(PS8-D82) Development of a Six-item Bilingual Form of the Depression Anxiety Stress Scales for Spanish-speaking Patients

The development of brief instruments that are psychometrically sound and clinically useful is critical for in-person and telehealth service provision. This study used data from 949 individuals who presented to a community-based organization that serves both rural and urban populations in North Carolina. Data from patients who were diagnosed with primary anxiety, depressive and stress disorders were used for developing a shortened form of the English and Spanish versions of the 21-item Depression, Anxiety, and Stress Scale (DASS). Moderated nonlinear factor analysis (MNLFA) was used for estimating (conditional) item parameters, scale score estimates and differential test functioning. Conditional item information curve values from the DASS-21 were then used to select the 6 items to balance maximum diagnostic classification accuracy with representation of item content across all 3 diagnoses. As part of this study, test-retest reliability of the DASS-21 and the shortened 6-item form (DASS-6) was examined with an independent patient sample (n=182). Provider acceptability for use of the DASS-6 in telehealth settings with patients from both rural and urban settings also was assessed. Results suggest that a) the overall DASS-21 performed in an equitable fashion across languages despite differential item functioning across multiple variables and b) the items selected for the DASS-6 optimally distinguished between diagnostic groups, even compared to the 6 items that solely had the highest IIC values. Moreover, test-retest reliability of the 6-item total score with the independent sample was high (ICC=.92) and reliability of individual items were also high (Κs ranged from .89 to .96). Finally, bilingual providers rated DASS-6 favorably for use telehealth or face to face settings. In sum, the six items that make up the DASS-6 in this paper a) can be administered in under 5 minutes, b) are relatively free of overall measurement bias (i.e., differential test functioning) across language and other factors (gender, age, diagnostic status) c) maximize clinical categorization across 3 diagnosis categories (i.e., stress, anxiety, and depression) despite being derived from a unidimensional negative affectivity construct and d) are well received by providers administering the instrument with restricted timelines. Implications are discussed in terms of use of MNLFA to reliably identify the most relevant symptoms to help researchers and practitioners continue work towards access to quality mental health services in lower-resource and telehealth settings.