Implications of the choice of method to identify major depressive disorder in large research cohorts

Clinical heterogeneity and variations in methods to identify major depressive disorder (MDD) across studies compromise replicability of research findings. This study evaluated potential implications of different MDD case definitions in a large biobank cohort.

Among Mayo Clinic Biobank participants, MDD was identified using two methods: self-report MDD in a participant questionnaire (PQ-MDD) and MDD ICD codes in the electronic health record (EHR-MDD). We examined agreement between these definitions and evaluated relationships between case agreement and participant characteristics, including MDD polygenic risk scores (PRS). Finally, we evaluated associations between different MDD case/control definitions and participant characteristics known to be related to MDD.

Among 55,656 participants, 23 % were identified as PQ-MDD cases and 17 % as EHR-MDD cases, with 85 % overall agreement (61 % case agreement) between these definitions. Among participants identified as MDD cases by one method, older and male patients, and those with lower measures of morbidity at enrollment, were less likely to be identified as cases by the other method. The strength of the associations between different MDD case/control definitions and participant characteristics varied depending on whether MDD definitions used the same source of information (i.e., EHR-only, self-report only)-resulting in stronger associations-versus different sources of information (i.e., one from EHR, one from self-report)-resulting in weaker associations.

Our results demonstrate how the methods used to identify patients with history of MDD can affect sample characteristics and risk factor associations, highlighting the importance of considering phenotype ascertainment in the interpretation of research results.