Logistic regression analyses were conducted to identify racial differences in the risk of problems at each of three levels of heavy drinking (none/low, moderate and high). Because racial/ethnic differences in heavy drinking indicators were observed even within certain levels of heavy drinking, we further adjusted for these differences by entering the heavy drinking factor score in a separate model, and in a third model, we controlled for demographics (gender, age, education, marital status, employment). To assess the role of social disadvantage, three additional models were developed, each including one indicator (poverty, unfair treatment or racial/ethnic stigma). These three models were compared to the base model to see whether the inclusion of a given social disadvantage indicator reduced the estimated adjusted odds ratio (AOR) for race/ethnicity.
