Screening accuracy for late-life depression in primary care: A systematic review
The CSDD is categorized by questions on mood, behavior, physical signs, diurnal patterns, and ideational disturbances. Each item is on a 3-point scale for a possible total score of 38, with higher scores indicating more depression. Most data generated about the CSDD have come from hospitalized patients, in whom it has demonstrated acceptable validity and reliability in demented and nondemented patients.29-31
We identified 1 study evaluating the CSDD that met our criteria. Vida et al22 screened outpatients from a family medicine clinic and found a sensitivity of 90% and specificity of 75% for detecting major depression.
Other instruments. Several very brief instruments have been validated in psychiatric or hospital settings where the prevalence of depressive symptoms is often high,32,33 but few have been tested in older primary care patients. Howe et al34 attempted to validate a 1-question screen (MHI-1) derived from the mental health component of the SF-36, asking elderly participants, “in the past month, how much of the time have you felt downhearted or sad?” (1=none, 6=all the time). They showed that as a “stand alone” screen, the MHI-1 did not perform well in the primary care setting, with a sensitivity of 67% and a specificity of 60%.
TABLE 1
Articles relevant to late-life depression screening
| Author | Test/cutpoint | Criterion standard | Avg. age | Dementia | Sn (%) (95% CI) | Sp (%) (95% CI) |
|---|---|---|---|---|---|---|
| D’Ath et al5 | GDS-15/5 | GMS/AGECAT | 74 | Not tested | 91 (86–96) | 72 (66–78) |
| Gerety et al6 | GDS/11 | SCID* | 79 | Avg MMSE 23 (SD4.7) | 89 (72–96) | 68 (58–77) |
| CES-D/16 | 74 (55–86) | 70 (60–79) | ||||
| Neal and Baldwin7 | GDS/11 | GMS/AGECAT | 77 | Not tested | 83 (72–94) | 80 (68–92) |
| Van Marjwick et al8 | GDS/7 | DIS | 74 | Mild/none | 79 (76–82) | 67 (63–71) |
| Arthur et al9 | GDS-15/3 | ICD-10 | 80 | None | 100 (98–102) | 72 (67–77) |
| Hoyl et al10 | GDS-15/5 | SCID | 75 | Avg MMSE 27 (SD 2.6) | 94 (89–99) | 82 (73–91) |
| Rait et al11 | GDS-15/4 | GMS/AGECAT* | >60 | Not tested | 92 (64–100) | 71 (63–79) |
| BASEDEC/6 | 92 (64–100) | 84 (78–91) | ||||
| CCSS/6 | 92 (64–100) | 79 (71–86) | ||||
| Abas et al12 | GDS-15/5 | GMS/AGECAT | >60 | Avg. MMSE 24 (SD 4.6) | 82 (62–92) | 82 (62–92) |
| CCSS/5 | 82 (62–92) | 68 (54–79) | ||||
| Beekman et al13 | CES-D/20 | DIS | 55-82 | None | 93 (91–95) | 73 (69–77) |
| Lewisohn et al14 | CES-D/12 | RDC, DSM-IIIR | 64 | Not reported | 76 (73–79) | 77 (74–80) |
| Lyness et al15 | CES-D/21 | SCID | 71 | Not tested | – | – |
| – Major depression | 92 (87–97) | 87 (81–93) | ||||
| – Minor depression | 40 (32–48) | 82 (75–89) | ||||
| GDS/10 | ||||||
| – Major depression | 100 (98–102) | 84 (78–90) | ||||
| – Minor depression | 70 (62–78) | 80 (73–87) | ||||
| Papassotiropoulos et al16 | CES-D/8 (demented excluded) | CIDI | >60 | Avg MMSE 27 (SD 6.0) | 75 (70–80) | 74 (67–81) |
| CES-D/9 (demented excluded) | CIDI | >60 | Avg MMSE 19 (SD 5.5)n demented sample | 75 (70–80) | 72 (67–77) | |
| 75 (70–80) | 72 (67–77) | |||||
| Papassotiropoulos et al17 | GHQ-12/0 Subthreshold depression | CIDI, DSM-IIIR; not reported | >60 | Avg. MMSE 28 (SD 2.0) | 46 (40-52) | 72 (67–77) |
| CES-D/9 Subthreshold depression | 39 (33-45) | 75 (70-80) | ||||
| Bird et al18 | SelfCARE(D)/5 | Independent psychiatric assessment* | 73 | Not tested | 77 (67–87) | 98 (95–101) |
| Upadhyaya and Stanley19 | SelfCARE(D)/5 | GMS/AGECAT* | 71 | Not tested | 95 (90–100) | 86 (78–94) |
| 74 (55–86) | 70 (60–79) | |||||
| Banerjee et al20 | SelfCARE(D)/8 | GMS/AGECAT | >65 | Not tested | 90 (86–94 | 53 (46–60) |
| Howe et al21 | MHI-1/2 | GMS/AGECAT | 81 | Excluded “organic mpairment” | 67 (58–76) | 60 (50–70) |
| Vida et al22 | Cornell Screen/7 | RDC* | 72 | Avg MMSE 19 (SD 7.8) | 90 (80–100) | 75 (60–90) |
| *These studies were blinded; all others were not reported. | ||||||
| GDS, Geriatric Depression Scale, 30-item; GDS-15, Geriatric Depression Scale, 15-item; GHQ, General Health Questionnaire; DIS, Diagnostic Interview Schedule; BASEDEC, Brief Assessment Schedule Depression Cards; CES-D, Center for Epidemiologic Study-Depression; MHI-1, single question from the Mental Health Inventory [“in the past month, how much have you felt downhearted or sad (1: none-6: all the time)”]; GMS, Geriatric Mental State/AGECAT computer program; CIDI, Composite International Diagnostic Interview; SCID, Structured Clinical Interview for DSM IIIR; CCSS, Caribbean Culture Specific Screen; RDC, Research Diagnostic Criteria; DSM IIIR, Diagnostic and Statistical Manual of Mental Disorders,3rd ed rev; MMSE, Mini Mental State Examination; ICD-10, International Classification of Diseases, 10th ed | ||||||
Discussion: late-life depression can be diagnosed accurately
Our systematic review shows that several instruments demonstrate good accuracy for detecting late-life major depression in primary care. The GDS, CES-D and SelfCARE(D) have comparable sensitivities and specificities. The CES-D and CCSD have similarly favorable accuracy in demented patients with an average MMSE score of 19.
A 1-question screen shows poor results, as do studies using the GHQ, CES-D, and GDS-15 to detect nonmajor depression. Finally, 2 studies demonstrate that a culturally specific screen in African-Caribbeans performs well, but no better than, the GDS.
The GDS has longstanding success in identifying major depression in psychiatric and hospital settings and now demonstrates accuracy in primary care, where the 15-item version in its yes/no self-administered format represents a realistic tool for use in the community or the clinic.
With a record of successful use in general adult research, the CES-D also has the benefit of a known track record and relative ease of administration. Evidence from this review suggests that it can be extended to the older primary care population. The SelfCARE(D) is comparably accurate in general practice, but has lower specificity in home care.
Our review highlights the need to further investigate the accuracy of screening tools for depression in patients with dementia, specifically where cognitive impairment may be severe. Using the CSDD, an instrument specifically designed for patients with dementia, Vida et al22 found good accuracy for detecting depression; however, they studied patients with relatively mild dementia. The prevalence of depression in dementia is 15% to 40%.35 Given the increasing incidence of dementia in our aging population, the availability of accurate screening tools that specifically account for the coexistence of these 2 common disorders is important.
