Search

Background: An innovative measurement system using a computerized adaptive testing technique based on the item response theory (CAT) has been expanding to measure mental health status. However, little is known about details in its measurement properties based on the empirical data. Moreover, the response time (RT) data, which are not available by a paper-and-pencil measurement but available by a computerized measurement, would be worth investigating for exploring the response behavior. Purpose: We aimed at constructing the CAT to measure depressive symptomatology in a community population and exploring its measurement properties. Also, we examined the relationships between RTs, individual item responses, and depressive levels. Method: For constructing the CAT system, responses of 2061 workers and university students to 24 depression scale plus four negatively revised positive affect items were subjected to a polytomous IRT analysis. The stopping rule was set for standard error of estimation < 0.30 or the maximum 15 items displayed. The CAT and non-adaptive computer-based test (CBT) were administered to 209 undergraduates, and 168 of them administered again after 1 week. Results: On average, the CAT was converged by 10.4 items. The θ values estimated by CAT and CBT were highly correlated ( r = 0.94 and 0.95 for the 1st and 2nd measurements) and with the traditional scoring procedures ( r's > 0.90). The test-retest reliability was at a satisfactory level ( r = 0.86). RTs to some items significantly correlated with the θ estimates. The mean RT varied by the item contents and wording, i.e., the RT to positive affect items required additional 2 s or longer than the other subscale items. Conclusion: The CAT would be a reliable and practical measurement tool for various purposes including stress check at workplace. [ABSTRACT FROM AUTHOR]