|| Checking for direct PDF access through Ovid
The current study extended previous work on the Health Behavior Inventory–20, a multidimensional measure of health risk and promotion behaviors, in several ways. First, participants were diverse in gender, including both cisgender and transgender persons. Second, dimensionality was reassessed using exploratory factor analysis, which revealed four factors: Proper Use of Health Care Resources, Diet, Anger and Stress, and Substance Use. Third, based on these results, we applied and optimized classical test theory to develop the 12-item Health Behavior Inventory–Short Form. Fourth, the dimensionality of the Health Behavior Inventory–Short Form was assessed using confirmatory factor analysis on a separate sample, finding that the items loaded on the four specific factors as expected, and that the common factors model fit better than bifactor, hierarchical, and unidimensional models. This finding supports the use of the four subscale scores but not the use of a total score. Fifth, measurement invariance was assessed across five gender identity groups: cisgender men and women, transgender men and women, and nonbinary individuals. Evidence emerged for full configural and metric invariance for all factors, and for partial scalar invariance. Sixth, concurrent evidence emerged for the validity of the four specific factors by examining relationships with the Duke Health Profile. The results are discussed in relationship to previous literature, future research directions, applications to practice, and limitations. Data (N = 1,233, 34.3% transgender) were from community and college participants who responded to an online survey.