Objective: To determine the validity of short Geriatric Depression Scale (GDS) versions for the detection of a major depressive episode according to ICD-10 criteria for research and DSM-IV.
Design: Cross-sectional evaluation of depressive symptoms in a sample of elderly subjects with short GDS versions. Different GDS cutoff points were used to estimate the sensitivity, specificity, positive predictive value and negative predictive value for the diagnosis of major depressive episode. Internal consistency of the scales was estimated with the Cronbach's alpha coefficient.
Setting: Mental Health Unit for the Elderly of 'Santa Casa' Medical School in São Paulo, Brazil.
Participants: Sixty-four consecutive outpatients aged 60 or over who met criteria for depressive disorder (current or in remission). Subjects with severe sensory impairment, aphasia or Mini-Mental State score lower than 10 were excluded from the study.
Measurements: ICD-10 Checklist of Symptoms, GDS with 15, 10, 4 and 1 items, Montgomery-Asberg Depression Rating Scale (MADRS), ICD-10 diagnostic criteria for research and DSM-IV diagnostic criteria.
Results: The use of the cutoff point 4/5 for the GDS-15 produced sensitivity and specificity rates of 92.7% and 65.2% respectively, and positive and negative predictive values of 82.6% and 83.3% respectively when ICD-10 diagnostic criteria for major depressive episode were used as the 'gold standard'. Similarly, rates of 97.0%, 54.8%, 69.6% and 94.4% were found when DSM-IV was the comparing diagnostic criteria. Sensitivity, specificity and positive and negative predictive values for the cutoff point 6/7 were 80.5%, 78. 3%, 86.8% and 69.2% according to ICD-10, and 84.8%, 67.7%, 73.7% and 80.8% respectively according to DSM-IV. Intermediate values were found for the cutoff point 5/6. The best fit for GDS-10 was the cutoff point 4/5, which produced a sensitivity rate of 80.5%, specificity of 78.3%, positive predictive value of 86.8% and negative predictive value of 60.2% according to ICD-10 diagnosis of a major depressive episode. Similarly, rates of 84.8%, 67.7%, 73.7% and 80.8% were found when DSM-IV criteria for major depression were used. GDS-4 cutoff point of 2/3 was associated with a sensitivity rate of 80.5%, specificity of 78.3%, positive predictive value of 86. 8% and negative predictive value of 69.2% when compared to ICD-10. Again, rates of 84.8%, 67.7%, 73.7% and 80.8% respectively were found when the criteria used were based on DSM-IV. GDS-1 had low sensitivity (61.0% and 63.6% for ICD-10 and DSM-IV respectively) and negative predictive value (56.7% and 67.6% for ICD-10 and DSM-IV respectively), suggesting that this question is of limited clinical utility in screening for depression. GDS-15 (rho=0.82), GDS-10 (rho=0.82) and GDS-4 (rho=0.81) scores were highly correlated with subjects' scores on the MADRS. Reliability coefficients were 0.81 for GDS-15, 0.75 for GDS-10 and 0.41 for GDS-4.
Conclusion: GDS-15, GDS-10 and GDS-4 are good screening instruments for major depression as defined by both the ICD-10 and DSM-IV. The shorter four- and one-item versions are of limited clinical value due to low reliability and failure to monitor the severity of the depressive episode. General practitioners may benefit from the systematic use of short GDS versions to increase detection rates of depression among the elderly. (c) 1999 John Wiley & Sons, Ltd.