying depressive- and anxiety symptoms in longitudinal, sparse data panels has received limited attention. Dynamic time warping (DTW) analysis may help to provide new insights into symptom network properties based on diagnostic- and disease-state stability criteria. MATERIALS AND METHODS: In the Netherlands Study of Depression and Anxiety depressive-, anxiety-, and worry symptoms were assessed four or five times over the course of 9 years using self-report questionnaires. The sample included 1,649 participants at baseline, comprising controls (n = 360), AD patients (n = 158), MDD patients (n = 265), and comorbid AD-MDD patients (n = 866). With DTW, 1,649 distance matrices were calculated, which yielded symptom networks and enabling comparison of network densities among subgroups. RESULTS: The mean age of the sample was 41.5 years (standard deviations, 13.2), of whom 66.4% were female. The largest distance was between worry symptoms and physiological arousal symptoms, implicating the most dissimilar dynamics over time. The network density in the groups, from lowest to highest, followed the order: controls, AD, MDD, and comorbid AD-MDD. The comorbid group showed strongly connected mood and cognitive symptoms, which contrasted with the more strongly connected somatic and arousal symptoms in the AD and MDD groups. Groups that showed more transitions in disease states over follow-up, regardless of the diagnoses, had the highest network density compared to more stable states of health or disease (beta for quadratic term = -0.095; P < 0.001). CONCLUSIONS: Symptom networks over time can be visualized by applying DTW methods on sparse panel data. Network density was highest in patients with comorbid anxiety and depressive disorders and those with more instability in disease states, suggesting that a stronger internal connectivity may facilitate "critical transitions" within the complex systems framework.