Psychometric assessment and exploratory scale refinement of the Generalized Anxiety Disorder 7-item scale among adolescents and young adults in a Swedish context

Abstract Background The Generalized Anxiety Disorder 7-item scale (GAD-7) could provide currently lacking data on anxiety levels among adolescents and young adults in Sweden. However, psychometric assessments of the scale are scarce in this population. Additionally, the structural validity of the scale needs to be examined further, given conflicting findings in other populations. The aim of this study was to psychometrically evaluate the GAD-7 among 15–29-year-olds using Rasch analysis and confirmatory factor analysis in a Swedish, general population sample. In addition, an iterative process explored if revised versions of the scale could be identified that fit the Rasch model in this sample. Methods A random, age- and sex-stratified sample of 15–29-year-olds from the Swedish population register was drawn, from which stratified groups were randomly invited to participate. Data ( n = 590) were collected through a study-specific app. Rasch analyses included investigation of item fit, local independence, principal components analysis of standardized Rasch model residuals, response category functioning, measurement invariance, scale-to-sample targeting, and reliability. In addition, confirmatory factor analyses (CFA) were conducted and interpreted using dynamic, data-adaptive cutoffs. An exploratory, iterative process was undertaken aiming at scale refinement based on results for the GAD-7. Results No Rasch analyses provided support for the unidimensionality of the GAD-7, which also displayed issues with disordered response categories and measurement invariance. The CFA found substantial misfit for the one-factor model. The exploratory scale refinement resulted in two abbreviated scales warranting further investigation: Items 1–3 fit the Rasch model as a unidimensional scale and demonstrated acceptable to good reliability. Items 5–7 had good fit after merging the two highest response categories, although demonstrated poor reliability. Conclusions No analyses provided support for the unidimensionality of the GAD-7 among 15–29-year-olds in Sweden. While these findings do not support the use of a single sum score, they require confirmation in independent samples. Future research should evaluate whether the abbreviated, candidate scale comprising items 1–3 could be useful to assess anxiety in this population, including examination of criterion validity and the potential benefits of adding a response category to enhance both reliability and scale-to-sample targeting.