Correspondence between common mental disorders assessed in diagnostic interviews and health registries: Implications for genetic studies

Background – An increasing number of studies use administrative health registries. It is unknown to what extent findings on common mental disorders are generalizable to the population at large. We study the correspondence... [ view full abstract ]

Background – An increasing number of studies use administrative health registries. It is unknown to what extent findings on common mental disorders are generalizable to the population at large. We study the correspondence between structured interviews of a population-based twin sample and primary and specialist health care registries for major depressive disorder (MDD), anxiety disorders (AD) and alcohol use disorder (AUD) with regard to prevalence, comorbidity and genetic influences.

Methods – Registry data from primary and specialist care were available for 11,727 twins and diagnostic interview data for 2,271 of these. We used logistic regression analyses and biometric twin modelling to investigate the diagnostic overlap between interview, primary care and specialist care. 

Results – Most individuals meeting criteria for common mental disorders according to interview data did not have a corresponding diagnosis in the health registries. The rates of registration were higher for MDD (36% in primary care and 15% in specialist care) and AD (21% and 18%) than for AUD (3% and 7%). Comorbidity estimates were higher in the registries than in the general population. Genetic factors for the disorders were highly correlated across data sources (median r = 0.81) and close to unity for MDD and AD. 

Conclusion – Only a subset of individuals with common mental disorders as assessed at interview receive treatment. Research relying on data from health registries may substantially underestimate true prevalence and overestimate comorbidity. Diagnoses based on health registries reflect the same genetic influences as common mental disorders in the general population, indicating generalizability across data sources.