DEVELOPMENT, ADMINISTRATION AND SCORING
The GDS was developed to address the limited applicability of existing depression scales with older adults1,2. One such problem with scales that were developed and validated with younger populations relates to somatic complaints. Though somatic complaints can be a useful indicator of depression in younger samples, it is common for older adults to report somatic problems that can skew the sensitivity and specificity of a screen where older adults are more likely to have false positive depression scores. As such, the GDS does not include many somatic components in the scale and also consists of questions that are appropriate for the ageing population both in content and format. Initially, 100 questions were selected by researchers and clinicians with geriatric experience and administered to both non-depressed and depressed older adults. Of the original 100 items, the 30 items that correlated most highly with the total score were included in the final scale.
The scale was designed so that it could be administered easily in a self-rating format that would not be cumbersome for patients or medical personnel2. Most individuals complete the GDS in under 10 minutes. The GDS has a forced-choice response format where individuals are asked to respond yes or no to each item. However, some patients have problems with the yes/no dichotomy, instead preferring to answer ‘sometimes’ by either writing it in or by circling the space between the yes and no responses, though neither is a scored answer. It can help if explicit instructions are provided to patients to respond to each item in the way that best answers the question. In the initial validation study, the examiner read the items orally and recorded responses for individuals who could not complete the form without assistance. However, later studies found significant differences between mean scores for self-administered and staff-administered modes of presentation, with higher scores reported for self-administered scales3,4. There is no significant difference between scores of a card-based administration and verbal administration of the GDS5.
For each of the 30 GDS items, the depressive response garners 1 point for a maximum total of 30 points. Regarding general cutoffs, total scores ranging from 0 to 9 indicate normal mood; scores ranging from 10 to 19 indicate mild depressive symptoms, and scores ranging from 20 to 30 indicate severe depressive symptoms. The measure is free to use and in the public domain. It can be accessed at www.stanford.edu/~yesavage/GDS.xhtml.
PSYCHOMETRIC PROPERTIES Reliability
The GDS has been shown to be internally consistent and reliable over time. In the original reliability study, split-half and alpha coefficients were both 0.942. Other studies have demonstrated similar findings with diverse populations, showing alpha and split-half coefficients ranging from 0.80 to 0.996-8. Several studies have examined reliability of scores over time, finding high test-retest correlations between 0.85 and 0.942,7-9.
Validity
An important psychometric property of a depression screen is its ability to differentiate between depressed and non-depressed individuals (i.e. criterion validity). The sensitivity of the GDS refers to its ability to correctly identify individuals who are depressed whereas the specificity refers to its ability to correctly identify those individuals who are not depressed. Brink et al.1 found an 84% sensitivity rate and a 95% specificity rate for the GDS among community-dwelling older adults. Several other studies have examined the sensitivity and specificity of the GDS with diverse populations, finding acceptable levels above 80%. These have included medical patients, stroke patients, individuals suffering from dementia, nursing home residents, and psychiatric inpatients (for a thorough review, see Stiles and McGarrahan10).
The correlation between the GDS and other screening measures for depression is an indicator of the scale’s construct validity. The GDS was initially validated against the Hamilton Rating Scale for Depression11 (HRS-D) and the Zung Self-Rating Depression Scale12 (SDS), finding high correlations suggesting that the GDS measures similar constructs as the HRS-D and SDS. Other comparisons with the GDS have also been conducted, including the Beck Depression Inventory13 (BDI) and the Center for Epidemiological Studies Depression Scale14 (CES-D). These comparisons also yielded significant correlations, adding to evidence supporting the construct validity of the GDS15-18. Furthermore, the GDS has non-significant correlations with measures of cognition, adding to the empirical evidence that the GDS measures depression and not another construct19-21.
Dimensionality
Factor studies on the 30 items of the GDS have yielded mixed results. Sheikh et al.22 proposed a five-factor solution among communitydwelling older adults. Abraham and colleagues23 proposed a sixfactor solution among nursing home residents. These studies contrast the findings from Parmelee et al.9 and Salamero and Marcos24, which both found the GDS items to be highly intercorrelated and suggested the scale has only one factor. Based on these results, both advised the GDS should only be interpreted from a single total score.
Several brief forms of the GDS have been designed for use in settings where time is limited for depression assessment and with frail individuals for whom fatigue and poor concentration are concerns. Sheikh and Yesavage25 developed a 15-item version of the GDS, commonly referred to as the Short Form (GDS-SF), by taking the 15 items from the original scale that had the highest correlation with depressive symptoms in earlier studies. Findings from their validation study suggested both forms were highly correlated (r = 0.84) and effectively differentiated depressed from non-depressed individuals. A subsequent study reported cut-off scores for the GDS-15 as normal (0-4), mild (5-9) and moderate to severe (10-15)26. Older adults rated the GDS-15 as an acceptable measure that was not difficult or stressful to complete27.
Evidence for the validity of the GDS-15 is mixed. Some studies suggest the GDS-15 is not a suitable substitute for the full scale GDS in community-dwelling older adults26,28 or cognitively impaired individuals29. It does appear effective for screening depression in older adults who are cognitively intact and medically ill29, have affective disorders30, or who are seen in primary care clinics27,31. The GDS-15 was shown to consistently identify depressed individuals in a VA nursing home population32. Chiang and colleagues33 suggest that the GDS-15 may be less effective as a screening tool, but that it could be most effectively used to detect changes in moderate levels of depression.
Hoyl and colleagues34 developed a five-item GDS for use in a geriatric outpatient population by selecting items from the GDS-15 that had the highest correlation with a clinical diagnosis of depression. Using a score of 2 or greater to indicate possible depression, the GDS-5 had a sensitivity of 97%, specificity of 85%, and alpha coefficient of 0.80. The GDS-5 was found to be as effective as the GDS-15 in an Italian sample of cognitively intact older adults across three settings: a geriatric acute care ward, a geriatric outpatient clinic and a nursing home35. In addition, validity of the GDS-5 has been demonstrated for older sedentary adults.36 Weeks and colleagues37 compared two versions of four-item GDS measures and the GDS-5 to the GDS-15 in a sample of acute care patients. They found the GDS-5 showed the highest sensitivity (97%), but led to a high number of false positives. The researchers re-ordered the GDS-15 items into a two-tiered instrument, named the GDS-5/15, so that the full GDS-15 would only be administered to individuals who scored a 2 or greater on the initial GDS-5 items37.
Many brief forms of the GDS have been compared to a single-item depression screen. For example, D’Ath et al?1 created three versions of the GDS containing 10 items, 4 items, or 1 item (‘Do you feel that your life is empty?’). While the 10-item form performed well using either 2/3 or 3/4 cut-off scores, the 4and 1-item versions had low sensitivity and specificity. Hoyl et al.34 found a single-item depression screen (‘Do you often feel sad or depressed?’) performed significantly worse than the 15-and 5-item GDS, with a sensitivity of 85% and specificity of 65%. Likewise, a comparison of different versions of the GDS in a Dutch population found the diagnostic value of 30-, 15-, 10-and 4-item scales did not differ significantly; however, the 1-item version performed no better than chance38. Almeida and Almeida39 determined the 15-and 10-item scales were good screening instruments for depression in a Brazilian population, but cautioned against using 4-and 1-item forms because of low reliability and failure to show severity of depressive episodes.
Nursing Home Forms
A 12-item GDS was created for use with individuals living in residential care settings, including those with cognitive impairment (GDS-12R)40. During administration of the longer GDS in residential care facilities, interviewers found difficulties with a few of the items that appeared irrelevant or ambiguous to residents (e.g. ‘Do you prefer staying in rather than going out and doing new things?’). Removal of these items increased internal reliability from 0.76 to 0.81, with no significant difference in internal reliability for those with cognitive impairment. A cut-off score of 4/5 maximized both sensitivity and specificity of the GDS-12R and was suggested for research purposes. A cut-off of 3/4 was suggested for clinical use40.
Similarly, Jongenelis et al.41
Related posts:
The Care Home Experience Alisoun Milne
The Nosology of Dementia Kenneth Rockwood
Behavioural Management: Non-Pharmacological Jiska Cohen-Mansfield
Depression After Stroke Peter Knapp and Allan House
Sleep and Ageing: Disorders and Management Helen Chiu and Joshua Tsoh
Training Requirements for Old Age Psychiatrists in the UK Susan Mary Benbow and Aparna Prasanna
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
