An Item Response Theory (IRT) study of the Geriatric Depression Scale-Short Form

Introduction: Depression is common in older adults and the main work of recognition and diagnosis has to be performed by non-psychiatric physicians. Thus, the use of self-report screening instruments turns out to be extremely useful for the identification of depressive disorders in old age. Among them the Geriatric Depression Scale - short form (GDS-SF; Sheik & Yesavage, 1986) is a largely employed 15-item instrument. Generally, a positive score of 5 or greater on the GDS-SF is indicative of the need for an in-depth psychiatric assessment. Objectives: The current study aimed at adding further evidence on the measurement precision of the GDS-SF as screening tool applying Item Response Theory (IRT) analyses. Specifically, IRT allowed investigating the reliability of the scale around the cutoff as well as the relative contribution of each item in measuring the trait level around the cutoff. Design/Methodology: Data from 1305 old people involved in gerontological research were used in this study (Male=42%, Mean age=76.9, SD=6.5, range 60-96). Unidimensional IRT analyses for dichotomous data were performed. The two-parameter (2PL) logistic model was tested in order to estimate location and slope parameters for each item, and the Test Information Function (TIF). Results: Nine out of 15 items of the GDS-SF were located around the trait level corresponding to the cutoff and the slopes showed they accurately differentiate among trait levels. Thus, the TIF peaked around the cutoff attesting the good reliability of the scale for this range of the trait. Conclusions: Results suggest that the GDS-SF could reliably differentiate depressed from non-depressed people in elderly population confirming its clinical utility as a screening test.