The analyses used all NESARC respondents rather than selecting those from NVDRS participating states to ensure sufficient statistical power. A separate analysis comparing the prevalence of alcohol use among NESARC participants by gender and age between participating NVDRS states and nonparticipating states revealed similar consumption patterns (Appendix 2). Logistic regression models were computed to test whether the rates of drinking in the deceased and the living sample were different after adjusting for race/ethnicity (white, black, American Indian/Alaska Native, Asian/xPacific Islander, and Hispanic) and AP. Models were stratified by age and gender. Estimates and standard errors from NESARC were weighted to reflect the complex sampling design and adjusted for nonresponse and poststratification using SUDAAN 11.0 [18].
