Shared structure of common mental disorders: Evidence from the ELSA-Brasil population study.

Abstract

The Clinical Interview Schedule-Revised (CIS-R) is widely used to assess common mental disorders (CMD) in population-based studies. However, the latent structure of CMD within contemporary psychopathology frameworks, as well as measurement invariance across sex, remains insufficiently characterized. To address this, we analyzed data from a large population-based sample from ELSA-Brasil (N = 14,764 adults aged 35-74 years), randomly split into exploratory (EFA; n = 7382) and confirmatory (CFA; n = 7382) subsamples. EFA supported a three-factor structure comprising depression-anxiety, fear-compulsive, and somatic-cognitive domains. Bifactor indices derived from the EFA solution indicated a reliable and dominant general internalizing factor. In CFA, the second-order model showed good fit but very high loadings of first-order factors on the general factor (.78-.99), indicating strong higher-order saturation. The EFA-informed symmetric bifactor did not converge. The bifactor S-1 model, anchored in the depression-anxiety domain, showed stable estimation, comparable fit, and greater interpretability. This model presented strong general-factor saturation (ECV = .80; categorical ω = .86; ωH = .79; PUC = .80), whereas residual fear-compulsive and somatic-cognitive reliability was modest (ωH-S = .28 and .16, respectively). The S-1 model yielded comparable results across sexes. Overall, supported by cross-validation in independent subsamples, these findings indicate that most CMD symptoms reflect a shared liability to emotional distress, with fear-related and somatic-cognitive symptoms providing additional but secondary information. This identified structure may be useful in future research examining associations between shared and specific CMD features and diverse clinical outcomes.