Cox proportional hazards (PH) regression analyses were conducted to predict alcohol use initiation as a function of childhood trauma exposure, with interpersonal trauma and more broadly defined trauma examined in separate models, using Stata [38]. This survival analysis approach accounts for the possibility that participants who have not yet initiated alcohol use may do so in the future. It also involves testing the PH assumption that risk remains constant over time, using the Grambsch and Therneau test of the Schoenfeld residuals [39]. To ensure that only traumatic events preceding alcohol use initiation were treated as predictors, childhood trauma was modeled as time-varying. A “person-year” data set was created with each line of data representing a single year in the participant’s life and childhood trauma was coded as absent in the years prior to its onset (based on calculated age at first exposure) and present from the year of first occurrence onward.
