Graded Response Model Analysis and Computer Adaptive Test Simulation of the Depression Anxiety Stress Scale 21:...

Area of Science:

• Psychometrics
• Mental Health Assessment
• Psychological Measurement

Background:

• Conflicting research exists on the factor structure of the Depression Anxiety Stress Scale 21 (DASS-21).
• No prior studies have developed a computer adaptive test (CAT) version of the DASS-21.

Purpose of the Study:

• To calibrate items for and simulate a DASS-21 CAT.
• To address inconsistencies in the DASS-21's factor structure.
• To evaluate the efficiency of a DASS-21 CAT.

Main Methods:

• Confirmatory factor analysis, Mokken analysis, and graded response modeling were used on an evaluation sample (n=580).
• A DASS-21 CAT was simulated using validation (n=248) and simulated (n=10,000) samples.
• Item response theory (IRT) was applied to model item performance.

Main Results:

• A bifactor (quadripartite) model demonstrated good fit for the DASS-21.
• All scales showed acceptable fit with the graded response model.
• Simulated unidimensional DASS-21 CATs achieved a 17%-48% reduction in administered items at 0.80 reliability.

Conclusions:

• The quadripartite model is the best fit for the DASS-21 item structure.
• IRT analysis indicates DASS-21 items are most effective for mild to moderate symptom severity (0θ to 3θ).
• The developed DASS-21 CAT model offers a more efficient screening method.