Table 4.

Criterion Validity of the TFI

Authors	Criterion Validity
Gobbens et al1	Concurrent using correlations: large for total frailty and quality of life domains physical, psychological, environmental, and medium to large for quality of life domain social assessed with the WHOQOL-BREF Concurrent using AUC: excellent for disability and reporting personal care, acceptable for reporting nursing and informal care, poor for reporting visits general practitioner and hospitalization
Metzelthin et al36	Concurrent using correlations: correlation between TFI and Groningen Frailty Indicator (GFI) was 0.76; the correlation between TFI and Sherbrooke Postal Questionnaire (SPQ) was 0.42
Gobbens et al42	Predictive, one and two years later, using multiple regression analyses: an increase in predictive accuracy of most adverse outcomes (disability, indicators of health care utilization, and quality of life) Predictive, one and two years later, using AUC: excellent for disability and reporting personal care, acceptable for reporting nursing, informal care, and facilities in residential care, poor for contacts with health care professionals and hospitalization, not significant for visits to a general practitioner
Gobbens et al40	Concurrent using sequential regression analyses: all components of the TFI together explained the scores on quality of life domains physical health, psychological, social relations, environmental assessed with the WHOQOL-BREF
Daniels et al37	Predictive, one year later, using OR unadjusted: disability 3.96, 95% CI = 2.48–6.30, mortality 3.08, 95% CI = 1.04–9.13, hospitalization 2.59, 95% CI = 1.36–4.90 Predictive, one year later, using AUC: poor for disability, mortality, and hospitalization
Theou et al46	Predictive, two and five years later, using AUC: acceptable for mortality
Gobbens and Van Assen41	Predictive, two and four years later, using sequential regression analyses): the items physical unhealthy, difficulty in walking, difficulty in maintaining balance, physical tiredness, feeling down, and lack of social support predicted quality of life scores assessed with the WHOQOL-BREF
Mulasso et al9	Concurrent using AUC: excellent for disability, poor for falls and visits to general practitioner
Gobbens et al39	Concurrent using regression analyses: all three domains (physical, psychological, social) together had an effect on disability, quality of life (physical health, psychological, social relationships, environmental), visits to a general practitioner, and falls; no effects were observed with contacts with health care professionals, hospitalization, receiving personal care, receiving nursing care, receiving informal care, and facilities in nursing home/rehabilitation center
Coelho et al10	Concurrent using multiple regression analyses: the TFI domains predicted 38.7% and 42.1% of quality of life variance, assessed with EUROHIS-QOL, and WHOQOL-OLD, respectivelyConcurrent using AUC: acceptable for disability in ADL, poor for disability in IADL, and health care utilizationConcurrent using AUC: discriminating ability was excellent regarding identifying frailty by the Groningen Frailty Indicator (GFI) (0.86, 95% CI = 0.85–0.93) and acceptable for frailty assessed with the Frailty Phenotype by Fried et al (0.75, 95% CI = 0.68–0.81)
Dong et al13	Concurrent using AUC: excellent for depression; acceptable for disability in ADL, and low social support; poor for disability in IADL, and for health care utilization (hospitalization, emergency use) Concurrent using AUC: discriminating ability was excellent regarding identifying frailty by the Frailty Phenotype by Fried et al (0.87, 95% CI = 0.87–0.93) and the Frailty Index (0.86, 95% CI = 0.82–0.91)
Renne and Gobbens45	Concurrent using sequential multiple linear regression analyses): all fifteen items together explained 36.5% of the variance of the score of quality of life
Santiago et al51	Predictive, 1 year later, using sequential logistic regression analyses: total frailty predicted mortality, adjusted for sex and age (HR = 2.72, 95% CI = 1.01–7.31); after controlling for sociodemographic variables the frailty domains (physical, psychological, social) improved the prediction of hospitalization (OR = 1.83, 95% CI = 1.10–3.06), falls (OR = 2.08, 95% CI = 1.21–3.58), disability in ADL (OR = 3.03, 95% CI = 1.45–6.29), disability in IADL (OR = 1.51, 95% CI = 1.05–2.17)
Vrotsou et al14	Concurrent using correlations: the correlation between total and the Frailty Phenotype by Fried et al was 0.49
Op Het Veld et al44	Predictive, 2 years later: positive predictive value 42.6% and negative predictive value 75.2% for disability in IADL
Op Het Veld et al43	Predictive, 2 years later, using AUC: poor for mortality, hospitalization, and disability in IADL
Gobbens et al38	Predictive, 1 year later, using linear and logistic regression analyses: the three frailty domains together predicted disability, visits general practitioner, contacts with health care professionals, receiving nursing; no effects were found on hospitalization, receiving personal care, falls (after controlling for sociodemographic characteristics and multimorbidity) Predictive, 1 year later, using AUC: excellent for total frailty with respect to disability and receiving personal care; poor for receiving nursing, falls, and hospitalization
Gobbens and Andreasen49	Predictive, 6 months later, using sequential logistic regression analyses: physical and social frailty predicted readmission and mortality; psychological frailty predicted only readmission
Zhang et al47	Concurrent using AUC: all AUC were excellent for SHARE-FI, and disability; all AUC were acceptable for limited function, poor mental health, and feeling lonely

Abbreviations: AUC, area under the curve; CI, confidence interval; ADL, activities of daily living; IADL, instrumental activities of daily living; OR, odds ratio; SHARE-FI, SHARE Frailty Instrument.