Mansfield et al.

, Physiotherapy, 2015

Title: Do measures of reactive balance control predict falls in people with stroke returning to the
community?

Authors: Avril Mansfield,a-c Jennifer S. Wong,a,b William E. McIlroy,a-d Louis Biasin,a,b Karen
Brunton,a.b Mark Bayley,a-c Elizabeth L. Innessa,b

Affiliations: aToronto Rehabilitation Institute – University Health Network, 550 University Ave,
Toronto, ON, M5G 2A2 Canada; bUniversity of Toronto, 500 University Ave, Toronto, ON, M5G 1V7,
Canada; cSunnybrook Health Sciences Centre, 2075 Bayview Ave, Toronto, ON, M4N 3M5, Canada;
d
University of Waterloo, 200 University Ave W, Waterloo, ON, N2L 3G1, Canada
Work conducted at the Toronto Rehabilitation Institute, Toronto, ON, Canada

Corresponding author: Avril Mansfield, Room 11-117, 550 University Ave, Toronto, ON, M5G 2A2,
Canada; tel: 416-597-3422 ext 7831; fax: 416-597-3031; e-mail: avril.mansfield@uhn.ca

Keywords: Accidental Falls; Stroke; Rehabilitation; Community; Postural Balance

Funding: This project has been generously funded by a grant from the Ontario Ministry of Health and
Long-Term Care, administered and supported by the Ontario Stroke Network (OSN1101-000117).
Equipment and space have been funded with grants from the Canada Foundation for Innovation,
Ontario Innovation Trust, and the Ministry of Research and Innovation. The views expressed do not
necessarily reflect those of the funders.

Conflict of interest: The authors declare no conflicts of interest.

Acknowledgements: We thank Shelley Makepeace and Svetlana Knorr for assistance with data
collection.

1
 Mansfield et al., Physiotherapy, 2015

ABSTRACT
Objective: To determine if reactive balance control measures predict falls after discharge from stroke
rehabilitation.
Design: Prospective cohort study.
Setting: Rehabilitation hospital and community.
Participants: Individuals with stroke who were discharged home after in-patient rehabilitation (n=95).
Interventions: Not applicable.
Main outcome measures: Balance and gait measures were obtained from a clinical assessment
immediately before discharge from in-patient stroke rehabilitation. Measures of reactive balance
control were obtained: 1) during quiet standing; 2) when walking; and 3) in response to large-
magnitude postural perturbations. Participants reported falls and activity levels up to six months post-
discharge. Logistic and Poisson regression were used to identify measures of reactive balance control
that were related to falls post-discharge.
Results: Decreased paretic limb contribution to standing balance control (rate ratio=0.8 [0.7, 1.0];
p=0.011), reduced between-limb synchronization of quiet standing balance control (rate ratio=0.9 [0.8,
0.9]; p<0.0001), increased step length variability (rate ratio=1.4 [1.2, 1.7]; p=0.0011), and inability to
step with the blocked limb (rate ratio=1.2 [1.0, 1.3]; p=0.013) were significantly related to increased
fall rates when controlling for age, stroke severity, functional balance, and daily walking activity.
Conclusions: Impaired reactive balance control in standing and walking predicted increased fall risk
post-discharge from stroke rehabilitation. Specifically, measures that revealed the capacity of both
limbs to respond to instability were related to increased fall risk. These results suggest that post-stroke
rehabilitation strategies for falls prevention should train responses to instability and focus on
remediating dyscontrol in the more-affected limb.

2
 Mansfield et al., Physiotherapy, 2015

INTRODUCTION
Fall risk post-stroke is high; up to 73% of community-dwelling stroke survivors fall in the six months
after discharge from hospital (1). This suggests that those at most risk are not identified or prepared for
the challenges they will encounter in their everyday living environments (2). Individuals with stroke
often have impaired balance control, which may increase the risk for falls. While several prospective
studies of community-dwelling adults with stroke demonstrated a link between functional balance
measures and falls (3-5), others found no differences in these balance measures between fallers and
non-fallers (6, 7).
There are several limitations to existing measures available to clinicians as fall risk predictors.
Clinical measures typically assign numerical values to varying levels of performance on tasks that
challenge an individual’s balance, but are neither able to quantify nor reveal the underlying sources of
dyscontrol that makes that task challenging. A key factor that ultimately determines whether an
individual will fall is the ability to react to a loss of balance (8). Individuals with stroke have impaired
reactive balance control (9-12); that is, the ability to execute appropriate and effective reactions to
recover from perturbations to balance. This study aimed to determine if measures of reactive balance
control (assessed during quiet standing and walking, and in response to external postural perturbations)
predict increased falls risk for stroke survivors returning to community living following discharge from
in-patient rehabilitation. We hypothesized that impaired reactive balance control in quiet standing,
walking, and in response to external postural perturbations would be predictive of increased risk for
falls, independent of age, stroke severity, daily walking activity, or functional balance measures.

METHODS
Study design
This was a prospective cohort study. Participants with stroke were recruited at discharge from in-
patient rehabilitation and followed for up to six months post-discharge. The study was approved by the
Toronto Rehabilitation Institute research ethics board (approval number: TRI REB #10-043).
Participants provided written informed consent prior to participation.

Participants
Individuals with stroke attending in-patient rehabilitation at the Toronto Rehabilitation Institute were
eligible for the study if they were: 1) assessed in a specialized balance clinic at discharge; 2) discharged
home; and 3) independently ambulating (i.e., without assistance/supervision of another person, with or
without a gait aid) at the time of discharge. The clinic assessment was completed as part of routine
practice by all individuals who had sufficient physical, communication, and cognitive function to
complete the assessment (as determined by the primary treating physiotherapist); therefore, no further
exclusion criteria were applied.

Predictor variables
Measures of reactive balance control were obtained from the clinic assessment. Variables focused on
three domains: 1) quiet standing, 2) walking, and 3) perturbation-evoked reactive stepping. Additional
data were obtained from participants’ hospital charts: age, sex, type of stroke, time post-stroke, affected
hemisphere, pre-morbid falls history, National Institutes of Health Stroke Scale (13) scores, and Berg
Balance Scale (14) scores. The National Institute of Health Stroke Scale is an 11-item scale that
provides a gross measure of the effects and severity of stroke, with higher scores indicating more
severe strokes. Items assess cognition (level of consciousness, orientation, and ability to follow
commands), gaze, visual fields, facial palsy, gross motor function in the arm and leg, ataxia, sensation,
language, speech, and extinction and inattention. The Berg Balance Scale is a 14-item observational

3
 Mansfield et al., Physiotherapy, 2015

rating scale that provides a measure of functional anticipatory balance control (rather than reactive
balance control). For participants with bilateral strokes, a more-affected side was identified.

Quiet standing
Participants stood in a standardized foot position (15) with one foot on each of two force plates for 30-
seconds, and were asked to stand as still as possible with eyes open. Forces and moments were
recorded from the force plates at 256Hz and filtered using a 10Hz low-pass zero phase lag Butterworth
filter prior to processing. The antero-posterior and medio-lateral centres of pressure (COP) were
calculated for each force plate separately and for both feet combined. The root mean square (RMS) of
total antero-posterior and medio-lateral COP were calculated to provide a measure of overall COP
variability. The contribution of the paretic limb to balance control was calculated by dividing the RMS
of antero-posterior COP under the more-affected limb by the sum of the RMS of antero-posterior COP
under each limb (16, 17); a value of 0.5 indicates that both limbs contribute equally to balance control,
<0.5 indicates that the less-affected limb contributes more to balance control, whereas >0.5 indicates
that the more-affected limb contributes more to balance control. Between-limb synchronization of
antero-posterior COP was calculated by determining the correlation coefficient between the left and
right antero-posterior COP (18, 19). We focused on antero-posterior COP for the contribution and
synchronization measures as individual-limb medio-lateral COP is less meaningful for overall bipedal
balance control (20).

Walking
Participants walked across a 4-m long pressure mat at their usual pace without a walking aid (whenever
possible). Participants walked 3-5 times across the mat such that at least 18 footfalls were recorded.
Step length, step width, and step time were calculated for each step. The standard deviations of step
length, width and time were calculated for each limb separately; variability was calculated as the
average of the standard deviations for the left and right limbs. Overall walking speed was also
calculated.

Reactive stepping
A lean-and-release system was used to study reactive stepping (9, 10, 12). Participants stood in a
standardized foot position (15), and wore a belt around their trunks attached to a beam via a cable.
Participants leaned forward such that approximately 10% of body weight was supported by the cable.
The cable released unexpectedly, causing them to fall forward; the magnitude of the perturbation was
so great that a reactive step was required to regain stability. Participants were closely supervised by a
physiotherapist who provided assistance, if necessary, and a safety harness attached to an overhead
track was worn such that participants would not fall to the floor. Five trials were completed in each of
two conditions: usual response, and encouraged use. In the usual-response condition, there was no
constraint on participants’ stepping reactions. The ‘preferred’ limb was the limb most frequently used
to initiate stepping in the usual response condition. In the encouraged-use condition, the preferred limb
was blocked by the physiotherapists’ hand or foot to encourage initiating stepping with the non-
preferred limb (11). Trials were video-recorded. For the usual response condition, the following
outcome measures were obtained: frequency of ‘assists’ (i.e., reliance on physiotherapist or harness to
prevent a fall); frequency of attempted reach-to-grasp reactions (i.e., participant reached for and
grasped the physiotherapist); number of steps taken to recover balance; and frequency of ‘slide’ steps
(i.e., the initial step involved foot movement without lifting the foot completely off the ground (10)).
For the encouraged use condition, we also determined the frequency of trials in which the participant
attempted to step with the blocked limb.

4
 Mansfield et al., Physiotherapy, 2015

Falls and daily activity

Participants completed a six-month falls monitoring period following discharge from in-patient
rehabilitation. Participants used fortnightly calendars printed on pre-stamped postcards to indicate if
they experienced a fall or near fall each day, which were mailed back to the study investigators when
each calendar was complete. Participants received a monthly study newsletter as a reminder to return
their postcards. If participants failed to return a postcard, a research assistant called them to ascertain if
they had fallen in the previous two weeks.
A fall was defined as “an event that results in a person coming to rest unintentionally on the
ground or other lower level” (21). Falls that occurred due to fainting or loss of consciousness were
excluded from analysis. Participants who reported a fall or near fall were asked to call the research
assistant to complete a short telephone questionnaire in order to determine the circumstances
surrounding the fall or near fall. Participants who reported a fall or near fall on the calendar but did not
call the research assistant were contacted to complete this questionnaire. Events were reclassified, as
determined necessary, based on the participants’ description (e.g., if a participant reported a near fall
but, upon interview, it was determined that s/he actually came to rest on a lower level).
Physical activity was evaluated with the Physical Activity Scale for Individuals with Physical
Disabilities (PASIPD (22)) three times (approximately every two months) during the falls monitoring
period to obtain an estimate of physical activity. As most falls post-stroke occur while walking (21), we
estimated the total time spent walking per day from the PASIPD. PASIPD scores and walking times
were averaged over the three time points to obtain an estimate of activity in the six-month period.

Statistical analysis
For descriptive purposes, participants were classified as non-fallers (no falls reported), single fallers (1
fall reported), or multiple fallers (>1 fall reported). Logistic and Poisson regression were conducted to
determine the ability of each measure of reactive balance control to predict risk of falls post-discharge.
Logistic regression was used to determine the relationship between measures of reactive balance
control and probability of being classified as a ‘faller’; that is, the dependent variable was falling status
(i.e., ‘faller’ versus ‘non-faller’). Poisson regression was used to determine the relationship between
measures of reactive balance control and fall rates. Poisson regression is appropriate when there are
multiple events per person, events are statistically rare, and there is a variable follow-up duration for
each participant. For Poisson regression, the dependent variable was normalized by determining the
number of falls divided by the monitoring duration. Univariate regressions were conducted for each
independent variable alone. For quiet standing and walking, the independent variables were: RMS of
antero-posterior and medio-lateral COP, contribution of the paretic limb to antero-posterior balance
control, between-limb synchronization, and step length, width and time variability. For reactive
stepping (usual-response condition), the independent variables were: frequency of trials where the
participant was unable to recover balance by stepping (i.e. either an ‘assist’ or reach-to-grasp response);
number of steps; and frequency of ‘slide’ steps. For the encouraged-use condition, the independent
variable was frequency of trials where the participant was unable to step with the unblocked limb (i.e.,
either an attempt to step with the blocked limb or the initial step was a ‘slide’ step with the unblocked
limb). Multivariate logistic and Poisson regressions were conducted for those reactive balance
measures that were significantly related to falls controlling for age, stroke severity (National Institutes
of Health Stroke Scale), functional balance (Berg Balance Scale), and daily activity (average time spent
walking per day). Odds ratios and rate ratios are presented in the paper with the corresponding 95%
confidence interval (CI) of the ratio in brackets. The Holm-Bonferroni method was used to correct for
multiple comparisons (23); separate corrections were completed for each analysis (initial α=0.0045).

5
 Mansfield et al., Physiotherapy, 2015

Table 1: Participant characteristics at the time of discharge from in-patient rehabilitation. Values are
means with standard deviations in parentheses (continuous/pseudo-continuous variables), or counts with
percentages in parentheses (categorical variables).
Non-fallers Single fallers Multiple fallers
(n=60) (n=21) (n=14)
Demographic & stroke information
Age (years) 62.9 (14.2) 61.8 (10.5) 62.7 (14.6)
Sex (number, %)
Women 20 (33) 8 (38) 7 (50)
Men 40 (67) 13 (62) 7 (50)
Type of stroke (number, %)
Ischaemic 48 (80) 15 (71) 13 (93)
Haemorrhagic 9 (15) 3 (14) 1 (7)
Transforming to haemorrhagic 2 (3) 2 (10) 0 (0)
Unknown 1 (2) 1 (5) 0 (0)
Time post-stroke at discharge (days) 45.1 (24.6) 52.1 (16.3) 56.3 (20.8)
Affected hemisphere (number, %)
Right 21 (35) 9 (43) 6 (43)
Left 32 (53) 8 (38) 5 (36)
Both 7 (12) 3 (14) 3 (21)
Unknown 0 (0) 1 (5) 0 (0)
National Institutes of Health stroke scale (score) 2.8 (2.4) 3.4 (2.6) 2.7 (2.4)
Falls history and functional balance and mobility
Falls prior to stroke (number, %) 10 (17.2) 6 (28.6) 4 (28.6)
Berg balance scale (score) 50.4 (8.0) 45.9 (9.6) 48.1 (4.0)
Prescribed gait aid (number, %)
None 31 (51.7) 7 (33.3) 2 (14.3)
Cane 12 (20.0) 5 (23.8) 5 (35.7.4)
Rollator or wheeled walker 12 (20.0) 8 (38.1) 4 (28.6)
Rollator and cane 5 (8.3) 1 (4.8) 3 (21.4)
Walking speed (m/s) 0.87 (0.35) 0.72 (0.40) 0.66 (0.23)
Quiet standing balance control
Antero-posterior RMS of COP (mm) 6.1 (3.0) 6.5 (3.9) 6.3 (2.9)
Medio-lateral RMS of COP (mm) 3.9 (2.8) 4.8 (3.2) 3.8 (2.5)
Paretic limb contribution 0.46 (0.13) 0.45 (0.10) 0.38 (0.12)
Between-limb synchronization 0.82 (0.14) 0.71 (0.33) 0.67 (0.31)
Spatio-temporal gait variability
Step length variability (cm) 2.9 (1.1) 3.6 (1.4) 3.2 (1.0)
Step width variability (cm) 2.3 (0.9) 2.2 (1.2) 1.7 (0.6)
Step time variability (s) 0.07 (0.19) 0.09 (0.10) 0.04 (0.02)
Reactive stepping (n=71)
Assists (% trials) 2.7 (6.9) 14.8 (23.1) 12.5 (31.7)
Reach-to-grasp reactions (% trials) 1.8 (6.8) 8.5 (20.3) 6.5 (14.2)
Number of steps 2.3 (0.9) 2.4 (0.9) 2.6 (0.8)
‘Slide’ steps (% trials) 4.4 (10.8) 9.0 (20.8) 2.0 (6.3)
No-step reactions (% trials) 0 0 0
Inability to step with the unblocked limb (% trials) 20.8 (35.1) 39.4 (43.4) 38.7 (21.8)
Daily activity
Time spent walking (hours per day) 0.9 (0.6) 1.0 (0.6) 0.5 (0.4)
PASIPD (score) 9.8 (5.9) 9.4 (8.3) 8.5 (4.9)
COP=centre of pressure; PASPID=Physical Activity Scale for Individuals with Physical Disabilities; RMS=root
mean square
6
 Mansfield et al., Physiotherapy, 2015

RESULTS
Between October 2010 and March 2013, 419 individuals were discharged from the hospital; 172 of
these (41%) met the inclusion criteria and were invited to participate and 100 agreed to participate in
the study. Five withdrew prior to completing any falls monitoring; therefore, 95 participants were
included in the analysis. However, only 71 of the 95 participants completed the reactive stepping
assessment; therefore, analysis of reactive stepping data is based on 71 participants. Table 1 details
participant characteristics at the time of discharge from in-patient rehabilitation. Seven participants
were lost to follow-up between 1.8 and 5.5 months after discharge and, therefore, did not complete the
full 6-month falls monitoring period. In the follow-up period, 60 of the 95 participants (63%) reported
no falls, 21 (22%) reported 1 fall, and 14 participants (15%) reported >1 fall. The 35 participants who
fell reported a total of 83 falls. The median time between the occurrence of the fall and the interview
with the research assistant was 16 days; 29% of falls (27/83) were reported within 7 days. The details
of 9 falls (11%) were reported with the assistance of a family member. Most falls occurred in
participants’ homes (51/83, 63%), while either walking (26/83, 32%) or transferring/transitioning
(23/83, 28%). For 37% of falls (30/83), participants required assistance to get up. Eighteen falls (18/83,
22%) resulted in injuries with five (5/83, 6%) requiring treatment from a healthcare professional. While
the majority of fallers (26/35, 74%) were recommended to use a gait aid, such as a cane or rollator, for
indoor and/or community mobility at discharge, only 22% of falls (18/83) occurred when a gait aid was
being used.
From logistic regression, no reactive balance control variables were related to increased odds of
falling when adjusted for multiple comparisons (p-values>0.0045; Table 2). Poisson regression
revealed that increased RMS of medio-lateral COP, decreased between-limb synchronization,
decreased contribution of the paretic limb to balance control, increased step length and width
variability, and increased frequency of encouraged-use trials in which the participant was unable to step
with the unblocked limb were related to increased fall rates when adjusted for multiple comparisons
(Table 3; p-values<0.01). Paretic limb contribution (rate ratio=0.8; 95% CI: [0.7, 1.0]; p=0.011),
between-limb synchronization (rate ratio=0.9; 95% CI: [0.8, 0.9]; p<0.0001), step length variability
(rate ratio=1.4; 95% CI: [1.2, 1.7]; p=0.0011), and inability to step with the unblocked limb (rate
ratio=1.2; 95% CI: [1.0, 1.3]; p=0.013) remained significantly related to fall rates when controlling for
age, stroke severity, functional balance, and daily walking activity.

Table 2: Reactive balance control and odds of falling. Values are point estimates and 95% confidence
intervals for the odds ratios (logistic regression) representing the change in odds of falling for each incremental
increase in each variable. The odds ratio is for the model including that independent variable alone. As no
independent variables were significantly related to increased odds of falling (adjusted alpha=0.0045),
multivariate analysis was not conducted.
Independent variable Increment Odds ratio p-value
Antero-posterior RMS of COP (mm) 1 1.0 [0.9, 1.2] 0.54
Medio-lateral RMS of COP (mm) 1 1.1 [0.9, 1.3] 0.21
Paretic limb contribution 0.1 0.8 [0.6, 1.1] 0.20
Between-limb synchronization 0.1 0.8 [0.6, 1.0] 0.020
Step length variability (cm) 1 1.5 [1.0, 2.3] 0.033
Step width variability (cm) 1 0.7 [0.4, 1.1] 0.14
Step time variability (s) 0.01 1.0 [0.1, 15.8] 0.95
Inability to recover by stepping (% trials) 20 2.2 [1.1, 4.2] 0.023
Number of steps 1 1.2 [0.7, 2.0] 0.51
Frequency of ‘slide’ steps (% trials) 20 1.3 [0.6, 2.5] 0.54
Inability to step with unblocked limb (% trials) 20 1.3 [1.0, 1.7] 0.048

7
 Mansfield et al., Physiotherapy, 2015

COP=centre of pressure; RMS=root mean square

Table 3: Reactive balance control and fall rates. Values are point estimates and 95% confidence intervals for
the rate ratios (Poisson regression), representing the change in rate of falls for each incremental increase in each
variable. The univariate rate ratio is for the model including that independent variable alone, whereas the
multivariate ratio is the ratio for the independent variable controlling for age, stroke severity (National Institutes
of Health Stroke Scale) functional balance (Berg balance scale), and daily walking activity (average time spent
walking per day). Statistically significant relationships are indicated with asterisks (α=0.01 for univariate
Poisson regression, and α=0.025 for multivariate Poisson regression).
Independent variable Increment Univariate p-value Multivariate p-value
rate ratio rate ratio
Antero-posterior RMS of COP (mm) 1 1.1 [1.0, 1.1] 0.016
Medio-lateral RMS of COP (mm) 1 1.1 [1.0, 1.2] 0.0066* 1.0 [0.9, 1.1] 0.38
Paretic limb contribution 0.1 0.7 [0.6, 0.9] 0.0008* 0.8 [0.7, 1.0] 0.011*
Between-limb synchronization 0.1 0.9 [0.8, 0.9] <0.0001* 0.9 [0.8, 0.9] <0.0001*
Step length variability (cm) 1 1.2 [1.1, 1.4] 0.0052* 1.4 [1.2, 1.7] 0.0011*
Step width variability (cm) 1 0.6 [0.4, 0.8] <0.0001* 0.7 [0.5, 1.0] 0.027
Step time variability (s) 0.01 0.6 [0.1, 3.7] 0.55
Inability to recover by stepping (% trials) 20 1.1 [0.9, 1.4] 0.42
Number of steps 1 0.9 [0.7, 1.2] 0.48
Frequency of ‘slide’ steps (% trials) 20 1.0 [0.7, 1.5] 0.83
Inability to step with unblocked limb (% trials) 20 1.3 [1.1, 1.4] 0.0002* 1.2 [1.0, 1.3] 0.013*
COP=centre of pressure; RMS=root mean square

DISCUSSION
While others have reported a link between functional balance measures and falls post-stroke (3-5), this
study demonstrated that specific measures of reactive balance control at discharge from in-patient
stroke rehabilitation predict increased fall rates in the six months post-discharge into the community
among independently ambulating individuals with stroke. These significant relationships were
independent of age, stroke severity, functional balance measures, and walking activity. Thus, this work
provides support for a hypothesized causal link between impaired reactive balance control and
increased fall risk. However, further support for this hypothesis is required from interventional studies.
During quiet standing, small centre of mass movements must be corrected to prevent excessive
postural sway and, potentially, a fall. Measurement of the location of reaction forces at the feet (i.e.,
COP) reveals reactions to maintain stability in quiet standing. Others have observed asymmetric
contribution of each lower limb to quiet standing balance control post-stroke; specifically, the non-
paretic limb appears to contribute more to balance control than the paretic limb (24). Previous work has
also demonstrated lower between-limb synchronization of COP among individuals with stroke
compared to healthy controls (18), and that this reduced synchronization is related to increased falls
experienced during in-patient rehabilitation (19). The current work demonstrates that reduced
contribution of the paretic limb to balance control and reduced between-limb synchronization are
related to increased risk of falls in the community. Combined, these results could suggest that, for feet-
in-place reactions to balance perturbations, both lower limbs need to work together in a synchronized
manner and contribute equally to prevent a fall.
During walking, individuals experience small perturbations to the desired trajectory of the
centre of mass. Varied placement and timing of steps during gait are thought to compensate for these
perturbations (25); thus stride-to-stride variability of spatio-temporal gait characteristics may represent
reactive balance control. However, it is unclear if increased gait variability indicates impaired reactive
balance control or an appropriate strategy given impaired anticipatory balance control. Alternatively,
variability may reflect impaired motor control and errors in limb placement. Individuals with stroke
8
 Mansfield et al., Physiotherapy, 2015

have increased spatio-temporal gait variability compared to healthy controls (26). Previous research
demonstrated that step length variability, but not step time or width variability, predicts increased fall
risk among older adults (25), in agreement with the current study. As most falls among individuals with
stroke occur while walking (21) it is logical to expect that impaired balance control during walking
would be related to increased fall risk.
Reactive steps are typically required to prevent a fall following large postural perturbations (e.g.
a trip or a slip). Individuals with stroke have difficulty executing effective reactive steps (10, 11).
Previous work reported that increased frequency of ‘falls’ into the safety harness following an external
postural perturbation were more prevalent among older adults with a history of falls (27). In the current
study, we also observed that increased ‘falls’ during the assessment were related to increased risk of
falling in the community; however, this effect was not statistically significant when adjusting for
multiple comparisons. A previous study reported that increased fall rates during in-patient
rehabilitation were linked to inability to recover balance by stepping and increased frequency of ‘slide’
steps, while inability to step with the unblocked limb was not linked to falls (10). However, inability to
step with the unblocked limb was the only measure significantly related to increased fall rates in the
current study. Differences in findings between the current and previous work may be partially due to
the fact that the community environment is less predictable than that of a rehabilitation hospital. Thus,
recovering from instability in less predictable environments in the community may pose a challenge for
those who are unable to initiate reactive stepping with either limb.
Balance training effectively prevents falls among older adults (28) but not among those with
stroke (29, 30). The current results suggest potential rehabilitation strategies for falls prevention post-
stroke. As measures that revealed dyscontrol of one or both limbs predicted increased falls risk (i.e.
reduced paretic limb contribution to and reduced between-limb synchronization of standing balance
control, and reduced ability to initiate a reactive step with either limb), rehabilitation strategies aiming
to improve control of both limbs, specifically the paretic limb, may help to prevent falls. Additionally,
rehabilitation focused on regaining stability following a loss of balance, such as perturbation-based
balance training (31), could be effective for preventing falls post-stroke. An important aspect of this
training is attention to balance reactions, such as stepping, rather than a more conventional focus on
voluntary or self-initiated movements; training the latter will likely have little transfer to reactive
balance control (32).
The assessments of reactive balance control were completed as part of routine clinical care.
While many of these measures relied on ‘research-grade’ technology (i.e. force plates and pressure
mats) that may not be available to clinicians in other settings, lower-cost technology may soon be able
to provide similar measures to those used in this work (33, 34). Additionally, the ‘lean-and-release’
system used to provide a large external postural perturbation does not depend on advanced technology
and the key outcome measures for this task can be obtained from observation.

Limitations
This study is potentially limited by including a relatively ‘high-functioning’ sample of stroke survivors
who were independently ambulatory at the time of discharge from in-patient rehabilitation. The sample
analysed represents only 23% (95/419) of all individuals discharged from the hospital. The primary
reasons for exclusion from the study were that patients were too low-functioning to complete the
balance clinic assessment or were not discharged home. Future work should determine falls risk factors
among lower-functioning individuals. Falls and activity data were collected by self-report. The method
of falls data collection used in this study (i.e., prospective collection with calendars, regular reminders,
and follow-up telephone calls in the event of failure to return a post-card) is considered the ‘gold
standard’ for collecting information on falls in daily life among community-dwelling individuals (35).
However, this method may still be prone to some error. Some events that met the definition of a ‘fall’
9
 Mansfield et al., Physiotherapy, 2015

may not be considered a fall by participants. To address this, we asked participants to also report near
falls, and re-classified near falls as falls (or vice versa) as necessary. However, in some cases,
participants might forget to report a fall or near fall, or might not recall the circumstances of the event,
which would limit our ability to determine if a fall had occurred. Thus, it is possible that some falls that
occurred were not recorded in this study. Additionally, in terms of activity data, individuals with stroke
may over-estimate physical activity when using self-report measures (36), although participants in the
current study were generally inactive (37).

Conclusions
Impaired reactive balance control predicts increased fall risk after discharge into the community
following in-patient stroke rehabilitation. Of note, measures that revealed capacity to use both paretic
and non-paretic lower limbs to maintain stability were particularly predictive of increased fall risk (i.e.
contribution of the paretic limb and between-limb synchronization of quiet standing balance control,
and ability to initiate reactive stepping with either limb). It may be possible to compensate for
persisting dyscontrol of one limb with the less-affected limb to show functional gains with recovery
and rehabilitation post-stroke. However, the results of this study suggest adequate control of both limbs
is required to respond to unexpected instability that precedes a fall.

10
 Mansfield et al., Physiotherapy, 2015

REFERENCES
1. Forster A, Young J. Incidence and consequences of falls due to stroke: a systematic inquiry.
BMJ. 1995;311:83-6.
2. Weerdesteyn V, de Niet M, van Duijhoven HJR, Geurts ACH. Falls in individuals with stroke. J
Rehabil Res Dev. 2008;45(8):1195-213.
3. Mackintosh SF, Hill KD, Dodd KJ, Goldie PA, Culham EG. Balance score and a history of falls
in hospital predict recurrent falls in the 6 months following stroke rehabilitation. Arch Phys Med
Rehabil. 2006;87:1583-9.
4. Simpson LA, Miller WC, Eng JJ. Effect of stroke on fall rate, location and predictors: a
prospective comparison of older adults with and without stroke. PLoS one. 2011;6(4):e19431.
5. Chin LF, Wang JY, Ong CH, Lee WK, Kong KH. Factors affecting falls in community-
dwelling individuals with stroke in Singapore after hospital discharge. Singapore Med J.
2013;54(10):569-75.
6. Lamb SE, Ferrucci L, Volapto S, Fried LP, Guralnik JM. Risk factors for falling in home-
dwelling older women with stroke: the women's health and aging study. Stroke. 2003;34:494-501.
7. Harris JE, Eng JJ, Marigold DS, Tokuno CD, Louis CL. Relationship of balance and mobility to
fall incidence in people with chronic stroke. Phys Ther. 2005;85(2):150-8.
8. Maki BE, McIlroy WE. Postural control in the older adult. Clin Geriatr Med. 1996;12(4):635-
58.
9. Inness EL, Mansfield A, Lakhani B, Bayley M, McIlroy WE. Impaired reactive stepping among
patients ready for discharge from inpatient stroke rehabilitation. Phys Ther. 2014;in press.
10. Mansfield A, Inness EL, Wong JS, Fraser JE, McIlroy WE. Is impaired control of reactive
stepping related to falls during inpatient stroke rehabilitation? Neurorehabil Neural Repair.
2013;27(6):526-33.
11. Lakhani B, Mansfield A, Inness EL, McIlroy WE. Compensatory stepping responses in
individuals with stroke: a pilot study. Physiother Theory Pract. 2011;27(4):299-309.
12. Mansfield A, Inness EL, Lakhani B, McIlroy WE. Determinants of limb preference for
initiating compensatory stepping post-stroke. Arch Phys Med Rehabil. 2012;93:1179-84.
13. Goldstein LB, Bertels C, Davis JN. Interrater reliability of the NIH Stroke Scale. Arch Neurol.
1989;46(6):660-2.
14. Berg K, Wood-Dauphinée S, Williams JI, Gayton D. Measuring balance in the elderly:
preliminary development of an instrument. Physiother Can. 1989;41:304-11.
15. McIlroy WE, Maki BE. Preferred placement of the feet during quiet stance: development of a
standardized foot placement for balance testing. Clin Biomech. 1997;12(1):66-70.
16. Hendrickson J, Patterson KK, Inness EL, McIlroy WE, Mansfield A. Relationship between
asymmetry of quiet standing balance control and walking post-stroke. Gait Posture. 2014;39(1):177-81.
17. Rougier PR, Genthon N. Dynamical assessment of weight-bearing asymmetry during upright
quiet stance in humans. Gait Posture. 2009;29:437-43.
18. Mansfield A, Danells CJ, Inness EL, Mochizuki G, McIlroy WE. Between-limb
synchronization for control of standing balance in individuals with stroke. Clin Biomech.
2011;26(3):312-7.
19. Mansfield A, Mochizuki G, Inness EL, McIlroy WE. Clinical correlates of between-limb
synchronization of standing balance control and falls during inpatient stroke rehabilitation.
Neurorehabil Neural Repair. 2012;26(6):627-35.
20. Winter DA, Prince F, Stergiou P, Powell C. Medial-lateral and anterior-posterior motor
responses associated with centre of pressure changes in quiet stance. Neurosci Res Commun.
1993;12(3):141-8.

11
 Mansfield et al., Physiotherapy, 2015

21. Hyndman D, Ashburn A, Stack E. Fall events among people with stroke living in the
community: circumstances of falls and characteristics of fallers. Arch Phys Med Rehabil. 2002;83:165-
70.
22. Washburn RA, Zhu W, McAuley E, Frogley M, Figoni SF. The Physical Activity Scale for
Individuals with Physical Disabilities: development and evaluation. Arch Phys Med Rehabil.
2002;83:193-200.
23. Holm S. A simple sequentially rejective multiple test procedure. Scand J Statist. 1979;6:65-70.
24. Roerdink M, Geurts ACH, de Haart M, Beek PJ. On the relative contribution of the paretic leg
to the control of posture after stroke. Neurorehabil Neural Repair. 2009;23(3):267-74.
25. Maki BE. Gait changes in older adults: predictors of falls or indicators of fear? J Am Geriatr
Soc. 1997;45:313-20.
26. Balasubramanian CK, Neptune RR, Kautz SA. Variability in spatiotemporal step characteristics
and its relationship to walking performance post-stroke. Gait Posture. 2009;29:408-14.
27. Wolfson LI, Whipple R, Amerman P, Kleinberg A. Stressing the postural response: a
quantitative method for testing balance. J Am Geriatr Soc. 1986;34:845-50.
28. Sherrington C, Whitney JC, Lord SR, Herbert RD, Cumming RG, Close JCT. Effective exercise
for the prevention of falls: a systematic review and meta-analysis. J Am Geriatr Soc.
2008;56(12):2234-43.
29. Batchelor F, Hill K, Mackintosh S, Said C. What works in falls prevention after stroke? a
systematic review and meta-analysis. Stroke. 2010;41(8):1715-22.
30. Verheyden GS, Weerdesteyn V, Pickering RM, Kunkel D, Lennon S, Geurts AC, et al.
Interventions for preventing falls in people after stroke. Cochrane Database Syst Rev.
2013;31(5):CD008728.
31. Mansfield A, Wong JS, Bryce J, Knorr S, Patterson KK. Does perturbation-based balance
training prevent falls? A review and meta-analysis of preliminary randomized controlled trials. Phys
Ther. 2014;doi: 10.2522/ptj.20140090.
32. Mansfield A, Peters AL, Liu BA, Maki BE. A perturbation-based balance training program for
older adults: study protocol for a randomised controlled trial. BMC Geriatr. 2007;7(1):12.
33. Clark RA, Bryant AL, Pau Y, McCrory P, Bennell K, Hunt M. Validity and reliability of the
Nintendo Wii Balance Board for assessment of standing balance. Gait Posture. 2010;31(3):307-10.
34. Stone EE, Skubic M. Passive in-home measurement of stride-to-stride gait variability
comparing vision and Kinect sensing. Conf Proc IEEE Eng Med Biol Soc. 2011;2011:6491-4.
35. Lamb SE, Jørstad-Stein EC, Hauer K, Becker C. Development of a common outcome data set
for fall injury prevention trials: The Prevention of Falls Network Europe consensus. J Am Geriatr Soc.
2005;53(9):1618-22.
36. Resnick B, Michael K, Shaughnessy M, Nahm ES, Kobunek S, Sorkin J, et al. Inflated
perceptions of physical activity after stroke: pairing self-report with physiologic measures. J Phys Act
Health. 2008;5(2):308-18.
37. Brown C, Fraser JE, Inness EL, Wong J, Middleton LE, Poon V, et al. Does participation in
standardized aerobic fitness training during inpatient stroke rehabilitation promote engagement in
aerobic exercise after discharge? A cohort study. Top Stroke Rehabil. 2014;21(Suppl 1):S42-51.

12