Comparison of the 15‐item Geriatric Depression Scale (Gds‐15) and the Gds‐4 During Screening for Depression in an In‐patient Geriatric Patient Group

To the Editor: After dementia, depression is one of the most common gerontopsychiatric disease patterns, especially in elderly persons living in nursing homes or receiving in-patient hospital treatment.1 Depression is associated with considerable suicide risk and greater general morbidity and mortality; it also has a negative effect on prognoses for concomitant diseases,2 but largely because of a frequently atypical presentation, depression in older adults often remains undetected and untreated.3 To this end, screening instruments such as the Geriatric Depression Scale (GDS)4 have been designed to identify elderly persons with depressive symptoms, enabling them to undergo more-extensive diagnosis. To test the correlation of results from a 4-item short version of the GDS5 with those from the validated 15-item version during initial screening for depressive symptoms in a random in-patient geriatric patient group, the following comparative study was performed. The results from a German translation of the GDS-15 for all patients admitted to and discharged from the Diakonissen-Krankenhaus (Frankfurt am Main, Germany) in January and February 2009 were compared with those from the 4-item short version (GDS-4) included within the GDS-15. A score greater than 5 on the GDS-15 and a score greater than 1 on the GDS-4 were selected as an indication of the presence of depression. Questionnaires containing fewer than 10 responses were excluded from the analysis as incomplete. With questionnaires containing 10 to 14 responses, the score was calculated by taking the actual score (e.g., 4 out of 12=1/3) and extrapolating the complete score (here 5 out of 15=1/3). Borderline scores (e.g., 5.45 out of 15) and ambiguous responses (“between yes and no”) were evaluated as positive. The Cohen κ-coefficient was calculated as a statistical measure of the correlation; the McNemars test was performed to clarify any differences between the two versions in the frequency of a positive screening. Two hundred forty-two patients were treated at the hospital during the period. The most frequent primary diagnosis was multifactorial gait disorder in 70 patients (29%). For other patient characteristics, see Table 1. No GDS questionnaires were available for 50 patients, and for four others, the questionnaires were incomplete (<10 responses), so that a total of 188 questionnaires were ultimately included in the evaluation. The GDS-15 returned a positive screening for depression in 83 patients (44%), the GDS-4 in 63 patients (34%). The Cohen κ-coefficient was 0.535 (95% confidence interval=0.396–0.675), corresponding to a clear and significant, although not strong, correlation. In addition, the McNemar test showed that the GDS-15 classifies significantly more patients with suspected depression (P=.003). This comparative study confirms our clinical impression that the GDS-15 and the GDS-4 classify different patients as suspected of being depressed and that the correlation between the results of the two versions is not a strong one. This finding contradicts the results of other positive correlation studies for the GDS-4.6,7 Limitations of our study were the small number of cases and the considerable number of unavailable or incomplete questionnaires (54 out of 242=22.3%). Nevertheless, use of the GDS is initiated in the Diakonissen-Krankenhaus without first considering cognitive status, so erroneous self-assessments are possible in conjunction with cognitive deficits, although a study has shown the GDS-15 and a 5-item GDS version to be acceptably sensitive and specific for cases of mild to moderate dementia.8 In summary, the GDS-4 seems to be an inadequate replacement for the GDS-15 during screening for depressive symptoms, at least in the patient group examined here. Conflict of Interest: RP: consultant for Vifor Pharma in the context of iron deficiency. EH: consultant for Novartis and Roche Pharma in the context of clinical trials for hepatitis B and C. Author Contributions: Kerstin Amadori and Rupert Püllen: study concept, acquisition of data, interpretation of data, and preparation of manuscript. Eva Herrmann: analysis and interpretation of data. Sponsor's Role: There was no sponsor.