A key research question was whether associations with polygenic scores for MDD and loneliness persisted in people without a clinical MDD diagnosis. We defined MDD by the phecode 296.22 “Major depressive disorder”, which included both single episode and recurrent major depressive episodes that are severe, moderate, or of unspecified severity. We also defined broad categories of mental illness, to adjust for the possibility that any degree of mental illness could increase CAD risk simply due to lifestyle factors that may be correlated with having the mental illness itself. We defined a milder depressive symptoms phenotype by one or more depression ICD-9 codes (296.2 and 296.3, including all fifth digit subclassifications, and 311), and as a final test of the robustness of associations, we excluded patients with any psychiatric symptoms (Supplementary Methods). Due to limitations of collecting medication data (e.g., incomplete records, noncompliance, etc.) we did not extract data on psychiatric medications but did model their downstream consequences (i.e., increased BMI).
