Content uploaded by Mona Kristin Aaslund
Author content
All content in this area was uploaded by Mona Kristin Aaslund on Feb 06, 2018
Content may be subject to copyright.
1
This is an Accepted Manuscript of an article published by Taylor &
Francis in Physiotherapy Theory and Practice on 17.08.17, available
online: http://dx.doi.org/10.1080/09593985.2017.1360424
TITLE
A longitudinal study investigating how stroke severity, disability and physical
function the first week post-stroke are associated with walking speed six
months post-stroke
AUTHORS
Mona Kristin Aaslund, PT, PhD, Rolf Moe-Nilssen , PT, PhD, Bente Bassøe Gjelsvik , PT, PhD,
Bård Bogen , PT, MSc, Halvor Næss , MD, PhD, Håkon Hofstad , MD, PhD & Jan Sture Skouen,
MD, PhD.
2
1. Introduction
Community walking, or the ability to confidently walk in outdoor environments, is
important to conduct daily life activities and to participate in social events and life roles
(Lord et al., 2004; Wing, Lynskey and Bosch, 2012). After stroke, walking speed (WS) is often
slow, and safe and effective outdoor ambulation may be a major challenge. Consequently, a
key goal in stroke rehabilitation is to improve or recover walking ability (Mayo, Korner-
Bitensky and Becker, 1991; Turnbull, Charteris and Wall, 1995; Mayo et al., 1999; Lord et al.,
2004; Tilson, Settle and Sullivan, 2008). Gait training has been found to be the most common
intervention in physiotherapy post-stroke (Jette et al., 2005), and specific gait training such
as treadmill walking has been found to be effective to improve WS in different stages post-
stroke (Peurala et al., 2014).
Community walking is multidimensional, and together with factors such as health
perceptions and fear of falling, it is largely determined by the combination of walking
distance and WS (van de Port, Kwakkel and Lindeman, 2008; Bijleveld-Uitman, van de Port
and Kwakkel, 2013; An, Lee, Shin and Lee, 2015; Barclay, Ripat and Mayo, 2015; Liu, Ng,
Kwong and Ng, 2015). The ability to walk distances between 318 and 368 m, and at speeds
between 0.66 and 0.87 m/s, have been identified as cut-off points for community walking in
patients post-stroke (van de Port, Kwakkel and Lindeman, 2008; Bijleveld-Uitman, van de
Port and Kwakkel, 2013; An, Lee, Shin and Lee, 2015). Correspondingly, Perry and colleagues’
Functional ambulation categories (FAC) are commonly used for determining community
walking ability in patients post-stroke, where a preferred walking speed (PWS) greater than
0.8 m/s indicates an ability to walk independently in the community (Perry, Garrett, Gronley
and Mulroy, 1995). Slow PWS (below 1 m/s) is also strongly associated with the onset of
disability, functional decline, frailty, future falls, cognitive impairment, poor health and
mortality in older people (Lusardi, 2012). A risk profile for future decline in health and
function has been proposed where PWS above 1 m/s indicates low risk, 0.6 – 1 m/s
moderate risk and PWS below 0.6 m/s high risk (Lusardi, 2012).
Subsequently, slow WS may be seen as a trainable and therefore modifiable risk factor
for poor community walking ability, functional decline and adverse health events. Therefore,
it is important early post-stroke to recognise the long-term prospect for a patient to regain
sufficient WS for community walking by using clinical measures with predictive properties.
3
Due to the multidimensionality of walking several types of measures may have this
capability. Knowledge about such predictors is sparse, and only a few longitudinal studies
investigating predictors for community walking at six months post-stroke have been
performed. One study identified leg strength and sitting balance within 72 hours after stroke
in 154 non-ambulatory patients as potential predictors of community walking six months
post-stroke (Veerbeek, Van Wegen, Harmeling-Van der Wel and Kwakkel, 2011). Another
recent study looked at 35 ambulatory patients within the first three months after stroke and
identified fast WS and less fear of falling as predictors of community walking at six months
post-stroke (Rosa, Marques, Demain and Metcalf, 2015).
Consequently, the main purpose of this study was to investigate to which degree typical
clinical measures of stroke severity, disability and physical function the first week post-
stroke are associated with PWS at six months.
2. Method
2.1 Design
This longitudinal observational study was done in the context of the randomised
controlled trial (RCT) “Early Supported Discharge after Stroke in Bergen” (ESD Stroke
Bergen), registered in ClinicalTrials.gov (NCT00771771) and described in a published
protocol (Hofstad, Naess, Moe-Nilssen and Skouen, 2013) as well as in three published
articles (Gjelsvik et al., 2014; Hofstad et al., 2014; Hofstad et al., 2016). In the present study,
data from baseline and six months post-stroke were used and all participants were analysed
independent of their RCT group allocation.
2.2 Sample
Participants were recruited from a stroke unit from December 2008 to December 2011.
The standard stroke unit care consisted of repeated physiologic monitoring (blood pressure,
heart rate, body temperature, oxygen saturation) and NIHSS scoring during the first 1-3 days.
Patients were referred early to the multi-professional rehabilitation team (mainly
physiotherapists, occupational therapists and speech therapists) and therapy was generally
scheduled five days a week, with amount of treatment time dedicated per session being
individual depending on the patients’ needs and tolerability. The aim for the stroke unit
team is to provide careful evaluation of possible complications, and to provide early
4
mobilisation and early rehabilitation of high standard. All patients admitted to the stroke
unit with suspected stroke were screened for inclusion. Stroke was confirmed by brain scans
using CT or MRI performed shortly after admission to the stroke unit. Since the focus of the
RCT study was to explore different models of early supported discharge, being home-
dwelling in the municipality of Bergen prior to the stroke was a requirement. The
participants had to be included within seven days after stroke onset and between six and
120 hours after admission to the stroke unit. They had to be conscious and score within 2-26
points on the National Institutes of Health Stroke Scale. This range was selected for the ESD
Stroke Bergen study in order to both include mildly, moderately and severely affected stroke
patients, and to exclude patients with none or minimal neurological deficits as well as
patients with very severe neurological deficits. Only the participants who had been assessed
using Functional Ambulation Categories (FAC) and PWS were included in the present study.
Exclusion criteria were serious comorbidity such as terminal cancer, current alcohol or
substance abuse, serious psychiatric disorders or insufficient Norwegian language skills. All
participants signed a written informed consent. If this was not possible, consent was given
by the next of kin until the participant could sign the written consent themselves. Cognitive
impairment (not including unconsciousness) was not an exclusion criterion. The study was
approved by the regional ethics committee (REK-2012/1793).
2.3 Data collection
The data collection was conducted at a university hospital. Baseline testing was done
within the first week after ictus. Neurologists at the stroke unit were responsible for use of
the National Institutes of Health Stroke Scale, and trained stroke nurses scored the Barthel
index and modified Rankin Scale. Four experienced neurophysiotherapists conducted the
measurements of physical function at baseline and did all testing at six months post-stroke.
All assessors underwent training of the test procedure to optimise standardisation and inter-
rater reliability in the study. Inter-rater reliability was not tested.
Descriptive data
Descriptive data such as age, gender, living arrangements pre-stroke, as well as type and
location of stroke were collected. Days in the stroke unit and discharge destination from the
stroke unit were also recorded.
5
Physical function
Preferred walking speed (PWS) was measured by timing a 5-meter walk with a
stopwatch. A distance of 1.5 meters at the beginning and the end of the walk was used to
avoid including the acceleration and deceleration phases (dynamic start/stop). The
participants were instructed to walk at their preferred speed. The participants where
allowed to use walking aids if necessary. WS has been found valid, reliable, sensitive and
specific, and correlates with functional ability and balance confidence (Fritz and Lusardi,
2009). PWS has also been found to be a consistent predictor of adverse health outcomes in
different populations (Abellan van Kan et al., 2009) and is commonly used for assessing
walking ability in patients post-stroke.
Postural Assessment Scale for Stroke (PASS) considers the patient’s postural control at
the activity dimension of the ICF (International Classification of Functioning, Disability and
Health). PASS contains 12 four-level items (0-36 points) that assess the ability to maintain a
given posture in sitting and standing and to maintain equilibrium during positional changes
in supine and standing and from sitting to standing. A higher score indicates better ability
(Benaim et al., 1999). PASS has been found to have good psychometric properties and
predictive ability for ADL function 1 year post-stroke (Benaim et al., 1999; Chien et al., 2007).
Trunk Impairment Scale-modified Norwegian version (TIS-modNV) considers the
patients’ level of trunk control at the body structure- and function dimension of the ICF. It is
a six-item scale (0 -16 points) used to assess impairments of trunk control in sitting post-
stroke, and has demonstrated adequate psychometric properties (Gjelsvik et al., 2012). A
higher score indicates less impairment.
Functional Ambulation Category (FAC) distinguishes six levels of walking ability based on
level of dependence (Holden et al., 1984; Kollen, Kwakkel and Lindeman, 2006). A score of 5
indicates completely independent walking and 0 indicates no functional walking. FAC has
been found to have excellent reliability, good concurrent and predictive validity, and good
responsiveness in patients with hemiparesis after stroke (Mehrholz et al., 2007).
Disability and stroke severity
Barthel Index (BI) is used to assess personal activities of daily living (ADL) and reflects
aspects of self-care such as continence, personal hygiene, dressing and basic mobility. Eight
of the 10 items represent activities related to personal care, and two are related to mobility.
6
The index yields scores from 0-100; a higher score indicates greater degree of functional
independence (Mahoney and Barthel, 1965; Salter et al., 2005). The test is considered to
have good reliability and validity for patients post-stroke (Salter et al., 2005).
Modified Rankin Scale (mRS) is a 7-point scale used to evaluate both limitations in
activity and changes in lifestyle. Grade 0 describes patients without symptoms, 5 indicate
severe disability, while grade 6 denotes death (van Swieten et al., 1988). The mRS is a simple
measure with well-studied reliability for patients post-stroke (Salter et al., 2005).
National Institutes of Health Stroke Scale (NIHSS) is used to assess stroke severity by
consciousness, vision, language, sensory function, ataxia, and arm and leg motor function on
an 11-item scale (0-34 points). Increasing scores represent increasing stroke severity. The
NIHSS is a validated scale that provides an overall stroke impairment score (Brott et al.,
1989; Goldstein and Samsa, 1997; Williams, Yilmaz and Lopez-Yunez, 2000).
2.4. Statistical analyses
Statistical analysis was performed using SPSS (IBM SPSS Statistics for Windows, Version
22.0. Armonk, NY: IBM Corp. US). Simple regression models were used to assess the
unadjusted effect of each baseline independent variable, and a stepwise multiple linear
regression (backwards) model was used to assess the adjusted effect and to find the variable
that best predicts PWS at six months. Variables were kept in the backwards model if p<0.1,
and variables with a p<0.05 were considered a significant predictor. The regression models
were run separately for the participants that could walk and produce a walking speed at
baseline, and for the participants that could not. Multicollinearity was assessed by inspecting
the variance inflation factor (VIF) and a VIF < 5 was considered acceptable.
Cut-off points for the significant predictor variables to walk at 0.8 and 1 m/s six months
post-stroke were analysed by Receiver Operating Characteristics (ROC). Optimal cut-off
points were found by using the highest sum of specificity and sensitivity from the curve. As
for the regression models, the ROC were run separately for the two sub-samples.
3. Results
Of the 306 participants originally included in the ESD Stroke Bergen-project, 132 (43%)
were included in this study. The reasons for non-inclusion were that 40 participants (13%)
had dropped out from the ESD Stroke Bergen-project and 16 participants (5%) died before
7
six months. This left 250 participants from the RCT potentially available for inclusion.
However, of these 118 participants lacked walking data from baseline and/or six months
post-stroke and could not be included. 31 of the included participants did not have sufficient
walking ability for PWS to be recorded at baseline. As a result we divided the sample in two
separate sub-groups for the analyses, namely the group that had sufficient walking ability for
PWS to be recorded at baseline (WSB group; n=101) and the group that did not (NoWSB
group; n=31).
The WSB group was tested on average 5 (±2) days after stroke onset. The sub-group
contained more men than women, and had a mean age of 69 years. At baseline 74
participants (74%) were categorised by FAC as having independent gait. Seventeen
participants (17%) used some type of walking aid, and the average PWS for the group was
0.95 m/s. 75 participants (75%) had a PWS faster than 0.8 m/s and 46 participants (46%)
walked faster than 1 m/s. At six months post-stroke 99 participants (99%) were categorised
as having independent walking ability. Sixteen participants (16%) used some type of walking
aid, and the average PWS for the group was 1.12 m/s. Eighty-four (84%) participants had a
PWS faster than 0.8 m/s and 67 (67%) participants walked faster than 1 m/s.
The NoWSB group was also tested on average 5 (±2) days after stroke onset, contained
more men than women, and the mean age was 71 years. At baseline 13 participants (42%)
had no walking ability, while the rest of the group had dependent gait. Twenty-eight
participants (90%) used some type of walking aid. At six months post-stroke 23 participants
(74%) were categorised as having independent walking ability and two participants (7%) had
no walking ability. Twenty participants (65%) used some type of walking aid, and the average
PWS for the group was 0.74 m/s. Fifteen (48%) participants had a PWS faster than 0.8 m/s
and nine (29%) walked faster than 1 m/s. Descriptives for the separate sub-samples are
summarised in Table 1 and outcome variables are summarised in Table 2. XY-plots of the
relation between the baseline variables and PWS at six months post-stroke are presented for
both sub-samples in Figure 1 A-M.
The simple regression models for the WSB group demonstrated that all baseline variables
(age, PASS, PWS, gender, TIS-modNV, mRS, and BI) except NIHSS were significantly
associated with PWS at six months, with age, PASS and PWS having the largest explained
variances. For the NoWSB group simple regression models established that only PASS, TIS-
modNV and BI were significantly associated with PWS at six months post-stroke with PASS
8
and BI having the largest explained variances. Detailed results from the regression analyses
are presented in Table 3.
All independent variables were included in stepwise multiple linear regression
(backwards) models. For the WSB group the final model gave an explained variance of 0.77.
The variables left were PASS, age and PWS, and all being significant predictors of PWS at six
months. For the NoWSB group only PASS remained in the final model and was a highly
significant predictor. The explained variance was 0.49. No VIF in any of the models were
found higher than 5 indicating that there was no concern about multicollinearity. Detailed
results from the regression analyses are presented in Table 3.
For the WSB group area under the ROC curves (AUC) for the predicted model (PASS, PWS
and age combined) was 0.95 for 0.8 m/s at six months, and 0.92 for 1 m/s at six months. Cut-
off values and AUC for PASS, PWS and age separately for the ability to walk 0.8 and 1 m/s at
six months post-stroke are presented in Table 4. To walk faster than 0.8 m/s at six months
post-stroke, PASS should be more than 30.5 points and PWS faster than 0.75 m/s for people
less than 73.5 years of age at baseline. To walk faster than 1 m/s at six months post-stroke
PASS should be more than 30.5 points and PWS faster than 0.93 m/s for people less than
75.5 years of age at baseline. For the NoWSB group to walk faster than 0.8 or 1 m/s at six
months post-stroke PASS should be more than 20.5 points. Cut-off values and AUC for PASS
for the ability to walk 0.8 and 1 m/s at six months post-stroke are presented in Table 4.
4. Discussion
In this study the aim was to investigate to which degree typical clinical measures of
stroke severity, disability and physical function the first week post-stroke are associated with
PWS at six months post-stroke. We found baseline PASS, PWS and age to significantly predict
PWS at six months post-stroke for participants in the WSB group, and the variables explained
77% of the variance in PWS. For the participants that did not have sufficient walking ability
to produce a walking speed (NoWSB group) in the first week post-stroke, only baseline PASS
significantly predicted PWS at six months. PASS alone explained 49% of the variance in PWS.
Compared to the WSB group, the NoWSB group was older, had a higher percentage of men
and a higher percentage had suffered a haemorrhagic stroke.
9
It is not surprising that age and PWS at baseline predict how fast a person walks six
months post-stroke for the WSB group. It has previously been reported that age is
associated with WS (Lusardi, 2012), and it could also be expected that the PWS in the first
week post-stroke is indicative of how fast that person walks six months later. It is interesting
however, that age was not a significant predictor in the NoWSB group. This suggests that
when walking ability is poor, age plays a lesser role for predicting long-term walking ability.
The present study is to our knowledge, the first study that has investigated and found an
association between baseline PASS and long-term PWS, indicating that PASS score at
baseline is predictive for walking ability long-term post-stroke. This is somewhat surprising
as walking is not included in any of the PASS items, but is of particular interest as PASS can
be used with all patients independent of walking ability in the early phase after stroke. Di
Monaco and colleagues (Di Monaco et al., 2010) have previously found that PASS was
predictive of functional ability and destination at discharge in patients post-stroke, and that
the TIS was almost as good a predictor. In this present study however, the TIS-modNV was
not found to be a significant predictor for PWS in the multiple regression models. TIS was
designed to measure ADL related selective trunk movements in sitting rather than the
involvement of the trunk in gross transfer movements (Verheyden et al., 2006). PASS
however, includes both transfers between postures and activities in a vertical position,
including standing on one leg on both sides. As walking is a vertical activity where single
stance is necessary to take steps and move forward, the ability to stand on one leg is of
special interest in persons post-stroke as they often present with hemiparesis. WS is
determined by the combination of step length and step frequency, and specifically step
length can be affected by difficulties standing on one leg. Also, the balance demands are
higher in a vertical position than in sitting, and PASS therefore may have a better ability to
predict balance difficulties during walking than TIS-modNV. These factors may be the reason
why PASS was found to be a better predictor for PWS than TIS-modNV. A strong relationship
between PASS and ambulation has also been found in previous studies. Lin and colleagues
(Lin et al., 2010) reported strong associations between functional ambulation and PASS at
baseline, two months and five months post-stroke respectively, while Benaim and colleagues
(Benaim et al., 1999) found strong associations between PASS and the locomotion part of
the Functional Independence Measure. O’Dell and colleagues (O'Dell et al., 2013) however,
found only weak associations between the admission PASS scores and discharge WS.
10
We used PWS as a continuous variable and FAC was only used for descriptive purposes.
Using continuous variables allows for more sensitive and sophisticated analyses, but a
challenge with using WS is how to treat the participants that could not walk. Clinically, this is
a particularly interesting sub-group, and excluding the participants from analyses would be
problematic. In this study we therefore performed separate analyses for those participants
who could and could not produce a walking speed at baseline. Inspecting the differences
between these sub-groups, it would seem that the NoWSB group improved more than the
WSB group on all parameters (Table 2). Also, 48% percent in the NoWSB group reached a
community ambulation walking speed of 0.8 m/s, and 29% reached 1.0 m/s, which is widely
regarded as a meaningful indicator of health and functioning. It stands to reason that those
with poor initial function had the greatest potential for improvement. However, a greater
proportion of the NoWSB group was discharged to rehabilitation units and not directly to
their own homes. This may have contributed to the improvement. This finding could
emphasize the importance of targeting stroke patients with low function.
We used PWS at six months post-stroke as an indication of community walking in this
study. This approach has been criticised as WS is only one of many aspects of community
walking, and indoor walking speed on short distances may overestimate community WS for
the slow walkers (Lord and Rochester, 2008; Carvalho, Sunnerhagen and Willen, 2010).
However, although WS is not the sole contributor to community walking, it has been found
to be the best predictor (van de Port, Kwakkel and Lindeman, 2008; An, Lee, Shin and Lee,
2015). WS is easily measured in almost any setting with few requirements for space and
equipment. Assessing community walking directly is more challenging and is commonly done
by using questionnaires or with accelerometer based activity monitors that provide
information about steps taken during real-life behaviour. As neither was used in this study,
we selected PWS at six months post-stroke as a measure to infer capability of community
walking. As presented in the introduction, these are often-used cut-offs or intercepts, where
speeds higher than 0.8 m/s indicate community walking ability, while speeds higher than 1
m/s indicate low risk for adverse health events. However, as intercepts for community
walking in different post-stroke populations differs between studies, and the health risk
profiles are predominantly based on research investigating elderly both with and without
disabilities, further studies are needed to investigate if 0.8 and 1 m/s are optimal intercepts
for community walking and health profile in persons post-stroke. Moreover, when assessing
11
cut offs for these intercepts we found that the WSB group had to score more than 30.5
points on the PASS both to be able to walk 0.8 and 1.0 m/s while the same cut-offs found for
the NoWSB group were 20.5 points on the PASS. This indicates that different cut-offs for
these parameters should be used for different sub-groups post-stroke. However, this may
also indicate that to find robust cut-offs for WS post-stroke; more diverse groups with a
variety of functional abilities should be analysed. In this study, as the sample was sub-
divided into two groups based on walking ability the groups were different with regards to
function, and therefore the cut-offs found for walking at 0.8 or 1.0 m/s had to be different.
The stroke severity measure (NIHSS) and the indexes for disability (mRS) and basic
functions (BI) are commonly used in stroke research as predictors for outcome. However,
these variables were not found to be significant predictors of PWS at six months post-stroke
in this study. The NIHSS is a measure of body structure and functions, and may be less
predictive than measures of activity such as PWS. The BI has often been found to have a
ceiling effect (Harrison, McArthur and Quinn, 2013), a tendency that can be observed also in
this study for the WSB group (Figure 1L) resulting in less sensitivity in predicting PWS at six
months post-stroke as participants scoring high on the BI at baseline ended up with a range
of different PWS’s at six months. For the NoWSB group however, there was not an
observable ceiling effect (Figure 1M), but BI was still not predictive of PWS at six months.
This study was done in the context of a larger RCT, but only the participants that were
still included in the RCT and that had been assessed with FAC and PWS could be recruited to
answer the research question in this present study. The sample therefore consisted of only
43% of the participants from the RCT. The main reason was that many participants (39%) in
the RCT lacked some or all of the walking speed data. The reason for this was a combination
of practical and clinical reasons. PWS was not included as an outcome measure in the ESD
Stroke Bergen-project initially; therefore the first 23 participants were not tested for PWS. In
addition, some participants could not or did not wish to come into the hospital for testing at
six months, and was tested in their homes. For these participants WS could not be tested
due to limitation in space (a total length of 8 meters was required). Also for a few
participants with very low functional ability, walking was not tested. However, comparing
the sample in this present study with the RCT as described in two previous articles (Hofstad
et al., 2014; Hofstad et al., 2016) the baseline characteristics of the samples are similar
although some differences can be seen. In the RCT sample (n=306) the participants were on
12
average 72 years, 55% men, 55% living with a partner and 88% had had an infarction type
stroke. On the NIHSS they scored 4.0, on the BI 77.3 and on the mRS they scored 2.6. This
indicates that both our sub-samples were slightly younger and more participants were living
with a partner. It was also more men in the NoWSB group compared both to the RCT sample
and the WSB group. For type of stroke, NIHSS, BI and mRS the WSB group in this study had a
higher proportion of infarction type strokes, lower NIHSS, higher BI and smaller mRS than
the RCT sample. The NoWSB group on the other hand had a higher proportion of
haemorrhage type strokes, higher NIHSS, lower BI and higher mRS than the RCT sample.
A strength of this study was few missing data for the 132 participants included, with only
NIHSS and BI each missing data from two participants. For the rest of the variables data were
complete (Table 3).
Further studies are needed to see if these results are replicable in other samples of
stroke survivors, and also to investigate whether the cut-offs found in this study in fact do
predict not only PWS, but also community walking and health. To gain such knowledge, PWS
at baseline could be explored as a predictor how much the people post-stroke walks based
on real-life activity monitors and for self-report variables concerning community walking.
5. Conclusion
Baseline PASS, PWS and age significantly predict PWS at six months post-stroke for
participants that had PWS at baseline (WSB group), and explained 77% of the variance in
PWS. For the participants that did not have sufficient walking ability to produce a walking
speed (NoWSB group) in the first week post-stroke, only baseline PASS significantly
predicted PWS at six months post-stroke. PASS explained 49% of the variance in PWS. In the
WSB group a PASS score higher than 30.5 indicated an ability to walk faster than 0.8 m/s
after six months. For the participants in the NoWSB group however, a PASS score higher
than 20.5 points was sufficient for reaching a PWS faster than 0.8 m/s. Clinically, it is
particularly interesting that PASS was a good predictor for long-time PWS independent of
walking ability at baseline. This indicates that PASS can be used by clinicians to recognise the
long-term prospect for a patient to regain sufficient WS for community walking and to guide
rehabilitation needs.
13
6. References
Abellan van Kan, G, et al. (2009). "Gait Speed at Usual Pace as a Predictor of Adverse
Outcomes in Community-Dwelling Older People An International Academy on
Nutrition and Aging (IANA) Task Force." The journal of nutrition, health and aging
13(10): 881-889.
An, S, et al. (2015). "Gait velocity and walking distance to predict community walking after
stroke." Nursing and Health Sciences 17(4): 533-538.
Barclay, R, J Ripat and N Mayo (2015). "Factors describing community ambulation after
stroke: a mixed-methods study." Clinical Rehabilitation 29(5): 509-521.
Benaim, C, et al. (1999). "Validation of a standardized assessment of postural control in
stroke patients: the Postural Assessment Scale for Stroke Patients (PASS)." Stroke
30(9): 1862-1868.
Bijleveld-Uitman, M, I van de Port and G Kwakkel (2013). "Is gait speed or walking distance a
better predictor for community walking after stroke?" Journal of Rehabilitation
Medicine 45(6): 535-540.
Brott, T, et al. (1989). "Measurements of acute cerebral infarction: a clinical examination
scale." Stroke 20(7): 864-870.
Carvalho, C, KS Sunnerhagen and C Willen (2010). "Walking speed and distance in different
environments of subjects in the later stage post-stroke." Physiotherapy Theory and
Practice 26(8): 519-527.
Chien, CW, et al. (2007). "A comparison of psychometric properties of the smart balance
master system and the postural assessment scale for stroke in people who have had
mild stroke." Archives of Physical Medicine and Rehabilitation 88(3): 374-380.
Di Monaco, M, et al. (2010). "The relationship between initial trunk control or postural
balance and inpatient rehabilitation outcome after stroke: a prospective comparative
study." Clinical Rehabilitation 24(6): 543-554.
Fritz, S and M Lusardi (2009). "White paper: "Walking speed: the sixth vital sign"." Journal of
Geriatric Physical Therapy 32(2).
Gjelsvik, B, et al. (2012). "The Trunk Impairment Scale - modified to ordinal scales in the
Norwegian version." Disability and Rehabilitation 34(16): 1385-1395.
Gjelsvik, BE, et al. (2014). "Balance and walking after three different models of stroke
rehabilitation: early supported discharge in a day unit or at home, and traditional
treatment (control)." BMJ Open 4(5): e004358.
Goldstein, LB and GP Samsa (1997). "Reliability of the National Institutes of Health Stroke
Scale. Extension to non-neurologists in the context of a clinical trial." Stroke 28(2):
307-310.
Harrison, JK, KS McArthur and TJ Quinn (2013). "Assessment scales in stroke: clinimetric and
clinical considerations." Clin Interv Aging 8: 201-211.
Hofstad, H, et al. (2014). "Early supported discharge after stroke in Bergen (ESD Stroke
Bergen): three and six months results of a randomised controlled trial comparing two
early supported discharge schemes with treatment as usual." BMC Neurology 14(1):
239.
Hofstad, H, et al. (2016). "Subjective health complaints predict functional outcome six
months after stroke." Acta Neurol Scand.
Hofstad, H, et al. (2013). "Early supported discharge after stroke in Bergen (ESD Stroke
Bergen): a randomized controlled trial comparing rehabilitation in a day unit or in the
14
patients' homes with conventional treatment." International Journal of Stroke 8(7):
582-587.
Holden, MK, et al. (1984). "Clinical Gait Assessment in the Neurologically Impaired: Reliability
and Meaningfulness." Physical Therapy 64(1): 35-40.
Jette, DU, et al. (2005). "Physical therapy interventions for patients with stroke in inpatient
rehabilitation facilities." Phys Ther 85(3): 238-248.
Kollen, B, G Kwakkel and E Lindeman (2006). "Time Dependency of Walking Classification in
Stroke." Physical Therapy 86(5): 618-625.
Lin, JH, et al. (2010). "Psychometric comparisons of 3 functional ambulation measures for
patients with stroke." Stroke 41(9): 2021-2025.
Liu, T-W, et al. (2015). "Fear Avoidance Behavior, Not Walking Endurance, Predicts the
Community Reintegration of Community-Dwelling Stroke Survivors." Archives of
Physical Medicine and Rehabilitation 96(9): 1684-1690.
Lord, S and L Rochester (2008). "Gait velocity and community ambulation: the limits of
assessment." Stroke 39(4): e75; author reply e76.
Lord, SE, et al. (2004). "Community ambulation after stroke: how important and obtainable
is it and what measures appear predictive?" Archives of Physical Medicine and
Rehabilitation 85(2): 234-239.
Lusardi, MM (2012). "Using Walking Speed in Clinical Practice: Interpreting Age-, Gender-,
and Function-Specific Norms." Topics in Geriatric Rehabilitation 28(2): 77-90
10.1097/TGR.1090b1013e31823d31827b31829f.
Mahoney, FI and DW Barthel (1965). "Functional Evaluation: The Barthel Index." Maryland
state medical journal 14: 61-65.
Mayo, NE, NA Korner-Bitensky and R Becker (1991). "Recovery time of independent function
post-stroke." American journal of physical medicine and rehabilitation 70(1): 5-12.
Mayo, NE, et al. (1999). "Disablement following stroke." Disability and Rehabilitation 21(5-6):
258-268.
Mehrholz, J, et al. (2007). "Predictive validity and responsiveness of the functional
ambulation category in hemiparetic patients after stroke." Archives of Physical
Medicine and Rehabilitation 88(10): 1314-1319.
O'Dell, MW, et al. (2013). "A comparison of two balance measures to predict discharge
performance from inpatient stroke rehabilitation." PM & R : the journal of injury,
function, and rehabilitation 5(5): 392-399.
Perry, J, et al. (1995). "Classification of walking handicap in the stroke population." Stroke
26(6): 982-989.
Peurala, SH, et al. (2014). "Evidence for the effectiveness of walking training on walking and
self-care after stroke: a systematic review and meta-analysis of randomized
controlled trials." J Rehabil Med 46(5): 387-399.
Rosa, MC, et al. (2015). "Fast gait speed and self-perceived balance as valid predictors and
discriminators of independent community walking at 6 months post-stroke--a
preliminary study." Disability and Rehabilitation 37(2): 129-134.
Salter, K, et al. (2005). "Issues for selection of outcome measures in stroke rehabilitation: ICF
activity." Disability and Rehabilitation 27(6): 315-340.
Tilson, JK, SM Settle and KJ Sullivan (2008). "Application of evidence-based practice
strategies: current trends in walking recovery interventions poststroke." Topics in
Stroke Rehabilitation 15(3): 227-246.
15
Turnbull, GI, J Charteris and JC Wall (1995). "A comparison of the range of walking speeds
between normal and hemiplegic subjects." Scandinavian Journal of Rehabilitation
Medicine 27(3): 175-182.
van de Port, IG, G Kwakkel and E Lindeman (2008). "Community ambulation in patients with
chronic stroke: how is it related to gait speed?" Journal of Rehabilitation Medicine
40(1): 23-27.
van Swieten, JC, et al. (1988). "Interobserver agreement for the assessment of handicap in
stroke patients." Stroke 19(5): 604-607.
Veerbeek, JM, et al. (2011). "Is Accurate Prediction of Gait in Nonambulatory Stroke Patients
Possible Within 72 Hours Poststroke? The EPOS Study." Neurorehabilitation and
Neural Repair 25(3): 268-274.
Verheyden, G, et al. (2006). "Trunk performance after stroke and the relationship with
balance, gait and functional ability." Clinical Rehabilitation 20(5): 451-458.
Williams, LS, EY Yilmaz and AM Lopez-Yunez (2000). "Retrospective Assessment of Initial
Stroke Severity With the NIH Stroke Scale." Stroke 31(4): 858-862.
Wing, K, JV Lynskey and PR Bosch (2012). "Walking Speed in Stroke Survivors: Considerations
for Clinical Practice." Topics in Geriatric Rehabilitation 28(2): 113-121.
16
Table 1. Descriptives
for the walking speed at baseline group (WSB group) and the no walking speed at
baseline group (NoWSB group).
Variables
WSB group
NoWSB group
Participants,
n
101
31
Age Years (SD)
69 (11.9)
71 (11.3)
Gender Male,
n (%)
55 (55)
21
(68)
Living with partner,
n (%)
64 (63)
20 (65)
Type stroke,
n (%)
Infarction
95 (94)
24 (77)
Haemorrhage
6 (6)
7 (23)
Location stroke,
n (%)
Right
39 (39)
11 (36)
Left
46 (45)
13 (42)
Bilateral
4 (4)
1 (3)
Brainstem
8 (8)
6 (19)
Cerebe
llum
4 (4)
0 (0)
Side of body affected,
n (%)
Right
58 (57)
15 (48)
Left
41 (41)
16 (52)
Bilateral
2 (2)
0 (0)
Days in stroke unit, mean (SD)
10 (5.4)
15 (6.5)
Initial discharge destination, n (%)
Home
78 (77)
5 (16)
Rehabilitation
12 (12)
13 (42)
Nursing home
11 (11)
12 (39)
Other institution
0 (0)
1 (3)
WSB group: participants with walking speed at baseline. NoWSB group: participants with no walking speed at
baseline. SD: standard deviation. n: number.
17
Table 2. Outcome measures at baseline and six months for the walking speed at baseline group (WSB
group) and the no walking speed at baseline group (NoWSB group).
WSB
-
group
NoWSB
-
group
Outcomes
Baseline
6 months
Baseline
6 months
Postural Assessment Scale, mean (SD)
31.5 (3.5)
32.
7 (3.4)
20.7 (7.8)
29.1 (5.2)
Preferred walking speed m/s, mean (SD)
0.95 (0.31)
1.12 (0.35)
NA
0.74 (0.46)
TIS m
-
NV, mean (SD)
10.6 (3.6)
13.0 (3.1)
5.5 (3.8)
9.8 (3.3)
Modified Rankin Scale, mean (SD)
2.0 (1.0)
1.8 (1.0)
3.5 (0.9)
2.4 (1.2)
0 No symp
toms n (%)
6 (6)
9 (9)
0 (0)
2 (7)
1 No significant disability n (%)
21 (21)
31 (31)
1 (3)
3 (10)
2 Slight disability
n (%)
48 (48)
34 (34)
5 (16)
13 (42)
3 Moderate disability
n (%)
19 (19)
25 (25)
2 (7)
6 (19)
4 Moderate severe disability
n (%)
7 (7)
2 (2)
23 (74)
7 (23)
5 Severe disability
n (%)
0 (0)
0 (0)
0 (0)
0 (0)
NIHSS points, mean (SD)
3.8 (3.1)
1.5 (2.0)
6.8 (4.9)
3.7 (3.1)
Barthel Index,
mean
(SD)
93 (14)
96 (10)
56 (26)
89 (19)
95
-
100 Independence , n (%)
75 (74)
86 (85)
3 (10)
21 (68)
<60
Major dependence, n (%)
20 (20)
13 (13)
13 (42)
7 (23)
60
-
94 Minor dependence, n (%)
5 (5)
2 (2)
14 (45)
3 (10)
Functional Ambulatory Category,
n (%)
5 Independent
48 (48)
91 (90)
0 (0)
19 (61)
4 Independent (level ground)
26 (26)
8 (8)
0 (0)
4 (13)
3 Dependent on supervision
23 (23)
1 (1)
0 (0)
3 (10)
2 Dependent 1
4 (4)
1 (1)
4 (45)
1 (3)
1 Dependent 2
0 (0)
0 (0)
14 (11)
1 (3)
0 No walking ability
0 (0)
0 (0)
13 (42)
2 (7)
Using walking aids Yes, n (%)
17 (17)
16 (16)
28 (90)
20 (65)
No walking aids
84 (83)
85 (84)
3 (10)
20 (65)
Stick or one crutch
5 (5)
5 (5)
1 (3)
4 (13)
Two crutches
3 (3)
2 (2)
0 (0)
0 (0)
Rollator
9 (9)
1 (1)
1 (3)
1 (3)
Wheelchair
0 (0)
1 (1)
26 (84)
6 (19)
WSB group: participants with walking speed at baseline. NoWSB group participants with no walking speed at
baseline. SD: standard deviation. n: number. NIHSS: National Institutes of Health Stroke Scale. TIS m-NV: Trunk
Impairment Scale modified-Norwegian version.
18
Table 3. Simple regression models and stepwise multiple linear regression (backwards) models to assess baseline variables as predictors of
walking speed at 6 months for the walking speed at baseline group (WSB group, grey area) and the no walking speed at baseline group (NoWSB group,
white area).
Baseline WSB group NoWSB group
n
R
2
Unadj
.
B
p
-
v
alue
Adj
.
B
95%CI
p
-
value
n
R
2
Unadj
.
B
p
-
v
alue
Adj
.
B
95%CI
p
-
value
PASS
101
0.
49
0.0
7
<.001
**
0.
02
(0.01
–
0.03)
.005
**
31
0.
52
0.0
4
<.001
**
0.
04
(0.02
–
0.06)
<.001
**
Age
101
0.
30
-
0.0
2
<.001
**
-
0.01
(
-
0.01
–
0.00)
.001
**
31
0.
00
0.0
0
.786
PWS
101
0.
68
0.
93
<.001
**
0.64
(0.50
–
0.79)
<.001
**
NA
Gender
101
0.0
9
-
0.
21
.0
03
**
31
0.0
5
-
0.
22
.214
TIS
m
-
NV
101
0.
1
4
0.0
4
<.001
**
31
0.
13
0.0
4
.044
*
mRS
101
0.1
0
-
0.1
2
.001
**
31
0.04
-
0.1
0
.314
NIHSS
99
0.0
0
0.0
0
.
949
31
0.0
6
-
0.0
2
.
176
BI
100
0.
20
0.01
<.001
**
30
0.
35
0.01
.001
**
n= 98, R
2
:
0.
77
n= 30 , R
2
: 0.49
Unadjusted effect is based on simple regression and adjusted B is from the final regression model from the backward analyses. **p<0.01. *p<0.05. WSB group: participants
with walking speed at baseline. NoWSB group participants with no walking speed at baseline. PASS: Postural Assessment Scale for Stroke. PWS: Preferred walking speed. TIS
m-NV: Trunk Impairment Scale modified-Norwegian version . mRS: Modified Rankin Scale. NIHSS: National Institutes of Health Stroke Scale. BI: Barthel Index. n: Number of
participants included in the model. R
2
: Explained variance. B: unstandardized beta. 95%CI: the confidence interval for the B in the final model. NA: Not applicable.
19
Table 4. Results from the Receiver Operating Curve (ROC)-analyses for finding cut-off points at baseline for predicting walking speed at six months for the
walking speed at baseline group (a) and the no walking speed at baseline group (b).
a) Walking speed at baseline group (WSB group) n=101
0.8 m/s at six months
(84 participants walked >0.8 m/s)
1 m/s at six months
(67 participants walked >1 m/s)
Baseline
measures
Cut
-
Off value
AUC (95% CI)
Sens %
Spes %
Cut
-
Off value
AUC (95% CI)
Sens %
Spes %
PASS
30
.5
points
0.9
1
(0.8
5
–
0.9
7
)
74
94
30.5
points
0.
90
(0.
8
3
–
0.9
7
)
87
85
PWS
0.
75
m/s
0.
96
(0.
92
–
1
.
00
)
8
9
9
4
0.
93
m/s
0.
93
(0.8
8
–
0.
9
8
76
94
Age
7
3
.5
years
0.
84
(0.
75
–
0.
93
)
7
1
88
7
5
.5
years
0.
83
(0.
74
–
0.
92
)
87
7
1
Predicted
model
0.
9
5
(0.
90
–
0.9
9
)
85
9
4
0.9
2
(0.8
7
–
0.9
8
)
87
88
PASS: Postural Assessment Scale for Stroke. PWS: Preferred walking speed. AUC: Area under the curve. Sens: Sensitivity. Spes: Specificity.
Predicted model: PASS, PWS and age all included in one model to calculate AUC, sensitivity and specificity for the model
b) No walking speed at baseline group (NoWSB group) n=31
0.8 m/s at six months
(15 participants walked >0.8 m)
1 m/s at six months
(9 participants walked >1 m/s)
Baseline measures
Cut
-
Off
value
AUC (95% CI)
Sens %
Spes %
Cut
-
Off value
AUC (95% CI)
Sens %
Spes %
PASS
20
.5 points
0.
87
(0.74
–
1.00
)
8
7
81
20
.5 points
0.
77
(0.
60
–
0.9
4
)
89
64
PASS: Postural Assessment Scale for Stroke. PWS: Preferred walking speed. AUC: Area under the curve. Sens: Sensitivity. Spes: Specificity.
20
WSB: Walking speed at baseline. NoWSB: No walking speed at baseline. TIS m-NV: Trunk Impairment Scale modified-
Norwegian version. NIHSS: National Institutes of Health Stroke Scale. PASS: Postural Assessment Stroke Scale. R2:
Explained variance.
WSB group
(n=101)
NoWSB group
(n=31)
A
) PASS
(R
2
=
0.49)
B
)
PASS
(R
2
=
0.52)
C
) Age
(R
2
=
0.30)
D
) Age
(R
2
=
0.00)
E
)
Walking speed
(R
2
=
0.68)
F
) TIS m
-
NV
(R
2
=
0.14)
G
) TIS m
-
NV
(R
2
=
0.13)
21
H
)
Modified Rankin Scale
(
mRS
)
(R
2
=
0.10)
I)
Modified Rankin Scale
(
mRS
)
(R
2
=
0.04)
J
) NIHSS
(R
2
=
0.00)
K
) NIHSS
(R
2
=
0.06)
L
)
Barthel Index (
BI
)
(R
2
=
0.20)
M
)
Barthel Index (BI)
(R
2
=
0.35)
Figure 1A-M. XY-plots of the relation between the baseline variables (x-axis) and walking speed at six
months (y-axis) for the walking speed at baseline group (left, red dots) and the no walking speed at
baseline group (right, blue dots).