An enormous well done to Daryl, who was awarded his PhD yesterday (subject to minor amendments). An excellent job. Many thanks also to the examiners for their excellent job.
Last but not least, this is a poster presentation that Daryl will be preseting at the Kuala Lumpur meeting.
EVALUATION OF THE GERIATRIC DEPRESSION SCALE (GDS) IN THE VISUALLY IMPAIRED USING RASCH ANALYSIS
Daryl R. Tabrett BOptom MCOptom 1A,B ; Keziah Latham PhD MCOptom 1A,B1. Anglia Ruskin University, Cambridge, UK A. Department of Vision & Hearing Sciences B. Vision & Eye Research Unit
Purpose: Depression is increasingly recognised as an important factor in low vision rehabilitation and an appropriate patient reported outcome measure of depressive symptoms for the visually impaired would be valuable. The 15 item GDS is a well established self-report instrument which is used to assess the degree of depressive symptoms. This study assesses the function of the GDS using Rasch analysis in a sample of adults with visual impairment.
Methods: One hundred individuals with visual impairment were administered the dichotomous scaled 15 item GDS as part of a face-to-face clinical interview. Rasch analysis was performed to: 1) evaluate item fit, 2) assess unidimensionality, 3) evaluate differential item functioning (DIF), 4) indicate item targeting, and 5) provide reliability indices.
Results: Mean age of the sample was 77.6±12.5 years (range 25-98) and mean visual acuity was 0.84±0.36 logMAR (range 0.08-1.64). The GDS demonstrated generally acceptable fit statistics (mean MNSQ infit 0.98±0.19, range 1.39-0.77; outfit 0.99±0.38, range 1.78-0.47). Principal component analysis of the residuals indicated unidimensionality (eigenvalue of 1st contrast 1.8). One item showed notable DIF (>1.0 logit) for gender, one for vision loss severity, and two different items for age. Despite relatively high GDS raw scores (23% scored ≥5 indicating at least mild depression) notable mistargeting between mean item and person measures was revealed (-2.09 logits). Although item reliability was good (0.92), person reliability (0.55) was suboptimal.
Conclusions: Despite displaying generally good fit to the unidimensional model of depression the targeting and person reliability of the GDS is not ideal, and the GDS could not sufficiently discriminate between individuals with different levels of depression. Formulation of items that assess milder symptoms of depression, or the selection of a suitable alternative measure targeted towards the less depressed is advisable for more ideal assessment of depressive symptoms in samples with similar characteristics.