Simultaneous Factor Analysis and Item Response Theory Analysis of Center for Epidemiological Studies Depression Scale (CES-D) and Quick Inventory of Depressive Symptomatology Self Report (QIDS-SR)

Abstract

The Center for Epidemiological Studies – Depression Scale (CES-D) and the Quick Inventory of Depressive Symptomatology – Self-Report (QIDS-SR) are psychological tests designed to assess depression symptoms. The psychometric properties of both scales have been examined using both Classical Test Theory (CTT) and Item Response Theory (IRT), and both have been demonstrated to be effective instruments. Previous evidence would suggest that each scale performs better at different depression symptom severities; the CES-D performed better at lower levels while the QIDS-SR performed better at higher levels. This study examined the viability of combining items from both scales to create a measure that would assess a broader range of depressive symptom severity using unidimensional IRT. In order to perform an IRT analysis, both scales would have to meet certain assumptions, namely unidimensionality and local independence. Previous studies have established that the QIDS-SR is a unidimensional scale, while the evidence for the CES-D have shown varying levels of multidimensionality. While problematic, previous studies have shown that a unidimensional IRT analysis can still be done with a multidimensional scale given that it adequately fits a one factor model. An Exploratory Factor Analysis (EFA) was conducted to determine if the combined CES-D & QIDS-SR scale met that assumption. A graded response model (GRM) IRT analysis was completed for the CES-D and QIDS-SR items to examine item level characteristics, namely the item’s ability to discriminate between individuals of varying levels of the higher order latent factor, as well as the amount of information that an item obtains relative to the respective scales. Results showed that the combined CES-D &QIDS-SR scale adequately fit a one factor model, enabling the use of unidimensional IRT analysis. The IRT analysis found that the combined scale did not seem to broaden the assessment range of depression levels. Furthermore, the IRT analysis showed that certain items from both scales performed poorly in its discriminatory function and amount of information obtained.