Table 3.
Psychometric properties and/or diagnostic accuracy reported*
| First author, year (reference) | Test being assessed | Reference (gold) standard | Reliability | Validity | Test performance | Other comments | ||
|---|---|---|---|---|---|---|---|---|
| Geriatric assessment studied in a prospective observational study design | ||||||||
| Aparicio, 2011 (19) | Compared MGA with a more traditional CGA.MGA consisted of a 1-item dementia screening, 1-item depression screening, 5 ADL and IADL items, 1-item malnutrition screening, 12 comorbidity items, screening of medications, 3 environment items (mobility and social support), and hemoglobin level and creatinine clearance | CGA consisting of cognition (MMSE), depression (modified GDS), ADL (Katz index), IADL (Lawton-Brody scale), nutrition (MNA), comorbidity (CIRS-G), polypharmacy, social intervention (no tool specified), hemoglobin level and creatinine clearance. Cutoffs used in CGA: MMSE ≤24, modified GDS ≥1, ADL<6, IADL <100%, MNA <23.5, CIRS ≥3, polypharmacy (1 anticoagulant or 2 cardiovascular drugs or 2 psychotropic drugs or ≥10 total medications), no cutoff listed for social intervention, hemoglobin level <10g/dL, creatinine clearance <50mL/min | Concordance between MGA and CGA was 86% for mental status; 73% for depression, dependence, and environment; and 66% for nutrition, comorbidity, and polypharmacy | NA | NA | NA | ||
| Presant, 2005 (37) | GOM was validated over a 12-mo period in 300 patients. GOM included a 5-point pain scale, a 5-point global quality-of-life scale, a 5-point energy scale, and 7 IADL and 8 ADL items | Each patient was interviewed by the physician after they completed the GOM; content of the interview not specified | NA | Compared physician interview results with the results of the pain, global quality of life, and energy scales and found >90% consistency (consistency not defined) | NA | NA | ||
| Geriatric assessment studied in a cross-sectional study design | ||||||||
| Extermann, 1998 (46) | CIRS-G was compared with the CCI to test the performance of both instruments and their relationship with functional status | NA | Interrater correlation (two raters) was good or very good for all subscales; the test–retest reliability was excellent for the following scales: CCI, CIRS-G categories, total CIRS-G score, and comorbidity severity grade 3 or 4. The interrater correlations for the following scales were: CCI rho = 0.74; CIRS-G total score rho = 0.76. Test–retest reliability excellent: CCI rho = 0.86, CIRS-G total score rho = 0.95. Correlation between the two comorbidity indices was fair (range = 0.25–0.39). The correlation between CIRS-G comorbidity severity grade 3 or 4 and ADL was 0.27. There was low or no correlation between comorbidity and functional status variables | NA | NA | NA | ||
| Kellen, 2010 (49) | Compared GFI, VES-13, and a newly developed aCGA to a CGA. The aCGA consisted of a 4-item GDS, 3-item ADL, 4-item IADL, and cognition.Cutoffs to define vulnerability used in the aCGA: GFI ≥4, VES-13 ≥3, aCGA GDS 2, ADL 1, IADL 1, MMS 6 | CGA consisted of ADL (Barthel index), IADL (Lawton scale), cognition (MMSE), and mood (GDS). Cutoffs to define vulnerability for the tests included in the CGA were: Barthel index of 2, MMSE ≤24, IADL 2, GDS ≥8. Vulnerability was defined as impairments in ≥2 domains (ADL and IADL) or cognitively impaired | NA | NA | GFI: sensitivity = 39%, specificity =86%, PPV = 86%, NPV = 40%; VES-13: sensitivity = 61%, specificity = 78%, PPV = 85%, NPV =48%; aCGA (aggregate): sensitivity = 51%, specificity = 97%, PPV = 97%, NPV = 48%; ADL, IADL questions in aCGA had high sensitivity; GDS, ADL had highest NPV | All three screening instruments missed cases of vulnerability that were identified using the CGA | ||
| Luciani, 2010 (51) | Compared VES-13 with CGA as a whole and CGA items. Cutoff used to classify vulnerable: VES-13 ≥3 | The CGA consisted of comorbidity (CIRS-G), cognition (MMSE), nutrition (Mini MNA), ADL (Katz index), IADL (Lawton scale), and mood (GDS). Cutoffs used to define vulnerability using the different scales included in the CGA: CIRS-G ≥3, MMSE <24, Mini MNA <12, ADL ≤5, IADL ≤5. For the GDS, no cutoff score listed. No cutoff for vulnerable provided based on the CGA as a whole | Spearman correlation between VES-13 and CGA was 0.4, and Spearman correlation between VES-13 ADL and IADL was 0.5. Weak correlations with individual items of the CGA and the VES-13; Cronbach’s alpha ranged from 0.1 (CIRS-G) to 0.3 (MMSE) | NA | The AUC comparing CGA as a whole to VES-13 was 0.83, sensitivity was 0.87, and specificity was 0.62. The AUCs comparing the whole CGA to VES-13 and CGA items were: VES-13 0.83, CIRS-G 0.58, Mini MNA 0.67, MMSE 0.81, GDS 0.56. The AUC of the VES-13 compared to ADL and IADL scale was 0.9, sensitivity was 0.9, and specificity was 0.7 | NA | ||
| Mohile, 2007 (78) | The VES-13 was compared with CGA. Cutoff: VES-13 ≥3 | CGA consisted of ADL (Katz index), IADL (Lawton scale), SPPB comorbidity (CCI), total number of medications, social support (RAND MOS Social Support Scale), and cognition (SPMSQ). Cutoffs: ADL ≤14, IADL ≤12, SPPB <9, CCI >10, number of medications ≥5, social support <4, and SPMSQ >3 | The reliability of the VES-13 using the Pearson correlation coefficient was 0.92 | NA | Using ≥2 of 7 deficits as the cutoff for CGA, the AUC was 0.90. Reliability: Pearson correlation coefficient = 0.92; with CGA: sensitivity = 0.73, specificity = 0.86, PPV = 88.9, NPV = 66.7; AUC = 0.9 | The patients who were impaired on the VES-13 performed statistically significantly worse on all tests of the CGA with the exception of social functioning (P < .05) | ||
| Molina- Garrido, 2011 (55) | Compared the BQ and the VES-13 with a CGA. Cutoffs: BQ >0 and VES-13 ≥3 | Geriatric assessment consisted of ADL (Barthel index), IADL (Lawton-Brody scale), comorbidity (CCI), social support (Guijon Social Scale), cognition (Pfeiffer test), nutrition (NSI), total number of medications, and PS (ECOG). Cutoffs: ADL ≤60, IADL ≥12, CCI >10, social support ≥8, Pfeiffer test >3, NSI ≥21, total number of medications >4, and ECOG PS ≥2. Frailty on the basis of the CGA was defined as deficits in ≥2 domains | NA | BQ, detecting risk of frailty: sensitivity 59.1%, specificity 78.9%, PPV 76.5%, NPV 62.5%, ICC 0.67 (95% CI = 0.46 to 0.81, P < .001).Detecting risk of frailty VES-13: sensitivity 54.6%, specificity 100%, PPV 100%, NPV 65.5%, ICC 0.81 (95% CI = 0.68 to 0.9, P < .001) | NA | The predictive ability for frailty: VES-13 AUC = 0.88; BQ AUC = 0.72 | ||
| Monfardini, 1996 (57) | Compared MACE with SIP. MACE consisted of the following domains: demographic, socioeconomic status, characteristics of neoplasia, comorbidity, symptoms, use of services, cognition (MMSE), depressive symptoms (GDS), balance (FICSIT), physical function (PPT), disability (IADL, ADL, and WHO PS. | NA | Using multivariable analysis, disability (using WHO PS) was associated with the SIP global score. Using multivariable analysis, disability (using WHO PS) and PPT were associated with physical SIP score. Using multivariable analysis, disability (using WHO PS) number of symptoms, GDS and balance were associated with psychosocial SIP | The ICCs for interrater reliability ranged from 0.4 (GDS) to 1 (household composition). The ICCs for test–retest reliability ranged from 0.25 (FICSIT balance score) to 1 (household composition). Cronbach’s alpha ranged from 0.4 (MMSE) to 1 (IADL and ADL) | NA | They also examined the variance by measures included in the MACE and found that WHO PS explained 70% of variance in SIP global score and 83% of the variance in SIP psychosocial score | ||
| Stauder, 2010 (65) | Used exploratory factor analysis to determine the number of individual domains assessed in the geriatric assessment (construct validity). The geriatric assessment consisted of the WHO PS, the KPS, ADL (Barthel index), IADL (Lawton-Brody scale), TUG, 7-item PPT, VES-13, GDS, FACT-G, MMSE, CIRS-G, CCI, and Social Support (F-SozU) | NA | NA | Factor analysis showed that six domains—functional status (KPS, ADL, PS, VES-13); health-related QOL (FACT-G, GDS); the variables comorbidities (as measured with both the CCI and CIRS), social support (as measured with the F-SozU and social well-being subscale FACT-G), cognition (as measured with the MMSE), and nutrition (as measured using BMI)—together explained 77% of total variance. Almost all correlations among (sub-) scales belonging to the same factor (domain) were at least moderately high; Spearman correlation coefficient >0.4 | NA | NA | ||
| Wedding, 2007 (99) | Compared physician’s judgment to the geriatric assessment. The physicians were hematologists and oncologists with >10 y of experience who classified patients as fit, vulnerable, or frail | Geriatric assessment consisted of ADL (Barthel index), IADL (Lawton-Brody Index), nutrition (MNA), cognition (MMSE), comorbidity (CCI), and mobility (Tinetti test).Cutoffs: Barthel index <100%, IADL <8, MNA continuous score used (no cutoff), MMSE continuous score used (no cutoff), CCI continuous score used (no cutoff), Tinetti test <20 | NA | Sensitivity was 0.88 and specificity was 0.31. They also compared the physician’s judgment to the Balducci classification: sensitivity was 0.43 and specificity was 0.80 | NA | NA | ||
| Geriatric assessment studied in chart reviews | ||||||||
| Overcash, 2005 (81)† | An aCGA consisting of 15 items was developed and compared with CGA. The aCGA and CGA consisted of the same scales, but the aCGA used only a few items of each scale whereas the CGA used the entire scales. The aCGA consisted of 3 ADL items (Katz index), 4 IADL items (Lawton scale), cognition (4 MMSE items), and mood (4 GDS items) | The CGA consisted of ADL 6 items (Katz index), IADL 10 items (Lawton scale), cognition (MMSE 10 items), and mood (GDS 15 items) | Items with the highest item–total correlation were selected for the aCGA. For ADL that included items 1, 3, and 4 (item–total correlations >0.70), for IADL, that included items 3–5 and 7 (item–total correlation >0.79), for GDS that included items 3, 7, 8, and 12 (item–total correlations >0.49), and for MMSE that included items 3 and 8–10 (item–total correlation >0.41). The Cronbach’s alpha coefficients for abbreviated aCGA scales were: ADL 0.84, IADL 0.930, MMSE 0.70 and GDS 0.70. The Cronbach’s alpha coefficients for CGA were: ADL 0.81, IADL 0.90, MMSE 0.65, and GDS 0.77. The Pearson correlation to assess construct validity between aCGA and CGA was 0.93 for ADL, 0.96 for IADL, 0.84 for MMSE, and 0.86 for GDS | NA | NA | NA | ||
| Overcash, 2006 (83)† | Developed cutpoints for the aCGA to indicate when a CGA is needed. The aCGA and the CGA consisted of the same scales, but the aCGA used only a few items of each scale whereas the CGA used the entire scales. The aCGA consisted of 3 ADL items (Katz index), 4 IADL items (no tool specified), cognition (4 MMSE items), and mood (4 GDS items) | The CGA consisted of ADL 6 items (Katz index), IADL 10 items (no tool specified), cognition (MMSE 10 items), and mood (GDS 15 items) | Cronbach’s alpha (internal consistency): abbreviated IADL = 0.93, both abbreviated MMSE and GDS = 0.7; abbreviated ADL = 0.84. Scores on aCGA and CGA highly correlated (ADL abbreviated–full ADL scale correlation = 0.93, abbreviated–full IADL scale correlation = 0.96, abbreviated–full MMSE scale correlation = 0.84, and abbreviated–full GDS scale correlation = 0.86) | The sensitivity for GDS using cutoff 2 was 0.81, specificity was 0.90. The sensitivity for MMSE using cutoff 6 was 0.82 and specificity 0.91 | NA | If a patient scores ≥2 on the abbreviated GDS then the full GDS needs to be administered. If a patient scores ≤6 on the abbreviated MMSE, the entire MMSE needs to be administered. If a patient scores a deficit on either of the ADL or IADL scales, the full scale needs to be administered | ||
*aCGA = abbreviated comprehensive geriatric assessment; AUC = area under the curve; BQ = Barber questionnaire; ADL= activities of daily living; BMI = body mass index; CGA = comprehensive geriatric assessment; CCI = Charlson comorbidity index; CI = confidence interval; CIRS-G = Cumulative Illness Rating Scale–Geriatric; ECOG = Eastern Collaborative Group Oncology; FACT-G=Functional Assessment of Cancer–General; FICSIT = Frailty and Injuries: Cooperative Studies of Intervention Techniques; F-SozU = Questionnaire for the Assessment of Social Support; PS = performance status; GDS = Geriatric Depression Scale; GFI = Groningen frailty indicator; GOM = Geriatric Oncology Module; HADS = Hospital Anxiety and Depression Scale; IADL = instrumental activities of daily living; ICC = intraclass correlation coefficient; IQCODE = Informant Questionnaire on Cognitive Decline in the Elderly; KPS = Karnofsky Performance Status; MACE = Multidimensional Assessment Protocol for Cancer in the Elderly; MGA = Mini Geriatric Assessment; MMSE = Mini Mental State Examination; MNA = Mini Nutritional Assessment; NA = not applicable; NPV = negative predictive value; NSI = nutritional risk screening; PPT = physical performance test; PPV = positive predictive value; QOL= quality of life; RAND MOS = Rand Corporation Medical Outcomes Survey; SIP = Sickness Impact Profile; SPMSQ = Short Portable Mental Screening Questionnaire; SPPB = Short Physical Performance Battery; TUG = Timed Up and Go test; VES-13 = Vulnerable Elder Survey-13 items; SIC = Satariano comorbidity index; WHO PS= World Health Organization performance status.
†Articles reporting on the same study.