Predicting ADL disability in community-dwelling elderly people using physical frailty indicators: a systematic review

Background Disability in Activities of Daily Living (ADL) is an adverse outcome of frailty that places a burden on frail elderly people, care providers and the care system. Knowing which physical frailty indicators predict ADL disability is useful in identifying elderly people who might benefit from an intervention that prevents disability or increases functioning in daily life. The objective of this study was to systematically review the literature on the predictive value of physical frailty indicators on ADL disability in community-dwelling elderly people. Methods A systematic search was performed in 3 databases (PubMed, CINAHL, EMBASE) from January 1975 until April 2010. Prospective, longitudinal studies that assessed the predictive value of individual physical frailty indicators on ADL disability in community-dwelling elderly people aged 65 years and older were eligible for inclusion. Articles were reviewed by two independent reviewers who also assessed the quality of the included studies. Results After initial screening of 3081 titles, 360 abstracts were scrutinized, leaving 64 full text articles for final review. Eventually, 28 studies were included in the review. The methodological quality of these studies was rated by both reviewers on a scale from 0 to 27. All included studies were of high quality with a mean quality score of 22.5 (SD 1.6). Findings indicated that individual physical frailty indicators, such as weight loss, gait speed, grip strength, physical activity, balance, and lower extremity function are predictors of future ADL disability in community-dwelling elderly people. Conclusions This review shows that physical frailty indicators can predict ADL disability in community-dwelling elderly people. Slow gait speed and low physical activity/exercise seem to be the most powerful predictors followed by weight loss, lower extremity function, balance, muscle strength, and other indicators. These findings should be interpreted with caution because the data of the different studies could not be pooled due to large variations in operationalization of the indicators and ADL disability across the included studies. Nevertheless, our study suggests that monitoring physical frailty indicators in community-dwelling elderly people might be useful to identify elderly people who could benefit from disability prevention programs.


Background
In ageing Western societies, the prevalence of frailty and its adverse outcomes increases [1]. Disability in Activities of Daily Living (ADL), which are the essential activities that a person needs to perform to be able to live independently [2], is an adverse outcome of frailty that places a high burden on frail individuals, care professionals and health care systems [3]. Frail elderly people have a higher risk of ADL disability compared to non-frail elderly people [4][5][6]. Effective interventions that prevent disability can diminish the burden caused by frailty. For the development of such interventions and the identification of people who might benefit from them, it is important to know which factors predict frailty-related ADL disability.
Frailty is a concept that has been defined in many different ways [7][8][9]. Various physical, cognitive, psychological, nutritional and social factors have been claimed to contribute to frailty [10]. A definition of frailty that is often used by geriatricians is the following: 'a biologic syndrome of decreased reserve and resistance to stressors, resulting from cumulative decline across multiple physiologic systems, and causing vulnerability to adverse outcomes' [11]. The well known frailty phenotype by Fried et al. [12] which classifies people into categories of robust, pre-frail or frail fits within this physiologic approach of frailty. The frailty phenotype postulates that five indicators of physical functioning (unintentional weight loss, exhaustion, slow walking speed, low grip strength, and low physical activity) are related to each other in a cycle of frailty. A person with none of the indicators is robust, a person with 1 or 2 indicators is pre-frail, and a person with 3 or more indicators is frail. Elderly people who are frail according to the phenotype have a higher risk of disability [4][5][6].
Although evidence exists that the phenotype predicts disability, it always involves a combination of the five indicators and provides no insight into the predictive value of the individual indicators. Besides that, the phenotype does not provide insight into the predictive value of other possible indicators of physical functioning that might relate to frailty. If individual indicators can predict ADL disability this could be clinically useful in identifying elderly people who might benefit from an intervention that prevents disability or increases physical functioning in daily life. A systematic literature review was conducted to investigate this in community-dwelling elderly people.

Search strategy
Potentially relevant articles were obtained by performing a search in three databases (PubMed, CINAHL, and EMBASE) from January 1975 until April 2010. This cutoff point was chosen because the term frailty was first introduced around the 1980's. To specify the study population the MESH term "aged" was combined with terms such as "frail*", "vulnerable", "low functioning", or "community-dwelling" where * denotes truncated terms. To specify the physical frailty indicators terms such as "grip strength", "weight loss", "balance", "exhaustion", "walking speed", "gait", "physical activity", and related MESH terms were combined with OR. To specify the outcome measure terms such as "disabil*", "Activities of daily living", "functional decline", and related MESH terms were combined with OR. To specify the study design terms such as "cohort studies", "longitudinal", "prognos*", "predict*", and related MESH terms were combined with OR. The searches for study population, physical frailty indicators, outcome, and study design were combined with AND, resulting in the final search. Reference lists of selected reviews and studies were screened for relevant publications that were not identified in the original search. Relevant studies found in these reference lists that met all inclusion criteria were also included in the review.

Study selection
Articles were eligible for inclusion if they met the following inclusion criteria: 1) written in English or Dutch, 2) a prospective longitudinal design, 3) involving community-dwelling elderly people aged 65 years or older, 4) at least 1 physical frailty indicator as independent variable, and 5) ADL disability as outcome measure. Most recent studies on disability in elderly persons focus on the ability or difficulty in carrying out ADL [13]. The fact that people who suffer from ADL disability, cannot live independently justifies the use of this measure as a key outcome [2]. Articles with only mobility disability as outcome variable were not included because this does not reflect the much broader concept of ADL disability. Studies that only focused on elderly patients with a disease such as Parkinson, depression, or stroke were excluded from the review.
All retrieved articles were first reviewed by two independent reviewers (JV & JCLN) based on their title. In case of disagreement or doubt, the article was included in the second phase of the selection process where all abstracts were assessed. Both reviewers independently labeled the remaining abstracts as 'include' or 'exclude'. Disagreement was resolved by consensus and if consensus could not be reached a third reviewer was consulted (MDS). In the third phase of the selection process, the full-text of the articles was retrieved and reviewed by both reviewers independently. Disagreement was resolved by consensus. In two cases the third reviewer had to be consulted. Agreement between the two independent reviewers in the second and third phase of the selection process was checked by calculating Cohen's Kappa.

Quality assessment & Data extraction
The quality of the included articles was assessed by both reviewers independently using a list of 27 criteria (see Table 1). This list was constructed based on previous research on methodological quality, quality of reporting criteria for observational research, and previous reviews regarding prediction of disability [14][15][16][17]. Each item was scored with 0 or 1 resulting in a possible range of 0 to 27 points per included study. A higher score indicated higher quality.
Data regarding design, duration of follow up, sample size, population characteristics, physical frailty indicators, outcome measures and results were extracted from the included studies. The extracted data were not pooled due to the fact that there was a large heterogeneity in the way physical frailty indicators and ADL disability were measured. In order to draw conclusions on the predictive strength of the different indicators, the number of articles reporting a significantly increased risk of ADL disability were counted for each indicator. The number of studies was then split up into studies that only included participants who were free of disability at baseline and studies that included participants not free of disability at baseline. Higher weight was given to studies that only included participants free of disability at baseline (++) compared to studies that included participants with and without disability at baseline (+). Negative weight was given to studies that reported no significant predictive value of the studied indicator (-).
In some cases, two different studies that reported positive findings for the same indicator used data from the same cohort. This was taken into account in the interpretation of the results by counting these findings as one.

Selection process
The search strategy yielded 3081 potentially relevant articles, after which 360 abstracts were scrutinized, leaving 64 full text publications for final review. After the selection process 28 studies were included in the review (see Figure 1 for details). The agreement between the two reviewers during the selection of abstracts and the selection of full-texts, as measured by Cohen's Kappa, was .74 and .82 respectively which is regarded as substantial to excellent. Was the follow-up of the cohort study 5 years or longer?

5
Were the key-elements of the study design described? 6 Were the setting, relevant dates and timeframe of the research described? 7 Were the eligibility criteria for participants described? 8 Were the participants free of disability at baseline? 9 Were the predictors and dependent variables described?
10 Were the measurement methods for the predictors and dependent variables described? 23 If sub-group analyses were performed, were these clearly described?
24 Were the key-results described in the discussion? 25 Were the limitations of the study reported? 26 Were previous research and the limitations of the study taken into account when an overall interpretation of the study results was provided?
27 Was the generalisability of the study results described?

Characteristics of included studies
The characteristics and details of the 28 included studies are presented in Table 2, ordered by year of publication. The main results from the included studies are presented in Table 3. All included studies were longitudinal cohort studies. Various studies reported on the same cohort data: 5 studies were based on the (Hispanic) Established Population for the Epidemiological Study of the Elderly [18][19][20][21][22], 3 studies on the Precipitating Events Project [23][24][25], 2 studies on the Jerusalem Longitudinal Study [26,27], 3 studies on the Longitudinal Interdisciplinary Study on Aging [28][29][30], 2 studies on the Cardiovascular Health Study [31,32], 2 studies on the Finland, Italy and The Netherlands Elderly Study [33,34], and 2 studies on the Project Safety cohort [35,36]. The other 9 included studies were based on other cohort studies [37][38][39][40][41][42][43][44][45]. The duration of follow-up of the studies varied from 1 year to 14 years (mean 5.4, SD 2.9 years). The  Balance, chair stands, and walking speed were significant predictors of progressive incident ADL disability. Walking speed was also a significant predictor of catastrophic incident disability.

Jacobs et al. (2008) [26]
Elderly people who go out less then daily at age 70 have a higher risk of incident dependence in ADL compared to elderly people who go out daily at age 70. RR 6.9 (1.4-34.0).

Ritchie et al. (2008) [43]
A history of unintentional weight loss at baseline predicts more rapid decline in ADL.
A variety of physical frailty indicators was measured in the included studies: weight loss, exhaustion, gait speed/ walking speed/gait, muscle strength/grip strength, physical activity, balance, lower extremity function, chair stand, 360°turn, bending over, foot taps, and hand signature. There was considerable variation in the way the same indicators were measured and operationalized in different studies. Also, different cut-off points were used in different studies. More detailed information regarding the measurement of the indicators is presented in Additional file 1.
The operationalization of ADL disability also varied across studies. Some studies defined disability as dependency in ADL at follow-up, others as difficulty in ADL at follow-up, and some studies used chronic ADL disability as an outcome measure. Some studies only measured disability in 4 different ADL, whereas others measured disability in 5, 6, or 7 ADL. More detailed information regarding the measurement of ADL disability is also presented in Additional file 1.

Predictive value of physical frailty indicators on ADL disability
For each individual physical frailty indicator the evidence regarding the predictive value is described below. The information is summarized in Table 4.

Weight loss
Four studies provided information regarding the predictive value of weight loss on ADL disability. These four studies were based on separate cohorts that only included participants who were free of disability at baseline [21,23,32,43]. All four studies concluded that elderly people who report (unintentional) weight loss have a significant higher risk to develop ADL disability.

Exhaustion
Only one study reported on the predictive value of exhaustion on ADL disability [23]. This study concluded that feelings of exhaustion are not a significant predictor of ADL disability in elderly people.

Gait speed
Twelve studies provided information about the predictive value of gait speed (walking speed) as an individual physical frailty indicator on ADL disability [18,19,23,25,[28][29][30]35,37,39,42,44]. All studies concluded that elderly people with slower gait speed have a higher risk of developing ADL disability. Nine studies were based on six separate cohort studies that only included participants free of ADL disability at baseline [18,19,23,25,28,29,31,35,42]. The other three studies were separate cohort studies that included participants with and without disability at baseline [37,39,44].

Physical activity
Nine studies reported on the predictive value of physical activity or exercise on ADL disability [23,26,27,34,[38][39][40][41]45]. All nine studies concluded that elderly people who are more physically active or who participate in exercise more regularly have a lower risk of developing ADL disability. Five out of these nine studies only included participants free of ADL disability at baseline [23,26,34,40,41]. These five studies were based on five separate cohort studies. The other four separate cohort studies included participants with and without disability at baseline [27,38,39,45].

Balance
Six studies provided information about the predictive value of balance [18,28,29,35,42,44]. Five out of these six studies concluded that elderly people with poorer balance have a higher risk of developing ADL disability [28,29,35,42,44]. These five studies were based on three separate cohorts that only included participants free of ADL disability at baseline [28,29,35,42]. The other study with a positive finding included participants with and without ADL disability at baseline [44].

Other physical frailty indicators
Eight studies reported on the predictive value of physical frailty indicators that were not mentioned above namely: lower extremity function, chair stands, 360°turn, bending over, foot taps, and hand signature.
Five of these studies reported on lower extremity function [19,20,24,36,41]. In all five studies, lower extremity function appeared to be a significant predictor of ADL disability. Elderly people with low lower extremity function had a higher risk of ADL disability at follow-up compared to people with moderate or high lower extremity function. Four of these studies were based on four separate cohorts that only included participants who were free of disability at baseline [19,20,24,36]. The other cohort study included participants with and without ADL disability at baseline [41].
Three studies investigated the predictive value of chair stands on ADL disability and concluded that this indicator is a significant predictor of ADL disability [18,35,42]. Two studies were based on two separate cohorts that only included participants who were free of disability at baseline [18,35]. The other cohort study included participants with and without ADL disability at baseline [42].
The study by Gill et al. [35] also investigated the predictive value of 360°turn, bending over, foot taps, hand signature and concluded that all indicators were predictors for ADL disability. The cohort study only included participants who were free of disability at baseline.

Discussion
This review provides evidence that physical frailty indicators are predictors of ADL disability in communitydwelling elderly people aged 65 years and older. Elderly people with unintended weight loss, slower gait speed, lower grip strength, lower physical activity, lower exercise, poor balance, or low lower extremity function have a higher risk of ADL disability in the future. Apparently, physical frailty indicators do not only predict disability when they are related in a frailty phenotype [12] but also independent of each other.
The number of studies that focused on the predictive value on ADL disability differed per physical frailty indicator. Almost half of the included studies investigated the predictive value of gait speed whereas only one study reported on exhaustion. Besides that, there were large variations in the measurement of frailty indicators and ADL disability across the 28 included studies. Therefore, it is difficult to draw firm conclusions regarding the predictive power of the different indicators compared to each other. Nevertheless, taking into account the number of studies per indicator that suggested a significantly increased risk of ADL disability for this indicator provides some insight into the predictive value. Slow gait speed and low physical activity or exercise seem to have the highest predictive power, followed by weight loss, lower extremity function, balance, muscle strength, and other indicators. These findings should be interpreted with caution because pooling of the data from different studies was not possible.
The follow-up period of the cohorts varied across the included studies. Three studies had a follow-up of 1 or 2 years, six studies had a follow-up of 3 years, and the rest of the studies had a follow-up longer than 3 years. From this can be concluded that certain indicators predict disability in the short-term, long-term or both. For example, gait speed and balance predict the development of ADL disability after a follow up of one year [35,44] and 6 years [29] and physical activity predicts the development of disability after a follow-up of 3 years [40,41] and 10 years [34]. For the identification of elderly people who could benefit from an intervention that prevents ADL disability, it is more useful to know the 'short-term' predictive value of the physical frailty indicators. It makes more sense to start with a preventive intervention when 'short-term predictors' are present in elderly people compared to a situation in which it will take another 6 years (or longer) before disability will develop.
A large part of the included studies had a relatively long follow-up period. It would be interesting to see whether indicators that predict disability after a long period of time, are also predictors of disability on the short term, e.g. 1 year. Besides that, it would also be useful to know how much the functioning of the physical frailty indicators would have to decrease before disability starts to develop in elderly people. Many of the included studies used quartile or quintile scores to define high or low physical functioning in the frailty indicators. As a result, many of these studies reported limited generalisability of their findings. Clear cutoff points have not been established yet for all indicators. This could be a focus of future research and should also be taken into account when developing interventions that can prevent disability in community-dwelling elderly people.
The only physical frailty indicator that appeared not to predict ADL disability was exhaustion. However, only one study included in this review focused on this [23]. Exhaustion is a feeling not only related to physical functioning but also to mental/psychological functioning. Since the search strategy focused strongly on physical functioning, some studies regarding exhaustion might not have been retrieved. Another possibility might be that hardly any studies focusing on the predictive value of exhaustion have been conducted.

Limitations of the review
Despite the effort of the authors to conduct a sensitive search strategy, some relevant studies or unpublished articles may not have been retrieved. It is also very remarkable that almost all selected studies showed positive results and were of (very) high quality. This may indicate publication bias.
A remark must be made regarding the quality scores of the included studies which were quite high. This is not necessarily a limitation of the study but rather an exceptional finding. The high quality scores might have been caused by the selection criteria which allowed only prospective cohort studies to be included. Another possible explanation could be that the criteria that were used to assess the quality of the studies did not only refer to the methodological quality but also to the quality of reporting. This might have elevated the quality scores compared to when the quality of reporting criteria would not have been taken into account.
The term frailty was first introduced in the 1980's. If earlier studies used different definitions or measurement methods for frailty or its adverse outcomes compared to more recent studies, this might have introduced the possibility of time-lapse bias. However, the probability of this type of bias is probably small due to the broad search terms that were used in the search strategy.
Many studies that were included in the review were based on secondary data-analyses. If measurement of the indicators or ADL disability was not the primary aim of the study, this might have resulted in the use of suboptimal measurement methods. However, the quality assessment of the included articles revealed that the majority of the studies used standardized or validated measurements for the indicators and outcome variables.