Description
Alcohol consumption amongst adolescents continues to be a major public health concern in the United States. State and federal policies have been designed and implemented with the intention of curbing problematic drinking behaviors amongst the adolescents identified as being at the highest risk. These policies are often informed by surveys such as the Youth Risk Behavior Survey (YRBS), sponsored by the Center for Disease Control (CDC). Previous research, however, has found that such surveys lack invariance (or measurement equivalency) across different groups, such as race/ethnicity and gender. A lack of invariance indicates that valid group comparisons cannot be made, since group membership determines perception of the survey question. This project examined the factorial invariance of the 2013 YRBS alcohol-related behaviors across race/ethnicity and gender. Multigroup confirmatory factor analysis statistically tested the psychometric and factorial invariance of the "Alcohol-Related Behaviors" latent variable across four ethnic and two gender groups. Invariance was examined across Hispanic/Latino (n = 1,734), white (n = 5,449), Asian (n = 491), and black/African American (n = 2,991) adolescents, and across males (n = 5,439) and females (n = 5,224). The latent variable was indicated by five questions, representing "drinking and driving," "lifetime alcohol use," "30 day alcohol use," "binge drinking," and "maximum number of drinks, 30 days." Invariance was first examined across gender groups to determine if subsequent tests could be done collapsing across gender. While the models achieved configural invariance, metric invariance could not be established. This indicated that associations between measured variables (e.g., binge drinking) and the latent construct (i.e., "Alcohol-Related Behaviors") differed between males and females. Subsequent tests of invariance were therefore done within gender groups, comparing race/ethnicity groups in pairs. Configural invariance was established across race/ethnicity groups for both genders. However, metric invariance could only be achieved between certain groups. Results suggest that comparisons made between gender groups on alcohol-related behaviors as a construct may be inappropriate. Additionally, even when making comparisons within gender, caution should be taken when making comparisons between race/ethnic groups. Findings highlight the need to tailor alcohol prevention and intervention efforts to particular groups based on group membership.
