Relationship of Balance and Mobility
to Fall Incidence in People With
Background and Purpose. People with stroke are at risk for falls. The
purpose of this study was to estimate the strength of the relationship of
balance and mobility to falls. Subjects. The participants were 99
community-dwelling people with chronic stroke. Methods. An inter-
view was used to record fall history, and physical performance assess-
ments were used to measure balance (Berg Balance Scale [BBS]) and
mobility (gait speed). Results. No differences were found between
subjects who fell once and subjects who did not fall or between subjects
who fell more than once and subjects who did not fall. Neither balance
nor mobility was able to explain falls in people with chronic stroke.
Discussion and Conclusion. Clinicians should be cautious when using
the BBS or gait speed to determine fall risk in this population. Falls
occurred frequently during walking; it may be necessary to focus on
reactive balance and environmental interaction when assessing indi-
viduals for risk of falls and devising fall prevention programs for
individuals with chronic stroke. The authors’ observations suggest that
the prescription of 4-wheel walkers for individuals with a low BBS score
(?45) may be a mobility aid that could reduce the risk of falls. [Harris
JE, Eng JJ, Marigold DS, et al. Relationship of balance and mobility to
fall incidence in people with chronic stroke. Phys Ther. 2005;85:150–
Key Words: Accidental falls, Balance, Cerebrovascular accident, Rehabilitation.
Jocelyn E Harris, Janice J Eng, Daniel S Marigold, Craig D Tokuno, Cheryl L Louis
150Physical Therapy . Volume 85 . Number 2 . February 2005
factors such as impaired balance,5–10declining cogni-
tion,3,10and presence of neurological disease.2,3,5,6,10,11
Stroke is considered one of the greatest risk factors for
falls among elderly people.12,13The majority of individ-
uals with stroke will have some degree of residual
impairment, but will regain walking ability and will be
discharged home following hospitalization.14Although
residual impairment is common, most people with
stroke will regain walking ability; however, poor balance
and impaired gait can persist.14
alls are the leading cause of injury-related deaths
among elderly people in North America.1For
elderly community-dwelling individuals, falls are
common2–4and can be attributed to multiple
Fall incidence rates between 23% and 50% have been
reported in studies of people with chronic stroke (?6
months poststroke).12,15–17This rate is much higher than
rates reported for older community-dwelling adults with-
out stroke (11%–30%)3,18,19but lower than rates for
people with subacute stroke (1–6 months poststroke)
(25%).20Injury is a frequent consequence of falls in
people with chronic stroke, with up to 28% reporting
an injury.15Over half of all reported falls occurred
indoors during walking activities.12,16Studies have shown
ment,13,20,23,25,26and impaired balance20,22,23are related
to fall incidence in people with acute stroke (up to 1
month poststroke) and subacute stroke. However, in
people with chronic stroke, factors that relate to falls and
fall risk are not as clear. To date, only a few studies12,15–17
have examined falls in people with chronic stroke.
Measures of balance,12,15,16motor status,12,15,16cognition/
mood,12,15–17strength (defined as muscle force),12,15
vision,12,15and activities of daily living (ADL)12,15–17have
been used to predict fall risk in people with chronic stroke.
These studies showed that only 3 of the factors—
ADL15–17—increased fall risk in people with chronic stroke.
In studies that examined balance, the results were
mixed, with Jorgensen et al12finding that balance was
not a risk factor for falls, whereas Lamb and colleagues15
and Hyndman and Ashburn17found that balance was a
risk factor. Even in these 2 studies,15,17the question of
balance as a risk factor was not clear. Lamb et al15used
measures of self-reported balance difficulties (eg, while
dressing) and performance measures (ie, people were
asked to stand with feet side by side, semitandem, and
tandem); only the self-reported measures predicted falls.
Hyndman and Ashburn17used the Berg Balance Scale
(BBS) and found a difference in scores between people
who fell more than once and people who did not fall, but
they found no difference in scores between people who
did not fall and people who fell once.
Only 2 studies15,16examined mobility (eg, gait speed,
ability to move in bed, transferring from one surface to
another) as a potential risk factor for falls in people with
chronic stroke. Lamb et al15found that gait speed and
transfer ability were related to falls in a bivariate analysis
but that only transfer ability was related to falls in a
multivariate analysis. Hyndman et al16found that mobil-
ity was not a factor in fall events. One of the reasons that
Lamb et al15found such a relationship but Hyndman et
al16did not may have been the method in which mobility
was measured. Lamb et al15used a measure of walking
(ie, gait speed) along with specific one-item functional
JE Harris, OT, MSc, is a graduate student in the School of Rehabilitation Sciences, University of British Columbia, Vancouver, British Columbia,
Canada, and in the Rehabilitation Research Laboratory, GF Strong Rehab Centre, Vancouver, British Columbia, Canada.
JJ Eng, PT/OT, PhD, is Associate Professor, School of Rehabilitation Sciences, University of British Columbia, T325-2211 Wesbrook Mall,
Vancouver, British Columbia, Canada V6T 2B5 (firstname.lastname@example.org), and Scientist, Rehabilitation Research Laboratory, GF Strong Rehab
Centre. Address all correspondence to Dr Eng.
DS Marigold is a graduate student in the Department of Neuroscience, University of British Columbia, and in the Rehabilitation Research
Laboratory, GF Strong Rehab Centre.
CD Tokuno, MSc, is Research Coordinator, Rehabilitation Research Laboratory, GF Strong Rehab Centre.
CL Louis is a student in the School of Rehabilitation Sciences, University of British Columbia, and a research assistant in the Rehabilitation
Research Laboratory, GF Strong Rehab Centre.
Ms Harris, Dr Eng, and Mr Marigold provided concept/idea/research design and writing. Mr Tokuno and Ms Louis provided data collection, and
Ms Harris, Mr Tokuno, and Ms Louis provided data analysis. Dr Eng provided project management.
This research was presented at the School of Rehabilitation Sciences, University of British Columbia, Research Symposium, May 11, 2004.
This study was supported by an operating grant (MOP-57862) from the Canadian Institute of Health Research to Dr Eng and by a career scientist
award to Dr Eng from the Canadian Institute of Health Research and the Michael Smith Foundation for Health Research.
This article was received April 25, 2004, and was accepted July 22, 2004.
Physical Therapy . Volume 85 . Number 2 . February 2005Harris et al . 151
tests such as rising from a chair, transfers, and bed
mobility, whereas Hyndman et al16used only one func-
tional measure (Rivermead Mobility Index). Taken sep-
arately, these one-item measures are related to fall
incidence, but once they are combined with other
variables that may be related to falls (eg, depression,
balance, ADL ability), their impact is diminished or
negated. This may suggest the importance of comprehen-
sive assessments that focus on the physical, cognitive/
psychological, and environmental factors.
Although cross-sectional studies have shown that stroke
and balance impairment are risk factors for falls and
fractures in older adults,2,27the evidence is not conclu-
sive as to whether balance or mobility are risk factors for
falls when the sample is limited to only individuals with
chronic stroke. The purpose of this study was to quantify
the relationship of balance (as measured by the BBS)
and mobility (gait speed) to fall incidence in a
community-dwelling sample of individuals with chronic
Gait difficulties have been shown to be associated with
falls in older adults.2,5,6,8Two of the studies involving
individuals with chronic stroke15,16examined mobility as
a potential risk factor for falls. Lamb et al15used a
3-point ordinal scale (1–3) to indicate walking ability
and may have lost important information (ie, variability)
by doing so. Only the rating of 1 (?0.25 m/s) was
defined in the article. We assumed that the distribution
was separated into thirds to indicate the ratings of 2 and
3. This study also was limited to female subjects only.
Because only 2 studies have used mobility as a depen-
dent variable, with mixed results, the role that mobility
might play in falls remains unclear. Several authors12,16,17
also contended that walking ability would be a factor in
falls because the majority of falls reported occurred
during walking. Gait speed is highly correlated with
measurements of mobility28and has been shown to be a
discriminate measure for studying locomotor recovery
because it is sensitive enough to reflect physiological and
Subjects and Design
One hundred eight people with stroke were recruited on
a voluntary basis from the community. Flyers were
posted in local senior centers and community centers,
and advertisements were placed in the local newspapers.
The inclusion criteria were: (1) over 50 years of age,
(2) first stroke, (3) at least 1 year poststroke, and
(4) reported ability to walk 8 m (with assistive device, if
required). This information was gained during a screen-
ing interview conducted over the telephone. The diag-
nosis and other inclusion criteria were confirmed by a
family physician and a therapist’s observation of motor
status. People with major musculoskeletal problems
(eg, amputation or recent joint replacement surgery) or
neurological disorders in addition to stroke were
excluded from the study. Nine participants were
excluded because their family physician confirmed addi-
tional neurological disorders (n?2), absence of stroke
(n?2), or more than one occurrence of stroke (n?5),
leaving 99 participants. Participant characteristics are
described in Table 1.
Sample size was calculated to determine the sample size
per group (no falls, 1 fall, more than 1 fall) required for
the outcome measures of BBS and gait speed. We used
data from a previous study,30which had a cohort of
subjects with a group mean of 45.3 for the BBS (SD?5.3)
and a mean gait speed of 0.68 m/s (SD?0.28). Using the
sample size calculation of Portney and Watkins,31an
80% power rate, an alpha of .05, and a group difference
of 3 points for the BBS and a 10% difference for gait
Participant Characteristics (N?99) and Mann-Whitney Test Comparisons Between Those Who (1) Did Not Fall and Fell Once and (2) Did Not
Fall and Fell More Than Oncea
Affected side (L/R/N)
Time since stroke (y)
Gait self-selected (m/s)
Gait fast paced (m/s)
69.09.550–9367.672.4 .0668.0 .86
aM/F?male/female, L/R/N?left/right/unable to detect a weaker side, BBS?Berg Balance Scale (0–56), MMSE?Mini-Mental Status Exam (0–30).
152 . Harris et alPhysical Therapy . Volume 85 . Number 2 . February 2005
speed, a sample of 50 individuals per group was
required. We were able to achieve an adequate sample
size for people with no falls (n?50), but not for people
with falls (n?49). Berg Balance Scale changes of at least
2 points have been shown to result in clinically mean-
The Functional Disability Scale of the American Heart
Association Stroke Outcome Classification (AHASOC)32
is a scale designed to measure residual impairment and
disability of stroke in the areas of basic activities of daily
living (BADL) and instrumental activities of daily living
(IADL). The scale consists of 5 levels (1–5), with 1
indicating independence in both BADL and IADL and 5
indicating complete dependence. The median score for
our sample was 2.0. Over half of our sample were
classified as either independent in all BADL and IADL
or independent in all BADL with partial assistance
required for IADL. Self-report information on comor-
bidity was collected with a checklist of 21 conditions
(eg, arthritis, congestive heart failure, depression). More
than 50% of our sample indicated more than 2 comor-
bidities. The most frequently cited comorbidities were
hypertension, a history of cardiac events, and a sedentary
In accordance with university and hospital policies,
informed consent was obtained from all participants
prior to their participation in the study. Ethics approval
was obtained from the local university and hospital
review boards. Participants took part in a 1-hour evalua-
tion session that consisted of a semistructured interview
to obtain a falls history and a clinical examination. All
testing was conducted at the rehabilitation research
laboratory located in a rehabilitation hospital. The
examiner was an occupational therapist ( JEH) who had
9 years of clinical experience in the area of neurology
and who had used the BBS and the Mini-Mental Status
Exam (MMSE) in clinical and research settings.
each participant from the interview: age, sex, time since
stroke, side of paresis, number of falls recalled over the
past 6 months, whether the individual sought medical
attention for any of the falls, and whether an injury
resulted from any of the falls. Data for fall history were
reported by participant
informed that a fall was defined as coming to rest on the
floor or another lower level but was not due to seizure,
stroke or myocardial infarction, or an overwhelming
displacing force (eg, earthquake).16Five individuals had
someone else confirm their fall history. Fall information
was classified into location of fall (indoor, outdoor) and
The following information was recorded for
activity (eg, walking, transferring) in which the individ-
ual was involved when he or she fell.
was examined using clinical measures such as the Rom-
berg test, Tinetti Fall Efficacy Scale, and Tinetti
Performance-Oriented Mobility Assessment. Similarly, in
the acute and subacute stages of stroke recovery,
impaired balance, assessed by clinical measures, has
been identified as a risk factor for falls. The BBS was
used in 2 of these studies.20,23All of the cited studies of
patients with chronic stroke and falls used a clinical tool
to measure balance impairment, except for the study by
Hyndman et al,16who did not measure balance. Of the
3 studies that involved balance, the study by Lamb et al15
showed that impaired balance was a predictor of falls in
older women with stroke, and the study by Hyndman
and Ashburn17showed that BBS scores of people who
fell were lower than scores of those who did not fall.
These findings suggest that clinical measures of balance,
including the BBS, are appropriate and sensitive mea-
sures to use in studies of falls in people with stroke.
In the studies cited for older adults, balance
Balance was measured using the BBS.33The BBS is a
14-item test (56 points maximum) using a 5-point (0–4)
scale to rate each item, with 0 indicating an inability or
need for maximal assistance to complete the task or
performs task with safety concerns and 4 indicating
independent and safe ability to perform task. The BBS
consists of tasks such as reaching, balancing on one limb,
and transferring. Concurrent validity of data for the BBS
has been examined in people with stroke. Correlations
with data for the Barthel Index (r ?.80), the Fugl-Meyer
Motor Impairment Scale (r ?.62–.94), measures of pos-
tural sway (r ?.55),33,34and gait speed (r ?.60)29have
been found. The BBS has been shown to yield data with
high interrater and intrarater reliability in elderly people
when using physical therapists as testers. Initial studies
by Berg and colleagues33,35produced intraclass correla-
tion coefficients (ICCs) of .98 for interrater reliability
and .71 to .99 for intrarater reliability in elderly people.
associated with falls in older adults.2,5,6,8Two of the
studies involving individuals with chronic stroke15,16
examined mobility as a potential risk factor for falls.
Gait difficulties have been shown to be
For the assessment of gait speed, participants were asked
to walk at their “most comfortable speed” (ie, self-
selected pace) using their usual assistive device along an
8-m distance 3 times and then to walk “as fast as possible
but safely” (ie, maximum pace) 3 times. The mean of the
3 trials (in meters per second) was calculated. Partici-
pants walked in their own shoes and used an orthosis
(n?8) (eg, ankle-foot orthosis) or an assistive device
(cane [n?34], walker [n?10]) if required. Infrared-
Physical Therapy . Volume 85 . Number 2 . February 2005Harris et al . 153
emitting diodes (IREDs) were attached to the distal
portion of the dorsal aspect of the participants’ foot and
the distal aspect of the Achilles tendon. An opto-
electronic sensor* was used to track the IREDs. Gait
speed was calculated using the distance covered by the
IREDs and the corresponding elapsed time during each
gait cycle. We have previously evaluated the test-retest
reliability (separate days) for 22 individuals with stroke
and found an ICC of .95 for self-paced gait speed.36
used to detect cognitive deficits in orientation, learning,
calculation, abstraction, memory, language, and spatial
relationships.37Each item is given a score of 1 (able to
fully complete the task) or 0 (unable to complete the
task), with a total possible score of 30. A score of below
24 is typically used to describe people who could be
experiencing cognitive deficits that would interfere with
daily living. The MMSE is a tool that is widely used
clinically and can be administered in 5 to 10 minutes.
Dick et al38examined the MMSE for test-retest reliability
in people with neurological conditions and reported an
ICC of .95, and construct validity was determined with a
correlation of r ?.64 with the Wechsler Memory Scale.
Scores for the MMSE have been associated with scores
for cognitive subscale of the Functional Independence
Measure (r ?.67) and the Loewenstein Occupational
Therapy Cognitive Assessment (r ?.59).39
The MMSE is a screening tool that can be
Descriptive statistics were used to describe the sample
and the content of the semistructured interview. Based
on fall history, participants were categorized as having 1
fall, repeated falls (?2 falls), or no falls.
The Mann-Whitney U test was used to detect mean
differences between groups for the continuous variables
of age, gait speed, MMSE score, and BBS score because
these variable distributions were not normal. The chi-
square test was used to detect proportional differences
for the dichotomous variables of sex and side of paresis.
Bivariate correlations were produced using the Pearson
product moment correlation to measure the strength of
the association among continuous variables and to
determine variable entry into the model; redundant
(ie, highly correlated) variables were removed. Binary
logistic regression was used to determine potential risk
factors for falls, with 95% confidence intervals calculated
for each of the independent variables entered into the
model. This type of regression is considered appropriate
for use when there is a combination of continuous and
categorical predictor variables.40Based on the results of
similar studies of people with chronic stroke12,15–17and
studies involving elderly people,2,5–10,11the model was
determined by blocked entry of all variables of interest
(BBS score, self-selected gait speed, MMSE score), con-
trolling for age and sex by first entry. Our statistics and
model did not include the number of falls (continuous
variable) per person because we had a small number of
individuals with large numbers of falls (eg, 10), which
would tend to overinflate the correlations. We used SPSS
statistical software 11.0 for Windows†in the analysis. A
value of P?.05 was considered significant in all compar-
isons. All statistical testing was 2-sided. Variable entry for
the regression was set at .05, and removal was set at .10.
A total of 117 falls were recorded. Forty-nine participants
(50%) experienced at least one fall over the past 6
months (Tab. 2). Falls that occurred indoors accounted
for 56% of the 117 falls, with 62% of the indoor falls
transpiring within the home. Falls that occurred outside
totaled 39%. We were unable to classify location for 7
(5%) of the reported falls. The most frequent activity at
the time of the fall was walking (51%). Further details
regarding fall circumstance are shown in Table 2. Of the
49 participants who reported a fall, 20 (41%) reported
an injury, with 17 (85%) of those seeking medical
attention. Women appeared to be more likely than men
* Northern Digital, 103 Randall Dr, Waterloo, Ontario, Canada N2V 1C3.
†SPSS Inc, 233 Wacker Dr, Chicago, IL 60606.
Fall Incidence, Location, Circumstance, and Injury by Faller
Fall incidence and injury
No injury reported
aPercentage derived by total number of participants who fell (n?49).
bPercentage derived by total number of falls (n?21) by participants who fell
cPercentage derived by total number of falls (n?96) by participants who fell
more than once.
154 . Harris et alPhysical Therapy . Volume 85 . Number 2 . February 2005
to be injured by a fall (?2?3.6, P?.06, likelihood
ratio?3.5) in the group of subjects who fell once.
Age, BBS score, gait speed, and MMSE score did not
differ between people who fell once and those who had
not fallen (Tab. 1). Chi-square tests showed no group
differences for sex or side of paresis between those who
did not fall and those who fell once. Of the 9 individuals
with MMSE scores below 20, 4 had fallen once and 5 had
not fallen. These individuals were not more likely to
fall compared with individuals with an MMSE score
There was no group difference on variables of age, BBS
score, gait speed, and MMSE score between subjects who
fell more than once and those who had not fallen
(Tab. 1). Chi-square tests showed no group differences
for sex or side of paresis between subjects who fell more
than once and those who had not fallen.
Significant bivariate correlations were found between
BBS score and MMSE score (r ?.24, P?.05), BBS score
and self-selected gait speed (r ?.74, P?.01), BBS score
and fast-paced gait speed (r ?.70, P?.01), and self-
selected gait speed and fast-paced gait speed (r ?.90,
P?.01). Due to the strong relationship between self-
selected gait speed and fast-paced gait speed, only self-
selected gait speed was included in the regression
model. A scatterplot of BBS scores against number of
falls illustrates the relationship between the 2 variables
When all 5 variables (age, sex, BBS score, gait speed, and
MMSE score) were block entered into the regression, no
significant model was produced for participants who fell
(5)?5.20, P?.39) or those who fell more than
(5)?1.90, P?.86) (Tab. 3).
Discussion and Conclusions
The purpose of this study was to determine whether
balance and gait speed could explain falls in individuals
with chronic stroke. The circumstances and characteris-
tics of falls in this sample also were described. We found
that falls are a common occurrence in older adults with
chronic stroke. One half of our sample reported at least
one fall in the past 6 months. In addition, fall-related
injuries were common, although serious injury was less
frequently reported in our sample. We found that more
than one half of falls occurred in the home during
In comparison with older adults (age ?65 years) living
in the community, our sample had a much higher rate of
fall incidence. Studies of older adults3,18,19have shown
6-month fall rates of 20% to 35%, whereas our sample
showed a 6-month fall rate of 50%. These studies3,18,19
examined falls in older adults, and individuals with
stroke accounted for less than 15% of the samples.
However, in studies where history of stroke was included
in the analyses, stroke was identified as a significant
correlate of fall incidence.2,3,5,6,10,11In addition, investi-
gators in many studies of falls in community-dwelling
older adults cited impaired balance5–10and cognition3,10
as factors in fall incidence and risk; these factors can be
residual impairments of stroke. These factors may
explain why our participants’ fall incidence rate was
higher than in general samples of older adults. The
percentage of our participants who fell and sought
medical attention (88%) also was higher than in Blake’s
study41of the epidemiology of falls in community-living
older adults, whose rate of fall incidence was 5%.
Our measures of balance and gait speed were not able to
explain falls in this population or to discriminate
between those who had not fallen, had fallen once, or
had fallen more than once. Our inability to find differ-
ences between participants who had fallen and those
who had not fallen may have been due to insufficient
power. We calculated needing a sample size of 50 people
in each group, and although we were able to achieve an
adequate sample size for participants who had not fallen,
we did not achieve an adequate sample size for those
who had fallen (n?49). Studies involving elderly people
without stroke have shown that age,6,9,11sex,5,11bal-
ance,6–10ability to perform ADL,2,5,6and cognitive
scores3,10are predictors of fall risk. Several predictors of
falls also are known for individuals in the acute and
subacute phases of stroke recovery, but determining fall
risk in the chronic stage remains difficult. Stroke can
affect different functions (motor, sensory, cognitive)
within an individual independently or in combination,
leaving people with different severity levels of residual
impairment and varying compensatory strategies. With
this level of outcome diversity, an individualized
Scatterplot of Berg Balance Scale (BBS) scores against number of falls
for all subjects.
Physical Therapy . Volume 85 . Number 2 . February 2005Harris et al . 155
approach to fall risk may account for impairment vari-
ability and be more effective in fall prevention than
group prediction models.
We hypothesized that BBS scores would be associated
with fall incidence and be a risk factor for falls in people
with chronic stroke. Our results did not support this
hypothesis. Given the surprising results from our
planned analyses, we examined our data further to
determine possible reasons for these findings. We noted
that all of the subjects who had a low BBS score (cutoff
score of ?45 suggested by the developers of the scale34
for fall risk) and either no falls or one fall were partici-
pants who used a wheelchair (n?9). All of our subjects
were ambulatory, but these 9 individuals described using
a wheelchair for outdoor mobility or for long distances,
and walking mainly in the home or for short distances. A
post hoc analysis revealed that when the data for these 9
subjects were excluded from the data set, a low relation-
ship between the BBS scores and falls was apparent
(r ?.37, P?.01). Further analysis showed that subjects
who used a 4-wheel walker (n?11) also had low fall
incidence (?1); however, of those who used a cane, 13
(37%) had ?2 falls, with some having as many as 10 falls.
Subjects who used a cane and had ?2 falls scored ?45
on the BBS. Thus, those individuals who had a low BBS
score (indicating risk for falling of ?45) but who used
either a walker or a wheelchair did not appear to be at
risk for falling. We believe that it would be useful in the
future to assess the impact of mobility aids on fall risk
because our post hoc observations suggest that the pre-
scription of mobility aids, especially 4-wheel walkers, may
reduce fall risk in individuals with a BBS score of ?45.
Clinical measures of function as predictors of falls in
people with chronic stroke remain elusive. Factors such
as vestibular function, sensation, perception, and home
environment have not been assessed in this population
and may add important information for fall risk. Berg
et al33defined 3 components of balance: static (mainte-
nance of posture), dynamic (adjustment to voluntary
movement), and reactive (reaction to external forces).
The majority of falls were reported during walking, an
activity that requires dynamic and reactive balance. The
measure we used to evaluate balance tests static balance
(eg, standing, sitting unsupported), with some activities
requiring dynamic balance (eg, turning 360°, transfer-
ring, placing alternate feet on stool) and to a lesser
degree reactive balance. It may be that the BBS does not
test the domains of balance required to prevent or
successfully recover from a fall, which may indicate that
a more sensitive measure of balance, including large
components of reactive and dynamic balance, is neces-
sary. Researchers may need to test situations in which
people are required to elicit reactive balance in response
to externally imposed perturbations such as recovering
from a push to the body, platform perturbation, or a
tripping paradigm (a tripping paradigm is constructed
using obstacle placement or the sudden introduction of
an obstacle that could produce tripping if the person is
unable to negotiate the obstacle). Furthermore, the
interaction of people within their environment, such as
approaching a curb and stepping up over it, may be
Our participants were a volunteer, community-based
sample, which could bias results because they may be
more mobile and cognitively intact than people living in
an institution; however, a wide range of balance and
mobility impairment was evident from the data. We also
relied on participant recall for fall history. Some
authors42,43have suggested that this method of informa-
tion gathering can produce recall bias and negatively
affect results. We attempted to control for recall bias by
having a caregiver or spouse present, if necessary, for
The clinical information (balance, mobility, and cogni-
tive status) collected at the time of the study may have
been different than at the time of the fall. During the
6-month time frame used for fall history (March–
August), participants could have experienced an illness
(eg, flu, cold) or an exacerbation or worsening of an
existing condition (eg, arthritis, dementia) that could
Logistical Regression: Risk Factors for Participants Who (1) Fell Once and (2) Fell More Than Oncea
One FallMore Than One Fall
P Odds Ratio95% CIPOdds Ratio 95% CI
aCI?confidence interval, BBS?Berg Balance Scale, MMSE?Mini-Mental Status Exam.
156 . Harris et alPhysical Therapy . Volume 85 . Number 2 . February 2005
have negatively influenced their functional status at the
time of the fall(s). However, at the time of examination,
these issues may have been controlled or resolved. In
contrast, participants might have been in worse physical
or mental condition at the time of the examination than
at the time of the fall(s). Activity level also may have
fluctuated during the 6-month period, because some
individuals might be more prone to indoor activities
during the winter months and more active outside
during the summer months.
We did not use a measure of activity level in our study.
Our sample may have been sedentary given the residual
effects of stroke. However, our measure of functional
status showed that participants were independent in
BADL and most IADL and were ambulatory. We believe
that future studies should use a measure of activity
level, as well as measures of vision, sensation, and
The results of our study suggest that people with chronic
stroke are a subgroup of older adults who are at risk for
falls. Because neither the BBS score nor gait speed were
able to explain falls, clinicians may be left wondering
how to assess fall risk. The majority of falls in our study
and in other studies12,16of stroke were reported during
walking, an activity that requires dynamic and reactive
balance. We recommend that clinicians examine indi-
viduals in situations requiring reactive balance, such as
recovering from a push to the body, a tripping para-
digm, or negotiating a curb. Examining individuals in
their community and home environments also may
provide valuable insight into potential hazards or diffi-
culties and may provide physical therapists with informa-
tion about the need for mobility aids in different
environments (eg, home, long distances, shopping,
negotiation of different terrain). In addition, clinicians
should not rely on only one assessment but rather a
battery of assessments that include physical, mental, and
environmental factors in attempting to examine which
individuals may fall and to prevent future falls.
The issue of mobility aids may be of particular impor-
tance in people with chronic stroke. Our observations
suggest that the prescription of 4-wheel walkers for
people with a low BBS score (?45) may reduce the risk
of falls. Of particular interest were individuals who used
a cane and fell at high rates. It is possible that these
individuals were assessed at discharge for a cane but have
re-evaluation of mobility status. Further research regard-
ing mobility aids and their association with falls and fall
risk would be of importance to clinicians in the area of
might benefit froma
1 National Trauma Registry 2002 Report: Hospital Injury Admissions.
Ottawa, Ontario, Canada: Canadian Institute for Health Information;
2 Campbell AJ, Borrie M, Spears GF. Risk factors for falls in a
community-based prospective study of people 70 years and older.
J Gerontology. 1989;44:112–117.
3 Graafmans WC, Ooms ME, Hofstee HMA, et al. Falls in the elderly: a
prospective study of risk factors and risk profiles. Am J Epidemiol.
4 Tinetti ME, Williams CS. Falls, injuries due to falls, and the risk of
admission to a nursing home. N Engl J Med. 1997;337:1279–1284.
5 Lipsitz LA, Jonsson PV, Kelley MM, Koestner JS. Causes and corre-
lates of recurrent falls in ambulatory frail elderly. J Gerontology. 1991;
6 Lord SR. Instability and falls in elderly people. In: Lafont C, Baroni
A, Allard M, et al. Falls, Gait, and Balance Disorders in the Elderly. New
York, NY: Springer Publishing Co; 1996:125–139.
7 Rigler SK. Instability in the older adult. Comp Ther. 1996;22:297–303.
8 Maki BE, McIlroy WE. Postural control in the older adult. Clin Geriatr
9 Ousset PJ, Henna AO, Faisant C, et al. A longitudinal study of falls,
gait, and balance disorders in 267 healthy elderly persons living in the
community. In: Lafont C, Baroni A, Allard M, et al. Falls, Gait, and
Balance Disorders in the Elderly. New York, NY: Springer Publishing Co;
10 Salgado R, Lord SR, Packer J, Ehrlich F. Factors associated with
falling in elderly hospital patients. Gerontology. 1994;40:325–331.
11 Yasumura S, Haga H, Nagai H, et al. Rate of falls and the correlates
among elderly people living in an urban community in Japan. Age
12 Jorgensen L, Engstad T, Jacobsen BK. Higher incidence of falls in
long-term stroke survivors than in population controls: depressive
symptoms predict falls after stroke. Stroke. 2002;33:542–547.
13 Mayo NE, Korner-Bitensky N, Kaizer F. Relationship between
response time and falls among stroke patients undergoing physical
rehabilitation. Int J Rehabil Res. 1990;13:47–55.
14 Jorgensen HS, Nakayama H, Raaschou HO, Olsen TS. Stroke:
neurologic and functional recovery—the Cogenhagen Stroke Study.
Phys Med Rehabil Clin N Am. 1999;10:887–906.
15 Lamb SE, Ferrucci L, Volapto S, et al. Risk factors for falling in
home-dwelling older women with stroke: the women’s health and
aging study. Stroke. 2003;34:494–501.
16 Hyndman D, Ashburn A, Stack E. Fall events among people with
stroke living in the community: circumstances of falls and characteris-
tics of fallers. Arch Phys Med Rehabil. 2002;83:165–170.
17 Hyndman D, Ashburn A. People with stroke living in the commu-
nity: attention deficits, balance, ADL ability, and falls. Disabil Rehabil.
18 Bogle Thorbahn LD, Newton RA, Chandler J. Use of the Berg
balance test to predict falls in elderly persons. Phys Ther. 1996;76:
19 Tinetti ME, Speechley M, Ginter SF. Risk factors for falls among
elderly persons living in the community. N Engl J Med. 1988;319:
20 Stapleton T, Ashburn A, Stack E. A pilot study of attention deficits,
balance control, and falls in the subacute stage following stroke. Clin
Physical Therapy . Volume 85 . Number 2 . February 2005Harris et al . 157
21 Byers V, Arrington ME, Finstuen K. Predictive risk factors associated
with stroke patient falls in acute care settings. J Neurosci Nurs. 1990;22:
22 Nyberg L, Gustafson Y. Patient falls in stroke rehabilitation: a
challenge to rehabilitation strategies. Stroke. 1995;26:838–842.
23 Teasall R, McRae M, Foley N, Bhardwaj A. The incidence and
consequences of falls in stroke patients during inpatient rehabilitation:
factors associated with high risk. Arch Phys Med Rehabil. 2002;83:
24 Tutuarima JA, van der Meulen JHP, de Haan RJ, et al. Risk factors
for falls of hospitalized stroke patients. Stroke. 1997;28:297–301.
25 Forster A, Young J. Incidence and consequences of falls due to
stroke: a systematic inquiry. Br Med J. 1995;311:83–86.
26 Sze K, Wong E, Leung HY, Woo J. Falls among Chinese stroke
patients during rehabilitation. Arch Phys Med Rehabil. 2001;82:
27 Ramnemark A, Nyberg L, Borssen B, et al. Fractures after stroke.
Osteoporos Int. 1998;8:92–95.
28 Olney SJ, Elkin ND, Lowe PJ, Symington DO. An ambulation profile
for clinical gait evaluation. Physiother Can. 1979:31:85–90.
29 Richards CL, Malouin F, Dumas F, Tardif D. Gait velocity as an
outcome measure of locomotor recovery after stroke. In: Craik RL,
Oatis CA, eds. Gait Analysis: Theory and Applications. St Louis, Mo: CV
Mosby Inc; 1995:355–364.
30 Eng JJ, Chu KS, Kim CM, et al. A community-based group exercise
program for persons with chronic stroke. Med Sci Sport Exerc. 2003;35:
31 Portney LG, Watkins MP. Foundations of Clinical Research. 2nd ed.
Upper Saddle River, NJ: Prentice Hall Health; 2000.
32 American Heart Association Classification of Stroke Outcome Task
Force. Stroke Outcome Classification. Circulation. 1998;97:2474–2478.
33 Berg K, Wood-Dauphinee S, Gayton D. Measuring balance in the
elderly: preliminary development of an instrument. Physiother Can.
34 Berg K, Wood-Dauphinee S, William, J, Maki B. Measuring balance
in the elderly: validation of an instrument. Can J Public Health.
35 Berg K, Wood-Dauphinee S, Williams J. The balance scale: reliability
assessment with elderly residents and patients with an acute stroke.
Scand J Rehabil Med. 1995;27:27–36.
36 Eng JJ, Chu KS, Dawson AS, et al. Functional walk tests in individ-
uals with stroke: relationship to perceived exertion and myocardial
exertion. Stroke. 2002;33:756–761.
37 Folstein MF, Folstein SE, McHugh PR. “Mini mental state”: a
practical method for grading the cognitive state of patients for the
clinician. J Psychiatr Res. 1975;12:189–198.
38 Dick JPR, Guiloff RJ, Stewart A, et al. Mini-mental state examination
in neurological patients. J Neurol Neurosurg Psychiatry. 1984;47:496–499.
39 Zwecker M, Levenkrohm S, Fleisig Y, et al. Mini-Mental State
Examination, cognitive FIM instrument, and the Loewenstein Occu-
pational Therapy Cognitive Assessment: relation to functional out-
come of stroke patients. Arch Phys Med Rehabil. 2002;83:324–345.
40 Cohen J, Cohen P, West SG, Aiken LS. Applied Multiple Regression/
Correlation for the Behavioral Sciences. 3rd ed. Mahwah, NJ: Lawrence
Erlbaum Associates; 2003.
41 Blake AJ. Falls in the elderly. Br J Hosp Med. 1992;47:268–272.
42 Lachenbruch PA, Reinsch S, macRae PG, Tobis JS. Adjusting for
recall bias with the proportional hazards model. Meth Inform Med.
43 Peel N. Validating recall of falls by older people. Accidental Analysis
and Prevention. 2000;32:371–372.
158 . Harris et al Physical Therapy . Volume 85 . Number 2 . February 2005