# Depression, Anxiety, and Stress Scale-21 (DASS-21): Further psychometric exploration using robust item response theory and classical theory measures among university students

## TL;DR

This study tested the DASS-21 mental health scale among Ethiopian university students and found it to be reliable and valid overall, though a few items need improvement.

## Contribution

The study provides the first psychometric evaluation of the DASS-21 in an Ethiopian university student population using both classical and item response theory.

## Key findings

- The DASS-21 showed a one-factor structure and strong reliability (ω and α of 0.93).
- Two items had sub-optimal psychometric properties in both classical and item response theory analyses.
- The scale demonstrated adequate divergent validity and item-level measurement invariance across sexes.

## Abstract

Psychometric validity of the 21-item Depression Anxiety and Stress Scale (DASS-21) has not been studied among Ethiopians. This study investigated the psychometric properties of the DASS-21 using Item Response Theory (IRT) and classical theory among Ethiopian university students.

A cross-sectional study was conducted with 364 Ethiopian university students. Participants completed a socio-demographic questionnaire, the Hamilton Anxiety Scale (HAM-A), and the DASS-21.

Factor analysis and measures of quality of factor estimates, construct stability, and robustness of unidimensionality supported a one-factor structure. Item parameters were satisfactory for all but three items using both CTT and IRT. Some CTT parameters indicated deviation from unidimensionality for three items. Two items had poor communality, and IRT infit/outfit values above 1.4. The findings indicated adequate divergent validity (weak-moderate correlation) for the general distress measure of DASS-21 with respect to HAM-A, excellent reliability (McDonald’s ω and Cronbach’s α both 0.93), and item-level measurement invariance across sexes.

The psychometric validity of the DASS-21 among Ethiopian university students was robust, as shown by analysis using classical and item response theory measures. However, some item-level psychometric characteristics for two items were sub-optimal in both CTT and IRT.