Data-driven subtypes of late-life depression-secondary analysis of a cluster-randomized RCT.

Objectives: Cognitive performance varies among depressed older patients and is known to affect treatment outcome. We used a data-driven approach to create subtypes of late-life depression and exploratively visualized the course of depression during either behavioural activation (BA) or treatment as usual (TAU).

Method: Among 161 depressed (PHQ ≥ 10) older (≥65 years) participants of a cluster randomized controlled trial in primary care (NL5436), we performed latent class analysis (LCA) on individual depressive symptoms and performance on several cognitive tests. The course of depressive symptoms during treatment was plotted to explore whether differences between the classes differed between BA and TAU.

Results: Five classes best fitted the data: (1) mild depression without cognitive deficits, (2) moderate depression with insomnia and cognitive deficits, (3) severe depression with cognitive deficits, (4) moderately severe depression with hypersomnia and cognitive deficits, and (5) moderately severe depression with cognitive-affective symptoms but no cognitive deficits. Graphs showed that depressive symptoms of subgroups with severe depressive symptoms improved more during BA compared to TAU, regardless of cognitive deficits.

Conclusions: We identified five subgroups. Graphs suggest that effectiveness of BA is similar across all subgroups, whereas TAU seems less effective in the more severely depressed subgroups. Replication is warranted.