Stops walking when talking a predictor of falls in older adults

European Journal of Neurology 2009, 16: 786–795

doi:10.1111/j.1468-1331.2009.02612.x

REVIEW ARTICLE

Stops walking when talking: a predictor of falls in older adults?
O. Beaucheta, C. Annweilera, V. Dubostb, G. Allalic, R. W. Kressigd, S. Bridenbaughd, G. Berrute,
F. Assalc and F. R. Herrmannf
a

Department of Geriatrics & Faculty of Medicine, Angers University Hospital and University of Angers, UNAM, France; bFormadep,

Korian, Paris, France; cDepartment of Neurology & Faculty of Medicine, Geneva University Hospitals, Switzerland; dDepartment of
Geriatrics, University of Basel & Basel University Hospital, Switzerland; eDepartment of Geriatrics & Faculty of Medicine, Nantes University
Hospital & University of Nantes, UNAM, France; and fDepartment of Rehabilitation and Geriatrics & Faculty of Medicine, Geneva
University Hospitals and University of Geneva, Switzerland

Keywords:

dual-task-related gait
changes, falling, gait,
older adult
Received 29 July 2008
Accepted 18 February 2009

The objective of this study was to systematically review all published articles examining the relationship between the occurrence of falls and changes in gait and attention-demanding task performance whilst dual tasking amongst older adults. An
English and French Medline and Cochrane library search ranging from 1997 to 2008
indexed under Ôaccidental fallsÕ, Ôaged OR aged, 80 and overÕ, Ôdual taskÕ, Ôdual taskingÕ,
ÔgaitÕ, ÔwalkingÕ, ÔfallÕ and ÔfallingÕ was performed. Of 121 selected studies, fifteen met
the selection criteria and were included in the final analysis. The fall rate ranged from

11.1% to 50.0% in retrospective studies and from 21.3% to 42.3% in prospective ones.
Amongst the three retrospective and eight prospective studies, two and six studies,
respectively, showed a significant relationship between changes in gait performance
under dual task and history of falls. The predictive value for falling was particularly
efficient amongst frail older adults compared with healthy subjects. Two prospective
studies challenged the usefulness of the dual-task paradigm as a significant predictor
compared to single task performance and three studies even reported that gait changes
whilst dual tasking did not predict falls. The pooled odds ratio for falling was 5.3 (95%
CI, 3.1–9.1) when subjects had changes in gait or attention-demanding task performance whilst dual tasking. Despite conflicting early reports, changes in performance
whilst dual tasking were significantly associated with an increased risk for falling
amongst older adults and frail older adults in particular. Description of health status,
standardization of test methodology, increase of sample size and longer follow-up
intervals will certainly improve the predictive value of dual-task-based fall risk
assessment tests.

Introduction
Simple and efficient detection of fall risk in older adults
is a major objective of geriatric medicine. Twelve years
ago, Lundin-Olsson et al. [1] published a seminal paper
that showed that Ôstops walking when talkingÕ could be
a predictor of falls and thus introduced a new approach
to fall prediction based on dual-task performance.
The principle of dual-task gait assessment is to
compare task performance whilst walking and simoultaneously executing an attention-demanding task to
performance of either one of single tasks [2,3]. Changes
Correspondence: Olivier Beauchet, MD, PhD, Department of
Geriatrics, Angers University Hospital, 49933 Angers Cedex 9,
France (tel.: ++33 2 41 35 45 50; fax: ++33 2 41 35 48 94; e-mail:
).

786

in performance whilst dual tasking are usually interpreted as interference because of competing demands
for attentional resources needed for both tasks [3,4] and
mainly depend on oneÕs capacity to properly allocate
attention between the two tasks [2,3].
Over the past years, dual-task-related gait changes
have frequently been reported amongst older adults [4,5].
However, published data are heterogeneous and show
that impaired dual tasking is [1,6–14] and is not associated with falls [15–17], or is even an irrelevant fall risk
indicator compared to impaired single task performance
[6,18,19]. Using a systematic review of the literature, we
selected and analysed all published data, which examined
the relationship between fall incidences and changes in
gait and/or attention-demanding task performance
whilst dual tasking amongst people aged 65 and older
and identified the reasons of those conflicting results.

Ó 2009 The Author(s)
Journal compilation Ó 2009 EFNS


Dual task and prediction of falls

Methods
Literature search

An English and French Medline literature search of all
articles published from March 1997 to April 2008 using
the Medical Subject Heading (MeSH) terms Ôaccidental

fallÕ and Ôaged or aged, 80 and overÕ combined with the
terms Ôdual taskÕ, Ôdual taskingÕ, ÔgaitÕ, ÔwalkingÕ, ÔfallÕ
and ÔfallingÕ was performed. The search also included
the Cochrane library and the references lists of the retrieved articles. To ensure a comprehensive approach,
additional key studies known to the authors that did
not meet the search criteria were also included.
Study selection and analysis

Abstract selection was based on the STrengthening the
Reporting of OBservational studies in Epidemiology
(STROBE) checklist that described items that should be
included in reports of cohort studies [20]. Abstracts
identified with the literature search were independently
evaluated by two reviewers. For those abstracts, which
fulfilled the inclusion criteria (observational studies,
retrospective or prospective data collection of falls,
number of falls and motor performance under single
and dual task as outcomes), full articles were obtained
for the final analysis. Final selection criteria were applied when mean age of 65 and older was reached, gait
performance under single and dual task were specified
as outcome measures, and enrolment methods, exact
procedures of dual tasking and discriminative or predictive values of falls provided. The study selection is
shown on a flow diagram (Fig. 1).
Twenty-eight of 121 identified abstracts were first
identified after screening using initial inclusion criteria.
Thorough revision further excluded 13 studies because

787

balance task was used as the main outcome measure

(n = 6) [21–26], mean age of the sample younger than
65 (n = 2) [27,28], the cognitive task not described
(n = 1) [29] or no measures of the discriminative or
predictive value for falls provided (n = 4) [30–33]. The
remaining 15 studies [1,6–19] were included in this review.
The association between dual-task-associated gait
changes and falls was determined based on the predictive value for falling, the odds ratio (OR) for falls with a
95% confidence interval (CI) and the positive and
negative predictive value (PPV and NPV) for falls along
with sensitivity and specificity.
All parameters were calculated for each study using
Dag-stat a spreadsheet for the calculation of comprehensive statistics for the assessment of diagnostic tests
and inter-rater agreement that provides a comprehensive range of statistics for 2 by 2 tables [34]. A metaanalysis was conducted using the STATA software version 10.1 (Stata Corp., College Station, TX, USA 2007)
to compute pooled OR associated with the risk of falls
according to the results from the dual-task procedure
either from the exact number of events and non-events
when available or to compute an estimated OR (ES)
from the natural logarithm transformation of the OR
and 95% CI. As a result of statistically significant heterogeneity amongst the studies, the DerSimonian and
Laird random effects method was used to estimate the
pooled OR. 95% CI for the ORs were computed with
the Woolf method.

Results
Table 1 summarizes the 15 studies included in this review. Number of participants ranged from 30 to 380
[6,18]. Fall rates ranged from 11.1% to 50.0% in retrospective studies of fall collection [6,14,15] and from

Initial references identified (n = 121)
Abstracts rejected because did not meet the initial criteria
selection (n = 70; Duplication [22], Interventional studies

[10], No dual-task [7], Review [5], Fall not outcome [49])
Meet criteria and were requested (n = 28)

Manuscripts rejected because did not meet the final criteria
selection (n = 13; balance task as outcome [6], mean age
under 65 [2], cognitive task not described [1], no measures
for discriminative or predictive value for falls [4])
Figure 1 Flow diagram of selection of
studies.

Manuscripts accepted for final inclusion (n = 15)

Ó 2009 The Author(s)
Journal compilation Ó 2009 EFNS European Journal of Neurology 16, 786–795


Community-dwelling
N = 95 (70 non-fallers and 25
fallers)
73.4 ± 1.7 yearsa
Women only
Community-dwelling
N = 370 (37 recurrent fallers and
333 non-fallers or single fallers)
78 ± 3 yearsa

Studies with a prospective data collection of falls
Lundin-Olsson et al.,
Senior housing facilities
1997 [1]

N = 58
80.1 ± 6.1 yearsa
Main comorbiditiesb [Dementia
(n = 26), Depression (n = 25),
Previous stroke (n = 20)]
Lundin-Olsson et al.,
Senior housing facilities
1998 [7]
N = 42
79.7 ± 6.1 yearsa
Main comorbiditiesb [Dementia
(n = 21), Depression (n = 18),
Previous stroke (n = 14)]
Lundin-Olsson
Senior housing facilities
et al., 2000 [8]
N = 78
82 (66-99) yearsc
Main comorbiditiesb [Dementia
(n = 38), Depression (n = 35),
Previous stroke (n = 25)]

Faulkner et al.,
2007 [14]

Vaillant et al.,
2006 [15]

Studies with a retrospective data collection of falls
Shumway-Cook

Community-dwelling
et al., 2000 [6]
N = 30 (15 non-fallers and 15
fallers)
78 ± 6 yearsa for non-fallers and
86.2 ± 6 for fallers

Participant

Conversation

To carry a glass of
water

To carry a glass of
water
Conversation

TUG
Normal pace

TUG
Walking from home
to an assessment
room
Normal pace

Visuo-spatial decision
task

Arithmetic task
(counting backward
by 2, 3 and 5)

Cognitive task:
Arithmetic task
(counting backward
by 3)
Manual task: To carry
a glass of water

Attention-demanding
task

Walking from home
to an assessment
room
Normal pace

Straight walk (20 m
course) and turn
walk (20 m course
with a turn at 10 m)
Normal pace

TUG
Normal pace

TUG
Normal pace

Walking task

Dual task

Table 1 Summary of main characteristics of the studies included in the systematic review

First fall indoors
Difference between
single and dual task
‡4.5 s (DiffTUG manual)
6 months
13 (31.0)
First fall indoors Stop
walking Difference between
single and dual task ‡4.5 s
(DiffTUG manual)
6 months
33 (42.3)

First fall indoors Stops
walking
6 months
21 (36.2)

Recurrent falls (‡2) Walking
time (s)
12 months
37 (11.1)

‡1 falls Walking time (s)
12 months
25 (26.3)

Recurrent falls (‡2) Walking
time (s)
6 months
15 (50.0)

Outcomes
Follow-up period
Fall rate, n (%)

Yes
28/33
37/45
28/36
37/42

(84.9)e
(82.2)e
(77.8)e
(88.1)e

Yes

Yes

(47.6)
(94.6)

(83.3)
(76.1)

Yes

No

Yes

Were dual-task-related
changes associated
with falls?

7/13 (53.9)
26/29 (89.7)
7/10 (70.0)
26/32 (81.3)

10/21
35/37
10/12
35/46

Data not available

Similar results for
cognitive and
manual task:
12/15 (80.0)
14/15 (93.3)

12/13 (92.3)
14/17 (82.4)
Data not available

Sensitivity
Specificity
Positive predictive
value
Negative predictive
value
n/n (%)

788
O. Beauchet et al.

Ó 2009 The Author(s)
Journal compilation Ó 2009 EFNS European Journal of Neurology 16, 786–795


Ó 2009 The Author(s)
Journal compilation Ó 2009 EFNS European Journal of Neurology 16, 786–795

Community-dwelling
N = 311
>70 yearsd

Community-dwelling and senior
housing facilities
N = 380
85 yearsd

Main comorbiditiesb [Cognitive
impairment (n = 40),
Depressive symptoms (n = 80),
Pain in lower limb (n = 195)]
Inpatients with stroke
N = 63
68.4 ± 10.6 yearsa

Inpatients with stroke
N = 159
74 (33-94) years

Stalenhoef et al.,
2002 [16]

Bootsma-van
Der Wiel et al.,
2003 [18]

Andersson et al.
2006 [10]

Hyndman et al.,
2004 [17]

Community-dwelling
N = 60
79.6 ± 6.3 yearsa
Main comorbiditiesb [Lower limb
neuropathy (n = 12), Previous

stroke (n = 2)]

Participant

Verghese et al.,
2002 [9]

Table 1 (Continued)

Walking from the
assessment room to
the lounge area
(30-m)
Straight walk
TUG
Normal pace

Turn walk 12-m ([3-m,
turn and 3-m return]
x 2)
As quickly as possible

Straight walk (3-m)
Normal pace

3-m (1.5-m, turn and
1.5-m return)
Normal pace

Walking task

Dual task

Conversation
To carry a glass of
water

Conversation

Reciting animals or
profession names aloud

Solve two simple
calculations

Reciting concecutively
letters of the
alphabet aloud
(simple dual task) or
alternatively
(complex dual task)

Attention-demanding
task

‡1 falls
Stop walking
6 months
30 (47.6)
First fall indoors

Stop walking
TUG difference between
single and dual task
(diffTUG) ‡4.5 s
12 months
68 (42.8)

Recurrent falls (‡2)
Abnormal score at dual task
9 months
95 (30.5)
‡1 falls
Stop walking
12 months
158 (41.6)

First fall Score for simple
dual task ‡ 20 s
Score for complex dual
task ‡ 33 s
12 months
13 (22.0)

Outcomes
Follow-up period
Fall rate, n (%)

16/30 (53.3)
23/33 (69.7)
16/26 (61.5)

23/37 (62.2)
Walking with
conversation:
7/48 (14.6)
65/67 (97.0)
7/9 (77.8)
65/106 (61.3)
Diff TUG manual:
5/29 (17.2)
56/59 (94.9)
5/8 (62.5)
56/80 (70.0)

71/158 (44.9)
121/222 (54.5)
71/172 (41.3)
121/208 (58.2)

Simple dual task:
6/13 (46.2)
41/46 (89.1)
6/11(54.6)
41/48 (85.4)
Complex dual task:
5/13 (38.5)
44/46 (95.7)
5/7 (71.4)
44/52 (84.6)
Data not available

Sensitivity
Specificity
Positive predictive
value
Negative predictive
value
n/n (%)

No

Yes

No

Yes

No

Yes

Were dual-task-related
changes associated
with falls?

Dual task and prediction of falls

789


Senior housing facilities

N = 187
84.8 ± 5.2 yearsa

Inpatients
N = 57
85.0 ± 6.6 yearsa

Senior housing facilties
N = 213
84.4 ± 5.5 yearsa

Beauchet et al.,
2008b [19]

Kressig et al.,
2008 [12]

Beauchet et al.,
2008a [13]

Backward counting
aloud starting from
50

Straight walk (10-m)
Normal pace

Backward counting
aloud starting from
50

Backward counting
aloud starting from
50

Straight walk (10-m)
Normal pace

Straight walk (10-m)
Normal pace

Backward counting
aloud starting from
50

Attention-demanding
task

Straight walk (10-m)
Normal pace

Walking task

Dual task

First Fall
Increase backward counting
performance
12 months
54 (28.9)

First Fall
Increase walking time whilst
dual tasking >19.6 s
12 months
54 (28.9)
First Fall
Coefficient of variation of
stride time variability whilst
dual tasking >10%
50 days
10 (21.3)
Recurrent fall (‡2 falls)
Decrease in walking speed
whilst dual tasking under the
lowest tertile
12 months
57 (26.8)

Outcomes
Follow-up period
Fall rate, n (%)

Yes

Yes

12/72 (16.7)
133/141 (94.3)
12/20 (60.0)
133/193 (69.9)

Yes

Yes

Were dual-task-related
changes associated
with falls?

7/10 (70.0)
41/47 (87.2)
7/13 (53.9)
41/44 (93.2)

35/54 (64.8)
76/133 (57.1)
35/92 (38.0)
76/95 (80.0)

45/52 (86.5)
117/130 (90.0)
46/54 (85.2)
120/133 (90.2)

Sensitivity
Specificity
Positive predictive
value
Negative predictive
value

n/n (%)

M, Meter; TUG, Timed Up & Go; aMean ± standard deviation; bMain comorbidities reported; cMedian (range); dAge at inclusion; eCombined results of both dual tasks which were performed
separately.

Senior housing facilities
N = 187
84.8 ± 5.2 yearsa

Participant

Beauchet et al.,
2007 [11]

Table 1 (Continued)

790
O. Beauchet et al.

Ó 2009 The Author(s)
Journal compilation Ó 2009 EFNS European Journal of Neurology 16, 786–795


Dual task and prediction of falls

21.3% to 42.3% in those with prospective ones [1,7–
13,16–19]. Periods of fall collection ranged from 6 to
12 months [1,6–19]. Six studies examined frail older
subjects living in senior housing facilities [1,7,8,11,13,19].
Six other studies examined healthy community-dwelling

older adults [6,9,14–16,18] and three studies focused on
inpatients [10,12,17]. Ten studies explored association
between dual-task-related changes with the first fall event
[1,7–12,15,17,19] and four with recurrent falls
[6,13,14,16], defined as ‡2 falls during the follow-up
period. The type of attention-demanding task substantially differed amongst studies, whereas 11 of them used a
conversation [1,8,10,17], an enumeration of the alphabet
[9] or backward counting [6,11–13,15,19] that corresponded to a spoken verbal task; four studies used a
manual task [6–8,10], one a visuo-spatial decision task
[14] and one a combination of both attention-demanding
tasks [8]. Normal paced gait was measured in all studies
[1,6–19] except one study [18] in which subjects walked as
quickly as possible. A straight walkway was used in all,
but three studies in which a turn around and return route
was used [9,14,18].
The occurrence of falls was significantly associated
with dual-task-related performance in 11 studies [1,614;19]. The highest odds ratio for falling was 56.0 [6]
and the lowest was 1.34 [14]. Of the studies with prospective data collection of falls, six [1,7–9,12] reported
that dual-task-related gait changes predicted falls with
specificity ranging from 82.2% to 95.7% but with sensitivity between 38.5% and 84.9%. Furthermore, both
PPV and NPV were above 70% in all but two studies,
which reported a PPV under 55% [9,12]. Prediction of
falls was based on gait changes in five studies [1,7–
10,12,13,19] and on attention-demanding task changes
in one study [11]. This last study showed the highest
sensitivity, specificity, PPV, NPV and OR of all other
studies with prospective data collection of falls (adjusted OR = 53.3, P < 0.001). In addition, two studies challenged the usefulness of the dual-task paradigm
as a predictor of falls compared to single task performance [18,19]. Although an increase in walking time
whilst dual tasking was significantly associated with the
occurrence of falls (respectively, OR 2.5 [1.3;4.7] and

OR 3.3 [1.3;8.6]), the dual-task paradigm did not bring
any additional information compared to the ambulation time under single task, i.e. walking alone [18,19].
Three studies did not find that changes in gait whilst
performing a cognitive task might predict falls
[10,16,17]. Inpatients with stroke had changes in gait
during a conversation [17] or whilst carrying a glass of
water [10], yet these changes were not associated with
falls. The used dual-task score based on three levels
(normal, doubtful and abnormal) was not associated
with falls in StalenhoefÕs study [16].

791

Figure 2 shows the specific OR and the pooled OR
computed with meta-analysis technique. Two studies
were not included in this analysis because of lack of
data [15,16]. The pooled odds ratio for falling was 1.62
(95% CI, 0.96–2.72) for retrospectives studies and 6.84
(95% CI, 3.06–15.28) for prospective studies, when
subjects had changes in gait or attention-demanding
task performance whilst dual tasking. The pooled odds
ratio for falling when analysis included all studies was
5.3 (95% CI, 3.1–9.1).

Discussion
Divergent findings on dual-task-related gait changes,
their relationship with falls and their usefulness for fall
prediction raise a number of issues. They mainly concern the methodology, demographics and/or sample
recruitment (institutionalized versus community-dwelling older adults, differences in health status) and lack of
consensus concerning dual-task paradigms.

From a methodological point of view, no information about the number of subjects required to predict
falls was reported and studies did not include a post-hoc
power analysis. As a consequence, equivocal or negative results could be the result of a lack of power [11–
15]. Furthermore, follow-up periods ranged from 6 to
12 months. Such short follow-up periods may underestimate the number of falls and thus explain the failure
to establish a relationship with dual-task-related gait
changes. In addition, primary outcome measures varied. Some studies examined the first fall following the
dual-task assessment [1,6–10,13,14], whilst others
examined the predictive value for recurrent falls defined
as ‡2 falls during the follow-up period [6,13,14,16]. Two
out of four studies with equivocal or negative results
used the latter outcome measure. Because recurrent
falls are more often related to intrinsic (i.e. subject-related) risk factors for falling than isolated falls [4,5] and
dual-task-related gait changes are closely associated
with intrinsic risk factors for falling [6], an association
between recurrent falls and dual-task-related gait
changes could be postulated. Intrinsic risk factors for
falling are those related to various cumulated effects of
chronic diseases and physiologic decline [35]. These effects become even more pronounced and diverse with
age, contributing to a vicious cycle of increasing frailty
and increasing risk of falling. The discrepancy between
this hypothesis and the results could be related to an
insufficient number of studies (i.e. 2), which specifically
explored the predictive value for recurrent falls.
Lastly, study samples were very heterogeneous. They
included frail older adults living in senior housing
facilities [1,7,8,11,13] as well as healthy communitydwelling older adults [6,8,11,13–15] or inpatients

Ó 2009 The Author(s)
Journal compilation Ó 2009 EFNS European Journal of Neurology 16, 786–795

792

O. Beauchet et al.

Study ID

ES (95%CI)

% Weight

Studies with retrospective data collection of falls
Shumway-Cook et al., 2000

56.00 (5.13, 611.71)

3.28

Faulkner et al., 2007*

1.34 (1.04, 1.74)

8.84

Faulkner et al., 2007†

1.42 (1.08, 1.85)

8.82

Subtotal (I-squared = 78.4%, p = 0.010)

1.62 (0.96, 2.72)

20.94

Lundin-Olsson et al., 1997

15.90 (3.02, 83.88)

4.89

Lundin-Olsson et al., 1998

10.11 (2.00, 50.98)

5.01

Lundin-Olsson et al., 2000

25.90 (7.64, 87.77)

6.19

Verghese et al., 2002*

7.03 (1.68, 29.43)

5.54

Verghese et al., 2002†

13.75 (2.26, 83.56)

4.52

Bootsma-Van Der Wiel et al., 2003

0.98 (0.65, 1.47)

8.58

Hyndman et al., 2004

2.63 (0.94, 7.38)

6.81

Andersson et al., 2006*

5.55 (1.10, 28.02)

5.01

Andersson et al., 2006†

3.89 (0.86, 17.59)

5.32

Beauchet et al., 2007

53.08 (20.65, 136.43)

7.09

Beauchet et al., 2008b

2.46 (1.28, 4.73)

7.98

Kressig et al., 2008

15.94 (3.22, 79.05)

5.06

Beauchet et al., 2008a

3.33 (1.29, 8.56)

7.08

Subtotal (I-squared = 86.9%, P = 0.000)

6.84 (3.06, 15.28)

79.06

5.30 (3.08, 9.13)

100.00

Studies with prospective data collection of falls

.
Overall (I-squared = 87.9%, P = 0.000)
NOTE: Weights are from random effects analysis

*: Simple attention-demanding task
†: Complex attention-demanding task

0.1 0.25 0.5

1

2

4

10

25

50 100

Figure 2 Forest plot of the pooled estimated OR (ES) associated with the risk of falls according to the results from the dual-task procedure
computed from the natural logarithm transformation of the OR and 95% CI. I-squared (variation in OR attributable to heterogeneity).

[10,12,17]. No association between dual-task-related
gait changes and fall incidence was found in the three
studies that focused on community-dwelling older
adults (i.e. relatively healthy older adults) [15,16,18]. In
contrast, the highest predictive values for falls based on
dual-task-related gait changes were found in institutionalized populations (i.e. frail older adults) [1,7,8] and
geriatric inpatients [12]. Worsened gait performance
whilst dual tasking was associated with polymedication,
a well-known marker of comorbidities and frailty [36].
Therefore, it might be suggested that gait changes
whilst dual tasking could be a particularly significant
predictor of falls amongst frail older adults.
The lack of standardization in dual-task paradigms
certainly explains many of the discrepancies listed
above. A consensus needs to be reached on test conditions and data collection. In contrast to a non-standardized attention-demanding task like engaging in a
conversation, a well-defined and quantitative verbal
fluency task, such as reciting letters of the alphabet or

backward counting would improve validity, reliability,
consistency and comparison of results [1,8,9,18,19].
Few studies gave instructions to participants prior to
dual tasking. It has been shown that the dual-task
interference depended on the priority given to either
one of both tasks [2,3], based on the assumption that
attentional resources are limited [3]. Therefore, interference suggests an overload of the central resources
associated with an inability to appropriately allocate
attention between two simultaneously performed tasks
or prioritization [2,3,37,38]. Thus, gait changes whilst
dual tasking, both their nature and their range, might

represent the ability to develop a strategy for maintaining optimal movement in terms of attentional demand and efficiency of gait control. The attentiondemanding tasks were qualitatively and quantitatively
too heterogeneous, some authors using a relatively
simple motor task [6–8,10] and others various cognitive
tasks based on spoken verbal responses with varying
degrees of difficulty [1,7–14]. As a more complex cog-

Ó 2009 The Author(s)
Journal compilation Ó 2009 EFNS European Journal of Neurology 16, 786–795


Dual task and prediction of falls

nitive dual task was related to a better fall prediction
than using a simpler one [9], failure to predict falls
might greatly depend on the dual task and be related to
a ceiling effect of task used.
Walking conditions also varied in the studies. As it
has been well-established that a walk route including
walking straight ahead, turning around and returning
required more attention as opposed to walking straight
alone [4,39], such conditions could also influence dualtask performance [39].
The right choice of dual-task-related parameters
seems to be crucial for fall prediction. All, but one of
the studies [11], which explored the predictive value for
falling, were based on changes in gait whilst dual
tasking. Most studies used visual observation and recorded ambulation time and the number of steps or
stops [1,6–11,13–19], although some authors suggested
its poor inter-rater reliability [40]. Only one study [12]
used a spatio-temporal gait analysis system that allowed quantified and objective gait measurements,
showing that increased stride time variability whilst

counting backwards was a significant indicator of fall
risk. In addition, unlike most of the previous studies,
one study also evaluated changes in the attentiondemanding task performance whilst counting backwards out loud from 50 [11]. Subjects were then divided
into two categories, according to their backward
counting performance. Subjects who counted more
numbers whilst walking than whilst seated had a higher
risk of falling than those who counted fewer numbers
whilst walking. The authors postulated that backward
counting, being both an attention and a rhythmic task,
would act as a regulator, comparable to a metronome
in subjects with an irregular walking pattern and,
therefore, at highest risk of falling [41]. Improved
counting performances in this case might be explained
by a phenomenon of the two rhythms, counting and
walking, being drawn toward each other and would be
an indirect way to identify subjects with irregular stride.
Despite the use of heterogeneous dual-task conditions,
the pooled OR shows a statistically significant increase
in the risk of falls when subjects had changes in gait or
attention-demanding task performance whilst dual
tasking, this is particularly obvious in prospective
studies.

793

formance [11,12], give a new impulse to the
understanding of dual tasking and its potential for
developing new approaches in fall prevention. A better
analysis of changes in attention-demanding tasks, gait
variability, their relationship with cognitive functions

and the search for their anatomical and neurochemical
correlates constitute a fascinating challenge [42]. To
improve the predictive value of dual task-based fall risk
assessment tests, more prospective population studies
are needed with larger sample sizes, better identification
of confounding variables, such as health status and
standardization of the methodology, mainly for the
attention-demanding task. Such a consensus will certainly improve fall prediction but also response to
cognitive treatments such as cholinergic enhancers and
other treatments that may be available in the future.

Acknowledgment
None.

Author contributions
Beauchet has full access to the data in the study and
takes responsibility for the integrity of the data and the
accuracy of the data analyses. Beauchet and Herrmann
were responsible for study concept and design. Dubost
and Annweiler performed the acquisition of data.
Beauchet, Allali, Herrmann, Kressig, Bridenbaugh,
Assal, Dubost and Annweiler performed the analysis
and interpretation of data. Beauchet, Allali, Annweiler
and Dubost was responsible for drafting of the manuscript. Annweiler, Herrmann, Kressig, Assal and
Bridenbaugh made critical revision of the manuscript
for important intellectual content. Herrmann was
responsible for statistical expertise. Dubost and
Annweiler gave administrative, technical, or material
support. Beauchet and Herrmann did study supervision. Funding is not applicable.

Conflict of interest
The authors have no relevant financial interest in this
manuscript.

References
Conclusions
Despite conflicting early reports, changes in performance whilst dual-tasking were significantly associated
with an increased risk for falling amongst older adults
(pooled odds ratio = 5.3 with 95% CI = 3.1–9.1).
Recent data, which focused on stride-to-stride variability and changes in attention-demanding task per-

1. Lundin-Olsson L, Nyberg L, Gustafson Y. ‘‘Stops walking when talking’’ as a predictor of falls in elderly people.
Lancet 1997; 349: 617.
2. Abernethy B. Dual-task methodology and motor skills
reseasch: some applications and methodologies constraints. J Hum Mov Stud 1988; 14: 101–132.
3. Pashler H. Dual-task interference in simple tasks: data
and theory. Psychol Bull 1994; 116: 220–244.

Ó 2009 The Author(s)
Journal compilation Ó 2009 EFNS European Journal of Neurology 16, 786–795


794

O. Beauchet et al.

4. Woollacott M, Shumway-Cook A. Attention and the
control of posture and gait: a review of an emerging area
of research. Gait Posture 2002; 16: 1–14.
5. Bloem BR, Steijns JA, Smits-Engelsman BC. An update

on falls. Curr Opin Neurol 2003; 16: 15–26.
6. Shumway-Cook A, Brauer S, Woollacott M. Predicting
the probability for falls in community-dwelling older
adults using the Timed Up & Go Test. Phys Ther 2000; 80:
896–903.
7. Lundin-Olsson L, Nyberg L, Gustafson Y. Attention,
frailty, and falls: the effect of a manual task on basic
mobility. J Am Geriatr Soc 1998; 46: 758–761.
8. Lundin-Olsson L, Nyberg L, Gustafson Y. The
mobility interaction fall chart. Physiother Res Int 2000; 5:
190–201.
9. Verghese J, Buschke H, Viola L, et al. Validity of divided
attention tasks in predicting falls in older individuals: a
preliminary study. J Am Geriatr Soc 2002; 50: 1572–1576.
10. Andersson AG, Kamwendo K, Seiger A, Appelros P.
How to identify potential fallers in a stroke unit: validity
indexes of 4 test methods. J Rehabil Med 2006; 38: 186–
191.
11. Beauchet O, Dubost V, Allali G, Gonthier R, Herrmann
FR, Kressig WR. ‘‘Faster counting while walking’’ as a
predictor of falls in older adults. Age Ageing 2007; 36:
418–423.
12. Kressig RW, Herrmann FR, Michel JP, Beauchet O. Gait
variability while dual-tasking: fall predictor in older
inpatients? Aging Clin Exp Res 2008; 20: 123–130.
13. Beauchet O, Annweiler C, Allali G, Berrut G, Herrmann
FR, Dubost V. Recurrent falls and dual-task related
decrease in walking speed: is there a relationship? J Am
Geriatr Soc 2008a; 56: 1265–1269.
14. Faulkner KA, Redfern MS, Cauley JA, et al. Multitasking: association between poorer performance and a history of recurrent falls. J Am Geriatr Soc 2007; 55: 570–

576.
15. Vaillant J, Martigne´ P, Vuillerme N, et al. Prediction of
falls with performance on Timed ‘‘Up-and-Go’’ and oneleg-balance tests and additional cognitive tasks. Ann
Readapt Med Phys 2006; 49: 1–7.
16. Stalenhoef PA, Diederiks JP, Knottnerus JA, Kester AD,
Crebolder HF. A risk model for the prediction of recurrent falls in community-dwelling elderly: a prospective
cohort study. J Clin Epidemiol 2002; 55: 1088–1094.
17. Hyndman D, Ashburn A. Stops walking when talking
as a predictor of falls in people with stroke living in the
community. J Neurol Neurosurg Psychiatry 2004; 75:
994–997.
18. Bootsma-van der Wiel A, Gussekloo J, de Craen AJ, van
Exel E, Bloem BR, Westendorp RG. Walking and talking as predictors of falls in the general population:
the Leiden 85-Plus Study. J Am Geriatr Soc 2003; 51:
1466–1471.
19. Beauchet O, Allali G, Annweiler C, et al. Does change
in gait while counting backward predict the occurrence
of a first fall in older adults? Gerontology 2008b; 54:
217–223.
20. Von Elm E, Altman DG, Egger M, Pocock SJ, Gøtzsche
PC, Vandenbroucke JP. STROBE initiative. The
Strengthening the Reporting of Observational Studies in
Epidemiology [STROBE] statement: guidelines for
reporting observational studies. J Clin Epidemiol 2008; 61:
344–349.

21. Shumway-Cook A, Woollacott M, Kerns KA, Baldwin
M. The effects of two types of cognitive tasks on postural stability in older adults with and without a history
of falls. J Gerontol A Biol Sci Med Sci 1997; 52: M232–
M240.

22. Shumway-Cook A, Woollacott M. Attentional demands
and postural control: the effect of sensory context.
J Gerontol A Biol Sci Med Sci 2000; 55: M10–M16.
23. Ashburn A, Stack E, Pickering RM, Ward CD. A community-dwelling sample of people with ParkinsonÕs disease: characteristics of fallers and non-fallers. Age Ageing
2001; 30: 47–52.
24. Lajoie Y, Gallagher SP. Predicting falls within the elderly
community: comparison of postural sway, reaction time,
the Berg balance scale and the Activities-specific Balance
Confidence (ABC) scale for comparing fallers and nonfallers. Arch Gerontol Geriatr 2004; 38: 11–26.
25. Ashburn A, Stack E, Pickering RM, Ward CD. Predicting
fallers in a community-based sample of people with ParkinsonÕs disease. Gerontology 2001; 47: 277–281.
26. Bergland A, Wyller TB. Risk factors for serious fall related injury in elderly women living at home. Inj Prev
2004; 10: 308–313.
27. Bloem BR, Grimbergen YA, Cramer M, Valkenburg VV.
‘‘Stops walking when talking’’ does not predict falls in
ParkinsonÕs disease. Ann Neurol 2000; 48: 268.
28. Bloem BR, Grimbergen YA, Cramer M, Willemsen M,
Zwinderman AH. Prospective assessment of falls in ParkinsonÕs disease. J Neurol 2001; 248: 950–958.
29. Dibble LE, Lange M. Predicting falls in individuals with
Parkinson disease: a reconsideration of clinical balance
measures. J Neurol Phys Ther 2006; 30: 60–67.
30. Camicioli R, Licis L. Motor impairment predicts falls in
specialized Alzheimer care units. Alzheimer Dis Assoc
Disord 2004; 18: 214–218.
31. Hyndman D, Ashburn A, Yardley L, Stack E. Interference between balance, gait and cognitive task performance
among people with stroke living in the community. Disabil
Rehabil 2006; 28: 849–856.
32. Springer S, Giladi N, Peretz C, Yogev G, Simon ES,
Hausdorff JM. Dual-tasking effects on gait variability: the
role of aging, falls, and executive function. Mov Disord

2006; 21: 950–957.
33. Toulotte C, Thevenon A, Watelain E, Fabre C. Identification of healthy elderly fallers and non-fallers by gait
analysis under dual-task conditions. Clin Rehabil 2006; 20:
269–276.
34. Mackinnon A. A spreadsheet for the calculation of comprehensive statistics for the assessment of diagnostic tests
and inter-rater agreement. Comput Biol Med 2000; 30:
127–134.
/>updated May 9, 2000 (accessed 25/03/2009).
35. Fried LP, Bandeen-Roche K, Kasper JD, Guralnik JM.
Association of comorbidity with disability in older
women: the WomenÕs Health and Aging Study. J Clin
Epidemiol 1999; 52: 27–37.
36. Beauchet O, Dubost V, Herrmann F, Rabilloud M,
Gonthier R, Kressig RW. Relationship between dual-task
related gait changes and intrinsic risk factors for falls
among transitional frail older adults. Aging Clin Exp Res
2005; 17: 270–275.
37. Yogev-Seligmann G, Hausdorff JM, Giladi N. The role of
executive function and attention in gait. Mov Disord 2008;
23: 329–342.

Ó 2009 The Author(s)
Journal compilation Ó 2009 EFNS European Journal of Neurology 16, 786–795


Dual task and prediction of falls

38. Verghese J, Kuslansky G, Holtzer R, et al. Walking while
talking: effect of task prioritization in the elderly. Arch
Phys Med Rehabil 2007; 88: 50–53.

39. Kressig RW, Beauchet O, European GAITRite Network
Group. Guidelines for clinical applications of spatiotemporal gait analysis in older adults. Aging Clin Exp Res
2006; 18: 174–176.
40. Eastlack ME, Arvidson J, Snyder-Mackler L, Danoff JV,
McGarvey CL. Interrater reliability of videotaped obser-

795

vational gait-analysis assessments. Phys Ther 1991; 71:
465–472.
41. Ebersbach G, Dimitrijevic MR, Poewe W. Influence of
concurrent tasks on gait: a dual-task approach. Percept
Mot Skills 1995; 81: 107–113.
42. Rosano C, Brach J, Studenski S, Longstreth WT Jr,
Newman AB. Gait variability is associated with subclinical brain vascular abnormalities in high-functioning older
adults. Neuroepidemiology 2007; 29: 193–200.

Ó 2009 The Author(s)
Journal compilation Ó 2009 EFNS European Journal of Neurology 16, 786–795