If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Address correspondence to Robbert J. J. Gobbens, RN (not practicing), MScN, PhD, FEANS, Inholland University of Applied Sciences, Faculty of Health, Sports, and Social Work, De Boelelaan 1109, 1081 HV Amsterdam, the Netherlands.
Faculty of Health, Sports and Social Work, Inholland University of Applied Sciences, Amsterdam, the NetherlandsZonnehuisgroep Amstelland, Amstelveen, the NetherlandsDepartment of Primary and Interdisciplinary Care, Faculty of Medicine and Health Sciences, University of Antwerp, Antwerp, Belgium
Department of Methodology and Statistics, Tilburg University, Tilburg, the NetherlandsDepartment of Sociology, Utrecht University, Utrecht, the Netherlands
To predict mortality with the Tilburg Frailty Indicator (TFI) in a sample of community-dwelling older people, using a follow-up of 7 years.
Design
Longitudinal.
Setting and Participants
479 Dutch community-dwelling people aged 75 years or older.
Measurements
The TFI, a self-report questionnaire, was used to collect data about total, physical, psychological, and social frailty. The municipality of Roosendaal (a town in the Netherlands) provided the mortality dates.
Results
Total, physical, and psychological frailty predicted mortality, with unadjusted hazard ratios of 1.295, 1.168, and 1.194, and areas under the receiver operating characteristic curves of 0.664, 0.671, and 0.567, respectively. After adjustment for age and gender, the areas under the curves for total, physical, and psychological frailty were 0.704, 0.702, and 0.652, respectively. Analyses using individual components of the TFI show that difficulty in walking and unexplained weight loss predict mortality.
Conclusions and Implications
This study has shown the predictive validity of the TFI for mortality in community-dwelling older people. Our study demonstrated that physical and psychological frailty predicted mortality. Of the individual TFI components, difficulty in walking consistently predicted mortality. For identifying frailty, using the integral instrument is recommended because total, physical, psychological, and social frailty and its components have proven their value in predicting adverse outcomes of frailty, for example, increase in health care use and a lower quality of life.
It is therefore important to identify frail community-dwelling older people at an early stage and to carry out appropriate interventions so that these adverse outcomes of frailty can be prevented or postponed. A well-known instrument for assessing frailty in community-dwelling older people is the Tilburg Frailty Indicator (TFI),
The TFI is a user-friendly self-report questionnaire, which aims to assess physical, psychological, and social frailty. Currently, the TFI has been translated into more than 10 languages, for example, Brazilian Portuguese,
Frailty and its prediction of disability and health care utilization: the added value of interviews and physical measures following a self-report questionnaire.
Frailty and its prediction of disability and health care utilization: the added value of interviews and physical measures following a self-report questionnaire.
Associations and predictions of readmission or death in acutely admitted older medical patients using self-reported frailty and functional measures. A Danish cohort study.
A Dutch study showed that frailty was associated with mortality within 3 years, adjusted for age, sex, and education, in a sample of 1042 community-dwelling people aged 65 years and older (odds ratio 3.27, P < .001).
In another Dutch cohort study including 2420 community-dwelling older people and with a 2-year follow-up, the area under the receiver operating characteristic curve (AUC) for predicting mortality by the TFI was 0.62, with a sensitivity of 80.6 and a specificity of 36.5.
In predicting mortality at 2 and 5 years, receiver operating characteristic curve analyses demonstrated AUCs of 0.74 and 0.73, respectively, using data of the Survey of Health, Aging and Retirement in Europe (SHARE), which included 27,527 community-dwelling adults (mean age 65.3 years, standard deviation 10.5).
The predictive validity of the TFI for mortality was also shown among 963 Brazilian people aged 60 years and older. Survival analysis of mortality according to frailty status during a 12-month follow-up period presented a hazard ratio of 2.72, adjusted for age and gender.
Finally, in a Danish sample consisting of 1328 acutely admitted older patients (≥65 years), frailty assessed with the TFI was associated with a higher risk of readmission or death within 6 months after discharge.
Associations and predictions of readmission or death in acutely admitted older medical patients using self-reported frailty and functional measures. A Danish cohort study.
The aforementioned studies, however, have some limitations for the prediction of mortality using the TFI. First, with regard to 2 of the aforementioned studies,
it should be noted that the original TFI was not used; the TFI was operationalized from data obtained by SHARE and the Longitudinal Aging Study Amsterdam (LASA).
Second, in the Danish, Brazilian, and the second Dutch study, the follow-up period can be considered as relatively short, being 6 months, 12 months, and 24 months, respectively.
Associations and predictions of readmission or death in acutely admitted older medical patients using self-reported frailty and functional measures. A Danish cohort study.
Associations and predictions of readmission or death in acutely admitted older medical patients using self-reported frailty and functional measures. A Danish cohort study.
The present study does not have any of these limitations. The aim of the present study was to determine the prediction of mortality by the TFI in a sample of community-dwelling people, using the original instrument and a follow-up period of 7 years.
Methods
Study Population and Data Collection
In June 2008, the TFI was sent to a sample of 1154 community-dwelling persons aged 75 years and older, randomly drawn from the register of the municipality in Roosendaal, a town of 78,000 inhabitants in the Netherlands. A total of 484 persons completed the questionnaire (42% response rate), which, in addition to the TFI, also contained measures for assessing disability and quality of life.
Part B of the TFI contains 15 components of frailty (total frailty): 8, 4, and 3 of these components belong to physical, psychological, and social frailty, respectively. The components of physical frailty are poor physical health, unexplained weight loss, difficulty in walking, difficulty in maintaining balance, poor hearing, poor vision, lack of strength in the hands, and physical tiredness. Psychological frailty consists of problems with memory, feeling down, feeling nervous or anxious, and unable to cope with problems. Finally, social frailty includes living alone, lack of social relations (loneliness), and lack of social support. The scores range from 0 to 15, 0 to 8, 0 to 4, and 0 to 3 for total and the physical, psychological, and social domains of frailty, respectively. Higher scores refer to a higher level of frailty.
Part A of the TFI includes 10 determinants of frailty. In this study, we present the data of 5 determinants: age, gender, marital status, ethnicity, and education (see Table 1), because these are well-known sociodemographic determinants of frailty.
In August 2015, referring to a follow-up period of approximately 7 years, the municipality of Roosendaal provided the mortality dates of the participants who completed the questionnaire in 2008.
Statistical Analysis
We used descriptive statistics to describe the baseline characteristics of the participants. Categorical variables were presented as counts and percentages. Continuous variables were presented as means with standard deviation. We defined a time-to-event outcome using the data of the mortality of the participants. The time in days was set to 0 at the time point the first participant died, and was set to 2613 for participants who were not dead.
In 2 sections, we first predicted mortality with a continuous frailty score (total, physical, psychological, or social as is common in studies using the TFI, and with a dichotomous frailty score using the defined cutoff points 5, 3, 2, and 2 for total, physical, psychological, and social frailty, respectively,
as this is how the TFI is commonly used in practice). In the subsequent section, we also predicted mortality with individual frailty components, examining which components contribute most to the (possible) predictive success of a frailty domain. In this last analysis, only the frailty components were included, with a bivariate P value < .20.
Finally, all aforementioned analyses were also conducted by adjusting for age and gender, as these are strongly associated to mortality. Before examining the prediction of mortality by individual components of total frailty, we imputed data for the missing values in the frailty components using the MICE package within the R software (m = 5, method: logreg).
For all analyses of survival, we used Kaplan-Meier analyses and Cox regression analyses to calculate hazard ratios (HRs) with 95% confidence intervals (CIs). The Kaplan–Meier survival curves with respect to subgroups were compared using the log-rank test. The predictive performance of the models was measured using the AUC. An AUC >0.7 was considered as an indication for good predictive performance.
For all analyses, we considered a P value < .05 as significant. For the analyses, we used R version 3.4.4 (R Foundation for Statistical Computing, Vienna, Austria).
Ethical Considerations
For this study, medical ethics approval was not necessary, as particular treatments or interventions were not offered or withheld from respondents. The integrity of respondents was not encroached on as a consequence of participating in this study, which is the main criterion in medical-ethical procedures in the Netherlands.
The analyses using frailty domain scores were carried out on the data of 455 participants (92.9%) with full data on all frailty components. The analyses using frailty components were carried out on the data of all 479 participants after imputing the missing data on components (0.54%).
Of the 479 participants at baseline, the mean age was 80.3 years (standard deviation = 3.8); 272 (56.8%) were women and 238 (49.8%) were married or cohabitating. The prevalence of frailty was 47.1%. Lack of social relations was the most prevalent component of frailty (282; 59.0%). Within the follow-up period of 7 years, 162 individuals died. See Table 1 for the descriptive statistics of the participants.
Prediction of Mortality by Total, Physical, Psychological, and Social Frailty
Table 2 presents the HRs for the total, physical, psychological, and social frailty scores with associated 95% CIs, P values, and AUCs, both unadjusted and adjusted for age and gender. Total, physical, and psychological frailty scores predicted mortality (all P values < .01), even after adjustment for age and gender.
Table 2Prediction of Mortality Using Frailty (Total Scores and Cutoffs): HRs and AUCs (P Values and CIs)
Social frailty did not predict mortality—the AUCs varied (unadjusted) from 0.551 (social frailty) to 0.671 (physical frailty).
Prediction of Mortality by Total, Physical, Psychological, and Social Frailty Based on Cutoff Points
Figure 1 presents the survival plots distinguishing frail from nonfrail older people using the defined cutoff points. The comparison of the survival plots with respect to the subgroups, defined by the cutoff points of total, physical, psychological, and social frailty using the log-rank test, showed P values of < .001, <0.001, 0.030, and 0.30, respectively. HRs and AUCs based on frailty cutoff scores showed (again) that total, physical, and psychological frailty predicted mortality (see Table 2).
Prediction of Mortality With All 15 Individual Components of Frailty
We excluded 4 components (poor hearing, poor vision, feeling down, and living alone) from the multivariate analyses because of P >.20 for these components in the bivariate analyses (results for bivariate analyses not shown). The Cox regression model with the 11 remaining components showed unadjusted and adjusted (for age and gender) AUCs of 0.716 and 0.747, respectively. Two components (unexplained weight loss and difficulty in walking) predicted mortality (unadjusted); in the adjusted model, these 2 components and lack of strength in the hands and physical tiredness predicted mortality. For more details, we refer to Table 3.
Table 3Prediction of Mortality Using Frailty Components: HRs (P Values and CIs)
Frailty Component
Unadjusted
Adjusted for Age and Gender
HR
95% CI
P Value
HR
95% CI
P Value
Poor physical health
1.169
0.786, 1.737
.44
1.183
0.789, 1.773
.42
Unexplained weight loss
1.917
1.157, 3.178
.012
1.804
1.078, 3.020
.025
Difficulty in walking
2.066
1.367, 3.122
.001
1.942
1.274, 2.960
.002
Difficulty in maintaining balance
1.439
0.986, 2.099
.06
1.415
0.963, 2.081
.08
Lack of strength in the hands
0.683
0.464, 1.005
.05
0.659
0.441, 0.986
.043
Physical tiredness
1.430
0.961, 2.128
.08
1.512
1.008, 2.267
.046
Problems with memory
1.565
0.990, 2.472
.06
1.460
0.910, 2.343
.12
Feeling nervous or anxious
0.906
0.626, 1.311
.60
0.905
0.614, 1.332
.61
Unable to cope with problems
1.110
0.721, 1.711
.64
1.248
0.798, 1.952
.33
Lack of social relations
1.327
0.931, 1.894
.12
1.325
0.917, 1.914
.13
Lack of social support
1.212
0.805, 1.823
.36
1.155
0.753, 1.772
.51
All components were entered into the model simultaneously.
Associations and predictions of readmission or death in acutely admitted older medical patients using self-reported frailty and functional measures. A Danish cohort study.
The aim of our study was to determine the predictive value of the TFI for mortality in a Dutch sample of 479 community-dwelling older people with a longer follow-up period (7 years).
Total, physical, and psychological frailty predicted mortality, with unadjusted HRs of 1.295, 1.168, and 1.194 and AUCs of 0.664, 0.671, and 0.567, respectively, and age-adjusted and gender AUCs were somewhat higher (0.704, 0.702, and 0.652, respectively). The unadjusted AUC for total frailty is comparable with the AUC of the TFI in the previously mentioned similar study, using a follow-up of 2 years (0.664 vs 0.620),
The AUCs of all these frailty instruments had higher scores in SHARE using a follow-up of 2 years, ranging from 0.72 (Groningen Frailty Indicator) to 0.77 (FI), and a follow-up of 5 years, ranging from 0.70 (Groningen Frailty Indicator, frailty phenotype) to 0.75 (FI).
The present study is the first to examine AUCs for mortality by the individual frailty domains, so a comparison of these AUCs with other studies is unfortunately not possible. In our study, the AUC of physical frailty was comparable with the AUC of total frailty. Social frailty did not predict mortality significantly. In a previous Dutch study carried out by the Netherlands Institute for Social Research, social frailty, assessed with the subscale of the TFI, did not increase the risk of death either.
We, therefore, conclude that the physical frailty domain is mostly responsible for the prediction of mortality. However, we do not recommend only using this domain for prediction, as other studies show that the other frailty domains (eg, psychological, social) do predict other future adverse outcomes, such as an increase in the use of care and a lower quality of life.
The AUC, using the individual components of the TFI for total frailty, was higher than the AUC determined with the total score (unadjusted: 0.716 vs 0.664). Consequently, if the researcher's aim is to identify very specific predictors of adverse outcomes, we recommend analyzing their associations with individual frailty components, but only in large samples when statistical power is less of an issue (eg, with a sample size of 1000 or more, small effects can be detected with power at least equal to 0.8 with α = 0.01). In smaller samples, and for practical purposes, we recommend using the total frailty score for predicting adverse outcomes, or the 3-domain scores.
The analyses using individual components of the TFI suggest that the 2 components difficulty in walking and unexplained weight loss predict mortality, even after adjustment for age and gender, and the effects of other components. Difficulty in walking is a well-known predictor of mortality. A prospective cohort study among British men, with a follow-up period of 2 up to 4 years, demonstrated that gait speed predicted mortality, with HR 2.90 (95% CI 1.84, 4.56) (physical performance) and HR 2.98 (95% CI 1.91, 4.66) (self-report).
Ability of self-reported frailty components to predict incident disability, falls, and all-cause mortality: Results from a population-based study of older British men.
In addition, low gait speed, assessed with the 6-m walking speed test, predicted mortality among 1176 older people living in the community, with an AUC value of 0.723.
Peak oxygen uptake, six-minute walk distance, six-meter walk speed, and pulse pressure as predictors of seven year all-cause and cardiovascular mortality in community-living older adults.
Finally, a study carried out in a large sample of inhabitants of Latin America, India, and China, found that slow walking speed predicted mortality, with 47,438 person-years of follow-up.
Unexplained weight loss is, like difficulty in walking, weakness (decreased grip strength), exhaustion, and low physical activity, a component of the phenotype of frailty,
and a frequently used measure of physical frailty. A systematic review and meta-analysis of prospective cohort studies, including 35,538 older people and 7994 deaths, revealed that frail older people, defined according to the phenotype of frailty, had a higher risk of mortality compared with pre-frail and robust older people.
The aforementioned study among British men showed that unintentional weight loss, assessed by physical examination and subjective assessment (self-report), predicted mortality, with HR 3.88 (95% CI 2.44, 6.16) and HR 3.14 (CI 2.02, 4.87), respectively.
Ability of self-reported frailty components to predict incident disability, falls, and all-cause mortality: Results from a population-based study of older British men.
In another study, men, as well as women (N = 4926), had higher mortality rates for weight loss (likely involuntary) with a ≥10-year follow-up period after adjustment for age, lifestyle, and medical factors.
In addition, a systematic review and meta-analysis including 15 observational studies concluded that unintentional weight loss had a significant effect on all-cause mortality.
However, findings of a randomized controlled trial showed that intentional weight loss was not associated with increased all-cause mortality over 12 years of follow-up in older people with overweight or obesity.
Another reason to be reluctant to interpret the effect of unintended weight loss is that it is not statistically significant after adjusting for multiple testing with the Bonferroni correction [ie, the P values (0.012 and 0.025) exceed 0.05/10 = .005]. This also holds for the effects of 2 physical components lack of strength in the hands and physical tiredness, which were only just statistically significant in the analysis adjusted for age and gender. Based on our findings and previous studies, we therefore recommend further research focusing on the role of weight loss, both unintentional and intentional, in the prediction of mortality among older people, especially stratified for gender and groups related to body mass index, but also on the effects of lack of strength in the hands and physical tiredness.
With regard to the predictive validity of the physical components of the TFI, it would be relevant to compare these findings with physical tests in the same population. Future research could examine if physical tests are necessary to be performed for the prediction of mortality, or that a self-report questionnaire such as the TFI is sufficient. This has been shown, for example, in the prediction of indicators of health care utilization (eg, visits to the general practitioner, receiving personal care, receiving nursing).
Frailty and its prediction of disability and health care utilization: the added value of interviews and physical measures following a self-report questionnaire.
Of the 15 components of the TFI, 4 were not included in the multivariate analyses: poor vision, poor hearing, feeling down, and living alone. However, removing these components from the TFI, on the basis of this study, is not recommended. The inclusion of sensory difficulties into a screening instrument such as the TFI has major consequences in terms of prevalence and the prediction of other adverse outcomes (eg, hospitalization).
Some limitations of this study should be noted. First, the sample consisted solely of people living in the municipality of Roosendaal; the generalizability of the findings can therefore be questioned. Second, the TFI is a frailty instrument based on self-report, so frailty is subjectively determined. However, the construct validity of the TFO has been extensively evaluated using objective measures. In addition, self-reported information concerning the components of the phenotype of frailty demonstrated superior predictive and discriminatory ability compared with objective measures of these components.
Ability of self-reported frailty components to predict incident disability, falls, and all-cause mortality: Results from a population-based study of older British men.
A strength of the present study is that the sample consisted of community-dwelling older people aged ≥75 years and older at baseline. Moreover, the mortality dates were correct, as they are derived from the register of the municipality, and the period of follow-up was long (7 years), in particular for this target group and in comparison with the other studies focusing on the prediction of mortality in community-dwelling older people by the TFI.
In conclusion, this study has shown the predictive value of the TFI for mortality in community-dwelling older people. Our study demonstrated that both physical and psychological frailty predicted death. Of the 15 individual TFI components, difficulty in walking consistently predicted mortality. However, for identifying multidimensional frailty, consisting of physical, psychological, and social frailty, it is important to use the integral instrument because all domains have proven their value in predicting other adverse outcomes of frailty, for example, disability, increase in health care use, and a lower quality of life.
Acknowledgments
We would like to thank the Dutch Public Health Services in West-Brabant and the municipality in Roosendaal, the Netherlands, for their support in making available the data.
References
Liu H.X.
Ding G.
Yu W.J.
et al.
Association between frailty and incident risk of disability in community-dwelling elder people: Evidence from a meta-analysis.
Frailty and its prediction of disability and health care utilization: the added value of interviews and physical measures following a self-report questionnaire.
Associations and predictions of readmission or death in acutely admitted older medical patients using self-reported frailty and functional measures. A Danish cohort study.
Ability of self-reported frailty components to predict incident disability, falls, and all-cause mortality: Results from a population-based study of older British men.
Peak oxygen uptake, six-minute walk distance, six-meter walk speed, and pulse pressure as predictors of seven year all-cause and cardiovascular mortality in community-living older adults.
30658-7&id=gr1.jpg){kind=link}