Using a hybrid subtyping model to capture patterns and dimensionality of depressive and anxiety symptomatology in the general population

Background: Researchers have tried to identify more homogeneous subtypes of major depressive disorder (MDD) with latent class analyses (LCA). However, this approach does no justice to the dimensional nature of psychopathology. In addition, anxiety and functioning-levels have seldom been integrated in subtyping efforts. Therefore, this study used a hybrid discrete-dimensional approach to identify subgroups with shared patterns of depressive and anxiety symptomatology, while accounting for functioning-levels.

Methods: The Comprehensive International Diagnostic Interview (CIDI) 1.1 was used to assess previous-year depressive and anxiety symptoms in the Netherlands Mental Health Survey and Incidence Study-1 (NEMESIS 1; n=5583). The data were analyzed with factor analyses, LCA and hybrid mixed-measurement item response theory (MM-IRT) with and without functioning covariates. Finally, the classes' predictors (measured one year earlier) and outcomes (measured two years later) were investigated.

Results: A 3-class MM-IRT model with functioning covariates best described the data and consisted of a 'healthy class' (74.2%) and two symptomatic classes ('sleep/energy' [13.4%]; 'mood/anhedonia' [12.4%]). Factors including older age, urbanicity, higher severity and presence of 1-year MDD predicted membership of either symptomatic class vs. the healthy class. Both symptomatic classes showed poorer 2-year outcomes (i.e. disorders, poor functioning) than the healthy class. The odds of MDD after two years were especially increased in the mood/anhedonia class.

Limitations: Symptoms were assessed for the past year whereas current functioning was assessed.

Conclusions: Heterogeneity of depression and anxiety symptomatology are optimally captured by a hybrid discrete-dimensional subtyping model. Importantly, accounting for functioning-levels helps to capture clinically relevant interpersonal differences.