The diagnostic accuracy of late-life depression is influenced by subjective memory complaints and educational level in an older population in Southern Italy

The prevalence of late-life depression (LLD) depends on the study sample, measurements, and diagnostic approaches. We estimated the 30 item-Geriatric Depression Scale (GDS-30) accuracy against the gold standard LLD diagnosis made with the Semi-structured Clinical Diagnostic Interview for DSM-IV-TR Axis I Disorders, focusing on the prevalence of a late-life major depressive disorder (MDD), in a population-based sample of 843 subjects aged>65 years, subdivided into three groups: normal cognition, subjective memory complaints, and mild cognitive impairment. At the optimal cut-off score (≥4), the GDS-30 showed 65.1% sensitivity and 68.4% specificity for LLD (63% and 66% for late-life MDD, respectively). Using the standard cut-off score (≥10), the GDS-30 specificity reached 91.2%, while sensitivity dropped to 37.7%, indicating a lower screening accuracy [area under the curve(AUC):0.728, 95% confidence interval(CI):0.67–0–78]. The GDS-30 performance was associated with educational level, but not with age, gender, cognition, apathy, and somatic/psychiatric multimorbidity. For subjective memory complaints subjects, at the optimal cut-off score (≥7), the GDS-30 showed better discrimination performances (AUC=0.792,95%CI:0.60–0.98), but again the educational level affected the diagnostic performance. In subjective memory complaints subjects, symptom-based scales like the GDS-30 may feature a better performance for diagnosing depression in older age, but the GDS-30 seems to require adjustment to the patient's educational level.