Single and dual task tests of gait speed are equivalent in the prediction of falls in older people a systematic review and meta analysis

Accepted Manuscript
Title: Single and dual task tests of gait speed are equivalent in
the prediction of falls in older people: a systematic review and
meta-analysis
Author: Jasmine C. Menant Daniel Schoene Stephen R. Lord
PII:
DOI:
Reference:

S1568-1637(14)00064-6
/>ARR 524

To appear in:

Ageing Research Reviews

Received date:
Revised date:
Accepted date:

26-2-2014
26-5-2014
2-6-2014

Please cite this article as: Menant, J.C., Schoene, D., Lord, S.R.,Single and
dual task tests of gait speed are equivalent in the prediction of falls in older
people: a systematic review and meta-analysis, Ageing Research Reviews (2014),
/>This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process

errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.


Single and dual task tests of gait speed are equivalent in the prediction of falls in older
people: a systematic review and meta-analysis

cr

ip
t

Jasmine C. Menanta,b, Daniel Schoenea,b, Stephen R. Lorda,b

Author affiliations:

Falls and Balance Research Group, Neuroscience Research Australia, PO Box 1165,

us

a

b

an

Randwick NSW 2031, Australia

School of Public Health and Community Medicine, University of New South Wales,

M

UNSW Sydney 2052, Australia

Ac
ce
p

te

d

E-Mail: , ,

Corresponding author:
Professor Stephen Lord,

Neuroscience Research Australia,

Barker Street, Randwick, NSW, 2031,
Australia.

Email:
Tel: +61 2 9399 1061
Fax: +61 2 9393 1204

Page 1 of 52


ABSTRACT

Although simple assessments of gait speed have been shown to predict falls as well as
hospitalisation, functional decline and mortality in older people, dual task gait speed

ip
t

paradigms have been increasingly evaluated with respect to fall prediction. Some studies
have found that dual task walking paradigms can predict falls in older people. A systematic

cr

review and meta-analysis was conducted to determine whether dual task walking paradigms
involving a secondary cognitive task have greater ability to predict falls than single walking

us

tasks. The meta-analytic findings indicate single and dual task tests of gait speed are

an

equivalent in the prediction of falls in older people and sub-group analyses revealed similar
findings for studies that included only cognitively impaired participants, slow walkers or used

te

Keywords

d

M

secondary mental-tracking or verbal fluency tasks.

Ac
ce
p

Accidental falls, older people, dual task paradigm, gait speed, cognition, prediction

Page 2 of 52


1. INTRODUCTION
Over the past two decades, an extensive body of research has shown that walking is not an

ip
t

automatic activity and that cognition, in particular attention and executive functioning,
contributes significantly to balance and locomotor control (Yogev-Seligman et al., 2008,

cr

Woollacott and Shumway-Cook., 2002). Much of this work has involved dual task paradigms
which typically require participants to divide their attention and concurrently undertake two

us

or more tasks, usually a cognitive and a motor task. In 1997, Lundin-Olsson et al. (1997) used

an

a dual task assessment, the “stops walking while talking test”, in a seminal study of fall risk
in residents of senior housing facilities. They found that residents who stopped walking in

M

order to engage in a conversation had significantly shorter times to a future fall than residents

d

who could talk without stopping walking.

te

Since then, many studies have investigated cognitive/ motor interference while walking and
its relationship to fall risk in elderly and clinical populations. Some of these findings have

Ac
ce
p

been pooled in previous systematic reviews (Al-Yahya et al., 2011, Beauchet et al., 2009,
Chu et al., 2013, Zijlstra et al., 2008). An early review (Zijlstra et al., 2008) included 19
studies with various protocols involving postural control, dynamic balance and gait tasks
performed with and without secondary motor or cognitive tasks. This led to a high level of
heterogeneity preventing a meaningful meta-analysis. The second review of 15 studies, by
Beauchet et al. (2009) found that large dual task costs (i.e. the percentage difference between
single and dual task performance in cognitive and/or motor tasks) were associated with a
significant increased odds of falling in older people living in community and residential aged

care settings. Al-Yahya and colleagues (2011) included 66 studies and focused on the effect
of cognitive task type on gait. The main conclusions were that gait performance was most

Page 3 of 52


affected when cognitive tasks involving internal interfering factors (such as mental tracking
tasks) were concurrently performed, and that older age and poorer cognitive function were
both strongly associated with gait speed reduction in the dual task conditions. Finally, a meta-

ip
t

analysis of 15 studies demonstrated that mental-tracking tasks yielded significant dual task
costs for fall prediction in older adults while verbal fluency or manual tasks did not (Chu et

cr

al., 2013).

us

Due to differing study aims, search strategies and review periods, the above systematic

an

reviews have contained relatively few as well as different sets of the studies now available on
the ability of dual task assessments to predict falls in older people. The findings of two

M

previous meta-analyses (Beauchet et al., 2009, Chu et al., 2013) should also be interpreted
with some caution as they pooled studies that used various walking tasks (some including

d

transfers and turns), differing secondary cognitive or motor tasks and study populations

te

varying with respect to residential settings and co-morbidities.

Ac
ce
p

Despite the methodological limitations, there appears to be good evidence that dual task
changes in gait performance are associated with significantly increased risk of falls in older
people. However, the ability of simple measures of walking speed to predict falls has also
been well-documented (Abellan van Kan et al., 2009, Deandrea et al., 2010), and the added
value of dual task protocols as opposed to single walking tasks in predicting falls in older
people has not been systematically evaluated in a meta-analysis.

The primary aim of this systematic review and meta-analysis, therefore, was to determine
whether dual task walking paradigms involving a secondary cognitive task are superior to
single walking tasks in predicting falls in older people. Additional aims were to compare the

Page 4 of 52

value of single and dual task walking speed measures for predicting falls in participant
subgroups (i.e. those with and without cognitive impairment, slower and faster walkers) and
whether a dual task: single task walking speed differential with respect to fall prediction is

ip
t

only apparent for particular secondary cognitive tasks (i.e. mental-tracking or verbal fluency
tasks). We focussed on walking speed as this is the most widely used gait measure in clinical

cr

settings as well as the gait measure most often described in published studies of fall risk in

an

us

older people.

2. METHODS

M

2.1 Literature Search

We conducted a systematic review of the literature to identify studies which had investigated

d

the relationship between gait and falls in older people using a dual task paradigm involving a

te

secondary cognitive task. Initially, we updated previous systematic reviews on the topic by
combining their searches (Al-Yahya et al., 2011, Beauchet et al., 2009, Chu et al., 2013,

Ac
ce
p

Zijlstra et al., 2008). We then searched for articles in the following seven electronic
databases: PubMed, Ovid MEDLINE, EMBASE, PsycINFO, CINAHL, Scopus and
Cochrane Central Register of Controlled Trials. Individual search strategies were tailored to
each database using the following Medical Subjects Headings (MeSH; in bold) and key
terms:

1. gait OR walking OR locomotion
2. Falls OR Accidental falls OR falling OR faller
3. aged OR aged, 80 and over OR aging OR ageing
4.

dual task* OR dual-task* OR cognition OR attention

Page 5 of 52


5. #1 AND #2 AND #3 AND #4

The search was performed without language restrictions and results were filtered to produce

ip
t

all publications from 2008 to February 2013 (inclusive of publications published
electronically ahead of print). To identify further possible studies, reference lists of the

cr

previous systematic reviews (Al-Yahya et al., 2011, Beauchet et al., 2009, Chu et al., 2013,

us

Zijlstra et al., 2008) and all other relevant articles were hand-searched.

an

2.2 Inclusion and exclusion criteria

To be included, articles had to describe studies which evaluated gait at self-selected speed

M

under single and dual task conditions in older people to either: (i) predict falls, or (ii)
discriminate between fallers and non-fallers based on retrospective data collection.

d

Articles were excluded if: (i) they were individual abstracts, case studies or reviews; (ii) the

te

focus was on patient groups (for example: Parkinson’s disease, stroke, etc) other than
cognitive impairment; (iii) participants’ mean age was less than 65 years or all participants

Ac
ce
p

were younger than 60 years of age; (iv) the walking task did not involve time or gait speed as
an outcome; (v) the secondary task was not a cognitive task; (vi) subjective scoring systems
were used to assess dual task performance; (vii) they were published in languages other than
Dutch, English, French or German.

2.3 Quality assessment

The Quality Assessment of Diagnostic Accuracy in Systematic Reviews (QUADAS) tool was
also used to provide a general standardised rating of methodological quality (Whiting et al.,
2003).

Page 6 of 52


2.4 Selection of papers and data extraction
The process of manuscripts selection is illustrated in Figure 1. Following the initial database
searches and after removal of duplicates, two independent reviewers (JM and DS) screened

ip
t

article titles and abstracts based on the inclusion and exclusion criteria defined above. The
following information was extracted from each included article: study design, sample size,

cr

residential setting, sample demographics, cognitive screening, fall outcome measure, fall

definition and months of follow-up, walking and cognitive tasks descriptions, prioritisation of

us

instructions, proportion of fallers and timed results for the single and dual task walking tests.

an

Means and standard deviations of gait speed for non-fallers and fallers in the single and dual
task tests were retrieved from each of the selected full-text articles or sought from the

M

authors.

d

2.5 Statistical analysis

te

To determine whether gait speed had a significantly better predictive value for falls in a dual
task versus a single task paradigm, we carried out random-effect meta-analysis models on all

Ac
ce
p

the studies retrieved from the search, using Stata software version X (Stata Corp., College
Station, TX, USA, 2007). A measure of effect size, the Mean Difference (MD) (95%
Confidence Intervals (CI)) between fallers and non-fallers groups’ gait speed was computed
for the single task and the dual task(s) for each study. Studies were weighed according to
sample sizes. The pooled MD (95% CI) for the single tasks was then compared to that of the
dual tasks using student t-tests (p<0.05). Forest plots displaying the MD and weight for all
the included studies in the single and dual task conditions were generated for gait speed.
Heterogeneity between studies was assessed using chi-square (χ2) tests (p < 0.1) and the I2
statistic which quantifies the proportion of variation that is a result of heterogeneity rather
than chance. The degree of inconsistency in the studies’ results was considered low, moderate

Page 7 of 52


and high for values of I2 (expressed in %) around 25%, 50% and 75%, respectively (Higgins
et al., 2003). These steps were repeated for each of the sub-analyses conducted: (i) sensitivity
analysis including only prospective studies, to determine the influence of study design; (ii)

ip
t

studies including only samples with cognitive impairment; (iii) studies including only straight
walking with no turn; (iv) studies including physically frailer samples based on mean gait

cr

speed for all participants in the single task ≥ 1.0 m.s-1 or < 1.0 m.s-1 (Tiedemann et al., 2008);
(v) studies including only mental tracking secondary tasks (requiring holding and

us

manipulating information (Al-Yahya et al., 2011, Chu et al., 2013) such as serial subtractions,

an

reciting alternate letters of the alphabet); (vi) studies including only verbal fluency tasks (for
example: generating as many animal names as possible, as many words starting with a given

3.1 Studies identified

te

3. RESULTS

d

M

letter) (Al-Yahya et al., 2011, Chu et al., 2013).

Ac
ce
p

Our search retrieved 1080 articles of which 33 articles met our inclusion criteria. We

requested additional data from the authors for 30 of these studies, of which 26 provided the
data. We were not able to obtain the gait speed data for four relevant articles, which where
therefore not included in the meta-analysis (Allali et al., 2010, Faulkner et al., 2007, Koskas
et al., 2010, Theill et al., 2011). Table 1 summarizes the 30 articles that were included in the
meta-analysis; three of them included two groups which were analysed separately (Camicioli
et al., 2006, De Melo et al., 2012, Montero-Odasso et al., 2012); thus 33 samples were
included in the overall meta-analysis (Figure 1).

3.2 Studies population

Page 8 of 52


The pooled studies yielded a total number of 4144 participants. Sample sizes ranged from 11
to 1308 participants. Twenty-three studies involved community-dwellers, one recruited
outpatients to a geriatrics department (Reelick et al., 2011). Four studies recruited

ip
t

participants from senior housing facilities or intermediate care hostels (Beauchet et al.,
2008a, 2008b, Kearns et al., 2012, Pichierri et al., 2012) and two other studies involved

cr

geriatrics and Alzheimer’s care unit inpatients (Camicioli et al., 2006, Kressig et al., 2008).

us

Fifteen studies clearly indicated that they included participants with no cognitive impairment

an

(Hall et al., 2011, Tombaugh and McIntyre, 1992, Verghese et al., 2002, Yamada et al.,
2011b). Three studies provided insufficient information regarding the cognitive level of their

M

participants (Bauer et al., 2010, Uemura et al., 2012, Yamada et al., 2011a). Five studies
included a small percentage (<27%) of participants with cognitive impairment (Donoghue et

d

al., 2013, Pichierri et al., 2012, Reelick et al., 2011, Trombetti et al., 2011, Verghese et al.,

te

2002). Seven studies (nine groups) were included in the cognitive impairment sub-analysis
(Beauchet et al., 2008a, 2008b, Camicioli et al., 2006, De Melo et al., 2012, Kearns et al.,

Ac
ce
p

2012, Kressig et al., 2008, Taylor et al., 2013). For all of them, either more than 50% of the
sample recorded MMSE scores <24 (Beauchet et al., 2008a, 2008b), the sample’s mean or
median MMSE score was less than 24 (Kearns et al., 2012, Kressig et al., 2008), or the
participants had a diagnosis of Alzheimer’s disease (Camicioli et al., 2006, De Melo et al.,
2012). Sixteen studies (18 groups) included participants who walked with mean gait speeds
slower than 1m.s-1 in the single task condition.

3.3 Methodological quality of included studies
Of the 30 studies included, 11 (12 samples) had a prospective design. Six of these had a 12month follow-up for falls (Beauchet et al., 2008a, 2008b, Liu-Ambrose et al., 2009, Nordin et

Page 9 of 52


al., 2010, Taylor et al., 2013, Verghese et al., 2002, Yamada et al., 2011a) and one had a 24month follow-up period (Herman et al., 2010). One study followed up participants until the
time of first fall, up to 12 months, using a wrist sensor to detect falls (Kearns et al., 2012).

ip
t

Two inpatient studies used bi-weekly chart reviews to identify falls and followed up patients
for approximately 12 months (Camicioli et al., 2006) or 1 month (Kressig et al., 2008). Six

cr

studies (Beauchet et al., 2008a, 2008b, Herman et al., 2010, Liu-Ambrose et al., 2009, Nordin
et al., 2010, Taylor et al., 2013, Yamada et al., 2011a) used monthly calendars or phone calls

us

to record falls, the gold standard method of falls ascertainment (Lamb et al., 2005); one study

an

used twice-yearly interviews at six and 12 months post baseline (Verghese et al., 2002).

M

Baseline gait data from control group participants of randomised controlled trials were used
in eight studies; one with prospective fall follow-up (Liu-Ambrose et al., 2009), and seven

d

with retrospective fall data (Halvarsson et al., 2011, Pichierri et al., 2012, Silsupadol et al.,

te

2009, Trombetti et al., 2011, Verghese et al., 2010, Yamada et al., 2010, Yamada et al.,
2011c). Two studies compared non-multiple fallers with multiple fallers (Reelick et al., 2011,

Ac
ce
p

Taylor et al., 2013), and the remainder compared non-fallers and fallers (i.e. people who
experienced one or more falls). Less than half the studies provided a detailed definition of a
fall and fall rates varied greatly between studies, ranging from 15% to 90%. Four studies had
a history of falls in the past 12 months as an inclusion criterion (Bauer et al., 2010,
Halvarsson et al., 2011, Siu et al., 2008, Trombetti et al., 2011).

Twenty-two studies included only one type of secondary cognitive task and nine studies (ten
samples) included two or more types of secondary cognitive task. Most studies used a mental
tracking task (n=28, 31 samples) or a verbal fluency task (n=6, 7 samples) as the cognitive

Page 10 of 52

task. Only two studies used discrimination and decision making secondary tasks (Kearns et
al., 2012, Siu et al., 2008).
In two studies (Liu-Ambrose et al., 2009, Verghese et al., 2002), two levels of complexity of

ip
t

a task of the same nature (reciting the alphabet and reciting alternate letters of the alphabet)
were used. For these studies, only gait speed data recorded in the most complex condition

cr

were included in our analyses.

us

Most studies recorded gait speed during straight line walking; in two studies (Liu-Ambrose et

an

al., 2009, Verghese et al., 2002) the gait protocol included a turn. Similarly, most studies
required participants to walk at a self-selected speed; two studies were unclear about this

M

specification (Liu-Ambrose et al., 2009, Uemura et al., 2012). Most studies (n=27) specified
the distance for which participants’ gait speed was recorded which ranged from 4.6 m to 25

d

m. Most studies used either an electronic walkway (n=16) or a stopwatch (n=11) to record

te

gait speed. All studies explained the nature of the secondary task to be performed and in 17
studies it was clearly indicated that participants were not to prioritise completing one task

Ac
ce
p

over the other.

3.4 Meta-analyses results

The single task and the dual task conditions across all domains investigated significantly
discriminated between fallers and non-fallers (p<0.05). Figure 2 displays the forest plots of
the gait speed variable for all of the studies included in the meta-analysis (n=30) in single
task (top) and dual task (bottom) conditions. The pooled MD (95% CI) for gait speed
between fallers and non-fallers in the single task (0.069 (0.045-0.094)) was not significantly
different to that in the dual task condition (0.074 (0.046-0.103)) (Z= 0.266, p=0.790). The

Page 11 of 52


heterogeneity was moderate for both paradigms (single task: I2=48.0%, χ2=61.5, DF 32,
p=0.002; dual task: I2=58.2%: χ2=76.6, DF 32, p=0.003).

ip
t

A sensitivity analysis including only prospective falls studies yielded the same nonsignificant result (single task 0.044 (0.015-0.072) vs. dual task 0.054 (0.025-0.083)) (Z=

cr

0.478, p=0.633), with moderate to low heterogeneity between studies ((single task: I2=32.1%,

us

χ2=16.19, DF 11, p=0.028; dual task: I2=28.1%: χ2=15.30, DF 11, p=0.001) (Figure 3).

an

Subgroup meta-analyses revealed no significant differences between pooled MD in gait speed
in the single task vs. the dual task conditions amongst people with fast gait speed (mean gait

M

speed in single task ≥ 1.0 m.s-1) (overall pooled MD (SEM): 0.009 (0.029), Z=-0.306,
p=0.760), slow gait speed (mean gait speed in single task < 1.0 m.s-1) (overall pooled MD

d

(SEM): 0.001 (0.022), Z=-0.045, p=0.964) (Figure 4), or people with cognitive impairment

te

(overall pooled MD (SEM): -0.030 (0.029), Z=-1.039, p=0.299) (Figure 5).

Ac

ce
p

The subgroup analyses exploring the type of secondary task indicated that dual task
conditions involving mental tracking tasks (overall pooled MD (SEM): 0.010 (0.018),
Z=0.544, p=0.587) and verbal fluency tasks (overall pooled MD (SEM): -0.666 (0.372), Z=1.791, p=0.073) were both not different to single task gait conditions in the prediction of falls
(Figures 6-7).

Restricting the analysis to studies that only included straight line walking showed a similar
pattern of no significant difference (single task 0.067 (0.042-0.092) vs. dual task 0.070
(0.040-0.099), (Z= 0.151, p=0.880), with moderate to low heterogeneity between studies

Page 12 of 52


(single task: I2=46.8%, χ2=56.36, DF 30, p=0.017; dual task: I2=58.2%: χ2=71.72, DF 30,
p=0.0030).

ip
t

Heterogeneity between studies in the single and dual task conditions dropped below 50% for
the slow gait speed sub-group, and reached minimal levels for the cognitive impairment

cr

group (0% vs. 13.8%) and in the verbal fluency tasks (0% for both single and dual task

us

conditions) sub-analyses.

an

4. DISCUSSION

4.1 Equivalence of single and dual task gait speed in the prediction of falls

M

This systematic review and meta-analysis of 31 studies revealed that slower gait speeds under
both single and dual task conditions significantly discriminate between fallers and non-

d

fallers. The differences in gait speed between fallers and non-fallers under both conditions

te

were virtually identical in the main and all sub-group analyses and a sensitivity analysis
confirmed this finding when calculations were restricted to studies with prospective designs.

Ac
ce
p

In other words, dual task gait speed offered no clinically or statistically significant benefit
over single task gait speed in either discriminating recent fallers from non-fallers or in
predicting future fallers and non-fallers. In all comparisons the difference between means
approached zero with virtually identical CIs. Heterogeneity was highest (but acceptable) for

the overall meta-analysis, which can be attributed to a range of between-studies differences
such as residential settings, cognitive and physical status, falls assessment, and dual task
protocols. Unsurprisingly, as the studies were pooled in more specific sub-groups,
heterogeneity declined. 

Page 13 of 52


Several potential reasons for these null findings were explored. First, it is possible that the
differential effect of dual task walking on fall risk may only be evident in either in fitter
people who walk quickly or in frailer / high risk groups who require more attentional

ip
t

resources for balance tasks. Contrary to the findings by Yamada et al (Yamada et al., 2011)
who showed that a manual task cost predicted falls in sub-groups of participants who

cr

performed the Timed Up and Go test quickly, we found no indication that dual task gait

speed discriminated fallers from non-fallers better than single task gait speed in the sub-set of

us

studies in our review that reported mean gait speeds ≥ 1. Further, the findings of equivalence
of single and dual task walking in discriminating between fallers and non-fallers were also

an

evident in the population groups with reduced mobility (gait speed < 1 m.s-1) and cognitive

M

impairment. Further, other studies that have included patient groups at high risk of falls have
reported similar findings (Hyndman et al., 2006; Smulders et al., 2012). In stroke patients, a

d

task requiring participants to remember a 7-item shopping list equally affected gait speed in

te

fallers and non-fallers (Hyndman et al., 2006) and in a large sample of Parkinson’s disease
patients, neither gait variables nor performance in an executive function (Stroop) test were

Ac
ce
p

associated with prospective falls in a dual task paradigm (Smulders et al., 2012).

Secondly, it may be that only certain secondary cognitive tasks adversely influence gait.
Other systematic reviews suggest that cognitive tasks that involve internal interfering factors,
such as mental tracking and verbal fluency tasks are more effective at disturbing gait
performance than cognitive tasks involving external interfering factors such as reaction time
tasks (Al-Yahya et al., 2011), and that only dual task costs of mental tracking tasks are
related to fall risk (Chu et al., 2013). However, in our sub-group analyses, we found that in
studies that used only mental tracking or verbal fluency tasks, no significant differences in

pooled MD in gait speed between fallers and non-fallers were evident for the single- and dual

Page 14 of 52


task conditions. There is some preliminary evidence for shared neural networks between
visuo-spatial tasks and balance (Barra et al., 2006, Sturnieks et al., 2008) which would
suggest that secondary tasks involving spatial working memory impair gait function to a

ip
t

greater extent than mental-tracking tasks. In a recent study, we found that older adults walked
slower with increased step time variability and poorer medio-lateral stability (increased

cr

harmonic ratio) when concurrently performing a visuo-spatial task as opposed to an

arithmetic mental tracking task (submitted manuscript). However, whether dual task

us

paradigms involving visuo-spatial secondary tasks would improve fall prediction is yet to be

an

verified.

M

Thirdly, it is possible walking speed is not a sufficiently sensitive measure to discriminate
between faller and non-faller groups in dual task paradigms (Beauchet et al., 2008a). Two

d

studies have reported that a dual task paradigm can improve the prediction of falls when gait

te

time variability measures are used. Herman et al found that swing time variability in a dual
task condition involving mental tracking was an independent predictor of falls status recorded

Ac
ce
p

prospectively over two-years in 262 older community-dwellers as well as a sub-group of 201
participants from this sample who did not report a past fall at baseline (Herman et al., 2010).
In both analyses, usual walking speed was not associated with faller status in either the single
or dual task conditions. Kressig et al. (2008) reported that stride time variability in dual task
but not single task conditions was significantly associated with the occurrence of a first fall in
57 aged care inpatients. Our analyses of this study data indicated no significant differences
between single and dual task gait speed MDs for the fallers and non-fallers (MD (95%CI):
0.020 (0.087), Z=0.230, p=0.818).

Page 15 of 52


Other studies, however, have not found gait stride or swing time variability measures to be

superior for fall prediction when conducted in dual task as opposed to single task conditions.
Springer et al reported a significant faller group x dual task condition interaction, but also

ip
t

reported significant differences between the faller and non-faller groups for usual gait
measures such as gait speed, swing time and swing time variability (Springer et al., 2006). In

cr

a retrospective study involving 60 participants, Reelick et al. (2011) analysed three gait
variability measures (stride length, stride time and stride width) and of these, only one

us

measure (stride length variability), discriminated significantly between the faller groups in

an

the dual task, and not the single task, condition. Finally, in a prospective study of 64 older
people with cognitive impairment, Taylor et al. (2013) found that gait variability measures

M

(swing time variability and stride length variability) discriminated similarly between fallers
and non-fallers. We initially planned to include gait variability measures in our review, but

d

found most studies did not collect these data and of those that did, the measures were very

te

heterogeneous and could not be pooled for a meta-analysis.

Ac
ce
p

Finally, it may be that the findings should be taken at face value, indicating that gait slowing
(due to underlying reduced balance, strength and coordination) and independent of cognitive
factors (such as poor attention and reduced executive functioning) is primarily important in
relation to fall risk and that the addition of secondary cognitive tasks affects fallers and nonfallers equally. Alternatively, it may be that reduced postural control when walking under
divided attention often results in falls in everyday life, but due to their high correlation,
measures of single or dual task gait speed are equally good indicators of this liability.

4.2 Implications for clinical practice

Page 16 of 52


Our data suggest that with regard to falls in older people, both single and dual task paradigms
are equivalent in their predictive and discriminative validity when gait speed is used. The
effect size of the largest single task - dual task difference is less than 0.15 (based on the

ip
t

largest single task - dual task difference (0.03) divided by the mean gait speed SD from the

four trials with samples with at least 100 fallers and non-fallers (0.22); this effect size is

cr

typically considered small and of dubious clinical utility. As gait speed provides a useful

measure of mobility and cut-off values for gait speed have been published with regard to fall

us

prediction in community-dwellers based on simple walking trials (Callisaya et al., 2011,

an

Tiedemann et al., 2008), we would suggest that clinicians could retain a simple test of fall risk
along with other pertinent fall history, medication, vision and physical performance measures

M

in fall risk screening batteries. However, if clinicians are seeking further information on
attention and executive functions of their patients, then dual task gait paradigms might be of

d

value, as the change in gait performance between the single and dual task allows exploring

te

the attentional requirement of balance during locomotion (Amboni et al., 2013). Such
assessment might in turn guide exercise and cognitive training interventions for older people

Ac
ce
p

to improve multitasking abilities when standing and walking, which are required in many
real-life situations.

4.3 Strengths and limitations of the analysis
This is the first meta-analysis focusing on the predictive ability of walking velocity in dual
task vs. single task on falls, as previous meta-analyses have focused on dual task costs
(Beauchet et al., 2009, Chu et al., 2013, Zijlstra et al., 2008). The present review required us
to request data from authors for the majority of the studies selected, i.e. gait speed data were
not presented for fallers and non-fallers but falls were mentioned as an outcome variable in
the article. The inclusion of these data, which was provided by most authors (87%), greatly

Page 17 of 52


helped minimise a common bias of systematic reviews to only include published data. The
main study limitation relates to the unavailability of similar gait variability data for inclusion
in any meta-analyses. Another limitation pertains to the inclusion of studies which included a

ip
t

turn in their protocol, even though a sensitivity analysis which included studies with only

cr

straight line walking yielded the same non-significant results as the primary one.

5. CONCLUSION

us

In conclusion, this meta-analysis indicates that both single and dual task paradigms were

an

equivalent in predicting falls and discriminating fallers from non-fallers based on gait speed
measurements. This finding implies that additional tests of dual task gait speed are not

M

required for assessing gait in clinical practice and that a simple gait speed measure along with
other pertinent fall history, medication, vision and physical performance measures could be

d

included in fall risk screening batteries. However, dual task gait paradigms may be helpful in

te

exploring specific attentional and execution function requirements of balance and walking,
which may assist in guiding interventions to maintain cognitive and physical health in older

Ac
ce
p

age.

ACKNOWLEDGEMENTS

We thank Mina Sarofim for his assistance with the literature search and data extraction.

REFERENCES

Abellan van Kan, G., Rolland, Y., Andrieu, S., Bauer, J., Beauchet, O., Bonnefoy, M., Cesari,
M., Donini, L.M., Gillette Guyonnet, S., Inzitari, M., Nourhashemi, F., Onder, G., Ritz, P.,
Salva, A., Visser, M., Vellas, B., 2009. Gait speed at usual pace as a predictor of adverse

Page 18 of 52


outcomes in community-dwelling older people an International Academy on Nutrition and
Aging (IANA) Task Force. J Nutr Health Aging 13(10), 881-889.
Allali, G., Dubois, B., Assal, F., Lallart, E., de Souza, L.C., Bertoux, M., Annweiler, C.,

ip
t

Herrmann, F.R., Levy, R., Beauchet, O., 2010. Frontotemporal dementia: pathology of gait?
Mov Disord 25(6), 731-737, doi: 10.1002/mds.22927.

cr

Al-Yahya, E., Dawes, H., Smith, L., Dennis, A., Howells, K., Cockburn, J., 2011. Cognitive

motor interference while walking: a systematic review and meta-analysis. Neurosci Biobehav

us

Rev 35(3), 715-728, doi: 10.1016/j.neubiorev.2010.08.008.

an

Amboni, M., Barone, P., Hausdorff, J.M., 2013. Cognitive contributions to gait and falls:
Evidence and implications. Mov Disord 28(11), 1520-1533, doi: 10.1002/mds.25674.

M

Barra, J., Bray, A., Sahni, V., Golding, J.F., Gresty, M.A., 2006. Increasing cognitive load
with increasing balance challenge: recipe for catastrophe. Exp Brain Res 174(4), 734-745.

d

Bauer, C., Gröger, I., Glabasnia, A., Bergler, C., Gassmann, K.G., 2010. First results of

te

evaluation of a falls clinic. International Journal of Gerontology 4(3), 130-136,
/>
Ac
ce
p

Beauchet, O., Allali, G., Annweiler, C., Berrut, G., Maarouf, N., Herrmann, F.R., Dubost, V.,
2008a. Does change in gait while counting backward predict the occurrence of a first fall in

older adults? Gerontology 54(4), 217-223, doi: 10.1159/000127318.
Beauchet, O., Annweiler, C., Allali, G., Berrut, G., Herrmann, F.R., Dubost, V., 2008b.
Recurrent falls and dual task-related decrease in walking speed: is there a relationship? J Am
Geriatr Soc 56(7), 1265-1269, doi: 10.1111/j.1532-5415.2008.01766.x.
Beauchet, O., Annweiler, C., Dubost, V., Allali, G., Kressig, R.W., Bridenbaugh, S., Berrut,
G., Assal, F., Herrmann, F.R., 2009. Stops walking when talking: a predictor of falls in older
adults? Eur J Neurol 16(7), 786-795, doi: 10.1111/j.1468-1331.2009.02612.x.

Page 19 of 52


Callisaya ML, Blizzard L, Schmidt MD, Martin, K.L., McGinley, J.L., Sanders, L.M.,
Srikanth, V.K., 2011. Gait, gait variability and the risk of multiple incident falls in older
people: a population-based study. Age Ageing 40(4), 481-487, doi: 10.1093/ageing/afr055.

ip
t

Camicioli, R., Bouchard, T., Licis, L., 2006. Dual-tasks and walking fast: relationship to
extra-pyramidal signs in advanced Alzheimer disease. J Neurol Sci 248(1-2), 205-209.

cr

Chu, Y.H., Tang, P.F., Peng, Y.C., Chen H.Y., 2013. Meta-analysis of type and complexity

13(2), 289-297, doi: 10.1111/j.1447-0594.2012.00893.x.

us

of a secondary task during walking on the prediction of elderly falls. Geriatr Gerontol Int

an

De Melo Coelho, F.G., Stella, F., De Andrade, L.P., Barbieri, F.A., Santos-GaldurÓz, R.F.,
Gobbi, S., Costa, J.L., Gobbi, L.T., 2012. Gait and risk of falls associated with frontal

M

cognitive functions at different stages of Alzheimer's disease. Neuropsychol Dev Cogn B
Aging Neuropsychol Cogn. 19(5), 644-656, doi: 10.1080/13825585.2012.661398.

d

Deandrea, S., Lucenteforte, E., Bravi, F., Foschi, R., La Vecchia, C., Negri, E., 2010. Risk

te

factors for falls in community-dwelling older people: a systematic review and meta-analysis.
Epidemiology 21(5), 658-668, doi: 10.1097/EDE.0b013e3181e89905.

Ac
ce
p

Donoghue, O.A., Cronin, H., Savva, G.M., O'Regan, C., Kenny, R.A., 2013. Effects of fear
of falling and activity restriction on normal and dual task walking in community dwelling
older adults. Gait Posture 38(1), 120-124, doi: 10.1016/j.gaitpost.2012.10.023.
Faulkner, K.A., Redfern, M.S., Cauley, J.A., Landsittel, D.P., Studenski, S.A., Rosano, C.,
Simonsick, E.M., Shorr, R.I., Ayonayon, H.N., Newman, A.B., Health, Aging, and Body
Composition Study, 2007. Multitasking: association between poorer performance and a

history of recurrent falls. J Am Geriatr Soc 55(4), 570-576.
Hall, C.D., Echt, K.V., Wolf, S.L., Rogers, W.A., 2011. Cognitive and motor mechanisms
underlying older adults' ability to divide attention while walking. Phys Ther 91(7), 10391050, doi: 10.2522/ptj.20100114.

Page 20 of 52


Halvarsson, A., Oddsson, L., Olsson, E., Faren, E., Pettersson, A., Ståhle, A., 2011. Effects
of new, individually adjusted, progressive balance group training for elderly people with fear
of falling and tend to fall: a randomized controlled trial. Clin Rehabil 25(11), 1021-1031.

ip
t

Herman, T., Mirelman, A., Giladi, N., Schweiger, A., Hausdorff, J.M., 2010. Executive
control deficits as a prodrome to falls in healthy older adults: a prospective study linking

cr

thinking, walking, and falling. J Gerontol A Biol Sci Med Sci 65(10), 1086-1092, doi:
10.1093/gerona/glq077.

us

Higgins, J.P., Thompson, S.G., Deeks, J.J., Altman, D.G., 2003 Measuring inconsistency in

an

meta-analyses. BMJ 327(7414), 557-560.

Hyndman, D., Ashburn, A., Yardley, L., Stack, E., 2006. Interference between balance, gait

M

and cognitive task performance among people with stroke living in the community. Disabil
Rehabil 28(13-14), 849-856.

d

Kearns, W.D., Fozard, J.L., Becker, M., Jasiewicz, J.M., Craighead, J.D., Holtsclaw, L.,

te

Dion, C., 2012. Path Tortuosity in Everyday Movements of Elderly Persons Increases Fall
Prediction Beyond Knowledge of Fall History, Medication Use, and Standardized Gait and

Ac
ce
p

Balance Assessments. J Am Med Dir Assoc 13(7), 665.e7-665.e13, doi:
10.1016/j.jamda.2012.06.010.

Koskas, P., Saad, S., Belqadi, S., Drunat, O., 2010. Gait analysis with simple and dual-task
paradigm in community-dwelling old people attending a geriatric outpatient clinic. Rev
Neurol (Paris) 166(3), 321-327, doi: 10.1016/j.neurol.2009.07.009.
Kressig, R.W., Herrmann, F.R., Grandjean, R., Michel, J.P., Beauchet, O., 2008. Gait
variability while dual-tasking: fall predictor in older inpatients? Aging Clin Exp Res ;20(2),
123-130.
Lamb, S.E., Jorstad-Stein, E.C., Hauer, K., Becker, C., Prevention of Falls Network Europe

and Outcomes Consensus Group, 2005. Development of a common outcome data set for fall

Page 21 of 52


injury prevention trials: the Prevention of Falls Network Europe consensus. J Am Geriatr Soc
53(9), 1618-1622.
Liu-Ambrose, T., Katarynych, L.A., Ashe, M.C., Nagamatsu, L.S., Hsu, C.L., 2009. Dual-

ip
t

task gait performance among community-dwelling senior women: the role of balance
confidence and executive functions. J Gerontol A Biol Sci Med Sci 64(9), 975-982, doi:

cr

10.1093/gerona/glp063.

Lundin-Olsson, L., Nyberg, L., Gustafson, Y., 1997. "Stops walking when talking" as a

us

predictor of falls in elderly people. Lancet 349(9052), 617.

an

Montero-Odasso, M., Muir, S.W., Speechley, M., 2012. Dual-task complexity affects gait in
people with mild cognitive impairment: the interplay between gait variability, dual tasking,

M

and risk of falls. Arch Phys Med Rehabil 93(2), 293-299, doi: 10.1016/j.apmr.2011.08.026.
Nordin, E., Moe-Nilssen, R., Ramnemark, A., Lundin-Olsson, L., 2010. Changes in step-

te

10.1016/j.gaitpost.2010.03.012.

d

width during dual-task walking predicts falls. Gait Posture 32(1), 92-7, doi:

Pichierri, G., Murer, K., de Bruin, E.D., 2012. A cognitive-motor intervention using a dance

Ac
ce
p

video game to enhance foot placement accuracy and gait under dual task conditions in older
adults: a randomized controlled trial. BMC Geriatr 12:74, doi: 10.1186/1471-2318-12-74.
Reelick, M.F., van Iersel, M.B., Kessels, R.P., Rikkert, M.G., 2009. The influence of fear of
falling on gait and balance in older people. Age Ageing 38(4), 435-340. doi:
10.1093/ageing/afp066.

Reelick, M.F., Kessels, R.P., Faes, M.C., Weerdesteyn, V., Esselink, R.A., Olde Rikkert, G.,
2011. Increased intra-individual variability in stride length and reaction time in recurrent
older fallers. Aging Clin Exp Res 23(5-6), 393-399, doi: 10.3275/7327.
Silsupadol, P., Shumway-Cook, A., Lugade, V., van Donkelaar, P., Chou, L.S., Mayr, U.,
Woollacott, M.H., 2009. Effects of single task versus dual-task training on balance

Page 22 of 52


performance in older adults: a double-blind, randomized controlled trial. Arch Phys Med
Rehabil 90(3), 381-387, doi: 10.1016/j.apmr.2008.09.559.
Siu, K.C., Lugade, V., Chou, L.S., van Donkelaar, P., Woollacott, M.H., 2008. Dual-task

ip
t

interference during obstacle clearance in healthy and balance-impaired older adults. Aging
Clin Exp Res 20(4), 349-354.

cr

Smulders, K., Esselink, R.A., Weiss, A., Kessels, R.P., Geurts, A.C., Bloem, B.R., 2012.

Assessment of dual tasking has no clinical value for fall prediction in Parkinson's disease. J

us

Neurol 259(9), 1840-1847.

an

Springer, S., Giladi, N., Peretz, C., Yogev, G., Simon, E.S., Hausdorff, J.M., 2006. Dualtasking effects on gait variability: the role of aging, falls, and executive function. Mov Disord

M

21(7), 950-957.

Sturnieks, D.L., St George, R., Fitzpatrick, R.C., Lord, S.R., 2008. Effects of spatial and

te

Sci Med Sci 63(10), 1063-1068.

d

nonspatial memory tasks on choice stepping reaction time in older people. J Gerontol A Biol

Taylor, M.E., Delbaere, K., Mikolaizak, A.S., Lord, S.R., Close, J.C.T., 2013. Gait parameter

Ac
ce
p

risk factors for falls under simple and dual task conditions in cognitively impaired older
people. Gait and Posture 37(1), 126-130, doi: 10.1016/j.gaitpost.2012.06.024.
Theill, N., Martin, M., Schumacher, V., Bridenbaugh, S.A., Kressig, R.W., 2011.
Simultaneously measuring gait and cognitive performance in cognitively healthy and
cognitively impaired older adults: the Basel motor-cognition dual-task paradigm. J Am
Geriatr Soc 59(6), 1012-1018, doi: 10.1111/j.1532-5415.2011.03429.x.
Tiedemann, A., Shimada, H., Sherrington, C., Murray, S., Lord, S., 2008. The comparative
ability of eight functional mobility tests for predicting falls in community-dwelling older
people. Age Ageing 37(4), 430-435, doi: 10.1093/ageing/afn100.

Page 23 of 52

Tombaugh, T.N., McIntyre, N.J., 1992. The mini-mental state examination: a comprehensive
review. J Am Geriatr Soc 40(9), 922-935.
Trombetti, A., Hars, M., Herrmann, F.R., Kressig, R.W., Ferrari, S., Rizzoli, R., 2011. Effect

randomized controlled trial. Arch Intern Med 171(6), 525-533, doi:

cr

10.1001/archinternmed.2010.446.

ip
t

of music-based multitask training on gait, balance, and fall risk in elderly people: a

Uemura, K., Yamada, M., Nagai, K., Shinya, M., Ichihashi, N., 2012. Effect of dual-tasking

us

on the center of pressure trajectory at gait initiation in elderly fallers and non-fallers. Aging

an

Clin Exp Res 24(2), 152-156.

Verghese, J., Buschke, H., Viola, L., Katz, M., Hall, C., Kuslansky, G., Lipton, R., 2002.

J Am Geriatr Soc 50(9), 1572-1576.

M

Validity of divided attention tasks in predicting falls in older individuals: a preliminary study.

d

Verghese, J., Mahoney, J., Ambrose, A.F., Wang, C., Holtzer, R., 2010. Effect of cognitive

te

remediation on gait in sedentary seniors. J Gerontol A Biol Sci Med Sci 65(12), 1338-1343,
doi: 10.1093/gerona/glq127.

Ac
ce
p

Whiting, P., Rutjes, A.W., Reitsma, J.B., Bossuyt, P.M., Kleijnen, J., 2003. The development
of QUADAS: a tool for the quality assessment of studies of diagnostic accuracy included in
systematic reviews. BMC Med Res Methodol 3:25, doi: 10.1186/1471-2288-3-25.
Woollacott, M., Shumway-Cook, A., 2002. Attention and the control of posture and gait: a
review of an emerging area of research. Gait Posture 16(1), 1-14.
Yamada, M., Aoyama, T., Arai, H., Nagai, K., Tanaka, B., Uemura, K., Mori, S., Ichihashi,
N., 2011a. Dual-task walk is a reliable predictor of falls in robust elderly adults. J Am Geriatr
Soc 59(1), 163-164, doi: 10.1111/j.1532-5415.2010.03206.x.
Yamada, M., Aoyama, T., Nakamura, M., Tanaka, B., Nagai, K., Tatematsu, N., Uemura, K.,
Nakamura, T., Tsuboyama, T., Ichihashi, N., 2011b. The reliability and preliminary validity

Page 24 of 52