To assess the effects of sociodemographic and clinical characteristics on probability of SUD treatment contact among those with lifetime diagnoses of SUD, Cox proportional hazards models with time-varying covariates were performed. Retrospective follow-up time started at age of disorder onset and terminated at age of first treatment contact or remission of the disorder. The probabilities of treatment-seeking for alcohol abuse, alcohol dependence, drug abuse, and drug dependence were first modeled separately for each individual sociodemographic and diagnostic predictor and again in a single model that controlled for the potentially confounding effects of sex, race/ethnicity, nativity, age at disorder onset, education years, marital status, and each of the other Axis I and II categories. Comorbid mental disorders, marital status, and educational level were also added as time varying variables. Comorbid disorders were coded as absent until their first occurrence, and then coded as present until the observation was censored. Marital status was coded as single until the individual was married for the first time (or coded as single until the observation was censored) and then modified each year the individual changed
