A Psychometric Study of the Geriatric Depression Scale

Abstract

Criterion-referenced (Livingston’s r) and norm-referenced (Cronbach’s α and Gilmer-Feldt coefficient) techniques were used to calculate the internal consistency reliability of the 30-item Geriatric Depression Scale (GDS) on a sample (N = 417) of elderly medical outpatients. For the widely used cutoff score of 10 for depression, the criterion-referenced reliability (Livingston’s r) of the GDS total score was found to be .896, .883, and .891 for the elderly patients with Mini Mental State Examination (MMSE) scores > 17, ≤ 17, and total sample, respectively, while the norm-referenced reliability Gilmer-Feldt coefficient (and Cronbach’s α) of the GDS total score was found to be .895 (α = .893), .877 (α = .875), and .890 (α = .888), respectively. Tests of significance between the GDS total score means and reliability coefficients for the patients with MMSE scores > 17 and ≤ 17 were found to be nonsignificant. Therefore, the cognitive status of our sample does not seem to have an effect on the GDS total score. Item-total score correlations showed that 29 out of the 30 GDS items were good discriminators. The use of true score confidence intervals around GDS observed scores is recommended when used to make decisions about depression in the elderly.