We used ICD-9 and ICD-10 codes to identify phenotypes. We used established algorithms to identify alcohol use disorder (AUD; e.g., ICD-9: 291*, 303*, and ICD-10: F10*, etc.),18 epilepsy (e.g., ICD-9: 345*, and ICD-10: G40*),19 hemorrhagic stroke (HS; e.g., ICD-9: 430, 431, 432*, and ICD-10: I60*, I61*, I62*),20 and traumatic brain injury (TBI; e.g., ICD-9: 850*, 851*, and ICD-10: S06*).21 We used a validated algorithm to identify Alzheimer’s disease (AD; e.g., ICD-9: 331.0, ICD-10: F00*, G30*).22 We used the R package comorbidity to identify AIDS, cancer, cerebrovascular disease, congestive heart failure, chronic pulmonary disease, diabetes, dementia, depression, hemiplegia or paraplegia, liver disease, myocardial infarction, peptic ulcer disease, peripheral vascular disease, other neurological disorder, psychoses, renal disease, and rheumatoid arthritis.23 We used the drug name look up table from the data source and RxNorm to assess drug exposure.24
