428
M. Alt Murphy et al.
sensor synchronization malfunction, were reconstructed using
uniform resampling of the accelerometer data.
The length of the available data for each participant was then
calculated and needed to be at least 20 h out of a possible 24 h
to be included in the analysis. This criterion was applied to the
individual sensors for weekdays and weekends separately. For
calculation of ratios, data needed to be available for both limbs.
The extracted measurement data were filtered using a Butterworth
bandpass filter (0.2–10 Hz passband). The activity level was ex-
pressed as the Signal Magnitude Area (SMA), which is computed
from the 1-norm of the acceleration vector averaged over a fixed
epoch length (T = 120s) and, thus, resulting in 1 value each epoch
1
SMA = T ǀa x ǀ+ ǀa y ǀ+ ǀa z ǀdt.
Furthermore, the SMA ratio (more affected/less affected in
stroke or non-dominant/dominant in controls) was computed for
arms and legs as a measure of asymmetry. The logarithm of the
SMA ratios was used to obtain a measure that was symmetrical
with respect to both limbs. The value zero indicates perfect
symmetry between limbs, while a negative value indicates that
the SMA is lower in the affected limb or non-dominant limb.
The SMA for trunk, both arms and legs along with ratios were
averaged over subsequent days of measurement to produce a
single value for each weekday and weekend session.
ſ
Clinical assessments
In stroke, the upper and lower extremity sensorimotor function
was assessed using Fugl-Meyer Assessment (FMA) (19–21).
The maximum score of 66 and 34 for the upper and lower
extremity, respectively, indicates normal function. The FMA
sensation, passive range of motion and pain were also assessed.
The muscle tone in elbow, wrist and ankle joint was assessed
using the Modified Ashworth Scale (22). Walking ability was
dichotomized to independent (scores 4–5) and dependent (scores
0–3) according to Functional Ambulation Categories (FAC, 0–5
pts) (23, 24). The stroke type was gathered from medical charts
and hand dominance by participants’ own reporting.
Feasibility evaluation
At the end of each measurement session, participants were as-
ked to rate whether the sensors were comfortable to wear on a
5-point scale (strongly agree, agree, agree partly, do not agree,
strongly disagree) and describe their experience and perception
of the measurement. When the participants with stroke had dif-
ficulties expressing themselves their next of kin or staff at the
rehabilitation ward that had assisted the participant was intervie-
wed. Comments regarding the practical management of sensors
were also collected. This data were summarized descriptively.
Statistical analysis
Statistical analyses were conducted using IBM Statistical
Package for Social Sciences™ (SPSS) version 24. The alpha
value was set to 0.05 (2-tailed). The activity logs were analy-
sed descriptively. Since some accelerometer measures showed
non-normal distribution and the sample size was not very large,
non-parametric statistics were used.
To verify whether there was a difference in activity levels
between weekdays and weekends, Wilcoxon signed-rank test
was used. Wilcoxon signed-rank test was used to determine
differences between the more-affected and less-affected limb
in stroke and between the dominant and non-dominant limb in
controls. The Mann–Whitney U test was used to test whether
the activity levels in individuals with stroke were different from
controls. Possible interaction effect of dominant/non-dominant
affected hand on arm activity and independence/dependence in
walking on leg activity were verified by using between-within
subjects analysis of variance.
The relative magnitude of the differences between groups was
calculated using effect size estimates for non-parametric data
(r = z/√N) and Cohen’s guidelines were followed while inter-
preting the effect sizes, where 0.1, 0.3 and 0.5 indicate small,
medium and large effect sizes, respectively (25).
RESULTS
The measurement protocol was followed in all 28 par-
ticipants with stroke and 10 healthy controls. Among
individuals with stroke, no data were recorded in one-
participant due to battery malfunction and in another due
to non-adherence related to cognitive impairment. These
measurements were fully missing (all 5 sensors, both
sessions) and therefore excluded from the data analysis
(Table I). For the remaining measurements from 26 par-
ticipants with stroke and 10 controls (36 measurements
from 5 sensors in 2 sessions = 360 measurements), 11.9%
had less than 20 h data and were signified as incomplete
Table I. The summary of data availability showing missing and incomplete data
Trunk
More affected
arm a
Less-affected
arm b
Participants with stroke, n
Total number measured in 2 sessions
28
28
28
Missing (≤ 20 h in all 5 sensors, excluded from analysis) 2
2
2
Incomplete data (≤ 20 h), weekday (n = 26)
3
1
3
Incomplete data (≤ 20 h), weekend (n = 26)
2
4
5
Healthy controls, n
Total number measured in 2 sessions
10
10
10
Incomplete (≤ 20 h) weekday, (n = 10)
3
1
1
Incomplete (≤ 20 h) weekend (n = 10)
2
1
1
Incomplete data (stroke and controls, n = 36; 43 out of 360 measurements), %
Technical failure (38 measurements)
Human factor (5 measurements)
Missing and incomplete data (stroke and controls, n = 38; 63 out of 380 measurements), %
Technical failure (48 measurements)
Human factor (15 measurements)
a
Non-dominant limb in controls; b dominant limb in controls.
www.medicaljournals.se/jrm
More affected
leg a Less-affected
leg b All 5
sensors
28
2
2
4 28
2
2
4 28
2
11
19
10
1
0 10
1
2 10
7
6
11.9
10.5
1.4
16.5
12.6
3.9