Which Version of the Geriatric Depression Scale is Most Useful in Medical Settings and Nursing Homes? Diagnostic Validity Meta-Analysis

Background: The Geriatric Depression Scale (GDS) has been evaluated in individual studies, but its validity and added value in medical settings and nursing homes is uncertain. Therefore, the authors conducted a meta-analysis, analyzing the diagnostic accuracy of long, short, and ultrashort versions of the GDS and stratified this into those with and without cognitive impairment. Methods: A comprehensive search identified 69 studies that measured the diagnostic validity of the GDS against a semistructured psychiatric interview, and of these, 43 analyses (in 36 publications) took place inmedical settings. Twenty-one studies examined the GDS(30), 12 studies examined the GDS(15), and 3 examined the GDS(4/5). For comparison, the authors also summarized studies examining unassisted clinical judgment. Heterogeneity was moderate to high; therefore, random effects meta-analysis was used. Results: Across all studies, the prevalence of late-life depression was 29.2% (95% confidence interval [CI] = 24.7%-33.9%), with no difference between inpatients, outpatients, and nursing homes. Diagnostic accuracy of the GDS30 aftermeta-analytic weighting was given by a sensitivity of 81.9% (95% CI = 76.4%-86.9%) and a specificity of 77.7% (95% CI = 73.0%-82.1%). For the GDS15, sensitivity was 84.3% (95% CI = 79.7%-88.4%) and specificity was 73.8% (95% CI = 68.0%-79.2%). For the GDS4/5, the sensitivity and specificity were 92.5% (95% CI = 85.5%-97.4%) and 77.2% (95% CI = 66.6%-86.3%), respectively. Results were not significantly influenced by the presence of dementia. Concerning added value, when identification using the GDS was compared with routine clinicians' ability to diagnose late-life depressions, at a prevalence of 30%, of every 100 attendees, the GDS30 would help correctly identify an additional 22 people as depressed but at a cost of 13 additional false positives. The GDS15 performed the same as GDS30 but with 15 false positives. The ultrashort form would help identify an additional 25 true positives with only 10 false positives. Thus, the best option when choosing between versions of the GDS seems to be the GDS4/5. Conclusion: All versions of the GDS yield potential added value in medical settings, but the GDS4/5 is the most efficient. In nursing homes, given an absence of data on the GDS4/5, the GDS15 may be preferred until more studies are reported. (Am J Geriatr Psychiatry 2010; 18: 1066-1077)