Customizing Computerized Adaptive Test Stopping Rules for Clinical Settings Using the Negative Affect Subdomain of...

Area of Science:

• Psychometrics
• Clinical Psychology
• Health Outcomes Research

Background:

• Patient-reported outcome measures (PROMs) are vital for clinical decisions and treatment evaluation.
• However, PROMs can be burdensome for patients and may lack the reliability needed for clear clinical interpretation.

Purpose of the Study:

• To evaluate how alternative stopping rules in computerized adaptive tests (CATs) can enhance reliability for clinical applications.
• To minimize the patient burden associated with CATs.

Main Methods:

• Computerized adaptive test (CAT) simulations were performed on three adult item banks from the NIH Toolbox Emotion Battery (Anger Affect, Fear Affect, Sadness).
• Original NIH Toolbox CAT stopping rules (score SE <0.3 before 12 items) were compared with a SE-change rule, fixed-length CATs (4-12 items), and a reduced maximum item rule (8 items).
• Measures included administered items (burden), proportion of assessments achieving reliability cutoffs (0.85, 0.90, 0.95) (precision), and accuracy of score recovery.

Main Results:

• Alternative stopping rules generally reduced burden and increased high reliability compared to original rules for adult banks.
• The SE-change rule and fixed-length CATs with ≤8 items increased assessments with reliability <0.85.
• The reduced maximum stopping rule demonstrated the best balance between precision and parsimony.

Conclusions:

• Reducing test burden while maintaining score precision for clinical use presents significant challenges.
• Stopping rules for CATs should be adapted based on the specific study population and objectives.