Measurement Invariance of the AUDIT among Gender Minority Populations

Abstract

Despite evidence for putative risk factors for hazardous alcohol use among gender minority (GM) populations, it remains unclear whether GM groups differ from cisgender groups in overall risk for hazardous alcohol use, and evidence is mixed regarding hazardous alcohol use prevalence in GM groups compared to cisgender groups. Various threats to alcohol measure validity among GM groups may contribute to this mixed evidence base, and validating these measures may help clarify this ambiguity. We used cross-sectional data from the Healthy Minds Study (final sample n = 32,843) to assess the measurement invariance of the Alcohol Use Disorders Identification Test, one of the most commonly used measures of hazardous alcohol use, across multiple cisgender and GM identity groupings. We tested both one-factor and two-factor models for model fit as well as configural, metric, and scalar invariance across all groups as well as within pairwise comparisons. The two-factor model of the AUDIT with correlated (1) consumption and (2) problems factors demonstrated superior fit to the one-factor model. For the one-factor model, we found evidence of configural noninvariance, indicating that the same factor structure could not be established across all groups. Pairwise comparisons similarly indicated poor configural fit. For the two-factor model, we were able to demonstrate configural and metric invariance (i.e., ability to equate factor structures and factor loadings) across groups, and we only found scalar noninvariance (i.e., threat to ability to equate item thresholds) on pairwise comparisons between cismasculine and cisfeminine groups as well as between cismasculine and genderqueer groups. However, with more lenient criteria for invariance, the two-factor model demonstrated adequate fit across all groups, supporting the use of the two-factor AUDIT to make comparisons among the tested gender identity groupings. Our study still contained many limitations, and both the AUDIT and alcohol measures more broadly remain mostly unvalidated among GM populations. More research is undoubtedly needed in this field.