Activity levels during inpatient stroke rehabilitation
to allow data comparison across devices, studies and
populations (13, 14). The feasibility of using activity
sensors that allow raw data handling in clinical settings
needs to be evaluated further.
The physical environment and support provided are
factors that influence activity levels early after stroke
(15, 16). In many rehabilitation wards scheduled th-
erapy sessions with physiotherapists and occupational
therapists are provided only on weekdays (2). Even
when the patients have the opportunity to carry out self-
training exercises with or without assistance and are
encouraged to be active at weekends, the activity levels
might be lower at weekends. In addition, weekends can
be perceived by patients and staff as days of rest, which
may influence the activity levels (12). A recent study
in chronic stroke found differences between weekday
and weekend data for walking time and step count, with
more activity occurring on weekdays, whereas time in
sitting, standing and in light or moderate PA did not
differ (17). Real-world activity and how this activity
is distributed over a week during stroke rehabilitation
are still not well described (12).
The primary aim of this study was to quantify arm,
leg and trunk activity in people with hemiparesis after
stroke during inpatient rehabilitation, and to determine
whether there were differences in activity levels bet-
ween weekdays and weekends. The hypothesis was
that the activity levels are lower at weekends when
scheduled activities are limited. It was also expected
that the activity levels would be lower in stroke compa-
red with healthy controls. The feasibility of the sensor
measurement in terms of comfort, acceptance and
management in the clinical setting was also explored.
METHODS
Participants
Participants were recruited consecutively from an inpatient
rehabilitation ward at Sahlgrenska University Hospital for
an 8-month period during 2015 and 2017. Preliminary power
analysis was not possible, since no data on the accelerometer
measure used in the current study have been published earlier.
Confirmatory power analysis was performed based on the ac-
celerometer data (m/s 2 ) from the arm and leg sensors after data
from 11 individuals was collected (18). To detect differences
between weekday and weekend sessions with 80% power and
alpha level equal to 0.05 and accounting for 10% missing data,
28 participants with stroke were required.
In total, 28 individuals with stroke were included in this cross-
sectional study. The inclusion criteria were: first-ever ischaemic
or haemorrhagic stroke, age 18 years or older, having the ability
to walk with or without assistance, and not receiving a full score
on the Fugl-Meyer Assessment of the affected arm or leg. Indi-
viduals with another condition affecting arm or leg function, or
having severe multi-impairment prior to stroke or malignancy,
and those unable to understand verbal instructions in Swedish
427
or English were excluded. All participants with stroke followed
an individual rehabilitation plan in accordance with the Natio-
nal Swedish Stroke Guidelines. Each patient had an individual
time schedule that included at least one 45-min session with a
physiotherapist and 1 with an occupational therapist per day, 5
days per week, together with group activities (walking, gaming)
and individual therapy with other rehabilitation team members
(e.g. speech therapist). In the ward, an equipped therapy room,
gaming room and common room for meals, as well as a therapy
garden and walking paths outside the hospital building, were
available for day-time use. Patients were also encouraged to
perform self-training with or without assistance from the nursing
staff or next of kin at weekends.
A convenience sample of 10 healthy individuals with varying
occupation, age and sex was included as a healthy control group
to allow better interpretation of the activity levels in individuals
with stroke. Healthy participants were included if they did not
report any medical, neurological or musculoskeletal disorders
affecting their motor function or everyday physical activity and
were able to communicate in Swedish or English.
Ethical approval for the study was provided by the Regional
Ethical Review Board in Gothenburg, Sweden (507-15). Written
and verbal informed consent was received from all participants
prior to participation in the study.
Activity monitoring procedure
Accelerometer data were collected using a set of 5 3-axial ac-
celerometers (Shimmer 3, Shimmer, Dublin, Ireland) during
2 separate 48-h sessions on weekdays and over a weekend,
respectively. The sensors were calibrated according to the
procedure from the supplier before use. The weekday measure-
ment was performed during any 2 consecutive days between
Monday and Friday. The sampling rate was set to 51.2 Hz, with
an accelerometer range of ± 8 g. Accelerometers (51 × 34 × 14
mm, weight 24 g) were fastened with customized Velcro straps
on the trunk, wrists and ankles. Participants were instructed to
wear the sensors during the entire session, including the night,
but the sensors could be removed at any time when necessary.
Since the sensors were not waterproof they had to be removed
for showering or swimming activities.
Participants were asked to record their main daily activities
using an activity log. For each hour between 08.00 h and
20.00 h participants recorded their main activities (e.g. eating,
taking a walk, training, preparing a meal, transport in a car,
working in the garden) and whether the activity was mostly
sitting, standing/walking, or lying/resting. Participants who
were unable to complete the log were interviewed during and
after the measurement period. The weekday schedule data,
including training sessions, were collected for the participants
with stroke. The logs were used to describe the participants’
main daily activities.
Acceleration data analysis
Data acquired from the sensors were extracted to Matlab soft-
ware (MathWorks Inc.) for custom-made analysis. Only the day-
time activity between 08.00 h and 20.00 h were extracted. All
data were visually inspected to identify segments in which the
sensor data contained no motion information or were missing.
Primarily, this occurred when the participant had removed one
or more sensors from their body, e.g. for taking a shower. Such
segments were documented and removed from the measurement
data. Corrupt measurement data time-stamps, possibly caused by
J Rehabil Med 51, 2019