Although the prevalence of measures of substance use across gender, race/ethnic group, and age group varies, it is markedly higher among individuals with psychotic illness within each group. We tested whether the association between substance measures and case/control status is stronger within each group (i.e. whether there is a statistical interaction between case/control status and group). All the items measured showed similar patterns of association. For illustrative purposes, Figure 1 illustrates select associations between substance use and severe psychosis. In groups where the controls had lower rates of substance use (Asians and Hispanics relative to European Americans, females relative to males, and individuals under age 30 relative to 30-50 year olds), the odds ratios were significantly higher than the reference odds ratios, leading to statistically significant interaction effects. This suggests that although belonging to certain groups may be protective for substance use in the control population, this protective effect is lost with the development of a severe mental illness. For example women are at lower risk for using recreational drugs in the general population (7% of women versus 18% of
