Objective: To re-examine the test characteristics of the Geriatric Depression Scale (GDS) and the Brief Carroll Depression Rating Scale (BCDRS) in elderly medical inpatients, simulating the procedure followed by clinicians when using screening instruments.
Design: Masked comparison of GDS and BCDRS with psychiatric interview.
Setting: Durham VA Medical Center.
Participants: 109 consecutively admitted persons aged 70 or over.
Measurements: Screening by a social worker using GDS and BCDRS on day one, followed the next day by an investigator's structured psychiatric interview to determine the presence of major depressive disorder (MDD).
Results: By this method, the sensitivity and specificity of the GDS (cutoff 11) were 82% and 76%, respectively; for the BCDRS (cutoff 6), they were 73% and 79%. Among those with a negative test, the likelihood of MDD dropped from an a priori probability of 10% to an a posteriori probability of 3% with the GDS and 4% with the BCDRS. Among those with a positive test, the likelihood of MDD was 27% for the GDS and 28% for the BCDRS. Excluding patients with cognitive impairment (MMSE < or = 25) only slightly improved test characteristics.
Conclusion: These estimates are considerably below those reported in earlier studies where concordant screening, two-stage screening, or other methods have been utilized and may impact the decision whether or not to screen for depression using these instruments.