Extending the Psychometric Information of the Leeds Dependence Questionnaire Using Item Response Theory and Differential Item Function Analysis.

摘要

The Leeds Dependence Questionnaire (LDQ) has been shown to tap a unitary construct hypothesized to be psychological dependence. The theoretical foundation for psychological dependence that the LDQ measures is rooted in social learning theory and allows the clinician to diagnose dependence in those without physiological dependence being present. Response data collected from 1681 people referred to community treatment centers in the United Kingdom have previously been examined using classical test theory and this response data is assumed to be substance and gender independent. In the present study, this same sample was examined using Item response theory (IRT) to extend the available psychometric information of the LDQ and to examine potential differential item functioning (DIF) by substance (opiate and alcohol) and gender. It was hypothesized that the LDQ response data would fit the graded response model and show no DIF between opiate and alcohol users as the psychological dependence construct was supposed to be global and not impacted by specific substance of use. The first hypothesis was not substantiated because the assumptions for the graded response model were met. The data was shown to have appropriate model-data fit with a 2-parameter logistic model. The second hypothesis was not substantiated as differential item functioning (DIF) was found to present in one item when examining DIF by substance (opiate and alcohol). The final hypothesis was substantiated as no DIF by gender was determined to be present. While the LDQ shows promise, the DIF analysis and lack of polytomous model fit suggest revising of some items would be appropriate.