Archives of Gerontology and Geriatrics 59 (2014) 280–287
Contents lists available at ScienceDirect
Archives of Gerontology and Geriatrics journal homepage: www.elsevier.com/locate/archger
The prediction of disability by self-reported physical frailty components of the Tilburg Frailty Indicator (TFI) R.J.J. Gobbens a,*, M.A.L.M. van Assen b, M.J.D. Schalk c,d a
Research & Development Centre Innovations in Care, Rotterdam University of Applied Sciences, Rotterdam, The Netherlands Department of Methodology and Statistics, Tilburg School of Social and Behavioral Sciences, Tilburg University, Tilburg, The Netherlands c Department of Tranzo, Scientific Center for Care and Welfare, Tilburg University, Tilburg, The Netherlands d Research Unit for People, Policy and Performance, Faculty of Economic and Management Sciences, North West University, South Africa b
A R T I C L E I N F O
A B S T R A C T
Article history: Received 22 February 2013 Received in revised form 27 June 2014 Accepted 30 June 2014 Available online 8 July 2014
Disability is an important health outcome for older persons; it is associated with impaired quality of life, future hospitalization, and mortality. Disability also places a high burden on health care professionals and health care systems. Disability is regarded as an adverse outcome of physical frailty. The main objective of this study was to assess the predictive validity of the eight individual self-reported components of the physical frailty subscale of the TFI for activities of daily living (ADL) and instrumental activities of daily living (IADL) disability. This longitudinal study was carried out with a sample of Dutch citizens. At baseline the sample consisted at 429 people aged 65 years and older and a subset of all respondents participated again two and a half years later (N = 355, 83% response rate). The respondents completed a web-based questionnaire comprising the TFI and the Groningen Activity Restriction Scale (GARS) for measuring disability. Five components together (unintentional weakness, weakness, poor endurance, slowness, low physical activity), referring to the phenotype of Fried et al., predicted disability, even after controlling for previous disability and other background characteristics. The other three components of the physical frailty subscale of the TFI (poor balance, poor hearing, poor vision) together did not predict disability. Low physical activity predicted both total and ADL disability, and slowness both total and IADL disability. In conclusion, self-report assessment using the physical subscale of the TFI aids the prediction of future ADL and IADL disability in older persons two and a half years later. ß 2014 Elsevier Ireland Ltd. All rights reserved.
Keywords: Disability Frailty Physical activity Slowness TFI
1. Introduction Disability is an important health outcome for elderly persons; it is associated with impaired quality of life, future hospitalization, and mortality (Boyd, Xue, Simpson, Guralnik, & Fried, 2005; Covinsky, Hilton, Lindquist, & Dudley, 2006; Walter et al., 2001). Disability also places a high burden on health care professionals and health care systems (Rochat et al., 2010). Although there is consensus on the relevance of the concept of disability, there is considerable variation in the way disability is defined (Tas, Verhagen, Bierma-Zeinstra, Odding, & Koes, 2007). Disability is defined mostly in terms of experiencing difficulty in carrying out activities that are essential to independent living, that is, difficulties in performing activities of daily living (ADL) and/or
* Corresponding author at: Rotterdam University of Applied Sciences, PO Box 25035, 3001 HA Rotterdam, The Netherlands. E-mail address: [email protected] (R.J.J. Gobbens). http://dx.doi.org/10.1016/j.archger.2014.06.008 0167-4943/ß 2014 Elsevier Ireland Ltd. All rights reserved.
instrumental activities of daily living (IADL) (Tas et al., 2007). ADL functions are essential for an individual’s self care (e.g., dress yourself and feed yourself), whereas IADL functions are more concerned with self-reliant functioning in a given environment (e.g., make the beds and do the shopping). ADL disability represents a more severe and later form of disability than IADL disability (Wong et al., 2010). The dynamic nature of disability has been shown in prior studies (Gill & Kurland, 2003; Hardy, Dubin, Holford, & Gill, 2005; Mendes de Leon et al., 1999). For instance, Hardy et al. (2005) demonstrated the dynamic nature of disability in community-dwelling elderly, aged 70 years or older. They concluded that multiple transitions among states of disability and independence are common among older persons. Frailty is also an important health-related characteristic of persons. There is no simple cause–consequence relationship between frailty and disability; both probably influence one another (Ahmed, Mandel, & Fain, 2007), and are likely to share some risk factors and pathophysiological mechanisms (Topinkova, 2008). Mostly, disability is regarded as an adverse
R.J.J. Gobbens et al. / Archives of Gerontology and Geriatrics 59 (2014) 280–287
outcome of frailty (Bergman et al., 2004; Gobbens, Luijkx, Wijnen-Sponselee, & Schols, 2010b; Gobbens, van Assen, Luijkx, Wijnen-Sponselee, & Schols, 2010); frailty is considered as a predisability state by the European, Canadian, and American Geriatric Advisory Panel (Abellan van Kan et al., 2008). There is not yet universal agreement on a definition of frailty. The debate has focused on whether frailty should be defined purely in terms of physical factors or whether psychological and social factors should be included as well (Gobbens, Luijkx, WijnenSponselee, & Schols, 2010a; Lally & Crome, 2007). However, researchers are becoming more and more convinced that frailty is a multidimensional concept (Hogan, MacKnight, & Bergman, 2003; Markle-Reid & Browne, 2003; Puts, Lips, & Deeg, 2005), having a physical, psychological, and social domain. Previous studies showed that of these three frailty domains only physical frailty predicted disability (Gobbens, van Assen, Luijkx, & Schols, 2012a, 2012b). Therefore, we focus on the relationship between physical frailty and disability in the present study. Physical frailty can be defined as ‘‘a biologic syndrome of decreased reserve and resistance to stressors, resulting from cumulative declines across multiple physiologic systems, causing vulnerability to adverse outcomes’’ (Fried et al., 2001). Previous research has shown that while there is some overlap between physical frailty and disability, they are separate concepts (Fried, Ferrucci, Darer, Williamson, & Anderson, 2004). Several operational definitions of physical frailty exist. Most frequently cited in the scientific literature is the operational definition by Fried et al. (2001), called a phenotype of frailty. According to the phenotype of frailty an individual is identified as frail if at least three of five of the following criteria are present: unintentional weight loss, weakness (decreased grip strength), poor endurance, slowness, and low physical activity (Fried et al., 2001). The phenotype of frailty combines time-consuming physical performance tests (weakness, slowness) and self-report assessment (weight loss, endurance, physical activity). The phenotype was first validated using data from the Cardiovascular Health Study (CHS) and has been shown to be predictive of decline in ADL ability in a three and seven years follow up by community-dwelling persons aged 65 years and older (Fried et al., 2001). In addition, the Women’s Health and Aging Study-I showed that the phenotype was associated with the development of dependence in ADL over 3 years of follow up, also after controlling for other established predictors of disability such as hospitalization, cognition and depressive symptoms (Boyd et al., 2005). Monitoring physical components of frailty for the prevention or delay of the onset of disability based only on self-report is less costly and more practical for large-scale studies. Despite the practical advantage of self-reported screening tools for physical frailty, few have been proposed and validated. Two exceptions that do not contain items referring to disability are an assessment tool developed by Barreto, Greig, and Ferrandez (2012) and the TFI (Gobbens, Luijkx, Wijnen-Sponselee, van Assen, & Schols, 2011; Gobbens, van Assen, et al., 2010). Barreto et al.’s tool includes four criteria, namely BMI, low level of physical activity, dissatisfaction with muscle strength and dissatisfaction with endurance. The physical subscale of the TFI includes five items referring to the phenotype of frailty (low physical activity, unintentional weight loss, weakness, poor endurance, and slowness) and three other physical components, poor balance, hearing and vision problems (Gobbens et al., 2011; Gobbens, van Assen, et al., 2010). A proper understanding of the contribution of individual physical frailty components in the prediction of disability is a requisite for preventive interventions. Recognizing frailty could be clinically useful for health care professionals in identifying older persons who may benefit from an intervention aimed at preventing or delaying disability. The main objective of this study was to
281
assess the predictive validity of the eight individual self-reported components of the physical frailty subscale of the TFI for total disability, ADL disability and IADL disability in older people. We distinguish between the predictive validity of the five components referring to the phenotype of frailty and the three other items of the TFI. We used a time span of two and a half years, which was shorter than studies by Fried et al. (2001) and Boyd et al. (2005). The predictive validity was assessed in a relatively young population of 65 years and older, for which a low expected prevalence of disability can be expected (Tas et al., 2007). Hence, our study examines if self-reported physical frailty improves prediction of disability of relatively young elderly in a relatively short time span.
2. Methods 2.1. Study population and data collection The ‘‘Senioren Barometer’’, initiated by the Academic Collaborative Center Policy for the Elderly and Informal Care (Tranzo, Tilburg University), is a web-based questionnaire to assess the opinion of a panel of Dutch older people (aged 50 years and older) about different aspects of life. Older people can volunteer and participation is always without obligation. In the period from December 2009 to January 2010 (T1), 1492 respondents completed at least part of the questionnaire, of whom 1031 filled out the part on background characteristics, frailty and disability; 723 (70.1%) of them also completed this part at follow-up, two and a half years later in May and June 2012 (T2). We selected the respondents of 65 years and older, of whom 429 completed all relevant parts at T1 and 355 (83%) at T2. The sample was invited to participate in the study in different ways and through multiple sources. Through the website www.seniorenbarometer.nl people could indicate that they wanted to participate in research among the target group. Organizations for the elderly in the Netherlands were asked to issue an announcement of the study on their websites so that persons who were interested in participating could register. A third major source of participants was persons who attended computer training courses for older persons given by a large training and educational institute in the Netherlands. The study was conducted according to the ethical guidelines laid down in the Declaration of Helsinki. 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 upon as a consequence of participating in the study, which is the main criterion in medicalethical procedures in the Netherlands (Central Committee on Research inv. Human Subjects, 2010). Informed consent, in terms of information-giving and maintaining confidentiality, was respected. 2.2. Measures 2.2.1. Physical frailty Physical frailty was assessed using the physical subscale of the TFI (Gobbens et al., 2011; Gobbens, van Assen, et al., 2010). Previous studies demonstrated that the psychometric properties of the physical subscale of the TFI were good (Gobbens et al., 2012a; Gobbens, van Assen, et al., 2010). The individual physical frailty components correlated as expected with other validated frailty measures (Gobbens, van Assen, et al., 2010). Another study demonstrated that physical frailty, measured with the TFI, is a powerful predictor of disability, hospitalization, and personal care assessed one and two years later (Gobbens et al., 2012a). In the
282
R.J.J. Gobbens et al. / Archives of Gerontology and Geriatrics 59 (2014) 280–287
Table 1 Operationalization of physical frailty by the TFI (physical subscale) and the phenotype of frailty used in the CHS. Components of physical frailty
TFI, physical subscale (Gobbens et al., 2011; Gobbens, van Assen, et al., 2010)
Phenotype of frailty used in the CHS (Fried et al., 2001)
Unintentional weight loss
Have you lost a lot of weight recently without wishing to do so? (’a lot’ is: 6 kg or more during the last six months, or 3 kg or more during the last month) Do you experience problems in your daily life due to lack of strength in your hands? Do you experience problems in your daily life due to physical tiredness?
Self-reported unintentional weight loss 10 pounds in prior year
Weakness Poor endurance; exhaustion
Slowness Low physical activity Poor balance Poor hearing Poor vision
Do you experience problems in your daily life due to difficulty in walking? Do you find that you can be sufficiently physical active? Do you experience problems in your daily life due to difficulty maintaining your balance? Do you experience problems in your daily life due to poor hearing? Do you experience problems in your daily life due to poor vision?
current study, the (unstandardized) Cronbach’s alpha at T1 was .67 for physical frailty. The score for physical frailty was determined by adding the responses to the items (Gobbens et al., 2011; Gobbens, van Assen, et al., 2010). The eight items have two response categories ‘yes’ and ‘no’. The scoring is as follows: problem ‘‘1’’, no problem ‘‘0’’. The maximum score for physical frailty is 8; this represents the highest level of physical frailty. Table 1 presents an overview of the operationalization of frailty by the TFI and the phenotype of frailty as used in the CHS (Fried et al., 2001). 2.2.2. Disability Disability was assessed using the Groningen Activity Restriction Scale (GARS) (Kempen, Miedema, Ormel, & Molenaar, 1996; Kempen & Suurmeijer, 1990; Suurmeijer et al., 1994). Studies showed that the GARS is an easy to administer, reliable, and valid measure for assessing disability (Kempen et al., 1996; Suurmeijer et al., 1994). The GARS comprises 18 items; eleven items refer to ADL and seven items refer to IADL. Each item has four response categories: 1, able to perform the activity without any difficulty; 2, able to perform the activity with some difficulty; 3, able to perform the activity with great difficulty; 4, unable to perform the activity independently. The GARS score ranges from 18 (no disability) to 72 (maximum disability), with persons scoring 29 or higher are considered to be disabled (Ormel, Rijsdijk, Sullivan, van Sonderen, & Kempen, 2002). ADL and IADL disability scores range from 11 (no ADL disability) to 44 (maximum ADL disability) and 7 (no IADL disability) to 28 (maximum IADL disability), respectively. The (unstandardized) Cronbach’s alpha at T1 was .88 for disability, .83 for ADL disability, and .79 for IADL disability. 2.2.3. Background characteristics: socio-demographic and multimorbidity Socio-demographic variables included age, sex, marital status, ethnicity, education level and income. We assessed multimorbidity by asking ‘Do you have two or more diseases and/or chronic disorders?’ 2.3. Analysis strategies First, we determined the characteristics of the participants using descriptive statistics, and compared our sample to the population of Dutch older persons. Subsequently, we tested the change in disability and frailty from T1 to T2 using a paired t-test. Variables were then coded for analysis. As in a previous study (Gobbens et al., 2012b), dummies were created for marital status
Grip strength: lowest 20% at baseline adjusted for sex and BMI Self-reported exhaustion identified by two questions from the CES-D scale: Everything I do is an effort; I cannot get going Walking time/15 feet; slowest 20% (by sex and weight) Kilocalories expended based on self-report of physical activities in the prior week: lowest quintile (by sex) – – –
(‘‘1’’ married or cohabiting and ‘‘0’’ rest), sex (‘‘1’’ woman, ‘‘0’’ man), and multimorbidity (‘‘1’’ yes, ‘‘0’’ no), and linear effects of age, education, and income were incorporated into the analyses. ‘Ethnicity’, was excluded because of the low frequency of nonDutch respondents (3.7%). Before assessing the association between frailty and disability we compared the drop-outs with those still in the sample at T2, using all variables. Bivariate associations between each background variable and physical frailty component at T1 on the one hand and total disability, ADL disability and IADL disability at T2 on the other hand, were tested using regression analyses. The assumption of a linear relationship was checked for all continuous predictors; in case of a nonlinear relationship a quadratic effect of the predictor was incorporated in the models. In additional regression analyses we assessed how much variance of each of the three disability variables was explained by all eight frailty components together, the phenotype of frailty, and the combination of the three remaining components of frailty. Sequential linear regression analyses were run to verify which physical frailty component of the TFI improved the prediction of total, ADL and IADL disability, after controlling for previous disability, the effects of the background variables, and the remaining components. The sequential analyses consisted of four blocks. The effect of previous disability (assessed at T1) was estimated in the first block. The second block contained the background variables assessed at T1. In the third block, the five physical components of frailty referring to the phenotype of frailty by Fried et al. (2001) were added to the model. In the fourth block, the remaining three physical components of the TFI (poor balance, hearing, and vision problems) were added to the model. We tested whether each block increased the prediction of each adverse outcome (disability, ADL and IADL disability) two and a half years later (T2), using the change in R2. Finally, we also ran a regression analysis in which the third and fourth block were interchanged; this additional analysis tells us if the components of the phenotype add to the explanation of disability after controlling for the effects of the other physical components of the TFI (and other variables). Power analysis using G*Power 3.1 (Faul, Erdfelder, Lang, & Buchner, 2007) showed that the regression analysis had a power of .41 (.59) to detect a small effect size ( f2 = .02) of a block with eight (three) frailty components on disability, with a sample size of 356. The power to detect a medium effect (f2 = .15) was 1.00. With three (eight) components a power of .8 is obtained with a small to medium effect size of f2 = .031 (f2 = .043). All statistical analyses were conducted using SPSS 20.0 (SPSS, IBM Corp., Somers, NY, United States of America). All p-values are two-tailed.
R.J.J. Gobbens et al. / Archives of Gerontology and Geriatrics 59 (2014) 280–287
283
3. Results
3.2. Correlations between eight physical frailty components of the TFI
3.1. Participant characteristics
Most correlations between the eight physical frailty components at T1 were weak (.3) to medium (.5). The highest correlation of .43 was between ‘slowness’ and ‘low physical activity’, followed by the correlation of .39 between ‘poor endurance’ and ‘low physical activity’, whereas the smallest correlation (.02) was between muscle strength and unintentional weight loss. Since none of the associations between the physical frailty components was strong, independent effects of some components on disability in the multiple regression analyses may be expected.
Table 2 provides an overview of the descriptive statistics of the sample at T1 (total, dropouts 2012, non-dropouts 2012) and for the participants who completed the questionnaire at both T1 and T2 (last column). At T1 the participants’ mean age was 72.6 years (SD = 5.4), which is slightly younger than the Dutch population of 65 years and older (with an average age of 74.5 years) (Statistics Netherlands, 2012); 66.7% were men, compared to 44.4% in the Dutch population (Statline, 2012); 43% was highly educated, which is more than the about 35% of the Dutch population of 15–64 years (Statistics Netherlands, 2013); 69.9% were married or cohabiting in the sample. The comparison of dropouts to those who did not dropout (third and fourth column of Table 2) showed lower education (t(427) = 3.49, p < .001) and more disabled persons among dropouts (18.9% versus 9.6%, exact x2-test, x2(1) = 5.65, p = .024). No significant differences were found with respect to the three continuous disability measures, and frailty, sex, multimorbidity, income, and marital status. Those who did not dropout showed significant increases from T1 to T2 in physical frailty (t(355) = 4.65, p < .001), IADL (t(355) = 2.66, p = .008) and total disability (t(355) = 2.39, p = .018), but not in ADL disability (t(355) = 1.27, p = .21) (fourth and last column of Table 2). The percentage of disabled persons increased from 9.6% to 13.0% (exact two-tailed McNemar’s test, p = .023).
3.3. Bivariate and sequential linear regression analyses Table 3 presents the results of the bivariate regression analyses on disability, ADL, and IADL disability, with p-values significant at .05 printed in bold. Of the background variables only higher age and multimorbidity were associated with total disability, ADL and IADL disability. The effect of age was quadratic, with its positive effect on disability increasing in age. All eight physical frailty components of the TFI were strongly associated with all three disability variables assessed two and a half years later. The five components of the phenotype of frailty together explained 44%, 36%, 39% of total, IADL, ADL disability, respectively. The other three components of the TFI (hearing, vision, poor balance) explained 15%, 11%, 15% of total, IADL, ADL disability, respectively. All these effects, not shown in Table 3, were highly significant (p < .001).
Table 2 Participants characteristics at T1 (December 2009–January 2010, N = 429) and T2 (May–June 2012, N = 355). Characteristic
T1 Total N = 429 n (%)
T1 Dropouts 2012 N = 74 n (%)
T1 Non-dropouts 2012 N = 355 n (%)
T2, n (%) N = 355 n (%)
Age, mean SD, range Sex, % of men Marital status Married or cohabiting Not married/single Divorced Widowed Living apart together Ethnicity Dutch Other Education None Primary Secondary Polytechnics and higher vocational training University Incomea s999 or less s1000–s1499 s1500–s1999 s2000–s2499 s2500–s2999 s3000–s3499 s3500–s3999 s4000–s4499 s4500 or more Multimorbidity, % Yes Physical frailty domain, mean SD Disability, mean SD ADL disability, mean SD IADL disability, mean SD 29 (range 0–72)
72.6 5.4, 65–87 286 (66.7)
73.7 5.8, 65–87 53 (71.6)
72.4 5.3, 65–87 233 (65.6)
74.8 5.3, 67–89 233 (65.6)
300 (69.9) 43 (10.0) 27 (6.3) 54 (12.6) 5 (1.2)
50 (67.6) 5 (6.8) 5 (6.8) 13 (17.6) 1 (1.4)
250 (70.4) 38 (10.7) 22 (6.2) 41 (11.5) 4 (1.1)
236 (66.5) 47 (13.2) 19 (5.4) 49 (13.8) 4 (1.1)
413 (96.3) 16 (3.7)
71 (95.9) 3 (4.1)
342 (96.3) 13 (3.7)
– –
24 (5.6) 37 (8.6) 193 (45.0) 141 (32.9) 34 (7.9)
9 (12.2) 11 (14.9) 31 (41.9) 19 (25.7) 4 (5.4)
15 (4.2) 26 (7.3) 162 (45.6) 122 (34.4) 30 (8.5)
– – – – –
7 (1.8) 59 (14.9) 72 (18.2) 97 (24.5) 51 (12.9) 44 (11.1) 29 (7.3) 18 (4.5) 19 (4.8) 166 (38.7) 1.3 1.6 21.7 6.1 12.5 3.1 9.3 3.5 48 (11.2)
– 13 (19.7) 17 (25.8) 17 (25.8) 5 (7.6) 3 (4.5) 3 (4.5) 4 (6.1) 4 (6.1) 27 (36.5) 1.5 1.7 22.8 7.1 13.0 4.0 9.9 3.7 14(18.9)
7 (2.1) 46 (13.9) 55 (16.7) 80 (24.2) 46 (13.9) 41 (12.4) 26 (7.9) 14 (4.2) 15 (4.5) 139 (39.2) 1.3 1.5 21.5 5.8 12.4 2.9 9.2 3.4 34 (9.6)
– – – – – – – – – 174 (49.0) 1.6 1.7 22.0 6.3 12.5 3.2 9.5 3.7 46 (13.0)
a
33 missing values at T1 Total, 8 missing values at T1 Dropouts 2012, 25 missing values at T1 Non-dropouts 2012.
284
R.J.J. Gobbens et al. / Archives of Gerontology and Geriatrics 59 (2014) 280–287
Table 3 Effect of previous disability, background characteristics, and physical TFI components on disability, ADL disability and IADL disability: results of bivariate and sequential linear regression analysis. Total disability
ADL
Bivariate se
p
0.894
0.033
Contents lists available at ScienceDirect
Archives of Gerontology and Geriatrics journal homepage: www.elsevier.com/locate/archger
The prediction of disability by self-reported physical frailty components of the Tilburg Frailty Indicator (TFI) R.J.J. Gobbens a,*, M.A.L.M. van Assen b, M.J.D. Schalk c,d a
Research & Development Centre Innovations in Care, Rotterdam University of Applied Sciences, Rotterdam, The Netherlands Department of Methodology and Statistics, Tilburg School of Social and Behavioral Sciences, Tilburg University, Tilburg, The Netherlands c Department of Tranzo, Scientific Center for Care and Welfare, Tilburg University, Tilburg, The Netherlands d Research Unit for People, Policy and Performance, Faculty of Economic and Management Sciences, North West University, South Africa b
A R T I C L E I N F O
A B S T R A C T
Article history: Received 22 February 2013 Received in revised form 27 June 2014 Accepted 30 June 2014 Available online 8 July 2014
Disability is an important health outcome for older persons; it is associated with impaired quality of life, future hospitalization, and mortality. Disability also places a high burden on health care professionals and health care systems. Disability is regarded as an adverse outcome of physical frailty. The main objective of this study was to assess the predictive validity of the eight individual self-reported components of the physical frailty subscale of the TFI for activities of daily living (ADL) and instrumental activities of daily living (IADL) disability. This longitudinal study was carried out with a sample of Dutch citizens. At baseline the sample consisted at 429 people aged 65 years and older and a subset of all respondents participated again two and a half years later (N = 355, 83% response rate). The respondents completed a web-based questionnaire comprising the TFI and the Groningen Activity Restriction Scale (GARS) for measuring disability. Five components together (unintentional weakness, weakness, poor endurance, slowness, low physical activity), referring to the phenotype of Fried et al., predicted disability, even after controlling for previous disability and other background characteristics. The other three components of the physical frailty subscale of the TFI (poor balance, poor hearing, poor vision) together did not predict disability. Low physical activity predicted both total and ADL disability, and slowness both total and IADL disability. In conclusion, self-report assessment using the physical subscale of the TFI aids the prediction of future ADL and IADL disability in older persons two and a half years later. ß 2014 Elsevier Ireland Ltd. All rights reserved.
Keywords: Disability Frailty Physical activity Slowness TFI
1. Introduction Disability is an important health outcome for elderly persons; it is associated with impaired quality of life, future hospitalization, and mortality (Boyd, Xue, Simpson, Guralnik, & Fried, 2005; Covinsky, Hilton, Lindquist, & Dudley, 2006; Walter et al., 2001). Disability also places a high burden on health care professionals and health care systems (Rochat et al., 2010). Although there is consensus on the relevance of the concept of disability, there is considerable variation in the way disability is defined (Tas, Verhagen, Bierma-Zeinstra, Odding, & Koes, 2007). Disability is defined mostly in terms of experiencing difficulty in carrying out activities that are essential to independent living, that is, difficulties in performing activities of daily living (ADL) and/or
* Corresponding author at: Rotterdam University of Applied Sciences, PO Box 25035, 3001 HA Rotterdam, The Netherlands. E-mail address: [email protected] (R.J.J. Gobbens). http://dx.doi.org/10.1016/j.archger.2014.06.008 0167-4943/ß 2014 Elsevier Ireland Ltd. All rights reserved.
instrumental activities of daily living (IADL) (Tas et al., 2007). ADL functions are essential for an individual’s self care (e.g., dress yourself and feed yourself), whereas IADL functions are more concerned with self-reliant functioning in a given environment (e.g., make the beds and do the shopping). ADL disability represents a more severe and later form of disability than IADL disability (Wong et al., 2010). The dynamic nature of disability has been shown in prior studies (Gill & Kurland, 2003; Hardy, Dubin, Holford, & Gill, 2005; Mendes de Leon et al., 1999). For instance, Hardy et al. (2005) demonstrated the dynamic nature of disability in community-dwelling elderly, aged 70 years or older. They concluded that multiple transitions among states of disability and independence are common among older persons. Frailty is also an important health-related characteristic of persons. There is no simple cause–consequence relationship between frailty and disability; both probably influence one another (Ahmed, Mandel, & Fain, 2007), and are likely to share some risk factors and pathophysiological mechanisms (Topinkova, 2008). Mostly, disability is regarded as an adverse
R.J.J. Gobbens et al. / Archives of Gerontology and Geriatrics 59 (2014) 280–287
outcome of frailty (Bergman et al., 2004; Gobbens, Luijkx, Wijnen-Sponselee, & Schols, 2010b; Gobbens, van Assen, Luijkx, Wijnen-Sponselee, & Schols, 2010); frailty is considered as a predisability state by the European, Canadian, and American Geriatric Advisory Panel (Abellan van Kan et al., 2008). There is not yet universal agreement on a definition of frailty. The debate has focused on whether frailty should be defined purely in terms of physical factors or whether psychological and social factors should be included as well (Gobbens, Luijkx, WijnenSponselee, & Schols, 2010a; Lally & Crome, 2007). However, researchers are becoming more and more convinced that frailty is a multidimensional concept (Hogan, MacKnight, & Bergman, 2003; Markle-Reid & Browne, 2003; Puts, Lips, & Deeg, 2005), having a physical, psychological, and social domain. Previous studies showed that of these three frailty domains only physical frailty predicted disability (Gobbens, van Assen, Luijkx, & Schols, 2012a, 2012b). Therefore, we focus on the relationship between physical frailty and disability in the present study. Physical frailty can be defined as ‘‘a biologic syndrome of decreased reserve and resistance to stressors, resulting from cumulative declines across multiple physiologic systems, causing vulnerability to adverse outcomes’’ (Fried et al., 2001). Previous research has shown that while there is some overlap between physical frailty and disability, they are separate concepts (Fried, Ferrucci, Darer, Williamson, & Anderson, 2004). Several operational definitions of physical frailty exist. Most frequently cited in the scientific literature is the operational definition by Fried et al. (2001), called a phenotype of frailty. According to the phenotype of frailty an individual is identified as frail if at least three of five of the following criteria are present: unintentional weight loss, weakness (decreased grip strength), poor endurance, slowness, and low physical activity (Fried et al., 2001). The phenotype of frailty combines time-consuming physical performance tests (weakness, slowness) and self-report assessment (weight loss, endurance, physical activity). The phenotype was first validated using data from the Cardiovascular Health Study (CHS) and has been shown to be predictive of decline in ADL ability in a three and seven years follow up by community-dwelling persons aged 65 years and older (Fried et al., 2001). In addition, the Women’s Health and Aging Study-I showed that the phenotype was associated with the development of dependence in ADL over 3 years of follow up, also after controlling for other established predictors of disability such as hospitalization, cognition and depressive symptoms (Boyd et al., 2005). Monitoring physical components of frailty for the prevention or delay of the onset of disability based only on self-report is less costly and more practical for large-scale studies. Despite the practical advantage of self-reported screening tools for physical frailty, few have been proposed and validated. Two exceptions that do not contain items referring to disability are an assessment tool developed by Barreto, Greig, and Ferrandez (2012) and the TFI (Gobbens, Luijkx, Wijnen-Sponselee, van Assen, & Schols, 2011; Gobbens, van Assen, et al., 2010). Barreto et al.’s tool includes four criteria, namely BMI, low level of physical activity, dissatisfaction with muscle strength and dissatisfaction with endurance. The physical subscale of the TFI includes five items referring to the phenotype of frailty (low physical activity, unintentional weight loss, weakness, poor endurance, and slowness) and three other physical components, poor balance, hearing and vision problems (Gobbens et al., 2011; Gobbens, van Assen, et al., 2010). A proper understanding of the contribution of individual physical frailty components in the prediction of disability is a requisite for preventive interventions. Recognizing frailty could be clinically useful for health care professionals in identifying older persons who may benefit from an intervention aimed at preventing or delaying disability. The main objective of this study was to
281
assess the predictive validity of the eight individual self-reported components of the physical frailty subscale of the TFI for total disability, ADL disability and IADL disability in older people. We distinguish between the predictive validity of the five components referring to the phenotype of frailty and the three other items of the TFI. We used a time span of two and a half years, which was shorter than studies by Fried et al. (2001) and Boyd et al. (2005). The predictive validity was assessed in a relatively young population of 65 years and older, for which a low expected prevalence of disability can be expected (Tas et al., 2007). Hence, our study examines if self-reported physical frailty improves prediction of disability of relatively young elderly in a relatively short time span.
2. Methods 2.1. Study population and data collection The ‘‘Senioren Barometer’’, initiated by the Academic Collaborative Center Policy for the Elderly and Informal Care (Tranzo, Tilburg University), is a web-based questionnaire to assess the opinion of a panel of Dutch older people (aged 50 years and older) about different aspects of life. Older people can volunteer and participation is always without obligation. In the period from December 2009 to January 2010 (T1), 1492 respondents completed at least part of the questionnaire, of whom 1031 filled out the part on background characteristics, frailty and disability; 723 (70.1%) of them also completed this part at follow-up, two and a half years later in May and June 2012 (T2). We selected the respondents of 65 years and older, of whom 429 completed all relevant parts at T1 and 355 (83%) at T2. The sample was invited to participate in the study in different ways and through multiple sources. Through the website www.seniorenbarometer.nl people could indicate that they wanted to participate in research among the target group. Organizations for the elderly in the Netherlands were asked to issue an announcement of the study on their websites so that persons who were interested in participating could register. A third major source of participants was persons who attended computer training courses for older persons given by a large training and educational institute in the Netherlands. The study was conducted according to the ethical guidelines laid down in the Declaration of Helsinki. 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 upon as a consequence of participating in the study, which is the main criterion in medicalethical procedures in the Netherlands (Central Committee on Research inv. Human Subjects, 2010). Informed consent, in terms of information-giving and maintaining confidentiality, was respected. 2.2. Measures 2.2.1. Physical frailty Physical frailty was assessed using the physical subscale of the TFI (Gobbens et al., 2011; Gobbens, van Assen, et al., 2010). Previous studies demonstrated that the psychometric properties of the physical subscale of the TFI were good (Gobbens et al., 2012a; Gobbens, van Assen, et al., 2010). The individual physical frailty components correlated as expected with other validated frailty measures (Gobbens, van Assen, et al., 2010). Another study demonstrated that physical frailty, measured with the TFI, is a powerful predictor of disability, hospitalization, and personal care assessed one and two years later (Gobbens et al., 2012a). In the
282
R.J.J. Gobbens et al. / Archives of Gerontology and Geriatrics 59 (2014) 280–287
Table 1 Operationalization of physical frailty by the TFI (physical subscale) and the phenotype of frailty used in the CHS. Components of physical frailty
TFI, physical subscale (Gobbens et al., 2011; Gobbens, van Assen, et al., 2010)
Phenotype of frailty used in the CHS (Fried et al., 2001)
Unintentional weight loss
Have you lost a lot of weight recently without wishing to do so? (’a lot’ is: 6 kg or more during the last six months, or 3 kg or more during the last month) Do you experience problems in your daily life due to lack of strength in your hands? Do you experience problems in your daily life due to physical tiredness?
Self-reported unintentional weight loss 10 pounds in prior year
Weakness Poor endurance; exhaustion
Slowness Low physical activity Poor balance Poor hearing Poor vision
Do you experience problems in your daily life due to difficulty in walking? Do you find that you can be sufficiently physical active? Do you experience problems in your daily life due to difficulty maintaining your balance? Do you experience problems in your daily life due to poor hearing? Do you experience problems in your daily life due to poor vision?
current study, the (unstandardized) Cronbach’s alpha at T1 was .67 for physical frailty. The score for physical frailty was determined by adding the responses to the items (Gobbens et al., 2011; Gobbens, van Assen, et al., 2010). The eight items have two response categories ‘yes’ and ‘no’. The scoring is as follows: problem ‘‘1’’, no problem ‘‘0’’. The maximum score for physical frailty is 8; this represents the highest level of physical frailty. Table 1 presents an overview of the operationalization of frailty by the TFI and the phenotype of frailty as used in the CHS (Fried et al., 2001). 2.2.2. Disability Disability was assessed using the Groningen Activity Restriction Scale (GARS) (Kempen, Miedema, Ormel, & Molenaar, 1996; Kempen & Suurmeijer, 1990; Suurmeijer et al., 1994). Studies showed that the GARS is an easy to administer, reliable, and valid measure for assessing disability (Kempen et al., 1996; Suurmeijer et al., 1994). The GARS comprises 18 items; eleven items refer to ADL and seven items refer to IADL. Each item has four response categories: 1, able to perform the activity without any difficulty; 2, able to perform the activity with some difficulty; 3, able to perform the activity with great difficulty; 4, unable to perform the activity independently. The GARS score ranges from 18 (no disability) to 72 (maximum disability), with persons scoring 29 or higher are considered to be disabled (Ormel, Rijsdijk, Sullivan, van Sonderen, & Kempen, 2002). ADL and IADL disability scores range from 11 (no ADL disability) to 44 (maximum ADL disability) and 7 (no IADL disability) to 28 (maximum IADL disability), respectively. The (unstandardized) Cronbach’s alpha at T1 was .88 for disability, .83 for ADL disability, and .79 for IADL disability. 2.2.3. Background characteristics: socio-demographic and multimorbidity Socio-demographic variables included age, sex, marital status, ethnicity, education level and income. We assessed multimorbidity by asking ‘Do you have two or more diseases and/or chronic disorders?’ 2.3. Analysis strategies First, we determined the characteristics of the participants using descriptive statistics, and compared our sample to the population of Dutch older persons. Subsequently, we tested the change in disability and frailty from T1 to T2 using a paired t-test. Variables were then coded for analysis. As in a previous study (Gobbens et al., 2012b), dummies were created for marital status
Grip strength: lowest 20% at baseline adjusted for sex and BMI Self-reported exhaustion identified by two questions from the CES-D scale: Everything I do is an effort; I cannot get going Walking time/15 feet; slowest 20% (by sex and weight) Kilocalories expended based on self-report of physical activities in the prior week: lowest quintile (by sex) – – –
(‘‘1’’ married or cohabiting and ‘‘0’’ rest), sex (‘‘1’’ woman, ‘‘0’’ man), and multimorbidity (‘‘1’’ yes, ‘‘0’’ no), and linear effects of age, education, and income were incorporated into the analyses. ‘Ethnicity’, was excluded because of the low frequency of nonDutch respondents (3.7%). Before assessing the association between frailty and disability we compared the drop-outs with those still in the sample at T2, using all variables. Bivariate associations between each background variable and physical frailty component at T1 on the one hand and total disability, ADL disability and IADL disability at T2 on the other hand, were tested using regression analyses. The assumption of a linear relationship was checked for all continuous predictors; in case of a nonlinear relationship a quadratic effect of the predictor was incorporated in the models. In additional regression analyses we assessed how much variance of each of the three disability variables was explained by all eight frailty components together, the phenotype of frailty, and the combination of the three remaining components of frailty. Sequential linear regression analyses were run to verify which physical frailty component of the TFI improved the prediction of total, ADL and IADL disability, after controlling for previous disability, the effects of the background variables, and the remaining components. The sequential analyses consisted of four blocks. The effect of previous disability (assessed at T1) was estimated in the first block. The second block contained the background variables assessed at T1. In the third block, the five physical components of frailty referring to the phenotype of frailty by Fried et al. (2001) were added to the model. In the fourth block, the remaining three physical components of the TFI (poor balance, hearing, and vision problems) were added to the model. We tested whether each block increased the prediction of each adverse outcome (disability, ADL and IADL disability) two and a half years later (T2), using the change in R2. Finally, we also ran a regression analysis in which the third and fourth block were interchanged; this additional analysis tells us if the components of the phenotype add to the explanation of disability after controlling for the effects of the other physical components of the TFI (and other variables). Power analysis using G*Power 3.1 (Faul, Erdfelder, Lang, & Buchner, 2007) showed that the regression analysis had a power of .41 (.59) to detect a small effect size ( f2 = .02) of a block with eight (three) frailty components on disability, with a sample size of 356. The power to detect a medium effect (f2 = .15) was 1.00. With three (eight) components a power of .8 is obtained with a small to medium effect size of f2 = .031 (f2 = .043). All statistical analyses were conducted using SPSS 20.0 (SPSS, IBM Corp., Somers, NY, United States of America). All p-values are two-tailed.
R.J.J. Gobbens et al. / Archives of Gerontology and Geriatrics 59 (2014) 280–287
283
3. Results
3.2. Correlations between eight physical frailty components of the TFI
3.1. Participant characteristics
Most correlations between the eight physical frailty components at T1 were weak (.3) to medium (.5). The highest correlation of .43 was between ‘slowness’ and ‘low physical activity’, followed by the correlation of .39 between ‘poor endurance’ and ‘low physical activity’, whereas the smallest correlation (.02) was between muscle strength and unintentional weight loss. Since none of the associations between the physical frailty components was strong, independent effects of some components on disability in the multiple regression analyses may be expected.
Table 2 provides an overview of the descriptive statistics of the sample at T1 (total, dropouts 2012, non-dropouts 2012) and for the participants who completed the questionnaire at both T1 and T2 (last column). At T1 the participants’ mean age was 72.6 years (SD = 5.4), which is slightly younger than the Dutch population of 65 years and older (with an average age of 74.5 years) (Statistics Netherlands, 2012); 66.7% were men, compared to 44.4% in the Dutch population (Statline, 2012); 43% was highly educated, which is more than the about 35% of the Dutch population of 15–64 years (Statistics Netherlands, 2013); 69.9% were married or cohabiting in the sample. The comparison of dropouts to those who did not dropout (third and fourth column of Table 2) showed lower education (t(427) = 3.49, p < .001) and more disabled persons among dropouts (18.9% versus 9.6%, exact x2-test, x2(1) = 5.65, p = .024). No significant differences were found with respect to the three continuous disability measures, and frailty, sex, multimorbidity, income, and marital status. Those who did not dropout showed significant increases from T1 to T2 in physical frailty (t(355) = 4.65, p < .001), IADL (t(355) = 2.66, p = .008) and total disability (t(355) = 2.39, p = .018), but not in ADL disability (t(355) = 1.27, p = .21) (fourth and last column of Table 2). The percentage of disabled persons increased from 9.6% to 13.0% (exact two-tailed McNemar’s test, p = .023).
3.3. Bivariate and sequential linear regression analyses Table 3 presents the results of the bivariate regression analyses on disability, ADL, and IADL disability, with p-values significant at .05 printed in bold. Of the background variables only higher age and multimorbidity were associated with total disability, ADL and IADL disability. The effect of age was quadratic, with its positive effect on disability increasing in age. All eight physical frailty components of the TFI were strongly associated with all three disability variables assessed two and a half years later. The five components of the phenotype of frailty together explained 44%, 36%, 39% of total, IADL, ADL disability, respectively. The other three components of the TFI (hearing, vision, poor balance) explained 15%, 11%, 15% of total, IADL, ADL disability, respectively. All these effects, not shown in Table 3, were highly significant (p < .001).
Table 2 Participants characteristics at T1 (December 2009–January 2010, N = 429) and T2 (May–June 2012, N = 355). Characteristic
T1 Total N = 429 n (%)
T1 Dropouts 2012 N = 74 n (%)
T1 Non-dropouts 2012 N = 355 n (%)
T2, n (%) N = 355 n (%)
Age, mean SD, range Sex, % of men Marital status Married or cohabiting Not married/single Divorced Widowed Living apart together Ethnicity Dutch Other Education None Primary Secondary Polytechnics and higher vocational training University Incomea s999 or less s1000–s1499 s1500–s1999 s2000–s2499 s2500–s2999 s3000–s3499 s3500–s3999 s4000–s4499 s4500 or more Multimorbidity, % Yes Physical frailty domain, mean SD Disability, mean SD ADL disability, mean SD IADL disability, mean SD 29 (range 0–72)
72.6 5.4, 65–87 286 (66.7)
73.7 5.8, 65–87 53 (71.6)
72.4 5.3, 65–87 233 (65.6)
74.8 5.3, 67–89 233 (65.6)
300 (69.9) 43 (10.0) 27 (6.3) 54 (12.6) 5 (1.2)
50 (67.6) 5 (6.8) 5 (6.8) 13 (17.6) 1 (1.4)
250 (70.4) 38 (10.7) 22 (6.2) 41 (11.5) 4 (1.1)
236 (66.5) 47 (13.2) 19 (5.4) 49 (13.8) 4 (1.1)
413 (96.3) 16 (3.7)
71 (95.9) 3 (4.1)
342 (96.3) 13 (3.7)
– –
24 (5.6) 37 (8.6) 193 (45.0) 141 (32.9) 34 (7.9)
9 (12.2) 11 (14.9) 31 (41.9) 19 (25.7) 4 (5.4)
15 (4.2) 26 (7.3) 162 (45.6) 122 (34.4) 30 (8.5)
– – – – –
7 (1.8) 59 (14.9) 72 (18.2) 97 (24.5) 51 (12.9) 44 (11.1) 29 (7.3) 18 (4.5) 19 (4.8) 166 (38.7) 1.3 1.6 21.7 6.1 12.5 3.1 9.3 3.5 48 (11.2)
– 13 (19.7) 17 (25.8) 17 (25.8) 5 (7.6) 3 (4.5) 3 (4.5) 4 (6.1) 4 (6.1) 27 (36.5) 1.5 1.7 22.8 7.1 13.0 4.0 9.9 3.7 14(18.9)
7 (2.1) 46 (13.9) 55 (16.7) 80 (24.2) 46 (13.9) 41 (12.4) 26 (7.9) 14 (4.2) 15 (4.5) 139 (39.2) 1.3 1.5 21.5 5.8 12.4 2.9 9.2 3.4 34 (9.6)
– – – – – – – – – 174 (49.0) 1.6 1.7 22.0 6.3 12.5 3.2 9.5 3.7 46 (13.0)
a
33 missing values at T1 Total, 8 missing values at T1 Dropouts 2012, 25 missing values at T1 Non-dropouts 2012.
284
R.J.J. Gobbens et al. / Archives of Gerontology and Geriatrics 59 (2014) 280–287
Table 3 Effect of previous disability, background characteristics, and physical TFI components on disability, ADL disability and IADL disability: results of bivariate and sequential linear regression analysis. Total disability
ADL
Bivariate se
p
0.894
0.033
