We used multivariate logistic regression models to test the association of SDs with suicidality outcomes (i.e., SI, SP, and SA). In the Yale-Penn cohort, this analysis was conducted in the full sample (N = 15,557), which included genotyped and non-genotyped individuals. Accordingly, the following covariates were considered: age, sex, and self-reported racial/ethnic groups. In the Army STARRS cohort, the logistic regression models were applied to a fully genotyped sample of participants of European descent (N = 11,235) and covariates considered were: age, sex, and the top 10 genetic PCs for population stratification adjustment. The different approaches used in Yale-Penn and Army STARRS cohorts are due to the sample characteristics and the data availability. The Yale-Penn cohort includes more than 15,000 participants but, to date, ~10,000 individuals have genome-wide data available. Approximately 80% of Yale–Penn participants report being Caucasian/White or African-American/Black not of Hispanic origin (Table 1). To avoid excluding individuals without genotype information or belonging to a racial/ethnic group not large to be analyzed separately, we decided to analyze Yale-Penn combining the full sample and correcting for self-reported racial/ethnic groups. The characteristics of the Yale–Penn participants stratified by the inclusion in the genetic analyses are reported in Supplemental Table 1.
