The second goal of the statistical analysis was to determine the impact of having specific child/adolescent diagnoses on subsequent development of substance use disorders in young adulthood controlling for familial risk status. To do this, stepwise Cox regression models (Cox and Oakes, 1984) were used to estimate the influence of potential moderators (presence of specific childhood disorders) that might influence the relationship between familial risk and the incidence rate of a psychiatric outcome in young adulthood (SPSS version 16; SPSS, Chicago, Illinois). The Cox models were run evaluating each possible predictor (Conduct Disorder, ADHD, MDD, and Anxiety Disorders) and each outcome: (1) Alcohol Abuse and Dependence; (2) Drug Abuse and Dependence; (3) Alcohol or Drug Abuse and Dependence (substance use disorder [SUD]). Models were tested controlling for risk to evaluate the increase in the odds of developing each of these three outcomes in association with the predictors.
