Using Item and Test Information to Optimize Targeted Assessments of Psychological Distress

Abstract

The need for efficient clinical assessment instruments has been growing during the past years. In the current application, the item information (item response theory) is used to evaluate and build fixed short versions. The method was applied to a questionnaire measuring psychological distress and data were collected from two mixed outpatient and general population samples. After fitting the partial credit model, two short versions were built: one to increase efficiency in screening applications; the other for the monitoring of high distress patients. A cross-validation bootstrap procedure is proposed to check whether the short versions are more efficient than alternative item selections. Using the partial credit model, the results from short and full versions can be compared on score level, which improves the flexibility of the assessment. The discussion focuses on the model selection and on how many items are realistically needed in routine assessments of psychological distress.