ABSTRACT Background: We currently use the depression subscale (HADD) of the Hospital Anxiety and Depression Scale (HADS) for depression screening in elderly inpatients. Given recent concerns about the performance of the HADD in this age group, we performed a quality-control study retrospectively comparing HADD with the diagnosis of depression by a psychiatrist. We also studied the effect of dementia on the scale's performance. Methods: HADS produces two 7-item subscales assessing depression or anxiety. The HADD was administered by a neuropsychologist. As "gold standard" we considered the psychiatrist's diagnosis based on the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, criteria. Patients older than 65 years, assessed by both the HADD and the psychiatrist, with a clinical dementia rating (CDR) score lower than 3, were included. The effect of dementia was assessed by forming three groups according to the CDR score (CDR0-0.5, CDR1, and CDR2). Simple and multiple logistic regression models were applied to predict the psychiatrist's depression diagnosis from HADD scores. Areas under the receiver operating characteristics curve (AUC) were plotted and compared by χ2 tests. Results: On both univariate and multiple analyses, HADD predicted depression diagnosis but performed poorly (univariate: p = 0.009, AUC = 0.60 (95% confidence interval (CI) = 0.53-0.66); multiple: p = 0.007, AUC = 0.65 (95% CI = 0.58-0.71)), regardless of cognitive status. Because mood could have changed between the two assessments (they occurred at different points of the hospital stay), the multiple analyses were repeated after limiting time interval at 28, 21, and 14 days. No major improvements were noted. Conclusion: The HADD performed poorly in elderly inpatients regardless of cognitive status. It cannot be recommended in this population for depression screening without further study.