Using network analysis to identify processes of change in low-intensity CBT interventions for depression and anxiety disorders

This study explores processes of change for individuals who responded to low-intensity Cognitive Behavioural Therapy (CBT) for depression, Generalised Anxiety Disorder (GAD), or panic disorder. Routinely collected data from NHS Talking Therapies for Anxiety and Depression (TTad) services (N = 11, 396, 69.2% female) were analysed using network analyses. Nine Gaussian graphical models (GGMs) were conducted: for each disorder, across three time phases (assessment to start-of-treatment; start to mid-point of treatment; mid-point to end-of-treatment). Each GGM included 19 nodes, based on PHQ-9, GAD-7,and NHS TTad phobia scores, using residuals as indices of change for each node. Networks of symptom change were largely similar. Estimated network matrix similarity ranged between r = .74 and r = .91 across disorders, with depression and GAD networks more similar to each other than to panic disorder. Networks varied over time within the same disorder, more so for panic disorder (r = .61–.63) than GAD (r = .86–.90) or depression (r = .87–.93). There were close links between changes in worry-related items and feeling nervous or anxious, and between depressed mood and anhedonia across all networks, as well as links between sleep disturbance, appetite, trouble relaxing and irritability. Findings suggest shared patterns of co-change across anxiety and depression. There is a potential indication that therapy may work by leveraging existing natural change mechanisms rather than by creating entirely new patterns of symptom interaction. Networks also show associations between symptom changes specific to certain disorders at certain points in therapy.