Frailty and The Prediction of Negative Health Outcomes

JAMDA 17 (2016) 1163.e1e1163.
e17
JAMDA
journal homepage: www.jamda.com
Review Article
Frailty and the Prediction of Negative Health Outcomes:

A Meta-Analysis
Soe Vermeiren MSc a, b, Roberta Vella-Azzopardi MD a, b, c, David Beckwe PhD a, b, d,
Ann-Katrin Habbig MSc b, e, Aldo Scafoglieri PhD b, f, Bart Jansen PhD g,
Ivan Bautmans PhD a, b, c, * on behalf of the Gerontopole Brussels Study grouph
a
Gerontology Department, Vrije Universiteit Brussel (VUB), Brussels, Belgium
b
Frailty in Ageing (FRIA) Research Department, Vrije Universiteit Brussel (VUB), Brussels, Belgium
c
Geriatrics Department, Universitair Ziekenhuis Brussel (UZ Brussel), Brussels, Belgium
d
Rehabilitation Sciences Research Department (RERE), Vrije Universiteit Brussel, Brussels, Belgium
e
Fundamental Rights and Constitutionalism Research Group (FRC), Vrije Universiteit Brussel (VUB), Elsene, Belgium
f
Experimental Anatomy (EXAN), Vrije Universiteit Brussel (VUB), Brussels, Belgium
g
Department of Electronics and Informatics ETRO, Vrije Universiteit Brussel (VUB), Elsene, Belgium
a b s t r a c t
Keywords: Introduction: Frailty is one of the most important concerns regarding our aging population. Evidence grows
Frailty that the syndrome is linked to several important health outcomes. A general overview of frailty concepts and
prospective risk outcomes a comprehensive meta-analysis of their relation with negative health outcomes still lacks in literature, making
community-dwelling
it difcult for health care professionals and researchers to recognize frailty and the related health risks on the
meta-analysis
one hand and on the other hand to appropriately follow up the frailty process and take substantiated action.
elderly
Therefore, this study aims to give an overview of the predictive value of the main frailty concepts for negative
health outcomes in community-dwelling older adults.
Methods: This review and meta-analysis assembles prospective studies regarding the relation be-
tween frailty and any potential health outcome. Frailty instruments were subdivided into frailty
concepts, so as to make comprehensive comparisons. Odds ratios (ORs), hazard ratios (HRs), and
relative risk (RR) scores were extracted from the studies, and meta-analyses were conducted in
OpenMeta Analyst software.
Results: In total, 31 articles retrieved from PubMed, Web of Knowledge, and PsycInfo provided
sufcient information for the systematic review and meta-analysis. Overall, (pre)frailty increased
the likelihood for developing negative health outcomes; for example, premature mortality (OR 2.34
[1.77e3.09]; HR/RR 1.83 [1.68e1.98]), hospitalization (OR 1.82 [1.53e2.15]; HR/RR 1.18 [1.10e1.28]),
or the development of disabilities in basic activities of daily living (OR 2.05 [1.73e2.44]); HR/RR 1.62
[1.50e1.76]).
Conclusion: Overall, frailty increases the risk for developing any discussed negative health outcome, with
a 1.8- to 2.3-fold risk for mortality; a 1.6- to 2.0-fold risk for loss of activities of daily living; 1.2- to 1.8-fold
risk for hospitalization; 1.5- to 2.6-fold risk for physical limitation; and a 1.2- to 2.8-fold risk for falls and
fractures. The analyses presented in this study can be used as a guideline for the prediction of negative
outcomes according to the frailty concept used, as well as to estimate the time frame within which these
events can be expected to occur.
2016 AMDA e The Society for Post-Acute and Long-Term Care Medicine.
h
Soe Vermeiren and Roberta Vella Azzopardi contributed equally to this article Members of the Gerontopole Brussels Study group: Ivan Bautmans (FRIA, VUB),
and should be indexed as shared rst authorship. Dominque Vert (Belgian Ageing Studies BAST, VUB), Ingo Beyer (Geriatric Medicine
This study was partly funded by an Interdisciplinary Research Program grant department, UZ Brussel), Mirko Petrovic (ReFrail, UGhent), Liesbeth De Donder
(number IRP3) from the research council of the Vrije Universiteit Brussel (VUB). The (Belgian Ageing Studies BAST, VUB), Tinie Kardol (Leerstoel Bevordering Active
authors have no other conict of interest to declare. Ageing, VUB), Gina Rossi (Clinical and Lifespan Psychology KLEP, VUB), Peter Clarys
* Address correspondence to Ivan Bautmans, PhD, Gerontology (GERO) and (Physical Activity and Nutrition PANU, VUB), Aldo Scafoglieri (Experimental Anat-
Frailty in Ageing Research (FRIA) Departments, Vrije Universiteit Brussel (VUB), omy EXAN, VUB), Eric Cattrysse (Experimental Anatomy EXAN, VUB), Paul de Hert
Laarbeeklaan 103, B-1090 Brussels, Belgium. (Fundamental Rights and Constitutionalism Research group FRC, VUB), Bart Jansen
E-mail address: ivan.bautmans@vub.be (I. Bautmans). (Department of Electronics and Informatics ETRO, VUB).
http://dx.doi.org/10.1016/j.jamda.2016.09.010
1525-8610/ 2016 AMDA e The Society for Post-Acute and Long-Term Care Medicine.
Descargado de ClinicalKey.es desde Universidad Nacional Autonoma de Mexico febrero 01, 2017.
Para uso personal exclusivamente. No se permiten otros usos sin autorizacin. Copyright 2017. Elsevier Inc. Todos los derechos reservados.
1163.e2 S. Vermeiren et al. / JAMDA 17 (2016) 1163.e1e1163.e17
Conceptually, frailty can be dened as a condition or syndrome were combined: frailty, frail elderly, aged, assessment, risk assessment,
that results from a multisystem reduction in reserve capacity, to the classication, diagnosis, prospective studies, longitudinal studies,
extent that a number of physiological systems are close to, or pass, the cohort studies. Prospective longitudinal studies were eligible for this
threshold of symptomatic clinical failure. As a consequence, the frail review if they investigated frailty in community-dwelling elderly per-
person is at increased risk of disability and death from minor external sons (aged 65 years and older) and if the occurrence of health outcomes
stresses.1 However, this frailty concept is operationalized in various was reported. Studies were included if they were written in English,
ways, complicating the identication of frail older persons in clinical Dutch, French, or German. No limit was set on publication date.
practice. In fact, the prevalence of frailty in community-dwelling older The screening process was performed by 2 reviewers indepen-
persons ranges between 4% and 59%, and for nursing home patients dently and blinded for each others results. First, articles were
between 19% and 76%, depending on the frailty denition used.2,3 screened based on title and abstract. Subsequently, full texts were
Moreover, some frailty instruments distinguish besides a frail and screened. In case of disagreement on the inclusion of an article, a
robust ( nonfrail) also a prefrail state, reecting a transition phase consensus was found by involving a third reviewer. In case insufcient
between robustness and frailty.4 More importantly, because frailty information was available to include or exclude a study, the corre-
status provides prospective information for negative health outcomes, sponding author was contacted.
it can be expected that the predictive value can be different according The following data were extracted from the included studies: the
to the operationalization of the frailty concept. number of participants tested, length of follow-up (categorized as
Several researchers have studied negative health outcomes 0e12 months, 12e24 months, 24e60 months, or >60 months), age,
possibly linked to frailty, with death, hospitalization, and institu- the frailty assessment tool and its concept (physical focus, multido-
tionalization as the most described end points.5,6 However, many main, or decit accumulation), the studied negative health outcomes
other outcomes have been put forward as possible repercussions of and their assessment method, the (adjusted) predictive statistical
frailty, including fractures, falls, disability in activities of daily living measures: odds ratio (OR), HR, or relative risk (RR). Articles were
(ADLs), mobility limitation, and cognitive decline.7,8 Yamada et al9 initially ordered according to the outcome of the study. When no ra-
found that their newly developed frailty screening index predicts tios were reported in the article, but frequencies were provided, the
healthy life expectancy in community-dwelling older people. They OR was calculated based on the frequencies of the group of robust
reported that prefrail (hazard ratio [HR] 8.4 [5.0e14.2]) and frail (HR subjects and the group of (pre)frail subjects. In case insufcient in-
22.7 [13.3e38.8]) older adults had a signicantly higher risk for formation was available in the article for calculating the ratio, the
making use of the long-term care insurance services compared with corresponding author was contacted to obtain the ratios or the fre-
robust older adults.9 This was supported by the ndings of Chang and quencies necessary to calculate the ratios. In case of multiple ratios per
Lin10 in the Cardiovascular Health Study4 for both a frail and prefrail frailty instrument (separate analysis per item), an overall predictive
status. According to Luo et al,11 the FRAIL NH-scale showed strong measure was calculated based on the total score of the instrument (as
predictive power for several negative health outcomes, such as inci- a whole).14,15
dent falls (HR 2.00 [1.41e2.83]), ADL-decline (HR 3.73 [2.69e5.16]), The systematic literature search yielded in January 2016 a total
hospitalization (HR 2.35 [1.57e3.54]), and death (HR 2.00 [1.41e2.83]) number of 1694 articles: 509 in PubMed, 464 in Web of Science, and
in nursing home residents. Almeida et al12 demonstrated that 721 in PsycINFO. Fifty-nine doubles were removed and after screening
community-dwelling men aged 75 years and older show a strong the remaining articles based on title and abstract, 291 were left for
relationship in the presence of frailty on the one hand, and an further analysis. The full texts, independently read and assessed by at
increased mortality associated with past depression on the other hand least 2 reviewers, were judged on content and methodological quality.
(HR 1.79 [1.21e2.62] after adjustment for frailty). Most studies found In total, 255 articles were excluded. In 12 cases, the rst 2 reviewers
in the literature focus on the relation between a certain frailty concept reported conicting results; a third reviewer was asked to assess the
and 1 or 2 negative health outcomes, providing relevant information article and a nal decision was reached based on consensus. Finally, 31
on the possible health risks induced by frailty. However, because of the articles were included for this systematic review and meta-analysis. In
various operationalizations of frailty concepts, results may vary Figure 1, the owchart of the literature selection is shown.
signicantly. A general overview of frailty concepts and a compre-
hensive meta-analysis of their relation with negative health outcomes Risk of Bias: Assessment of Methodological Quality
still lacks in literature, which makes it difcult for health care pro-
fessionals and researchers to recognize frailty and the related health Methodological quality was assessed by 2 reviewers by using
risks on the one hand and on the other hand to appropriately follow methodology checklists of the National Institute for Health and Care
up the frailty process and take substantiated action. Therefore, this Excellence.16 An overview of the applied checklists and the results per
study aims to give an overview of the prospective predictive value of study can be found in Appendix 2.
the main frailty concepts for negative health outcomes in community-
dwelling older adults. As prospective evidence will systematically be Meta-Analysis
quantied in this meta-analysis, we expect to provide important in-
formation for clinical practice so as to develop adequate interventions Meta-analyses were performed separately for the OR and the HR/RR.
and to counter the development and outcomes of frailty. The latter (HR and RR) were analyzed together, given their similarity.17
Meta-analyses were conducted in OpenMeta[Analyst] software for
Methods advanced meta-analysis from the Brown University Evidence-Based
Practice Center.18 Subgroup-analyses were performed based on (1) the
This systematic review was written according to the PRISMA frailty assessment tool category (physical-oriented scales, multidomain
guidelines for transparent reporting of systematic reviews and meta- scales, decit accumulation methods), (2) length of follow-up (0e12,
analyses.13 12e24, 24e60, and >60 months) and (3) the level of frailty (prefrail or
frail). I2 (heterogeneity) values with signicance level were reported as
Literature Search a measure of the degree of inconsistency in the studies results. I2 values
may range from 0% (no observed heterogeneity) to 100% (complete
A literature search was performed in PubMed, Web of Knowledge heterogeneity) and values of 25%, 50%, and 75% can be considered as
and PsycInfo (last search in January 2016). The following search terms respectively low, moderate, and high.19
S. Vermeiren et al. / JAMDA 17 (2016) 1163.e1e1163.e17 1163.e3
For each negative health outcome, overall OR HR/RR were calcu- model,4 other frailty instruments with a physical focus were devel-
lated, as well as subgroup results according to the instrument used, oped. Chang et al22 suggested the Short Physical Performance Battery
frailty status, and follow-up. (containing balance tests, gait speed, and the 5 times chair stand test)
to be an effective instrument for assessing frailty.
Results and Discussion
Multidomain
Data Extraction and Synthesis Sixteen different instruments describe the multidimensional char-
acter of frailty, containing medical, physical, cognitive, psychosocial, and/
Articles were initially ordered according to the outcome of the study. or environmental factors. The Conselice Study of Brain Aging Index
Multiple negative health outcomes were possible for 1 article. In total, 24 (CSBA), for example, is a multidimensional tool consisting of 9 possible
studies described the relation between frailty and mortality, 11 articles predictors for negative health outcomes, clustered in 6 domains: socio-
looked at hospitalization, 11 at disability in basic ADLs (BADLs), 7 at demographic, lifestyle, nutrition, physical function, medical status, and
disability in instrumental ADLs (IADLs), 6 looked at falls, 5 at institu- mood/cognitive status.23 To collect data in a straightforward manner,
tionalization, 5 at fractures, 3 looked at physical limitation, 2 at emer- questionnaires can be a solution. In the frailty context, several re-
gency department (ED) visits, 2 at cognitive decline, and 1 each looked at searchers developed self-reported questionnaires so as to determine the
life satisfaction, body composition, and dependency. Figure 2 shows the presence or absence of frailty. All questionnaires selected for this review
proportion of the different frailty instruments studied per outcome. contain several domains: physical, medical (medication), psychological,
Twenty-nine different frailty instruments were identied and social, and/or environmental (eg, Tilburg Frailty Indicator).24,25
categorized based on their concept (focus and calculation method).
Three overall groups could be composed: (1) physical focus, (2) Decit accumulation
multidomain, and (3) decit accumulation. Seven different instruments were identied. A third approach to
assess frailty is the method of decit accumulation. Usually this type
Physical focus of index consists of a varying number of decits covering several do-
Seven frailty assessment tools use a frailty concept with a physical mains like the medical, physical, cognitive, psychological, and/or so-
focus. Linda Fried,4 one of the important frailty researchers, devel- cial. Each decit is given a score ranging between 0 and 1 depending
oped a physical frailty phenotype based on the CHS (Cardiovascular on the presence and the severity of the decit. A total score is obtained
Health Study) consisting of 5 components: unintentional weight loss, by dividing the sum-score on present decits by the total number of
muscle weakness, exhaustion (low energy level), slowness (slow gait), decits.26 Most researchers consider a score of 0.25 or higher to be
and low physical activity. Self-reported questions and noninvasive indicative for frailty.27 Several researchers used this methodology, and
tests dene the level of frailty: having none of the 5 previous com- the (number of) decits varies between the different scales.28,29
ponents indicates robustness, having 1 or 2 components expresses Frailty instruments were thus classied according to the 3 pre-
prefrailty (an in-between state), and frailty is dened as having 3 or viously mentioned categories. The rationale for this classication lies
more of the 5 components. Frieds approach4 is applied in many within the domains these instruments refer to in their assessment.
studies; however, not always using the same assessment tools and/or First and foremost, a distinction was made between assessment tools
construct. Woods et al,20 for example, used the RAND-36 Physical looking at 1 domain (only instruments with a physical focus were
Function Scale instead of hand grip strength (weakness) and walking found) or more than 1 domain (multidomain). The rst category
speed. Buchman et al,21 however, removed the component physical (physical focus) seemed to be coherent, whereas the latter (multi-
activity and developed a continuous composite measure rather than domain) contained diverse tools. Therefore, a third category was
a categorical classication. Notwithstanding that many researchers dened including those instruments operationalizing frailty as an
have applied an adjusted model of this phenotypic approach, the main accumulation of decits. This categorization has been used previ-
focus remains the physical aspect of frailty. Next to Frieds phenotypic ously in other studies, but also other combinations are possible (eg,
Fig. 1. Literature owchart.
Fig. 2. Number of frailty instruments per type and per outcome.
self-report versus objective versus mixed frailty measures; ques- RR 2.10 [1.38e3.19]), 12 to 24 months (OR 2.31 [1.55e3.45]), 24 to
tionnaires versus decit accumulation).30 The choice of the classi- 60 months (OR 3.25 [2.14e4.94], RR 1.57 [1.43e1.72]), as well as when
cation that we used here may have inuenced the results of the followed up more than 60 months (OR 2.17 [1.22e3.86], RR 2.14
meta-analysis shown in this article. [1.60e2.87]) (see Figure A3 in Appendix 1). It seems that a frail status
(OR 2.55 [1.76e3.70], RR 2.01 [1.82e2.22]) and a prefrail status (OR
Results and Interpretation per Outcome(s) 1.76 [1.36e2.28], RR 1.47 [1.32e1.62]) show a similar elevated risk of
mortality (see Figure A1 in Appendix 1). It needs to be mentioned that
For each negative health outcome, studies have been ordered a statistically signicant overall heterogeneity was found for OR
according to the ratio provided (OR and HR/RR). Age category, (I2 95%, P < .001) and HR/RR (I2 98%, P < .001). This may be
number of subjects, the frailty instrument and its concept categori- because different population groups, different follow-up periods, and
zation, follow-up (in months and category), the outcome assess- several frailty concepts were included in the meta-analysis.
ment, and the retrieved or calculated results are described. Per Table 1 shows all studies regarding frailty and mortality that are
negative health outcome and ratio, individual study results and included in the meta-analysis. In all studies, mortality was dened as
overall results are shown. In each table, rst all ORs are shown, fol- whether the participant died during the follow-up period. Individual
lowed by all HR/RRs. study results are provided in the last column, whereas the overall ratio
is shown below these studies. The overall OR and RR/HR have been
Mortality calculated and are shown separately.
Twenty-four prospective studies, comprising 25 different frailty
instruments and 150,763 older adults, described the association be- Hospitalization, institutionalization, ED visits
tween frailty and mortality and were included in the meta-ana- Hospitalization. Eleven studies encompassing 16 different frailty
lysis.5,15,20,24,26,28,31e48 Studies included frailty scales with a physical instruments and 59,297 subjects described the association between
focus (n 5 different scales), a multidomain focus (n 14), and decit frailty and hospitalization and were included in the meta-anal-
accumulation approach (n 6). ysis.5,20,24,25,28,33,34,38,46,47,49 Overall, these studies show that frailty
Overall, the presence of frailty signicantly increases the likelihood increases the risk for hospitalization (see Table 2 and Figure B1,
of premature mortality (OR 2.34 [1.77e3.09]; RR 1.83 [1.68e1.98]) (see Appendix 1). However, these ndings seem to be heterogeneous
Figure A1 in Appendix 1). Subanalysis per frailty concept shows that for OR as well as for HR/RR (for both ratios I2 81%, P < .001). As
the likelihood for premature mortality in frail subjects is signicantly shown in Table 2, the prefrail subgroup described by Hogan et al28
elevated regardless when applying scales are with a physical focus (OR and established with the Armstrong Frailty Index (decit accu-
2.58 [1.83e3.64], RR 1.70 [1.49e1.95]), multidomain instruments (OR mulation), presents no signicantly elevated risk for hospitaliza-
2.13 [1.38e3.29], RR 1.32 [1.22e1.43]), as well as with decit accu- tion. This analysis concerns a follow-up of less than 1 year. This
mulation methods (OR 1.85 [1.30e2.63] [only 1 study/variable/ratio nding is supported by Forti et al,46 who established (pre)frailty
observed so no meta-analysis], RR 3.64 [1.72e7.72]) (see Figure A2 in according to the SOF (Study of Osteoporotic Fractures, physical-
Appendix 1). When divided into time to follow-up categories, meta- oriented scale), but with a follow-up period of 3 years. Overall
analysis shows that the risk for mortality is signicantly elevated in subgroup analysis of all studies shows that having a frail status
frail persons within a follow-up of 0 to 12 months (OR 1.33 [1.11e1.60], exhibits a signicantly elevated risk for hospitalization (OR 1.97
Table 1
Predictive Value of Frailty on Mortality
Author Description of Participants Frailty Instrument and Category Follow- up, mo Outcome Assessment Results
Mortality
Woo et al (2012) Age 65 n 4000 CHS (Fried) (1) 48 (c) Died yes/no PF vs R: OR 1.89 [1.52e2.34]*
FRAIL scale (1) F vs R: OR 4.41 [3.22e6.05]*
PF vs R: OR 1.82 [1.44e2.31]*
F vs R: OR 1.59 [0.08e33.26]*
Forti et al (2012) Age 65 n 1007 mSOF (1) 48 (c) Died yes/no F vs R: OR 7.13 [4.67e10.89]
84 (d) F vs R: OR 5.92 [3.92e8.95]
Forti et al (2014) Age 65 n 766 SOF (1) 84 (d) Died yes/no PF vs R: OR 0.97 [0.69e1.37]
F vs R: OR 2.12 [1.34e3.36]
Romero-Ortuno et al (2014) Older women 75 n 2402 mCHS (Fried) (1) 24 (b) Died yes/no PF vs R: OR 1.70 [1.10e2.70]*
Older men 75 n 1907 F vs R: OR 2.20 [1.20e3.80]*
PF vs R: OR 1.90 [1.30e2.80]*
F vs R: OR 4.20 [2.60e6.80]*
OCaoimh et al (2014) Age 65 n 803y CFS (2) 12 (a) Died yes/no F vs R: OR 1.18 [0.92e1.51]
RISC (2) F vs R: OR 1.58 [1.20e2.08]
Ma et al (2009) Age 65 n 230 VES-13 (2) 72 (d) Died yes/no F vs R: OR 1.16 [0.98e1.37]
mVES-13 (2) F vs R: OR 1.10 [1.02e1.17]
Woo et al (2012) Age 65 n 4000 Hubbard Scale (2) 48 (c) Died yes/no PF vs R: OR 3.99 [2.01e7.94]*
S. Vermeiren et al. / JAMDA 17 (2016) 1163.e1e1163.e17

F vs R: OR 2.41 [1.35e4.30]*
Forti et al (2012) Age 65 n 1007 CSBA (2) 48 (c) Died yes/no F vs R: OR 8.11 [5.34e12.32]
84 (d) F vs R: OR 7.33 [5.41e9.92]
Daniels et al (2012) Age 70 n 430 GFI (2) 12 (a) Died yes/no F vs R: OR 1.35 [0.32e5.76]
TFI (2) F vs R: OR 1.05 [0.24e4.60]
SPQ (2) F vs R: OR 0.92 [0.20e4.33]
Woo et al (2012) Age 65 n 4000 Frailty Index (47 items) (3) 48 (c) Died yes/no F vs R: OR 1.85 [1.30e2.62]*
Overall OR for mortality Overall OR 2.34 [1.77e3.09]
Hogan et al (2012) Age 65 n 1066z CHS (Fried) (1) 12 (a) Died yes/no PF vs R: RR 0.88 [0.51e1.54]
F vs R: RR 1.74 [1.07e2.81]
Kulminski et al (2008) Age 65 n 4721 CHS (Fried) (1) 48 (c) Died yes/no PF vs R: RR 2.08 [1.63e2.66]
F vs R: RR 4.81 [3.53e6.55]
Graham et al (2009) Age 65 n 1996 mCHS (Fried) (1) 120 (d) Died yes/no PF vs R: HR 1.25 [1.07e1.46]
F vs R: HR 1.81 [1.41e2.31]
Avila-Funes et al (2009) Age 65 n 6030 mCHS (Fried) (1) 48 (c) Died yes/no PF vs R: HR 1.23 [0.94e1.60]
F vs R: HR 1.30 [0.83e2.04]
Ensrud et al (2009) Older men 65 n 3132 SOF (1) 10 (a) Died yes/no PF vs R: HR 1.31 [0.94e1.83]
mCHS (Fried) (1) F vs R: HR 2.53 [1.75e3.66]
PF vs R: HR 1.77 [1.17e2.68]
F vs R: HR 3.51 [2.21e5.57]
Ensrud et al (2007) Older women 65 n 6724 CHS (Fried) (1) 108 (d) Died yes/no PF vs R: HR 1.32 [1.18e1.48]
F vs R: HR 1.82 [1.56e2.13]
Woods et al (2005) Age 65 n 40,657x mCHS (Fried) (1) 69 (d) Died yes/no PF vs R: HR 1.25 [1.11e1.41]
F vs R: HR 1.71 [1.48e1.97]
Buchman et al (2011) Age 65 n 949k mCHS (Fried) (1) 60 (c) Died yes/no F vs R: HR 1.48 [1.27e1.73]
Fried et al (2001) Age 65 n 5317 CHS (Fried) (1) 36 (c) Died yes/no PF vs R: HR 1.49 [1.11e1.99]
84 (d) F vs R: HR 2.24 [1.51e3.33]
PF vs R: HR 1.32 [1.13e1.55]
F vs R: HR 1.63 [1.27e2.08]
Lee et al (2014) Age 65 n 12,087 mCHS (Fried) (1) 36 (c) Died yes/no PF vs R: HR 1.60 [1.26e2.05]
F vs R: HR 2.28 [1.61e3.22]
Jotheeswaran et al (2015) Age 65 n 47,438 mCHS (Fried) (1) 36 (c) Died yes/no F vs R: HR 1.18 [1.06e1.33]
Sirola et al (2011) Older men 70 n 1125 mCHS (Fried) (1) 96 (d) Died yes/no PF vs R: HR 2.15 [1.46e3.16]
F vs R: HR 3.55 [2.17e5.79]
Hogan et al (2012) Age 65 n 1066z CHESS (2) 12 (a) Died yes/no PF vs R: RR 1.38 [0.98e1.94]
F vs R: RR 1.87 [1.35e2.59]
Lucicesare et al (2010) Age 65 n 1016 CSBA (2) 48 (c) Died yes/no F vs R: HR 1.52 [1.28e1.81]
1163.e5
(continued on next page)
Table 1 (continued )
1163.e6
Author Description of Participants Frailty Instrument and Category Follow- up, mo Outcome Assessment Results
Garca-Garca et al (2014) Age 65 n 1972 Frailty Trait Scale (2) 42 (c) Died yes/no F vs R: HR 1.02 [1.01e1.03]
Jotheeswaran et al (2015) Age 65 n 47,438 Multidimensional phenotype (2) 36 (c) Died yes/no F vs R: HR 1.38 [1.24e1.54]
Rothman et al (2008) Age 70 n 745 Rothman scale (2) 96 (d) Died yes/no F vs R: HR 1.45 [1.23e1.70]{
Rockwood et al (2005) Age 75 n 2305x CFS (2) 60 (c) Died yes/no F vs R: HR 1.30 [1.27e1.33]
Rules based def. of Frailty (2) F vs R: HR 1.17 [1.13- 1.20]
Hogan et al (2012) Age 65 n 1066z Armstrong Index (3) 12 (a) Died yes/no PF vs R: RR 1.17 [0.59e2.29]
Full Frailty Index (83 items) (3) F vs R: RR 1.94 [1.02e3.70]
PF vs R: RR 2.00 [1.33e3.00]
F vs R: RR 2.35 [1.56e3.54]
Kulminski et al (2008) Age 65 n 4721 Frailty Index (48 items) (3) 48 (c) Died yes/no PF vs R: RR 1.94 [1.45e2.61]
F vs R: RR 4.45 [3.26e6.08]
Armstrong et al (2015) Age 70 n 3845** Frailty Index (3) 62 (d) Died yes/no F4 vs R: RR 4.68 [3.80e5.76]
F5 vs R: RR 5.81 [4.76e7.09]
F6 vs R: RR 6.38 [5.29e7.69]
Lucicesare et al (2010) Age 65 n 1016 Frailty Index (43 items) (3) 48 (c) Died yes/no F vs R: HR 5.26 [1.05e26.42]
Rockwood et al (2005) Age 75 n 2305x Frailty Index (70 items) (3) 60 (c) Died yes/no F vs R: HR 1.26 [1.24e1.29]
Overall HR/RR for mortality Overall HR/RR 1.83 [1.68e1.98]
CARTS, Community Assessment of Risk and Treatment Strategies Study; CFS, Clinical Frailty Scale; CHESS, Changes in Health, End-stage disease, Signs, and Symptoms scale; CHS, Cardiovascular Health Study (m modied

version); CSHA, Canadian Study of Health and Aging; F, frailty; GFI, Groningen Frailty Indicator; (m)SOF, (modied) Study of Osteoporotic Fractures; mVES, modied Vulnerable Elders Survey; PF, prefrailty; R, robustness; RISC,
Risk Instrument for Screening in the Community; SPQ, Sherbrooke Postal Questionnaire; TFI, Tilburg Frailty Indicator; VES, Vulnerable Elders Survey.
Please see Appendix 3 for alphabetical list of references. Description of study characteristics and OR/HR/RRs of the individual studies, and overall calculations. Instrument conceptual categories: 1 physical focus;
2 multidomain focus; 3 decit accumulation. Follow-up category in months: a 0e12; b 12e24; c 24e60; d >60.
*Self-calculated according to frequencies provided by the article or author.
y
Participants from the CARTS study.
z
Older adults in assisted living (AL) and Supported housing facilities (DAL).
x
Participants from the CSHA study.
k
Participants from the Rush Memory and Aging Project.
{
Self-calculated according to the HR per assessment item from the article> an overall score was calculated.
**Older men participating in the HAAS (Honolulu Asia Aging Study).
Table 2
Predictive Value of Frailty on Hospitalization, Institutionalization, and ED Visits
Author Description of Participants Frailty Instrument and Category Follow-up, mo Outcome Assessment Results
Hospitalization
Forti et al (2012) Age 65 n 1007 mSOF (1) 48 (c) Hospitalization F vs R: OR 1.88 [0.97e3.63]
yes/no
Kiely et al (2009) Age 65 n 765 SOF (1) 10 (a) Hospitalization PF vs R: OR 2.64 [1.74e4.01]
mCHS (Fried) (1) yes/no F vs R: OR 3.49 [1.53e7.98]

PF vs R: OR 1.97 [1.37e2.84]
F vs R: OR 4.45 [2.42e8.18]
Avila-Funes et al (2009) Age 65 n 6030 mCHS (Fried) (1) 48 (c) Hospitalization PF vs R: OR 1.15 [0.99e1.32]
yes/no F vs R: OR 1.41 [1.06e1.87]
Woods et al (2005) Age 6 n 40,657* mCHS (Fried) (1) 69 (d) Hospitalization PF vs R: OR 1.30 [1.17e1.45]
yes/no F vs R: OR 1.95 [1.72e2.22]
Forti et al (2014) Age 65 n 766 SOF (1) 36 (c) Hospitalization PF vs R: OR 0.97 [0.98e5.63]
yes/no F vs R: OR 2.35 [0.98e5.63]
Daniels et al (2012) Age 70 n 430 GFI (2) 12 (a) Hospitalization F vs R: OR 1.33 [0.73e2.41]
TFI (2) yes/no F vs R: OR 2.59 [1.36e4.90]
SPQ (2) F vs R: OR 2.42 [1.27e4.62]
Forti et al (2012) Age 65 n 1007 CSBA (2) 48 (c) Hospitalization F vs R: OR 2.89 [1.88e4.43]
yes/no

y
OCaoimh et al (2014) Age 65 n 803 CFS (2) 12 (a) Hospitalization F vs R: OR 1.05 [0.85e1.30]
RISC (2) yes/no F vs R: OR 1.28 [1.06e1.54]
Gobbens et al (2012) Age 75 n 484 TFI (2) 12 (a) Hospitalization F vs R: OR 2.59 [1.30e5.13]
24 (b) yes/no F vs R: 2.77 [1.31e5.84]
Overall OR for hospitalization Overall OR 1.82 [1.53e2.15]
Hogan et al (2012) Age 65 n 1066z CHS (Fried) (1) 12 (a) Hospitalization PF vs R: RR 1.06 [0.82e1.38]
yes/no F vs R: RR 1.45 [1.15e1.83]
Fried et al (2001) Age 65 n 5317 CHS (Fried) (1) 36 (c) Hospitalization PF vs R: HR 1.13 [1.03e1.25]
84 (d) yes/no F vs R: HR 1.29 [1.09e1.54]
PF vs R: HR 1.11 [1.03e1.19]
F vs R: HR 1.27 [1.11e1.46]
Hogan et al (2012) Age 65 n 1066z CHESS (2) 12 (a) Hospitalization PF vs R: RR 1.27 [1.07e1.51]
yes/no F vs R: RR 1.25 [1.05e1.50]
Garca-Garca et al (2014) Age 65 n 1972 Frailty Trait Scale (2) 42 (c) Hospitalization F vs R: HR 1.02 [1.01e1.02]
yes/no
Hogan et al (2012) Age 65 n 1066z Armstrong Index (3) 12 (a) Hospitalization PF vs R: RR 0.91 [0.68e1.22]
Full Frailty Index (83 items) (3) yes/no F vs R: RR 1.16 [0.87e1.53]
PF vs R: RR 1.37 [1.13e1.66]
F vs R: RR 1.28 [1.04e1.57]
Overall HR/RR for hospitalization Overall HR/RR 1.18 [1.10e1.28]
Institutionalization
Forti et al (2012) Age 65 n 1007 mSOF (1) 48 (c) Institutionalization F vs R: OR 2.53 [0.71e10.01]
yes/no
OCaoimh et al (2014) Age 65 n 803y CFS (2) 12 (a) Institutionalization F vs R: OR 1.03 [0.76e1.38]
RISC (2) yes/no F vs R: OR 1.43 [1.09e1.88]
Forti et al (2012) Age 65 n 1007 CSBA (2) 48 (c) Institutionalization F vs R: OR 5.52 [2.07e14.70]
yes/no
Overall OR for institutionalization Overall OR 1.69 [1.02e2.81]
Hogan et al (2012) Age 65 n 1066z CHS (Fried) (1) 12 (a) Institutionalization PF vs R: RR 1.49 [0.91e2.43]
yes/no F vs R: RR 2.17 [1.38e3.41]
Hogan et al (2012) Age 65 n 1066z CHESS (2) 12 (a) Institutionalization PF vs R: RR 1.33 [0.98e1.82]
yes/no F vs R: RR 1.87 [1.39e2.50]
Rothman et al (2008) Age 70 n 745 Rothman scale (2) 96 (d) Institutionalization F vs R: HR 1.73 [1.26. 2.37]x
yes/no
k
Rockwood et al (2005) Age 75 n 2305 CFS (2) 60 (c) Institutionalization F vs R: HR 1.46 [1.39e1.53]
Rules-based denition of Frailty (2) yes/no F vs R: HR 1.27 [1.19- 1.35]
1163.e7
[1.58e2.46]; HR/RR 1.23 [1.07e1.40]), as well as having a prefrail
CFS, Clinical Frailty Scale; CHESS, Changes in Health, End-stage disease, Signs, and Symptoms scale; F, frailty; GFI, Groningen Frailty Indicator; (m)CHS, (modied) Cardiovascular Health Study; (m)SOF, (modied) Study of
Overall HR/RR 1.65 [1.48e1.84]
Overall HR/RR 1.03 [0.82e1.29]

status (OR 1.53 [1.19e1.96]; HR/RR 1.15 [1.06e1.24]) (see
MF vs R: HR 1.10 [0.74e1.64]
PF vs R: HR 0.93 [0.62e1.39]
PF vs R: OR 2.19 [1.43e3.33]
PF vs R: OR 1.34 [0.95e1.89]
PF vs R: RR 2.21 [0.98e4.99]
PF vs R: RR 1.87 [1.27e2.75]
Overall OR 2.16 [1.39e3.37]

F vs R: HR 1.56 [1.48e1.65]
F vs R: HR 1.06 [0.73e1.54]
F vs R: OR 3.54 [1.43e8.79]
F vs R: OR 3.10 [1.64e5.86]
F vs R: RR 4.14 [1.87e9.14]
F vs R: RR 3.30 [2.29e4.76] Figure B1, Appendix 1). It seems that frailty is linked to a signif-
icantly elevated risk for hospitalization when identied with
multidomain instruments (OR 1.84 [1.35e2.51]; HR/RR 1.26
[1.12e1.42]), instruments with a physical focus (OR 1.83
[1.47e2.28]; HR/RR 1.16 [1.06e1.27]), as well as with decit
accumulation scales (OR: no studies; HR/RR 1.20 [1.02e1.41]) (see
Results
Figure B2, Appendix 1). Frailty predicts hospitalization within

1 year or less (OR 2.03 [1.51e2.74]; HR/RR 1.24 [1.14e1.35]), as
well as within 1 to 2 years (OR 2.77 [1.31e5.84]). However, it
should be noted that the latter analysis is based on only 1 study25
(see Figure B3, Appendix 1).
One or more ED visits
Outcome Assessment
A repeat outpatient
during follow-up
Osteoporotic Fractures; PF, prefrailty; R, robustness; RISC, Risk Instrument for Screening in the Community; SPQ, Sherbrooke Postal Questionnaire; TFI, Tilburg Frailty Indicator.
hospitalization)
ED visit (so not
Institutionalization. Five studies, encompassing 11 different

followed by
frailty instruments and 5926 older adults, discussed the relation

between frailty and institutionalization.5,15,26,28,47 As presented
yes/no
yes/no
in Table 2, the overall analysis shows a signicantly increased risk

for institutionalization in relation to frailty (OR 1.69 [1.02e2.81],
HR/RR 1.65 [1.48e1.84]) (see Figure B4, Appendix 1). This applies
to both a frail status (OR 1.69 [1.02e2.81]; HR/RR 1.67
[1.47e1.89]), and a prefrail status (HR/RR 1.55 [1.26e1.91]). No
Follow-up, mo
ORs were available regarding the prefrail status. When looking at

the type of frailty instrument applied, risk for institutionalization
12 (a)
10 (a)
60 (c)
is high when instruments with decit accumulation are used

1 (a)
(HR/RR 2.30 [1.54e3.43]). This effect remains for the HR/RR

calculated for physical-oriented tools (HR/RR 1.82 [1.26e2.63])
and instruments with a multidomain focus (HR/RR 1.44
[1.28e1.62]). However, this effect is not statistically signicant
Decit Accumulation Index (44 items) (3)
when considering the OR for physical-oriented tools (OR 2.53

[0.71e9.02] [only 1 study/variable/ratio observed so no meta-
Frailty Instrument and Category
analysis]) or instruments with a multidomain focus (OR 1.62

Full Frailty Index (83 items) (3)
[0.93e2.81]) separately (see Figure B5 in Appendix 1). Several

Frailty Index (70 items) (3)
follow-up periods have been observed across the different

studies. We found no data for a follow-up period of 12 to
Armstrong Index (3)
24 months. Nevertheless, an elevated risk for institutionalization

mCHS (Fried) (1)
has been established within a follow-up of 0 to 12 months (OR

1.22 [0.89e1.69]; HR/RR 2.04 [1.58e2.63]), a follow-up of 24 to
60 months (OR 4.13 [1.90e8.97]; HR/RR 1.43 [1.28e1.59]), as well
SOF (1)
as a follow-up of 60 months or longer (OR no data, HR/RR 1.73

*Older women participating in the Womens Health Initiative Observational Study.
[1.26e2.37] [only 1 study/variable/ratio observed so no meta-

analysis]) (see Figure B6 in Appendix 1).
Older adults in assisted living (AL) and Supported housing facilities (DAL).
Description of Participants
ED visits. Two articles including in total 2616 participants

described the risk of frailty for ED visits.49,50 Only physical-
Age 75 n 2305k
Age 65 n 1066z
oriented instruments and decit accumulation methods are

Age 65 n 1851
Age 65 n 765
applied in these studies. ED visits are dened as a repeated

outpatient ED visit (ie, not followed by hospitalization)50 or as
any ED visit during follow-up.49 Overall analysis shows that
frailty signicantly increases the risk for ED visits in the near
future in one study (OR 2.16 [1.39e3.37]),49 whereas this is not
statistically signicant in the other study (HR 1.03 [0.82e1.29])50
(Table 2 and Figure B7 in Appendix 1). Given the slightly diverse
denition of ED visits in both studies, comparing both articles is
Overall HR/RR for institutionalization
Participants from the CARTS study.
Participants from the CSHA study.
difcult. However, the frequency of ED visits as well as the ne-

cessity for ED help, do have a similar character. Both studies have
a follow-up period of less than 1 year. One article applies a
Overall HR/RR for ED visits
physical frailty assessment scale,49 and the other one a decit

Rockwood et al (2005)
Overall OR for ED visits

Hastings et al (2008)
accumulation method.50 Because only 2 different studies were

Hogan et al (2012)
Kiely et al (2009)
found providing different ratios (OR vs HR), no meta-analysis

could thus be performed for the length of follow-up or for the
frailty instrument applied.
ED visits
The Decit Accumulation Index used by Hastings et al50 divides

Author
the cohort into 3 subgroups: prefrail, midfrail, and frail. This cate-
y
gorization attributes a different meaning to the previously used
Table 3
Predictive Value of Frailty on the Development of Disabilities (Basic and Instrumental), Physical Limitation, and Dependency
Development of Disabilities (decline in BADL)
Forti et al (2012) Age 65 n 1007 mSOF (1) 48 (c) At least one new disability F vs R: OR 1.52 [0.72e3.21]
Avila-Funes et al (2009) Age 65 n 6030 mCHS (Fried) (1) 48 (c) BADL index (Katz et al, 1963) PF vs R: OR 0.83 [0.46e1.50]
F vs R: OR 3.28 [1.61e6.67]
Woods et al (2005) Age 65 n 40,657* mCHS (Fried) (1) 69 (d) At least one new disability PF vs R: OR 1.64 [1.31e2.04]
F vs R: OR 3.15 [2.47e4.02]
Forti et al (2014) Age 65 n 766 SOF (1) 36 (c) Any new disability PF vs R: OR 1.80 [1.15e2.82]
F vs R: OR 1.53 [0.61e3.87]
Gobbens et al (2014) Age 65 n 355 TFI physical subscale (1) 18 (b) GARS (Kempen et al, 1990) F vs R: OR 1.39 [1.28e1.51]
Romero-Ortuno et al (2015) Older women 75 n 3325 mCHS (Fried) (1) 48 (c) BADL index (Katz et al, 1963) PF vs R: OR 2.73 [2.27e3.29]y
Older men 75 n 2587 F vs R: OR 3.84 [2.85e5.18]y
PF vs R: OR 2.36 [1.77e3.15]y
F vs R: OR 2.83 [1.65e4.85]y
Forti et al (2012) Age 65 n 1007 CSBA (2) 48 (c) At least one new disability F vs R: OR 3.45 [2.32e5.12]
Daniels et al (2012) Age 70 n 430 GFI (2) 12 (a) GARS (Kempen et al, 1990) F vs R: OR 2.62 [1.48e4.64]
TFI (2) F vs R: OR 2.00 [1.18e3.57]
SPQ (2) F vs R: OR 2.49 [1.35e4.61]
Gobbens et al (2012) Age 75 n 484 TFI (2) 12 (a) GARS (Kempen et al, 1990) F vs R: OR 1.39 [1.28e1.51]

24 (b) F vs R: OR 1.27 [1.19e1.36]
Overall OR for BADL disability Overall OR 2.05 [1.73e2.44]
Buchman et al (2011) Age 65 n 949z mCHS (Fried) (1) 60 (c) BADL index (Katz et al, 1963) F vs R: HR 1.44 [1.21e1.71]
Fried et al (2001) Age 65 n 5317 CHS (Fried) (1) 36 (c) BADL index (Katz et al, 1963) PF vs R: HR 1.67 [1.41e1.99]
84 (d) F vs R: HR 1.98 [1.54e2.55]
PF vs R: HR 1.55 [1.38e1.75]
F vs R: HR 1.79 [1.47e2.17]
Rothman et al (2008) Age 70 n 745 Rothman scale (2) 96 (d) Questions on the need for help F vs R: HR 1.59 [1.21. 2.09]x
for certain tasks
Overall HR/RR for BADL disability Overall HR/RR 1.62 [1.50e1.76]
Development of Disabilities (decline in IADL)
Kiely et al (2009) Age 65 n 765 SOF (1) 10 (a) IADL scale (Lawton et al, 1969) PF vs R: OR 2.88 [1.81e4.58]
mCHS (Fried) (1) F vs R: OR 5.38 [2.34e12.35]
PF vs R: OR 2.73 [1.69e4.40]
F vs R: OR 7.68 [4.01e14.74]
Avila-Funes et al (2009) Age 65 n 6030 mCHS (Fried) (1) 48 (c) IADL scale (Lawton et al, 1969) PF vs R: OR 1.50 [1.17e1.82]
F vs R: OR 2.20 [1.47e3.24]
Ensrud et al (2009) Older men 65 n 3132 SOF (1) 10 (a) Any new IADL disability PF vs R: OR 2.47 [1.87e3.25]
PF vs R: OR 2.61 [1.89e3.62]
F vs R: OR 7.52 [5.14e11.02]
Gobbens et al (2014) Age 65 n 355 TFI physical subscale (1) 18 (b) GARS (Kempen et al, 1990) F vs R: OR 1.39 [1.28e1.51]
Carrire et al (2005) Older women 75 n 545 Physical Frailty Score (1) 84 (d) IADL scale (Lawton et al, 1969) F vs R: OR 1.86 [1.54e2.25]x
Romero-Ortuno et al (2015) Older women 75 n 3325 mCHS (Fried) (1) 48 (c) IADL scale (Lawton et al, 1969) PF vs R: OR 2.41 [2.03e2.86]y
Older men 75 n 2587 F vs R: OR 3.22 [2.41e4.30]y
PF vs R: OR 2.32 [1.77e3.04]y
F vs R: OR 2.28 [1.34e3.86]y
Daniels et al (2012) Age 70 n 430 GFI (2) 12 (a) GARS (Kempen et al, 1990) F vs R: OR 2.62 [1.48e4.64]
TFI (2) F vs R: OR 2.00 [1.18e3.57]
SPQ (2) F vs R: OR 2.49 [1.35e4.61]
Overall OR IADL disability Overall OR 2.52 [2.08e3.06]
Physical Limitation
Avila-Funes et al (2009) Age 65 n 6030 mCHS (Fried) (1) 48 (c) Rosow-Breslau scale (1966) PF vs R: OR 1.34 [1.13e1.57]
F vs R: OR 1.60 [0.88e2.87]
1163.e9
1163.e10
Woo et al (2012) Age 65 n 4000 CHS (Fried) (1) 60 (c) 2 questions about mobility: climbing stairs and PF vs R: OR 1.95 [1.68e2.26]y
FRAIL scale (1) performing household activities F vs R: OR 6.22 [4.19e9.23]y
PF vs R: OR 1.97 [1.60e2.44]y
F vs R: OR 7.32 [0.35e52.52]y
Woo et al (2012) Age 65 n 4000 Hubbard Scale (2) 60 (c) 2 questions about mobility: climbing stairs and PF vs R: OR 3.08 [1.47e6.48]y
performing household activities F vs R: OR 2.86 [1.61e5.08]y

Woo et al (2012) Age 65 n 4000 Frailty Index (47 items) (3) 60 (c) 2 questions about mobility: climbing stairs and F vs R: OR 4.83 [3.23e7.21]y
performing household activities PF vs R: OR 1.95 [1.68e2.26]y
Overall OR for physical limitation Overall OR 2.58 [1.85e3.62]
Fried et al (2001) Age 65 n 5317 CHS (Fried) (1) 36 (c) Time to walk 15 ft and maximal grip strength in PF vs R: HR 1.58 [1.41e1.76]
84 (d) the dominant hand F vs R: HR 1.50 [1.23e1.82]
PF vs R: HR 1.41 [1.29e1.54]
F vs R: HR 1.36 [1.15e1.62]
Overall HR/RR for physical limitation Overall HR/RR 1.46 [1.37e1.56]
Dependency
Jotheeswaran et al (2015) Age 65 n 30,689 mCHS (Fried) (1) 36 (c) Series of open-ended questions regarding F vs R: HR 1.28 [1.10e1.48]
received and needed help and care
Jotheeswaran et al (2015) Age 65 n 30,689 Multidimensional phenotype (2) 36 (c) Series of open-ended questions regarding F vs R: HR 1.36 [1.18e1.57]
received and needed help and care

Overall HR/RR for dependency Overall HR/RR 1.32 [1.19e1.47]
F, frailty; GFI, Groningen Frailty Indicator; (m)CHS, (modied) Cardiovascular Health Study; (m)SOF, (modied) Study of Osteoporotic Fractures; PF, prefrailty; R, robustness; SPQ, Sherbrooke Postal Questionnaire; TFI, Tilburg
Frailty Indicator.
Please see Appendix 3 for alphabetical list of references. Description of study characteristics and OR/HR/RR of the individual studies, and overall calculations. Instrument conceptual categories: 1 physical focus;
*Older women participating in the Womens Health Initiative Observational Study.
y
Self-calculated according to frequencies provided by the article or author.
z
Participants from the Rush Memory and Aging Project.
x
Table 4
Predictive Value of Frailty on the Occurrence of Falls and Fractures
Falls
Forti et al (2012) Age 65 n 1007 mSOF (1) 48 (c) Two or more falls F vs R: OR 0.95 [0.28e3.21]
Ensrud et al (2009) Older men 65 n 3132 SOF (1) 10 (a) Two or more falls PF vs R: OR 1.56 [1.23e1.97]
PF vs R: OR 1.62 [1.24e2.11]
F vs R: OR 3.56 [2.58e4.93]
Ensrud et al (2007) Older women 65 n 6724 CHS (Fried) (1) 108 (d) Two or more falls PF vs R: OR 0.90 [0.72e1.12]
F vs R: OR 1.38 [1.02e1.88]
Forti et al (2012) Age 65 n 1007 CSBA (2) 48 (c) Two or more falls F vs R: OR 1.49 [0.69e3.22]
Overall OR for falls Overall OR 1.70 [1.18e2.44]
Kiely et al (2009) Age 65 n 765 SOF (1) 10 (a) Two or more falls PF vs R: HR 1.62 [1.14e2.32]

PF vs R: HR 1.10 [0.80e1.50]
F vs R: HR 1.90 [1.17e3.10]
Fried et al (2001) Age 65 n 5317 CHS (Fried) (1) 36 (c) At least 1 fall PF vs R: HR 1.16 [1.00e1.34]
84 (d) F vs R: HR 1.29 [1.00e1.68]
PF vs R: HR 1.12 [1.00e1.26]
F vs R: HR 1.23 [0.99e1.54]

Rothman et al (2008) Age 70 n 745 Rothman scale (2) 96 (d) At least one injurious fall F vs R: HR 1.21 [0.95. 1.53]*
Overall HR/RR for falls Overall HR/RR 1.24 [1.12e1.37]
Fractures
Forti et al (2012) Age 65 n 1007 mSOF (1) 48 (c) Any fracture F vs R: OR 5.79 [2.90e11.55]
Forti et al (2014) Age 65 n 766 SOF (1) 36 (c) Any fracture (except PF vs R: OR 1.22 [0.54e2.74]
those due to cancer or F vs R: OR 5.51 [1.87e16.26]
road accidents)
Forti et al (2012) Age 65 n 1007 CSBA (2) 48 (c) Any fracture F vs R: OR 1.76 [0.99e3.15]
Overall OR for fractures Overall OR 2.78 [1.30e5.99]
Ensrud et al (2009) Older men 65 n 3132 SOF (1) 10 (a) Any fracture PF vs R: HR 1.30 [0.91e1.84]
PF vs R: HR 1.39 [0.94e2.06]
F vs R: HR 2.30 [1.43e3.71]
Ensrud et al (2007) Older women 65 n 6724 CHS (Fried) (1) 108 (d) Hip fractures PF vs R: HR 1.27 [1.04e1.56]
Nonspine fractures F vs R: HR 1.40 [1.03e1.90]
PF vs R: HR 1.11 [0.99e1.24]
F vs R: HR 1.25 [1.05e1.49]
Woods et al (2005) Age 65 n 40,657y mCHS (Fried) (1) 69 (d) Hip fractures PF vs R: HR 1.31 [1.00e1.71]
F vs R: HR 1.57 [1.11e2.00]
Overall HR/RR for fractures Overall HR/RR 1.37 [1.21e1.54]
F, frailty; (m)CHS, (modied) Cardiovascular Health Study; (m)SOF, (modied) Study of Osteoporotic Fractures; PF, prefrailty; R, robustness.
*Self-calculated according to the HR per assessment item from the article> an overall score was calculated.
y
Older women participating in the Womens Health Initiative Observational Study.
1163.e11
Table 5
Predictive Value of Frailty on Cognitive Decline
Author Description of Participants Frailty Instrument and Follow-up, mo Outcome Assessment Results
Category
Cognitive Decline
Avila-Funes et al (2009) Age 65d n 6030 mCHS (Fried) (1) 48 (c) Analysis of dementia (3- PF vs R: HR 1.29 [0.86
step procedure: 3C Study e1.93]
Group, 2003) and F vs R: HR 1.14 [0.58e2.21]
subjective cognitive
complaints (questions)
Buchman et al (2011) Age 65 n 949* mCHS (Fried) (1) 60 (c) Diagnosis of Alzheimer F vs R: HR 1.56 [1.27e1.92]
disease
Overall HR/RR for cognitive Overall HR 1.47 [1.23e1.76]
decline
F, frailty; (m)CHS, (modied) Cardiovascular Health Study; PF, prefrailty; R, robustness.

Please see Appendix 3 for alphabetical list of references. Description of study characteristics and HRs of the individual studies, and overall calculations. Instrument conceptual
categories: 1 physical focus; 2 multidomain focus; 3 decit accumulation. Follow-up category in months: a 0e12; b 12e24; c 24e60; d >60.
*Participants from the Rush Memory and Aging Project.
terms prefrail and frail, complicating the comparison with frailty (OR 1.80 [1.45e2.22]; HR/RR 1.59 [1.21e2.09]) (see Figure C2,
stages as described for other frailty instruments. Appendix 1). No studies applying decit accumulation methods
were found. Regardless the time of follow-up, frailty is signicantly
Disability and dependency linked to a higher risk for BADL disability (0e12 months OR 1.92
BADL disability. The relation between frailty and the development [1.32e2.80]; HR/RR no data; 12e24 months OR 1.27 [1.19e1.36]
of BADL disability was described by 11 articles comprising in total [only 1 study/variable/ratio observed so no meta-analysis]; HR/
60,065 older adults and 10 different frailty RR no data; 24e60 months OR 2.17 [1.60e2.95]; HR/RR 1.65
instruments.15,20,24,25,34,35,38,45e47,51 BADL disability is generally [1.40e1.95]; >60 months OR 2.27 [1.20e4.30]; HR/RR 1.61
described as a decline of at least 1 daily activity (bathing, dressing, [1.46e1.77]) (see Figure C3 in Appendix 1).
transferring from bed to chair, toileting, and feeding)20,38 based on Because the article of Rothman et al15 provides HRs for each in-
the ADL-scale developed by Katz et al.52 Two studies use GARS dividual assessment item (slow gait speed, low physical activity,
(Groningen Activity Restriction Scale) to assess BADL (but also weight loss, self-reported exhaustion, weakness, cognitive impair-
IADL) disability. This instrument represents 18 items for measuring ment, and depressive symptoms), it was not feasible to compare these
disability encompassing 11 BADL and 7 IADL items, but no separate scores with other studies. Therefore, an overall score has been
measures for BADL and IADL are available. Therefore, the total GARS calculated and used for the meta-analysis. The sum of these individual
score is used for the assessment of BADL disability, as well as for items has been categorized as a multidomain scale.15
IADL disability.24,51 One study interprets a disability in BADL as the
need for help for certain tasks.15 These different approaches of IADL disability. The relation between frailty and IADL disability was
BADL disability might inuence the interpretation and results of described by 7 studies, encompassing 3 different physical-oriented
comparison between studies, because it could instinctively be and 3 multidomain frailty instruments assessed on 14,582 older
assumed that frailty shows a higher prediction for 1 BADL disability subjects.14,24,36,38,45,49,51 No decit accumulation methods were
than more disabilities or an IADL disability (GARS). applied. All studies provide exclusively OR values, enabling a
Overall analysis shows that frailty increases the risk for devel- straightforward meta-analysis. Disabilities in IADL are generally
oping BADL disabilities. Heterogeneity is signicantly high for OR assessed by using the IADL scale encompassing the ability to use
(I2 91%, P < .001) but low for HR/RR (I2 15%, P .317). When the telephone, shopping, food preparation, housekeeping, laundry,
prefrailty is present, one is at higher risk for developing disabilities in mode of transportation, responsibility for own medications, and
BADL (OR 1.86 [1.35e2.56]; HR/RR 1.59 [1.44e1.75]); similar results the ability to handle nances.53 Two studies assess IADL by using
are found for a frail status (OR 2.13 [1.76e2.59]; HR/RR 1.67 the GARS. The purpose of this instrument is to dene BADL and
[1.45e1.92]) (see Figure C1, Appendix 1). Only 1 study does not show IADL disabilities. Because only total GARS scores are available, the
a signicantly higher risk for BADL disability when being prefrail.38 ORs are used for both outcomes (BADL and IADL).24,51
The presence of frailty is accompanied with a signicantly Compared with people who are robust, the presence of frailty
increased risk for the development of BADL disabilities when map- signicantly increases the risk for developing disabilities in IADL
ped with either a physical-oriented instrument (OR 2.11 [1.61e2.76]; (OR 2.73 [2.19e3.42]) (see Figure C4, Appendix 1). In analog with
HR/RR 1.63 [1.49e1.79]) or instruments with a multidomain focus the analysis for BADL disabilities, frailty status objectied using
Table 6
Predictive Value of Frailty on BC
Body Composition
Jung et al (2014) Age 65 n 341 CHS (Fried) (1) 62 (d) Decline of >5% in Lean Mass PF vs R: OR 1.25 [0.83e1.89]
Index (Total body lean F vs R: OR 3.41 [1.57e7.40]
mass/height2)
Overall OR for BC Overall OR 1.95 [0.73e5.19]
F, frailty; PF, prefrailty; R, robustness.

Please see Appendix 3 for alphabetical list of references. Description of study characteristics and ORs, and overall calculation. Instrument conceptual categories: 1 physical
focus; 2 multidomain focus; 3 decit accumulation. Follow-up category in months: a 0e12; b 12e24; c 24e60; d >60.
Table 7
Predictive Value of Frailty on LS
Life Satisfaction
St. John (2013) Age 65 n 1751 Brief Frailty Instrument (2) 12 (a) Terrible-delightful scale PF vs R: OR 1.94 [0.94e4.00]
(Andrews, 1974) F vs R: OR 3.88 [1.61e9.35]
Overall OR for LS Overall OR 2.62 [1.34e5.13]
F, frailty; PF, prefrailty; R, robustness.

Please see Appendix 3 for alphabetical list of references. Description of study characteristics and ORs, and overall calculation. Instrument conceptual categories: 1 physical
focus; 2 multidomain focus; 3 decit accumulation. Follow-up category in months: a 0e12; b 12e24; c 24e60; d >60.
instruments with a physical focus (OR 2.81 [2.20e3.58]), as well as compared with 4 different physical-oriented instruments applied in 5
instruments with a multidomain focus (OR 2.33 [1.68e3.23]) is studies. Given the different predictive values reported for the multi-
related to a higher likelihood for a deterioration in IADLs (see domain instruments (1 OR and 1 HR), it was impossible to compute a
Figure C5, Appendix 1). In line with results of the previous out- meta-analysis. Interestingly, frailty status as identied using in-
comes, having both a frail and a prefrail status is related to a struments with a physical focus is signicantly related to increased
signicantly elevated risk for IADL disabilities (respectively OR 3.06 risk for falls (OR 1.72 [1.16e2.54]; HR/RR 1.26 [1.12e1.41]), whereas
[2.13e4.39] and OR 2.30 [1.95e2.72]) (see Figure C4, Appendix 1). frailty assessed with multidomain instruments shows no statistically
However, both meta-analyses show moderate to high heterogeneity signicant relation to falls (OR 1.49 [0.69e3.22]; HR/RR 1.21
(I2 94%, P < .001 for frail and I2 57%, P .032 for prefrail status). [0.95e1.53]) (see Figure D2, Appendix 1). No studies assessing frailty
Three studies demonstrate an average follow-up of 1 year or by decit accumulation have been found. Within a short follow-up
less.24,36,49 Frail older persons show a signicantly increased risk for period (less than 1 year) frailty is already signicantly predictive for
IADL disability regardless of the time of follow-up (12 months OR the occurrence of falls (OR 2.27 [1.51e3.41]; HR/RR 1.56 [1.15e2.11]).
3.49 [2.62e4.65], 12e24 months no data found; 24e60 months OR Results within a longer period of follow-up are less consistent,
2.08 [1.58e2.75]; >60 months OR 1.86 [1.62e2.12]) (see Figure C6, showing a high heterogeneity and wide condence intervals, espe-
Appendix 1). cially for the ORs: 24 to 60 months (OR 1.31 [0.68e2.52]; HR/RR 1.19
[1.05e1.36]) and longer than 60 months (OR 1.10 [0.72e1.67]; HR/RR
Physical limitation. Three studies were included for this analysis, 1.15 [1.05e1.26]) (see Figure D3, Appendix 1) Table 4 provides an
describing results from 15,347 older adults. One article denes overview of the general description and the retrieved ratios per study
mobility problems as having difculties doing heavy housework or regarding falls..
walking half a mile and going up the stairs,54 another study considers Although the occurrence of falls is dened in different ways, it seems
mobility problems when participants give a positive answer to 2 convenient to include all studies into the meta-analysis so as to have a
mobility questions,44 and the third study denes physical limitation more comprehensive look on this common outcome. Results from the
as a long time to walk 15 feet and low maximal grip strength in the article of Rothman et al15 have been generated to an overall score,
dominant hand.34 Despite a high heterogeneity (I2 75%e80%), classied as a multidomain scale and used for the meta-analysis.
overall meta-analysis shows a signicantly increased risk for physical
limitations when being frail (OR 3.63 [2.14e6.16]) or prefrail (OR 1.81 Fractures. A total of 5 articles studied the relation between frailty
[1.41e2.33]) (see Figure C7, Appendix 1).38,44 This nding is supported and fractures in 52,286 older adults.20,36,37,46,47 Five different frailty
by the HR as reported by Fried et al34: HR Frail 1.42 (1.25e1.61); HR instruments are applied: 4 physical-oriented instruments and 1
prefrail 1.48 (1.33e1.67) (forest plots: see Figure C7 in Appendix 1). with a multidomain focus. The type of fracture varies among the
different studies: 1 group of authors investigates frailty in relation
Dependency. Only 1 study examined the relation between frailty to the risk for any fracture,46,47 whereas others focus on hip frac-
and dependency, dened by a series of open-ended questions tures or nonspine fractures.20,36,37
regarding received or needed help and care.31,55 As shown in Overall meta-analysis shows that frailty induces a signicantly
Table 3, for the physical-oriented (CHS) (HR 1.28 [1.10e1.48]), as greater risk for fractures (OR 2.78 [1.30e5.99]; HR/RR 1.37 [1.21e1.54])
well as for the multidomain instrument (HR 1.36 [1.18e1.57]), (see Figure D4 in Appendix 1 and Table 5). The study assessing frailty
frailty is associated with an increased risk for dependency (see with a multidomain-focused instrument does not demonstrate a sig-
Figure C8 in Appendix 1). nicant prediction for fractures (OR 1.76 [0.99e3.13]),47 contrary to the
studies using physical-oriented instruments (OR 3.35 [1.18e9.55]) (see
Falls and fractures Figure D5, Appendix 1). Having a frail status signicantly increases the
Falls. In total, 6 articles, describing 6 different frailty instruments and risk for fractures (OR 3.64 [1.53e8.67]; HR 1.59 [1.27e1.20]), a nding
17,690 older subjects, studied the relation between frailty and that is less consistent for having a prefrail status (OR 1.22 [0.54e2.76];
falls.15,34,36,37,47,49 Falls are usually dened as 2 or more falls during HR 1.18 [1.08e1.29]) (see Figure D4, Appendix 1). All follow-up periods
1 year of follow-up.36,49 One article denes a fall as 1 injurious fall over show a signicantly high risk for the occurrence of fractures when (pre)
a follow-up of 8 years.15 Overall, frailty signicantly increases the risk frailty is measured: less than 1 year of follow-up HR 1.69 (1.27e2.24); 2
of falling (OR 1.70 [1.18e2.44]; HR/RR 1.24 [1.12e1.37]). Having a frail to 5 years of follow-up OR 2.78 (1.29e5.99); more than 5 years of
status (OR 2.06 [1.28e3.34]; HR/RR 1.34 [1.14e1.58]), as well as having follow-up HR 1.23 (1.13e1.35) (see Figure D6, Appendix 1).
a prefrail status (OR 1.31 [0.89e1.93]; HR/RR 1.17 [1.05e1.30]) both Heterogeneity is low to moderate for the HRs (overall I2 52%,
increase the risk of falling (see Figure D1, Appendix 1). However, this P .022), but moderate to high for the ORs (overall I2 75%, P .007),
nding is not conrmed by all individual studies. According to Ensrud which might be due to a difference in dening fractures.
et al,37 the prefrail group assessed with a physical-focused tool, shows
no signicant risk of falling (OR 0.90 [0.72e1.13]), and Forti et al47 Cognitive decline
shows similar results for the frail group (OR 0.95 [0.28e3.22]). Two Two studies examined whether frailty increased the risk for
different multidomain instruments are described in 2 articles, cognitive decline.35,38,56 This outcome is dened as signs of dementia
and subjective cognitive complaints38 or as a diagnosis of Alzheimer were scarcely reported in literature despite their major impact on the
disease.35 Although cognition is not a physical function, no multido- older adults well-being.
main or decit accumulation instruments are applied for the deter- Physical-oriented instruments do not take cognitive function into
mination of frailty. Both studies do apply a physical-oriented frailty consideration, whereas multidomain or decit accumulation in-
scale consisting in an adapted version of the CHS tool, according to struments sometimes do cover this domain (eg, Groningen Frailty
Fried and colleagues.34 Overall meta-analysis shows that frailty Indicator, Rockwood Frailty Index). There are still debates ongoing
signicantly increases the likelihood of cognitive decline (HR 1.47 whether cognition should be part of the frailty denition, and study
[1.23e1.76]) (see Figure E1, Appendix 1). results and methods vary too much to reach a conceptual consensus or
its operationalization.60-62 Surprisingly, the studies investigating the
Body composition risk of cognitive decline related to frailty assessed frailty from a
Only 1 included study examined frailty in relation to changes in physical-oriented approach.20,37 This discrepancy suggests the need
body composition (BC) in 341 older adults. Bio-impedance is used for for further research on the implementation of a cognitive domain in
the assessment of BC at baseline and after 5 years of follow-up. frailty instruments.
Changes in BC are dened as a decline of more than 5% in the Lean Depending on the health outcome studied, different predictive
Mass Index ( total body lean mass/height2).56 Frailty is assessed values were found depending on the type of frailty instrument that
according to Frieds phenotype (CHS), a physical-oriented instrument. was used. Although ratios ranged from 1.32 to 3.64, depending on
Frailty is signicantly associated with an increased risk for decline in the scale used, all 3 methods (physical, multidomain, and decit
body lean mass. However, this nding is not statistically signicant for accumulation) signicantly predicted mortality. Despite the fact that
the prefrail group (see Table 6 and Figure F1, Appendix 1). More hospitalization and institutionalization are often considered as
research into the predictive value of frailty for a change in BC is similar outcomes, results in predicting those outcomes differed.
desirable. Prediction of the risk for hospitalization was signicant for all 3
methods ranging from 1.16 (1.06e1.27) (physical focus) to 1.84
Lower life satisfaction (1.35e2.51) (multidomain focus), whereas results for predicting the
St John et al57 examined the relation between frailty and changes risk of institutionalization were not always signicant (OR 2.53
in life satisfaction (LS), which was assessed using a 12-question scale [0.71e9.02] for physical-oriented instruments and OR 1.62
concerning various aspects of a persons life: the terrible-delightful [0.93e2.81] for instruments with a multidomain focus). On the
scale.58 A frail status, but not a prefrail status, signicantly increases contrary, decit accumulation methods showed a signicantly high
the likelihood of decline in LS (frail OR 3.88 [1.61e9.35]; prefrail OR risk for institutionalization when looking at the HR/RR (2.30
1.94 [0.94e4.00]) (see Table 7 and Figure G1, Appendix 1 [only 1 [1.54e3.43]). Prediction rates for falls, fractures, or BADL disabilities
study/variable/ratio observed so no meta-analysis]). also ranged (1.21e1.72; 1.37e3.35, and 1.59e2.11, respectively) with
overlapping intervals, emphasizing the predictive ability of all 3
General discussion types of instruments. For the prediction of the latter 3 outcomes, no
This systematic review and meta-analysis aimed to explore the decit accumulation methods were found for (meta-)analysis. For
prospective predictive value of the main frailty concepts for negative the development of IADL disabilities, only physical tools were
health outcomes in community-dwelling older adults. Because frailty explored in this analysis. It should be further researched whether
is described as a syndrome implying a decrease in reserve capacity and multidomain instruments or decit accumulation methods appear to
thus implying physiological decrements, the occurrence of negative be better predictors for this outcome.
health outcomes can be expected.59 The underlying mechanism of this The presence of frailty increased the likelihood for developing any
process is a complex interaction of i.a. undernutrition, weight loss, of the negative health outcomes studied. High overall ratios were
sarcopenia, a lower activity level, and difculty to maintain homeo- established when having a prefrail status, but with a frail status, ratios
stasis.34 This imbalance of multiorgan systems is reected in poor seemed to be somewhat higher. Data retrieved from Forti et al46
energy metabolism and neuromuscular changes,60 and may therefore showed no convincing evidence that a prefrail status can predict
lead to negative health outcomes. Results from our meta-analyses are hospitalization (OR 0.97 [0.98e5.63]) or premature mortality (OR 0.97
potentially of great importance for clinical practice, as they expand the [0.69e1.37]). Similarly, Avila-Funes et al38 failed to conrm the pre-
knowledge of the frailty process, enabling a more adequate estimation dictive value of prefrailty toward the development of BADL disabilities
of its overall negative health outcomes. As clinicians, one should be (OR 0.83 [0.46e1.50]). Both authors do acknowledge the fact that a
aware of the potential risks of frailty, as well as which instruments are prefrail status is less predictive for negative health outcomes
able to predict the outcome of interest. Furthermore, our systematic compared with a frail status. When looking at the studies of Jung
analysis of the follow-up period in the different studies may guide et al56 and St John et al,57 a frail status signicantly predicted the risk
clinicians regarding the urgency for initiating interventions in (pre) for a change in BC (OR 3.41 [1.57e7.40]) and a lower LS (OR 3.88
frail older adults. [1.61e9.35]), whereas a prefrail status did not conrm this signicance
An extensive literature search resulted in a comprehensive over- (respectively, OR 1.25 [0.83e1.89] and OR 1.94 [0.94e4.00]). According
view of 31 prospective studies. This is the rst meta-analysis that to Gill et al,63 frailty is a dynamic process among community-dwelling
quanties the risk related to frailty for the prospective development of older persons. During 54 months, they followed a cohort of 754 adults
negative health outcomes. The most frequently reported outcomes aged 70 years and older to study their current frailty status and to gain
were mortality, hospitalization, institutionalization, and disability in insight into the transitions between frailty states. More than 50% of
BADL and IADL. Frailty was prospectively linked to all outcome mea- the participants evolved toward another frailty state during follow-
sures, with a signicantly elevated risk for mortality (OR 2.34 up: most developed an increased frailty level, but some (23%)
[1.77e3.09], HR/RR 1.83 [1.68e1.98]), hospitalization (OR 1.82 evolved to a lesser frailty state.63 Because the presence of frailty in-
[1.53e2.15], HR/RR 1.18 [1.10e1.28]), institutionalization (OR 1.69 creases the risk for negative health outcomes, it might be possible that
[1.02e2.81], HR/RR 1.65 [1.48e1.84]), BADL disability (OR 2.05 these transitions between frailty states might inuence interpretation
[1.73e2.44], HR/RR 1.62 [1.50e1.76]), and IADL disability (OR 2.52 of the retrieved results. More research on the dynamic process and
[2.08e3.06]). These outcomes are major health concerns for older reversibility of frailty are essential for the development of a proactive
adults and their environment, emphasizing the importance of this approach, allowing prevention and early interventions to overcome or
review. Other health outcomes, such as decline in cognition and LS, delay the onset of frailty.
When looking at the frailty assessment tools reported, 3 older adults aged 65 years and older. Frailty literature also encom-
different types of instruments were identied: those with a physical passes many articles regarding the predictive power of frailty in-
focus, multidomain focus, and decit accumulation method. It struments in hospitalized patients, or institutionalized older
seems that the multidomain instruments are used more often than adults.11,67 Conceptually, it can be assumed that hospitalized and/or
the physical instruments (eg, for the outcome mortality 14 institutionalized older adults are possibly at an already more
different types of multidomain instruments, and 5 different types of advanced frailty stage. To this extent, results retrieved from those
physical instruments were used). However, while reading frailty populations may bias the predictive power of frailty in the occurrence
literature, physical-oriented approaches seem to be used very of these outcomes at a moment when they are not yet present.
frequently. This contrast may be because the physical group consists Therefore, results of this study are not to be translated to the whole
of only 6 distinct instruments (mainly Frieds approach and adapted elderly population. Analog research into the prospective predictive
versions of this model4), causing a seemingly imbalance in the value of the main frailty concepts for negative health outcomes in
classication of frailty models. hospitalized and institutionalized older adults is desirable.
It can be assumed that time to follow-up is an important factor that Results retrieved from the studies were maximally adjusted for
determines the risk for an outcome to occur. Follow-up differed be- confounders, such as age, sex, stroke, education level, sociodemo-
tween studies, which may inuence the meta-analyses. Therefore, 4 graphic covariates, and baseline health status. However, when meta-
subgroups were formed according to the length of follow-up (0e12, analyses were performed, heterogeneity was sometimes high. This
12e24, 24e60, >60). Analyses showed that, according to the negative may be explained by differences in study population, confounders
health outcome, different results were obtained per follow-up cate- taken into account, sample size, and follow-up period of the individual
gory. For a follow-up of 1 year or less, ratios ranged between 1.22 studies. Also, articles focusing on physical-oriented instruments often
(institutionalization) and 3.49 (IADL disabilities), a period of 24 to presented data for the prefrail, whereas usually multidomain in-
60 months gives a range between 1.11 (hospitalization) and 4.13 struments or decit accumulation methods only make the distinction
(institutionalization), and for a follow-up of >60 months this becomes between frail and robust older persons. Hastings et al50 described 3
1.10 (falls) to 2.14 (mortality). Only 2 researchers considered a follow- levels of frailty, which were dened in this article as prefrail, midfrail,
up period of 12 to 24 months, making it difcult to dene a range for and frail.
this period.24,44,50 To gain more insight into the importance of the When only frequencies were provided in the articles, the OR was
follow-up period for a certain negative health outcome, not only the calculated based on the frequencies of the group of robust subjects
range of the ratio, but also the 95% condence interval (CI) should be and the group of (pre)frail subjects. In case insufcient information
considered. These intervals are very narrow for some outcomes, but was available in the article for calculating the ratio, the corresponding
very wide for others. Also within certain outcomes, there is a large author was contacted to obtain the frequencies necessary to calculate
discrepancy between the CI (eg, for the development of IADL dis- the ratio. In case of multiple ratios per frailty instrument (separate
abilities, the ratio with the smallest CI: OR 1.39 [1.28e1.31]; the ratio analysis per item), an overall predictive measure was calculated based
with the largest CI: OR 5.38 [2.34e12.37]). Comparing ratios should on the total score of the instrument (as a whole).13,14 The dissimilarity
thus always be done carefully, because CI may overlap, making it in ratios, confounders, and frailty assessment tools may somewhat
difcult to speak about a genuine difference. complicate the interpretation of the results. However, overall meta-
The OR and HR/RR values calculated and analyzed in this review analysis for all negative health outcomes shows convincing evidence
should not be used interchangeably. An OR describes the probability that (pre-)frailty signicantly predicts their occurrence.
that a particular event will occur in relation to the probability that it Although primary prevention is currently a hot topic among cli-
will not occur, whereas RR and HR represent the ratio of the risk of an nicians and researchers, frailty is often looked at as a starting point.
event in 2 different groups (ie, the (pre)frail versus the robust). Very little research has been done in looking at frailty as an endpoint.
This review shows convincing evidence for the predictive power In the literature search performed, only a few articles looked at frailty
of frailty toward several negative health outcomes. The results of our as an outcome.68,69 All other studies compared the nonfrail elderly
meta-analyses can be used as a guideline for the prediction of with the prefrail or frail elderly in developing several negative health
negative outcomes according to the frailty concept used as well as to outcomes that may be linked to frailty. A growing attention for the
estimate the time frame within which these events can be expected robust elderly and more research into this eld is essential so as to
to occur. better understand the onset of frailty. Risk factors and predictors for
frailty are important research topics to fully understand the syn-
Strengths and Limitations drome70 and to proactively address the development of frailty. A
longitudinal prospective cohort study exploring the potential pre-
Given the large scope of this review (31 articles studying 13 dictors for the development of frailty would be useful in this regard.
important negative health outcomes and 158,764 subjects in total),
the generalizability of the results is expected to be high. This is the rst
comprehensive meta-analysis providing evidence stratied by frailty
status, frailty instrument, and duration of follow-up. Although various Conclusion
operationalizations for the discussed negative health outcomes were
used, studies were grouped to improve the readability of this review. The results of this extensive review and meta-analysis, based on 31
Because the age limit of the participants for the inclusion of articles articles studying 13 negative health outcomes and 158,764 subjects in
was set on 65 years or older, some studies regarding frailty and its total, show that overall, frailty signicantly increases the risk for
negative health outcomes in younger adults were not included. developing several negative health outcomes. Taken together, frailty is
However, most articles with a lower age cutoff show similar associated with a 1.8- to 2.3-fold risk for mortality, a 1.6- to 2.0-fold
ndings.64e66 Malmstrom et al,66 for example, explored how 4 risk for loss of ADLs, a 1.2- to 1.8-fold risk for hospitalization, a 1.5-
different frailty models predicted future disability and mortality in a to 2.6-fold risk for physical limitation, and a 1.2- to 2.8-fold risk for
population-based cohort of people aged 49 to 65 at baseline and falls and fractures. The analyses presented in this study can be used as
showed that, overall, the Frailty Index and FRAIL scale contain the a guideline for the prediction of negative outcomes according to the
strongest predictive power for those outcomes over 3 and 9 years of frailty concept used as well as to estimate the time frame within which
follow-up. This review, however, focuses on community-dwelling these events can be expected to occur.
Supplementary Data 28. Hogan DB, Freiheit EA, Strain LA, et al. Comparing frailty measures in their
ability to predict adverse outcome among older residents of assisted living.
BMC Geriatr 2012;12:56.
Supplementary data related to this article can be found online at 29. Rockwood K, Andrew M, Mitnitski A. A comparison of two approaches to
http://dx.doi.org/10.1016/j.jamda.2016.09.010. measuring frailty in elderly people. J Gerontol A Biol Sci Med Sci 2007;62:
738e743.
30. Bouillon K, Kivimaki M, Hamer M, et al. Measures of frailty in population-based
References studies: An overview. BMC Geriatr 2013;13:64.
31. At J, Bryce R, Prina M, et al. Frailty and the prediction of dependence and
1. Campbell AJ, Buchner DM. Unstable disability and the uctuations of frailty. mortality in low- and middle-income countries: a 10/66 population-based
Age Ageing 1997;26:315e318. cohort study. BMC Med 2015;13:138.
2. Kojima G. Prevalence of frailty in nursing homes: A systematic review and 32. Lee JH, Fulp W, Wells KJ, et al. Effect of patient navigation on time to diagnostic
meta-analysis. J Am Med Dir Assoc 2015;16:940e945. resolution among patients with colorectal cancer-related abnormalities.
3. Collard RM, Boter H, Schoevers RA, Voshaar RCO. Prevalence of frailty in J Cancer Educ 2014;29:144e150.
community-dwelling older persons: A systematic review. J Am Geriatr Soc 33. Garcia-Garcia FJ, Carcaillon L, Fernandez-Tresguerres J, et al. A new operational
2012;60:1487e1492. denition of frailty: The Frailty Trait Scale. J Am Med Dir Assoc 2014;15:371.
4. Fried LP, Tangen CM, Walston J, et al. Frailty in older adults: Evidence for a e7e371.e13.
phenotype. J Gerontol A Biol Sci Med Sci 2001;56:M146eM156. 34. Fried LP, Tangen CM, Walston J, et al. Frailty in older adults: Evidence for a
5. OCaoimh R, Gao Y, Svendrovski A, et al. Screening for markers of frailty and phenotype. J Gerontol A Biol Sci Med Sci 2001;56:M146eM156.
perceived risk of adverse outcomes using the Risk Instrument for Screening in 35. Buchman AS, Leurgans SE, Boyle PA, et al. Combinations of motor measures
the Community (RISC). BMC Geriatr 2014;14:104. more strongly predict adverse health outcomes in old age: The rush memory
6. Lahousse L, Maes B, Ziere G, et al. Adverse outcomes of frailty in the elderly: and aging project, a community-based cohort study. BMC Med 2011;9:42.
The Rotterdam Study. Eur J Epidemiol 2014;29:419e427. 36. Ensrud KE, Ewing SK, Cawthon PM, et al. A comparison of frailty indexes for the
7. de Vries OJ, Peeters GM, Lips P, Deeg DJ. Does frailty predict increased risk of prediction of falls, disability, fractures, and mortality in older men. J Am Geriatr
falls and fractures? A prospective population-based study. Osteoporos Int Soc 2009;57:492e498.
2013;24:2397e2403. 37. Ensrud KE, Ewing SK, Taylor BC, et al. Frailty and risk of falls, fracture, and
8. Vermeulen J, Neyens JCL, van Rossum E, et al. Predicting ADL disability in mortality in older women: The study of osteoporotic fractures. J Gerontol A Biol
community-dwelling elderly people using physical frailty indicators: A sys- Sci Med Sci 2007;62:744e751.
tematic review. BMC Geriatr 2011;11:33. 38. Avila-Funes JA, Amieva H, Barberger-Gateau P, et al. Cognitive impairment
9. Yamada M, Arai H. Predictive value of frailty scores for healthy life expectancy improves the predictive validity of the phenotype of frailty for adverse health
in community-dwelling older Japanese adults. J Am Med Dir Assoc 2015;16: outcomes: The three-city study. J Am Geriatr Soc 2009;57:453e461.
1002.e7e1002.e11. 39. Graham JE, Al Snih S, Berges IM, et al. Frailty and 10-Year mortality in
10. Chang SF, Lin PL. Frail phenotype and mortality prediction: A systematic review community-living Mexican American older adults. Gerontology 2009;55:
and meta-analysis of prospective cohort studies. Int J Nurs Stud 2015;52: 644e651.
1362e1374. 40. Lucicesare A, Hubbard RE, Fallah N, et al. Comparison of two frailty measures in
11. Luo H, Lum TY, Wong GH, et al. Predicting adverse health outcomes in nursing the Conselice Study of Brain Ageing. J Nutr Health Aging 2010;14:278e281.
homes: A 9-year longitudinal study and development of the FRAIL-Minimum 41. Sirola J, Pitkala KH, Tilvis RS, et al. Denition of frailty in older men according to
Data Set (MDS) Quick Screening Tool. J Am Med Dir Assoc; 2015:1042e1047. questionnaire data (RAND-36/SF-36): The Helsinki Businessmen Study. J Nutr
12. Almeida OP, Hankey GJ, Yeap BB, et al. Depression, frailty, and all-cause mor- Health Aging 2011;15:783e787.
tality: A cohort study of men older than 75 years. J Am Med Dir Assoc 2015;16: 42. Armstrong JJ, Mitnitski A, Launer LJ, et al. Frailty in the Honolulu-Asia Aging
296e300. Study: Decit accumulation in a male cohort followed to 90% mortality.
13. Liberati A, Altman DG, Tetzlaff J, et al. The PRISMA statement for reporting J Gerontol A Biol Sci Med Sci 2015;70:125e131.
systematic reviews and meta-analyses of studies that evaluate healthcare in- 43. Kulminski AM, Ukraintseva SV, Kulminskaya IV, et al. Cumulative decits better
terventions: Explanation and elaboration. BMJ 2009;339:b2700. characterize susceptibility to death in elderly people than phenotypic frailty:
14. Carriere I, Colvez A, Favier F, et al. Hierarchical components of physical frailty Lessons from the Cardiovascular Health Study. J Am Geriatr Soc 2008;56:
predicted incidence of dependency in a cohort of elderly women. J Clin Epi- 898e903.
demiol 2005;58:1180e1187. 44. Woo J, Leung J, Morley JE. Comparison of frailty indicators based on clinical
15. Rothman MD, Leo-Summers L, Gill TM. Prognostic signicance of potential phenotype and the multiple decit approach in predicting mortality and
frailty criteria. J Am Geriatr Soc 2008;56:2211e2216. physical limitation. J Am Geriatr Soc 2012;60:1478e1486.
16. The guidelines manual: Appendices B-I. 2012. Available at: https://www.nice. 45. Romero-Ortuno R, Soraghan C. A frailty instrument for primary care for those
org.uk/process/pmg6/resources/the-guidelines-manual-appendices-bi-1967364/ aged 75 years or more: Findings from the Survey of Health, Ageing and
chapter/1-appendix-b-methodology-checklist-systematic-reviews-and-meta- Retirement in Europe, a longitudinal population-based cohort study (SHARE-
analyses Accessed October 1, 2015. FI75). BMJ Open 2014;4:e006645.
17. Deeks JJ, Higgins JP, Altman DG. Analysing data and undertaking meta-ana- 46. Forti P, Maioli F, Lega MV, et al. Combination of the clock drawing test with the
lyses. In: Higgins JP, Green S, editors. Cochrane Handbook for Systematic Re- physical phenotype of frailty for the prediction of mortality and other adverse
views of Interventions: Cochrane Book Series. Chichester, UK: John Wiley & outcomes in older community dwellers without dementia. Gerontology 2014;
Sons, Ltd; 2008. Available at: http://handbook.cochrane.org/chapter_9/9_ 60:204e211.
analysing_data_and_undertaking_meta_analyses.htm. Accessed April 5, 2016. 47. Forti P, Rietti E, Pisacane N, et al. A comparison of frailty indexes for prediction
18. Wallace BC, Schmid CH, Lau J, Trikalinos TA. Meta-Analyst: Software for meta- of adverse health outcomes in an elderly cohort. Arch Gerontol Geriatr 2012;
analysis of binary, continuous and diagnostic data. BMC Med Res Methodol 54:16e20.
2009;9:80. 48. Ma SL, Oyler J, Glavin S, et al. Self-reported frailty is associated with low
19. Higgins JP, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in calcaneal bone mineral density in a multiracial population of community-
meta-analyses. BMJ 2003;327:557e560. dwelling elderly. Osteoporos Int 2009;20:1837e1846.
20. Woods NF, LaCroix AZ, Gray SL, et al. Frailty: Emergence and consequences in 49. Kiely DK, Cupples LA, Lipsitz LA. Validation and comparison of two frailty in-
women aged 65 and older in the Womens Health Initiative Observational dexes: The MOBILIZE Boston Study. J Am Geriatr Soc 2009;57:1532e1539.
Study. J Am Geriatr Soc 2005;53:1321e1330. 50. Hastings SN, Purser JL, Johnson KS, et al. Frailty predicts some but not all
21. Buchman AS, Leurgans SE, Boyle PA, et al. Combinations of motor measures adverse outcomes in older adults discharged from the emergency department.
more strongly predict adverse health outcomes in old age: The rush memory J Am Geriatr Soc 2008;56:1651e1657.
and aging project, a community-based cohort study. BMC Med 2011;9:42. 51. Gobbens RJ, van Assen MA, Schalk MJ. The prediction of disability by self-
22. Chang SF, Yang RS, Lin TC, et al. The discrimination of using the Short Physical reported physical frailty components of the Tilburg Frailty Indicator (TFI).
Performance Battery to screen frailty for community-dwelling elderly people. Arch Gerontol Geriatr 2014;59:280e287.
J Nurs Scholarsh 2014;46:207e215. 52. Katz S, Ford AB, Moskowitz RW, et al. Studies of illness in the aged. The Index of
23. Ravaglia G, Forti P, Lucicesare A, et al. Development of an easy prognostic score ADL: A standardized measure of biological and psychosocial function. JAMA
for frailty outcomes in the aged. Age Ageing 2008;37:161e166. 1963;185:914e919.
24. Daniels R, van Rossum E, Beurskens A, et al. The predictive validity of three 53. Lawton MP, Brody EM. Assessment of older people: Self-maintaining and
self-report screening instruments for identifying frail older people in the instrumental activities of daily living. Gerontologist 1969;9:179e186.
community. BMC Public Health 2012;12:69. 54. Rosow I, Breslau NA. Guttman health scale for the aged. J Gerontol 1966;21:
25. Gobbens RJ, van Assen MA, Luijkx KG, Schols JM. The predictive validity of the 556e559.
Tilburg Frailty Indicator: Disability, health care utilization, and quality of life in 55. Woo J, Goggins W, Sham A, Ho SC. Public health signicance of the frailty in-
a population at risk. Gerontologist 2012;52:619e631. dex. Disabil Rehabil 2006;28:515e521.
26. Rockwood K, Song X, MacKnight C, et al. A global clinical measure of tness and 56. Jung HW, Kim SW, Lim JY, et al. Frailty status can predict further lean body
frailty in elderly people. CMAJ 2005;173:489e495. mass decline in older adults. J Am Geriatr Soc 2014;62:2110e2117.
27. Song X, Mitnitski A, Rockwood K. Prevalence and 10-year outcomes of frailty in 57. St John PD, Tyas SL, Montgomery PR. Life satisfaction and frailty in community-
older adults in relation to decit accumulation. J Am Geriatr Soc 2010;58: based older adults: Cross-sectional and prospective analyses. Int Psychogeriatr
681e687. 2013;25:1709e1716.
58. Andrews FM. Social indicators of perceived life quality. Soc Indic Res 1974;1: 65. Wu YH, Liu LK, Chen WT, et al. Cognitive function in individuals with
279e299. physical frailty but without dementia or cognitive complaints: Results from
59. Fried LP, Ferrucci L, Darer J, et al. Untangling the concepts of disability, frailty, the I-Lan Longitudinal Aging Study. J Am Med Dir Assoc 2015;16:899.
and comorbidity: Implications for improved targeting and care. J Gerontol A e9e899.e16.
Biol Sci Med Sci 2004;59:255e263. 66. Malmstrom TK, Miller DK, Morley JE. A comparison of four frailty models. J Am
60. Fedarko NS. The biology of aging and frailty. Clin Geriatr Med 2011;27:27e37. Geriatr Soc 2014;62:721e726.
61. Rodriguez-Manas L, Feart C, Mann G, et al. Searching for an operational de- 67. Kaehr EW, Pape LC, Malmstrom TK, Morley JE. FRAIL-NH Predicts Outcomes in
nition of frailty: A Delphi method based consensus statement: The frailty Long Term Care. J Nutr Health Aging 2016;20:192e198.
operative denition-consensus conference project. J Gerontol A Biol Sci Med 68. Yeap BB, Paul Chubb SA, Lopez D, et al. Associations of insulin-like growth
Sci 2013;68:62e67. factor-I and its binding proteins and testosterone with frailty in older men. Clin
62. Gobbens RJ, Luijkx KG, Wijnen-Sponselee MT, Schols JM. Towards an integral Endocrinol 2013;78:752e759.
conceptual model of frailty. J Nutr Health Aging 2010;14:175e181. 69. Gruenewald TL, Seeman TE, Karlamangla AS, Sarkisian CA. Allostatic load and
63. Gill TM, Gahbauer EA, Allore HG, Han L. Transitions between frailty states frailty in older adults. J Am Geriatr Soc 2009;57:1525e1531.
among community-living older persons. Arch Intern Med 2006;166:418e423. 70. Walston J, Hadley EC, Ferrucci L, et al. Research agenda for frailty in older adults:
64. Morley JE, Malmstrom TK, Miller DK. A simple frailty questionnaire (FRAIL) Toward a better understanding of physiology and etiology: Summary from the
predicts outcomes in middle aged African Americans. J Nutr Health Aging American Geriatrics Society/National institute on aging research conference on
2012;16:601e608. frailty in older adults. J Am Geriatr Soc 2006;54:991e1001.

Frailty and The Prediction of Negative Health Outcomes

Diunggah oleh

Informasi Dokumen

Judul Asli

Hak Cipta

Format Tersedia

Bagikan dokumen Ini

Bagikan atau Tanam Dokumen

Opsi Berbagi

Apakah menurut Anda dokumen ini bermanfaat?

Apakah konten ini tidak pantas?

Hak Cipta:

Format Tersedia

Frailty and The Prediction of Negative Health Outcomes

Diunggah oleh

Hak Cipta:

Format Tersedia

JAMDA 17 (2016) 1163.e1e1163.

Frailty and the Prediction of Negative Health Outcomes:

Fig. 1. Literature owchart.

Fig. 2. Number of frailty instruments per type and per outcome.

S. Vermeiren et al. / JAMDA 17 (2016) 1163.e1e1163.e17

S. Vermeiren et al. / JAMDA 17 (2016) 1163.e1e1163.e17

mCHS (Fried) (1) yes/no F vs R: OR 3.49 [1.53e7.98]

S. Vermeiren et al. / JAMDA 17 (2016) 1163.e1e1163.e17

[1.58e2.46]; HR/RR 1.23 [1.07e1.40]), as well as having a prefrail

Overall HR/RR 1.03 [0.82e1.29]

Overall OR 2.16 [1.39e3.37]

Figure B2, Appendix 1). Frailty predicts hospitalization within

Institutionalization. Five studies, encompassing 11 different

frailty instruments and 5926 older adults, discussed the relation

in Table 2, the overall analysis shows a signicantly increased risk

ORs were available regarding the prefrail status. When looking at

is high when instruments with decit accumulation are used

(HR/RR 2.30 [1.54e3.43]). This effect remains for the HR/RR

when considering the OR for physical-oriented tools (OR 2.53

analysis]) or instruments with a multidomain focus (OR 1.62

[0.93e2.81]) separately (see Figure B5 in Appendix 1). Several

follow-up periods have been observed across the different

24 months. Nevertheless, an elevated risk for institutionalization

has been established within a follow-up of 0 to 12 months (OR

as a follow-up of 60 months or longer (OR no data, HR/RR 1.73

[1.26e2.37] [only 1 study/variable/ratio observed so no meta-

ED visits. Two articles including in total 2616 participants

oriented instruments and decit accumulation methods are

applied in these studies. ED visits are dened as a repeated

Participants from the CARTS study.

Participants from the CSHA study.

difcult. However, the frequency of ED visits as well as the ne-

physical frailty assessment scale,49 and the other one a decit

Overall OR for ED visits

accumulation method.50 Because only 2 different studies were

found providing different ratios (OR vs HR), no meta-analysis

The Decit Accumulation Index used by Hastings et al50 divides

gorization attributes a different meaning to the previously used

S. Vermeiren et al. / JAMDA 17 (2016) 1163.e1e1163.e17

performing household activities F vs R: OR 2.86 [1.61e5.08]y

S. Vermeiren et al. / JAMDA 17 (2016) 1163.e1e1163.e17

mCHS (Fried) (1) F vs R: HR 2.19 [1.19e4.03]

S. Vermeiren et al. / JAMDA 17 (2016) 1163.e1e1163.e17

F, frailty; (m)CHS, (modied) Cardiovascular Health Study; PF, prefrailty; R, robustness.

F, frailty; PF, prefrailty; R, robustness.

F, frailty; PF, prefrailty; R, robustness.

Anda mungkin juga menyukai