1
Frailty as a Predictor of Disabilities among Community-Dwelling Older People: A
Systematic Review and Meta-analysis
Gotaro Kojima, MD
1
1
Japan Green Medical Centre, London, United Kingdom.
Keywords: frailty, disability, activities of daily living, instrumental activities of daily living,
community-dwelling older people.
ABSTRACT
Background: Frailty has been shown to be associated with disability in the previous studies.
However, it is not clear how consistently or to how much degree frailty is actually associated
with the future disability risks.
Methods: A systematic review of the literature was conducted using Embase, MEDLINE,
CINAHL, PsycINFO and the Cochrane Library for any prospective studies published from
2010 to September 2015 examining associations between baseline frailty status and
subsequent risk of developing or worsening disabilities among community-dwelling older
people. A meta-analysis was performed to synthesize pooled estimates.
Results: Of 7,012 studies identified through the systematic review, 20 studies were included
in the meta-analysis. 12 studies examined ADL disability risks, two studies examined IADL
disability risks, and six studies examined both ADL and IADL disability risks. Overall, frail
older people were more likely to develop or worsen disabilities in ADL (12 studies, pooled
OR=2.76, 95%CI=2.23-3.44, p<0.00001; 5 studies, pooled HR=2.23, 95%CI=1.42-3.49,
p<0.00001) and IADL (6 studies, pooled OR=3.62, 95%CI=2.32-5.64, p<0.00001; 2 studies,
pooled HR=4.24, 95%CI=0.85-21.28, p=0.08). Prefrailty was also associated with incident or
worsening disability risks to a lesser degree in most pooled analyses. High heterogeneity
observed among 12 studies with OR of ADL disability risks for frailty was explored using
subgroup analyses, which suggested methodological quality and mean age of the cohort were
the possible causes.
Conclusion: This systematic review meta-analysis quantitatively showed that frail older
people are at higher risks of disabilities. These results are important for all related parties
given population aging worldwide. Interventions for frailty are important to prevent disability
and preserve physical functions, autonomy, and quality of life.
2
INTRODUCTION
Frailty and disability were once used interchangeably due to the similarity, the high co-
existence rate, and the lack of standardized definitions to operationalize frailty against
disability.
1-3
Frailty has now been conceptualized as a distinct state of decreased
physiological reserve and compromised capacity to maintain homeostasis when exposed to a
stressor resulting from age-related multiple accumulated deficits, thereby predisposing frail
individuals to high vulnerability to adverse health outcomes.
1-3
The adverse health outcomes
include falls, fractures, hospitalization, institutionalization, dementia, and mortality.
3-8
Prevalence of frailty is 10.7% among community-dwelling older people aged 65 and older
and generally increases as people age
9
and more than 90% of institutionalized people are
frail.
10
Frailty is a dynamic state and can transition to worse as well as better status over
time.
11
Given that frailty can be potentially prevented or reversed with interventions, such as
exercise or nutritional supplementation,
3
it was recommended by a consensus group of
experts from international societies to screen older people aged 70 years or older with
significant weight loss due to chronic disease.
12
Although a number of definitions and criteria for frailty have been proposed, international
consensus has yet to be reached partially because of the multidimensional and heterogeneous
nature of the concept.
3
Among a number of frailty definitions proposed, the ones described by
Fried et al. in the Cardiovascular Health Study (CHS) have been most frequently used in the
literature.
2
They defined frailty as a clinical syndrome using a combination of five physical
components: weight loss, exhaustion, weakness, slow walking speed, and low physical
activity.
2
In their original study, weakness was defined as having grip strength of less than 29-
32 Kg for men and 17-21 Kg for women depending on BMI quartiles, slow walking speed
was defined as taking more than 6-7 seconds to walk 15 feet stratified by gender and height,
and low physical activity was defined as having less than 383 Kcals/week for men and 270
Kcals/week for women for physical activity based on the short version of the Minnesota
Leisure Time Activity questionnaire.
2
The Frailty Index is another popular conceptualization
of frailty.
13
While CHS criteria involve mainly physical components, this method defines
frailty according to accumulation of much broader health deficits including cognitive,
psychological, and social factors in addition to physical aspects.
14
Multiple studies have examined associations between frailty and disability and mostly found
frail individuals were significantly more likely to develop or worsen disabilities than the non-
frail,
2,3
but a few did not.
15-17
Some researchers even state that frailty may be a physiologic
precursor and etiologic factor in disability.
2
It may feel obvious and not surprising that frailty
is closely related to disability and predicts disability. However, since no systematic review or
meta-analysis on the association was identified in the literature, it is not clear how
consistently or to how much degree frailty is actually associated with the future disability
risks.
One review paper has examined activities of daily living (ADL) disability risks predicted by
not frailty but frailty components, such as weight loss or gait speed, and showed these frailty
components individually predicted ADL disability.
18
As described earlier, frailty is a
multidimensional complex state, and its features of predicting disability risks cannot be
completely evaluated by examining only an individual component of frailty. Furthermore,
this review did not include instrumental activities of daily living (IADL) and did not conduct
a meta-analysis to synthesize pooled risk estimates.
18
Therefore, it was considered that
quantifying the disability risks according to frailty by pooling the findings of the previous
studies is new and confirms what was already known. The objectives of the current study
3
were to perform a systematic review of the literature and to conduct a meta-analysis to
synthesize pooled estimates of future disability risks predicted by baseline frailty status
among community-dwelling older people.
METHOD
Data Sources and Search Strategy
A systematic review of the literature was conducted by one researcher based on a protocol
developed in accordance with Preferred Reporting Items for Systematic Review and Meta-
Analyses (PRISMA)
19
and Meta-analysis of Observational Studies in Epidemiology
(MOOSE)
20
statements using five electronic databases (Embase, MEDLINE, CINAHL Plus,
PsycINFO, and the Cochrane Library) in September 2015. Any prospective studies in any
language published in 2000 or later on associations between baseline frailty status and a
subsequent risk of developing or worsening disabilities among community-dwelling older
people were potentially eligible. Explosion functions were used if available. The Medical
Subject Heading (MeSH) and keywords used were as follows: ((Activities of daily living
(MeSH)) OR (Daily life activity (MeSH)) OR (Disability (MeSH)) OR (Disabilities (MeSH))
OR (ADL disability (MeSH)) OR (Physical disability (MeSH)) OR (Disabled (MeSH)) OR
(Disabled person(s) (MeSH)) OR (Disabled personnel (MeSH)) OR (disable*) OR
(disabilit*) OR (“activities of daily living”) OR (ADL) OR (IADL)) AND ((Frailty syndrome
(MeSH)) OR (Frail elderly (MeSH)) OR (frailty)). The systematic search was repeated for
update in June 2016 using the same strategy for the newer citations published from 2015 to
June 2016. The reference lists of the retrieved articles were manually reviewed for any
additional studies.
Study Selection
Studies were included if they met following inclusion criteria;
Involved community-dwelling older people with mean age of 60 and older.
Prospectively examined a risk of developing new disability or of worsening disability
according to baseline frailty status defined by validated criteria originally designed to
define frailty or its modified versions.
Defined disability measured by activities of daily living (ADL) or instrumental
activities of daily living (IADL).
Provided odds ratio (OR) or hazard ratio (HR) as a risk measure for incidence or
worsening of disability.
Studies were excluded with the following exclusion criteria;
Defined frailty by slow walking speed or being certified for long-term care insurance,
or used individual components of frailty criteria.
Defined frailty status as a continuous measure, rather than categorizing as frail or non-
frail.
Review papers, randomized controlled trials, conference abstracts, comments, or
editorials.
When the same cohort was used for the same disability outcome, the study defining three
categories: frail, prefrail, and robust, instead of two: frail and robust, or the study using the
largest number of the individuals was included. When different frailty definitions were used
in one study, the results based on CHS criteria, which is the most frequently used in the
literature,
3
or the results from the largest sample size were included. When different follow-
up periods were used, the results of the longest follow-up period were used. These criteria
were defined a priori.
4
Studies potentially eligible for meta-analysis selected through systematic review of title,
abstract, and full-text were examined for methodological quality using the Newcastle-Ottawa
scale for cohort studies.
21
A study was considered to have adequate quality to be included in
the meta-analysis if they met half or more of the criteria.
Data Extraction
Data extracted were first author, cohort name if any, publication year, location (country),
sample size of a cohort used for an analysis of interest or the entire cohort, proportion of
female participants, age (mean or age criterion for inclusion), frailty criteria, disability
outcome, effect measure, and follow-up period. OR and HR with corresponding 95%
confidence intervals (95%CI) of disability risk for frailty and prefrailty compared with non-
frailty were extracted from each study. Adjusted risk measures were preferred to unadjusted
ones.
Statistical Analysis
When two or more studies presented the same disability risk (incident ADL, worsening ADL,
combined ADL, incident IADL, worsening IADL, or combined IADL) using the same type of
risk estimates (OR or HR), the risk measures were combined to synthesize pooled estimates
using the generic inverse variance method. Random-effects models were used since
significant heterogeneity was expected given different methodology and various frailty and
disability definitions used across the included studies.
Studies were pooled according to types of disability (ADL vs. IADL), type of risk (incident
vs. worsening), effect measure (OR vs. HR), and frailty status (frail vs. prefrail) separately.
Effects measures of incident and worsening disability were also pooled for the same type of
disability and effect measure, for which estimates of worsening disability were selected over
incidence disability when a study provided both estimates. Heterogeneity across the studies
was assessed using Cochran’s Q statistic and the degree of the heterogeneity was assessed
using the I
2
statistic. I
2
values of 25%, 50%, and 75% were considered as low, moderate, and
high heterogeneity, respectively.
22
When high heterogeneity was detected, subgroup analysis,
sensitivity analysis, and random effects meta-regression were conducted according to
location, sample size, female proportion of the cohort, mean age, frailty criteria, adjustment
for an effect measure, follow-up period, and methodological quality to explore the potential
source of heterogeneity. Publication bias was assessed using Begg-Mazumdar’s and Egger’s
tests and visually examining funnel plots.
All analysis were performed using Review Manager 5 (version 5.2, The Cochrane
Collaboration, Copenhagen, Denmark), IBM SPSS Statistics (version 22, IBM Corporation,
New York, USA), and StatsDirect (version 2.8, StatsDirect, Cheshire, UK).
RESULTS
Selection Processes
The initial literature search using the five databases yielded 7,012 studies and 3 additional
studies were identified from reference lists of relevant articles. Of a total of 7,015 studies,
2,892 duplicate studies were excluded and 4,085 studies were excluded because the titles or
abstracts were considered as not relevant, leaving 38 studies for full-text review. Of these, 18
studies were excluded because nine studies did not provide OR or HR of disability risk for
frailty status, three studies did not used ADL or IADL to measure disability, two studies had
cohorts with mean age of less than 60 years, two studies used the same cohorts, and two
5
studies were review or cross-sectional studies. Twenty studies were left and further assessed
for methodological quality using the Newcastle-Ottawa scale for cohort studies. No
additional studies were identified by the updated systematic review from 2015 to June 2016.
All the 20 studies were considered to have adequate quality (Table 1) and were included in
this review and meta-analysis. Figure 1 shows a flowchart of the study selection with
numbers of studies at each stage.
Study Characteristics
Study characteristics of the included 20 studies are summarized in Table 1. More than half of
the included studies were published within the last four years (2012-2015).
16,17,23-31
Nine
studies were from the USA,
2,29,32-38
five from Europe,
15-17,23,30
two from Mexico,
24,25
one each
from Australia
27
and Korea,
26
and two studies used cohorts from multiple countries.
28,31
Sample size varies from 226
16
to 40,332.
31
Six studies used female-only cohorts
27,31,33,34,36,38
and one study used an all-male cohort.
35
Although not provided by all studies, mean age
ranged from 68.7
24
to 79.4
23
years old. Most of the included studies (80%, 16/20) used
modified or original CHS criteria to define frailty.
2,15,23,24,26,28-38
Prevalence of frailty varied
substantially ranging from 4.3%
32
to 37.2%.
24
Disability outcomes were either incident or
worsening ADL or IADL disability. As many as 90% (18/20) of the studies examined ADL
disability risks while IADL was examined by eight studies; two studies used IADL alone
35,36
and six studies used both ADL and IADL.
15,23,24,27,28,33
Twelve studies reported adjusted
OR
15,17,24,25,27-29,31,35-38
and two studies reported only unadjusted OR.
16,26
All HR reported by
six studies were adjusted for covariates.
2,23,30,32-34
Follow-up periods were from 1 year
16,31
to
11 years.
24
Frailty as a Predictor of Incident or Worsening ADL Disability
The meta-analysis of the associations between frailty and ADL disability included 12 studies
with OR
15-17,24-29,31,37,38
and five studies with HR.
2,23,30,32,34
Both frailty and prefrailty were
significant predictors of ADL disability (incident, worsening, and combined) persistently in
all meta-analysis. Eight
15,17,24,25,28,31,37,38
and six
16,17,26-29
studies provided effect sizes as OR of
incident and worsening ADL disability, respectively, for frailty compared with non-frailty (8
studies: pooled OR=2.85 95%CI=2.18-3.71, p<0.00001, 6 studies: pooled OR=2.84
95%CI=1.85-4.37, p<0.00001, respectively). There was no significant difference between
these two groups (p=0.99). A pooled risk of combined incident and worsening ADL disability
from 12 studies
15-17,24-29,31,37,38
was also calculated (12 studies: pooled OR=2.76,
95%CI=2.22-3.44, p<0.00001). High heterogeneity was observed among these three meta-
analysis on incident, worsening, and combined ADL disability risks (I
2
=84-94%). ADL
disability risks for prefrailty were also calculated and showed to be significantly higher, but
to a smaller degree than frailty, pooled disability risk estimates compared with non-frailty: 8
studies
15,17,24,25,28,31,37,38
for incident ADL disability: pooled OR=1.64, 95%CI=1.44-1.87,
p<0.0001, 4 studies
17,26,28,29
for worsening ADL disability: pooled OR=1.82, 95%CI=1.52-
2.17, p.00001 (p=0.35 for group difference), 10 studies
15,17,24-26,28,29,31,37,38
for combined ADL
disability: pooled OR=1.70, 95%CI=1.52-1.91, p<0.0001). Heterogeneity was low to
moderate (I
2
=30-43%) for these analyses for prefrailty. (Figure 2)
A total of five studies used HR to measure disability risks according to frailty.
2,23,30,32,34
Similarly to the studies with OR, among studies with HR, frailty was significantly associated
with all incident, worsening, and combined ADL disability risks (2 studies: pooled HR=2.09,
95%CI=1.56-2.80, p<0.00001, 3 studies: pooled HR=2.38, 95%CI=1.13-4.99, p<0.00001,
and 5 studies: pooled HR=2.23, 95%CI=1.42-3.49, p<0.00001, respectively). Prefrailty was
also shown to be a significant predictor of incident, worsening, and combined ADL disability