ArticlePDF Available

Abstract and Figures

Background: The objective and accurate assessment of children’s sedentary and physical behavior is important for investigating their relation to health. The purpose of this study is to validate a simple and robust method for the identification of sitting, standing, walking, running and biking performed by preschool children, children and adolescents in the age from 3 to 16 years from a single thigh-worn accelerometer. Method: A total of 96 children were included in the study and all subjects followed a structured activity protocol performed in the subject’s normal kindergarten or school environment. Thigh acceleration was measured using the Axivity AX3 (Axivity, Newcastle, UK) device. Method development and accuracy was evaluated by equally dividing the subjects into a development and test group. Results: The sensitivity and specificity for identifying sitting and standing was above 99.3% and for walking and running above 82.6% for all age groups. The sensitivity and specificity for identifying biking was above 85.8% for children and adolescents and above 64.8% for the preschool group using running bikes. Conclusion: The accurate assessment of sitting, standing, walking, running and biking from thigh acceleration and with children in the age range of 3 to 16 is valid, although not with preschool children using running bikes.
Content may be subject to copyright.
children
Article
Simple Method for the Objective Activity Type
Assessment with Preschoolers, Children
and Adolescents
Jan Christian Brønd 1, * , Anders Grøntved 1, Lars Bo Andersen 2, Daniel Arvidsson 3
and Line Grønholt Olesen 1
1Center for Research in Childhood Health/Unit for Exercise Epidemiology, Department of Sports Science
and Clinical Biomechanics, University of Southern Denmark, 5000 Odense, Denmark;
agroentved@health.sdu.dk (A.G.); lgolesen@health.sdu.dk (L.G.O.)
2Faculty of Teacher Education and Sport, Sogn og Fjordane University College, 6851 Sogndal, Norway;
lars.bo.andersen@hvl.no
3Department of Food and Nutrition and Sport Science, Center for Health and Performance,
University of Gothenburg, 43141 Gothenburg, Sweden; daniel.arvidsson@gu.se
*Correspondence: jbrond@health.sdu.dk
Received: 4 May 2020; Accepted: 23 June 2020; Published: 2 July 2020


Abstract: Background:
The objective and accurate assessment of children’s sedentary and physical
behavior is important for investigating their relation to health. The purpose of this study is to validate
a simple and robust method for the identification of sitting, standing, walking, running and biking
performed by preschool children, children and adolescents in the age from 3 to 16 years from a
single thigh-worn accelerometer.
Method:
A total of 96 children were included in the study and all
subjects followed a structured activity protocol performed in the subject’s normal kindergarten or
school environment. Thigh acceleration was measured using the Axivity AX3 (Axivity, Newcastle,
UK) device. Method development and accuracy was evaluated by equally dividing the subjects into
a development and test group.
Results:
The sensitivity and specificity for identifying sitting and
standing was above 99.3% and for walking and running above 82.6% for all age groups. The sensitivity
and specificity for identifying biking was above 85.8% for children and adolescents and above 64.8%
for the preschool group using running bikes.
Conclusion:
The accurate assessment of sitting, standing,
walking, running and biking from thigh acceleration and with children in the age range of 3 to 16 is
valid, although not with preschool children using running bikes.
Keywords: classification; activity type; acceleration; thigh
1. Introduction
Children and adolescents spend, on average, 6–8 h of their time awake on sedentary behavior
(SB) [
1
], and engaging in excessive amounts of SB may be adversely associated with physical and
mental health [
2
]. Dierent determinants of SB have been investigated and active transportation seems
to be one modifiable component that may reduce sedentary time in the transition from childhood into
adolescence [
3
]. The accurate and objective assessment of SB and biking activities with children and
adolescents would be valuable to further investigate the determinants of SB in this population.
Assessing physical activity and SB in children and adolescents is challenging [
4
], and objective
assessment using activity monitors worn at the wrist or hip has been shown to be an attractive method
when compared to both direct observation and self-report methods [
5
]. The validity of using the hip [
6
] or
wrist [
7
,
8
], respectively, for the assessment of SB has been investigated, and significant dierences have
been reported [
9
]. Using the wrist has been shown to provide an excellent wear comfort, which increases
Children 2020,7, 72; doi:10.3390/children7070072 www.mdpi.com/journal/children
Children 2020,7, 72 2 of 15
subjects adherence to the measurement protocol [
10
], but thigh-worn accelerometers have also been
shown to provide high compliance rates with tape-mounted devices [
11
,
12
]. The ActivPAL (PAL
Technologies Ltd., Glasgow, UK) activity monitors are taped to the thigh and have been shown to
provide accurate estimates of SB based on the assessment of postural allocation with children and
adolescents [
13
15
]. The underlying algorithms for identifying SB with the ActivPAL device are not
available to the researcher, and the lack of transparency makes it dicult for researchers to fully
understand the strength and limitations of the selected methodology. Furthermore, the economic
burden of using the ActivPAL device might make it more applicable with studies including a small
number of subjects. These limitations with the ActivPAL device have previously been addressed by
Skotte et al. (2014) proposing an open source method for the identification of the following body
positions and activities: Sitting, standing, walking, running and biking with adults using a single
ActiGraph GT3X accelerometer worn on the thigh [
16
]. However, the size and bulky design of the
ActiGraph GT3X device does not make the proposed method by Skotte et al. (2014) applicable for use
with smaller children. In the study by Stewart et al. (2018), a smaller open source Axivity AX3 (Axivity,
Newcastle, UK) device was used for the identification of children’s activity types using accelerometers
worn at the thigh and trunk in addition to the use of a more advanced machine learning method in
the data processing [
17
]. However, besides the use of an advanced method, the study only provided
data for children 7–15 years of age and did not investigate the identification of biking. Moreover,
the underlying implementation of this method, and also the method proposed by Skotte et al. (2014),
have never been made public available, which is a major limitation. Currently, there are no simple
and public methods available for the assessment of PA behavior—specifically SB and biking—with
preschoolers, children and adolescents using a single thigh-worn activity monitor.
The aim of this study is to investigate the accuracy and validity of a simple method for the
identification of six common activity types with preschoolers, children and adolescents using a single
accelerometer worn on the thigh, and to make this method publicly available as open source technology.
2. Methods
2.1. Participants
A total of 96 Danish children were included in the study. Twenty-nine preschool children (age three
to six) were recruited from a local kindergarten in the Aarhus Municipality in 2018, whereas 36 children
(third and fourth grade) and 31 adolescents (eight and nine grade) were recruited from a local school
in the Odense Municipality. Written information about the study and an invitation to join the oral
information meeting arranged for the children at the preschool were communicated out using the
preschool–parent communication internet platform and handouts. During the study period, it was
secured that the preschool children having the parents
´
written consent also gave their assent. This was
secured being aware of the children
´
s reactions, and whether they seemed to be comfortable or not
with the test situation.
The children and adolescents were invited by email through the school oce and word of mouth.
Informed consent was provided by their parents The Ethics Committee of the Region of Southern
Denmark approved this additional preschool study, as a part of the Motor skills in PreSchool study
(S-2015-0178) as well as by the Danish Data protection Agency (10.450). The Ethics Committee of the
Region of Southern Denmark also approved the children and adolescent study (S-20140068).
2.2. Procedure
The protocols for the preschool children and the school children are not identical but in both
studies all children performed the activities in the same order as presented in Table 1.
Children 2020,7, 72 3 of 15
Table 1. Activities performed during the structured protocol at the school and preschool study.
Order Intensity Category Activity Description of Activity
School Study
Description of Activity
Preschool Study
1 Sedentary Sitting Sitting on a chair close to a
table with arms in the lap
* Sitting on the buttocks on
the floor playing with Lego
or Geomac toys.
2 Sedentary Sitting playing
Playing the Fruit Ninja game
on an iPad
Sitting on a chair close to a
table playing with Lego or
Geomac toys at the table.
3 Light Standing playing Playing a game on the iPad
while standing
* Standing close to a table
playing with Lego or
Geomac toy at the table.
4 Light Slow walking Slow walking speed
Walks at the child’s
preferred walking speed
together with the instructor.
5 Moderate Brisk walking Brisk walking speed
Walks fast trying to catch up
with the instructor without
running.
6 Vigorous Running Running at the subjects own
preferred running speed
Runs at the child’s preferred
running speed together with
the instructor.
7 Very vigorous Basketball One-to-one competitive
basketball game play Not performed
8 Very vigorous Playground
** Running and walking
around the school
playground in a follow my
leader activity
Walking, crawling, jumping
and running through a
predefined obstacle course
9 Moderate/vigorous Biking Commuting cycling on
subjects’ own bike
Cycling on an adjustable
child running bike at the
child´s preferred speed.
10 Sedentary Sitting Sitting close to a table with
arms in the lap
11 Light Swing Not performed ***Sitting on a swing in
self-selected eort
* The child is placed in a hula hoop and told not to step out of it while playing ** This activity was only performed by
nine subjects from the children’s group *** The activity was only performed by nineteen subjects from the preschool
children´s group.
All subjects performed the protocol using the same order (1–11) and the basketball activity was
not performed by the preschool children and the swing activity was not performed by the children and
adolescents group. The intensity classification is based on the measurement of energy expenditure
of similar activities in young children (Butte et al. 2018). The total duration of each activity was
1.5–5 min. A shorter duration (1.5–3 min) was used with the preschool children than with children
and adolescents (5 min). This was to ensure the preschool children could sustain the activity for the
full duration. The intensity of each of the movement activities was self-selected, but subjects were
also encouraged to adapt to an intensity they could complete the suggested distance (preschool study)
or full duration of each activity (school study). All sedentary activities were performed indoors and
all other activities outdoors at or around the preschool or school area, respectively. An activity log
was used to record activity start times (timestamps using seconds accuracy) during protocol execution
and subsequently used to extract the activity specific information with the analysis. The log time
was determined from the researcher’s smart phone which was synchronized with the computer that
was used to initialize the accelerometers. The activity start times were manually validated after data
download using visual inspection of the raw acceleration. If the start time was more than 5 s from
the onset of the activity, it was adjusted to the exact start of the activity. The activities basketball,
playground and swing are included to assess the performance of the method with activities that are a
complex mixture of fast transitions but also movement behaviors that are dicult to define.
Children 2020,7, 72 4 of 15
2.2.1. Preschool Children
Most children were tested one at a time, but some were tested two at the time. No height, weight
or anthropometric values were measured, as this was not specified in the protocol accepted by the
ethics committee. All children willing to participate was included in the study. During the assessment
process, attention was made to the children’s visual appearance (height and weight) to assess if this
was a reasonable homogeneous preschool group. The indoor activities were carried out in dierent
locations at the preschool throughout the test period. Thus, the children were sitting and standing
at dierent chairs and tables, but arrangements were made to secure somewhat appropriate chair
and table sizes. To avoid too many disturbances, the outdoor activities were carried out on a nearby
flat, less busy walking and bicycle path with hard surface running through a small recreative area.
The activities were carried out on a circular path with a distance of 75 m/246 feet. Each child was joined
by the researcher during the activities and encouraged to perform 1–4 laps at a time depending on
activity type, age and selected speed. One child performed all the outdoor activities at the preschool
playground. During the indoor sitting and standing activities, the child could choose his/her own
sitting position. The time was stopped if the child changed body position—e.g., from sitting to standing
position—but not if the child adjusted sitting position. In the standing position, the time was stopped
if the child stepped outside the predefined area (a hula hoop) or picked up toys dropped on the floor. If
the child stopped or changed activity during the outdoor activities the time was stopped and continued
after a small break. The position of the thigh belt was checked before and after performing an activity,
since the monitors were not mounted with tape. In total, the whole procedure for each child took, on
average, one hour (range 48–85 min), and the children completed 9–10 activities with each activity
lasting around 1.5–3 min.
2.2.2. Children and Adolescent
The measurements presented in Table 1were performed right after school. The participants
were introduced to the activity protocol right after arrival to an empty classroom following the
measurement of their body weight on a calibrated digital scale and body height using a stadiometer.
Anthropometric measurements of arm, leg and waist circumference were measured with a soft tape
measure. Arm length was measured from the acromion to the tip of the middle finger, and leg length as
the distance between the iliac crest to the floor while standing in an upright position with legs together.
All equipment was attached to the body and the measurement started after an adaption period of
10 min. The walking, fast walking and running activities were performed consecutively in order of
intensity without breaks. A 2–5-min natural break was used between all other activities. The duration
of each activity was 5 min and the first 60 s of data was not used in the analysis.
2.3. Instrumentation
The Axivity AX3 (Axivity, Newcastle UK) is a small (23
×
32.5
×
7.6 mm) MEMS based triaxial
accelerometer with 512Mb of onboard memory weighing only 11 g. The device is IPx8 certified
providing water resistance at 1.5 m for 1 h. The Axivity AX3 (Axivity, Newcastle UK) provides a
configurable sampling frequency (12.5–3200Hz), measurement range (
±
2,
±
4,
±
8,
±
16 g) and 12-bit
resolution (13-bit using the
±
16 g measurement range). The battery capacity of the Axivity AX3
provides measurement of 14 days at a 100 Hz sampling frequency. The OMGui (current version
1.0.0.30) used with instrument initialization and data download is available online [
18
]. The subjects
included wore the accelerometer on the right front thigh at the same location as used in the original
study by Skotte et al. (2014). The orientation of the device for the preschool children was with the y
axis pointing towards the knee and xaxis lateral (label side of the device visible and USB connection
lateral); whereas for the children and adolescents, the orientation was with the xaxis pointing towards
the knee and yaxis lateral (label side of the device visible and USB connection pointing towards knee).
The monitors were attached using belts made of OEKO-TEX-certified materials with the preschool
Children 2020,7, 72 5 of 15
children and tape mounted on the skin using a hospital dressing commonly used with wound treatment
for children and adolescents [
11
]. The alternative device orientation used with the preschool children
was needed to reduce belt width, and thus improve wear comfort. Device orientation is specified with
the classification method implemented in Matlab, which will automatically account for the orientation.
All instruments were initialized to measure acceleration using 100 Hz sampling frequency and
±
8 g
measurement range. The acceleration was resampled after data download to the same 30 Hz sampling
frequency used in the study by Skotte et al. (2014). The downsample function available in Matlab was
used to resample the data.
2.4. Classification Method
The identification of sitting, standing, walking, running and biking using the acceleration measured
with a single thigh-worn accelerometer is implemented as a simple decision tree. This is the same
method used in the study by Skotte et al. (2014), although not including stair walking. It is not the
purpose of this study to evaluate the accuracy of using the adult thresholds with children but to assess
new thresholds and the corresponding accuracy. A movement category is introduced in the method to
categorize activities performed in an upright posture, which include minor movements. This could be
preschool children’s playing in the sandbox making sand-cake at the outdoor child kitchen, or washing
dishes or folding laundry by either adolescents or children. The outline of the decision tree is presented
in Figure 1.
Children 2020, 7, x FOR PEER REVIEW 5 of 15
towards knee). The monitors were attached using belts made of OEKO-TEX-certified materials with
the preschool children and tape mounted on the skin using a hospital dressing commonly used with
wound treatment for children and adolescents [11]. The alternative device orientation used with the
preschool children was needed to reduce belt width, and thus improve wear comfort. Device
orientation is specified with the classification method implemented in Matlab, which will
automatically account for the orientation. All instruments were initialized to measure acceleration
using 100 Hz sampling frequency and ±8 g measurement range. The acceleration was resampled after
data download to the same 30 Hz sampling frequency used in the study by Skotte et al. (2014). The
downsample function available in Matlab was used to resample the data.
2.4. Classification Method
The identification of sitting, standing, walking, running and biking using the acceleration
measured with a single thigh-worn accelerometer is implemented as a simple decision tree. This is
the same method used in the study by Skotte et al. (2014), although not including stair walking. It is
not the purpose of this study to evaluate the accuracy of using the adult thresholds with children but
to assess new thresholds and the corresponding accuracy. A movement category is introduced in the
method to categorize activities performed in an upright posture, which include minor movements.
This could be preschool children’s playing in the sandbox making sand-cake at the outdoor child
kitchen, or washing dishes or folding laundry by either adolescents or children. The outline of the
decision tree is presented in Figure 1.
Figure 1. Decision tree implemented for identifying sitting, standing, walking, running, biking and
moving (T = true, F = false). SDX is the standard deviation of the acceleration in the x direction, Θ is
the backward/forward angle, Inc is the inclination angle and SDMAX is the maximum standard
deviation of the acceleration across all axes.
A total of five conditions are required to identify the six activity types. The signal features used
with the five conditions are: The standard deviation of the acceleration in the x-axis (SDx—along the
thigh direction), inclination angle (Inc), maximum standard deviation across all axes (SDMAX) and
backward/forward angle (Θ). All signal features are generated for each 2-s window using a 50%
overlap across the complete recording, which provides a temporal classification resolution of 1 s. The
first condition in the decision tree discriminates into either a stationary or dynamical movement
branch using the SDx signal feature. The first condition in the stationary branch discriminates
between the sitting and standing postural allocation using the inclination angle, which subsequently
discriminates into standing still and moving using the SDMAX feature. The first condition in the
Stationary
(SDX)
Sitting
(Inc)
Biking
(Θ)
FT
Biking
T
Walk
Running
(SDX)
Run
Moving
(SDMAX)
F
Sitting
Move Stand
T
T
F
F
TF
Figure 1.
Decision tree implemented for identifying sitting, standing, walking, running, biking and
moving (T =true, F =false). SD
X
is the standard deviation of the acceleration in the x direction,
Θ
is the
backward/forward angle, Inc is the inclination angle and SD
MAX
is the maximum standard deviation of
the acceleration across all axes.
A total of five conditions are required to identify the six activity types. The signal features used
with the five conditions are: The standard deviation of the acceleration in the x-axis (SDx—along the
thigh direction), inclination angle (Inc), maximum standard deviation across all axes (SD
MAX
) and
backward/forward angle (
Θ
). All signal features are generated for each 2-s window using a 50% overlap
across the complete recording, which provides a temporal classification resolution of 1 s. The first
condition in the decision tree discriminates into either a stationary or dynamical movement branch
using the SDx signal feature. The first condition in the stationary branch discriminates between the
sitting and standing postural allocation using the inclination angle, which subsequently discriminates
into standing still and moving using the SD
MAX
feature. The first condition in the dynamical movement
Children 2020,7, 72 6 of 15
branch uses the angle feature to discriminate between the biking and locomotion, and subsequently
the locomotion into walking or running using the SDx. The median filtering used in the original study
by Skotte et al. (2014) to eliminate sporadic misclassification is also implemented in this proposed
method. The identification of lying was implemented as proposed in the study by Skotte et al. (2014)
combining the inclination angle of the back (>65 degrees) in combination with the identification of
sitting with the thigh. The identification accuracy of lying was not evaluated. Multiple assignments of
activity types caused by the median filtering are not permitted and reduced to the activity that seem
most obvious. Thus, if lying or sitting is identified, then any other activity is removed. If biking and
other activities are identified, then biking is selected. If either walking or running in combination with
either stand or move is identified, then either walking or running is selected.
2.5. Statistical Considerations
The conditional thresholds of the decision tree and accuracy of the method were investigated by
dividing all subjects into a development or validation group. One half of each age group was randomly
selected for the development group, with the rest of the subjects assigned to the validation group.
The activity log recorded during the measurement protocol was used to extract the activity specific
signal features.
The threshold used to discriminate between dynamical movement and stationary activities using
SD
X
was evaluated using the pooled distribution of all sitting and standing activities with respect
to all activities requiring dynamical movements of the whole body (walking, running basketball,
playground). All pooled distributions were plotted with the ggplot2 R package and specifically the
geom_density function. The default setting of this function was used. The threshold for identifying
biking from locomotion using the backward/forward angle (
Θ
) was determined by evaluating the
pooled distribution of biking with respect to all locomotion activities. The threshold for the identification
of running from walking using SD
X
was evaluated by the pooled distribution of running in relation
to both the normal and brisk walking activities. The threshold for identifying sitting from standing
activities using the Inc angle was determined from the pooled distribution of all sitting activities with
respect to all standing activities. The threshold for identifying standing still from moving around
(shuing) using SD
MAX
was determined by evaluating the pooled distribution of all standing activities
with respect to the normal walking activity. The accuracy of identifying the activities sitting, standing,
walking, biking, and running was assessed by evaluating the sensitivity and specificity determined
from the agreement between the expected activity type and estimated activity (tabulated data) using
the second by second level data. For evaluating method performance with basketball, playground,
and swing, the relative amount of individual activities is presented. All accelerometry data processing
and activity type identification was implemented in Matlab (Mathworks Inc. Version R2019a 9.6.0) and
is publicly available for download on Github (https://github.com/jbrond/SkotteChild). All statistical
analyses were performed using R (Version 3.5.1).
3. Results
3.1. Descriptive Statistics
The age of the preschool children ranges from 3 to 6 years, 9 to 12 years in the children group,
and 13 to 16 in the adolescent group. The height and weight of the preschool children were not
measured, but none of the included subjects seemed to have a stature or weight that was substantially
dierent from the general norm of Danish preschool children. The mean (SD) weight of the children
and adolescents was 38.7 (7.0) kg and 59.2 (9.6) kg, and height was 145 (7.3) cm and 170.8 (11.1) cm,
respectively. A pairwise T-test showed no significant dierence in the age, weight, height or any of the
anthropometric values between the development and validation group.
Children 2020,7, 72 7 of 15
3.2. Algorithm Development
Figures 26illustrate the calculated signal features to discriminate between the tested activities
based on data from the developmental group. The original adult thresholds determined by
Skotte et al. (2014) are presented in all figures (black vertical line) for comparison. The optimal
threshold to be used with the preschool children, children and adolescents was made by the visual
inspection of the distribution in combination with the threshold known from the adults. Dierent
analytical methods were explored (Receiver Operating Characteristics, Naïve Bayesian classification).
However, the thresholds determined from these methods were very dierent from those of the adults,
which indicated that an analytical method is not optimal. The SD
x
distribution for the stationary and
dynamical movements activities is presented in Figure 2. The 0.1G SD
x
threshold determined with
adults is similarly applicable to preschool children, children and adolescents.
The angle distribution for biking and locomotion activities is presented in Figure 3, with only
children and adolescents in left plot and all the age groups included in the right plot. From the right
plot it is clear that the proportion of angle below 25 degrees is substantially increased by adding the
preschool children. However, the optimal angle threshold for discriminating biking from locomotive
activities is 22.5 degrees, which is slightly lower than the 24 degrees used with adults. Setting the
detection of biking at this threshold seems to suggest an increased misclassification with biking in the
preschool group as compared to the children and adolescents’ groups. This is most likely caused by
the use of running bikes with the preschool children and is further addressed in the discussion.
The SD
x
distribution for walking and running is presented in Figure 4. The SD
x
threshold for
discriminating walking from running is 0.65G, which is lower than the 0.72G determined with adults.
The inclination angle distribution for the sitting and standing activities is presented in Figure 5.
The optimal inclination angle discriminating sitting from standing is 47.5 degrees, which is slightly
higher than the 45 degrees used with adults.
The SD
MAX
distribution for standing still and locomotion activities is presented in Figure 6.
The SD
MAX
threshold for discriminating standing still from moving/shuing is 0.13G, which is slightly
higher than the 0.1 found with adults.
Children 2020, 7, x FOR PEER REVIEW 7 of 15
3.2. Algorithm Development
Figures 2–6 illustrate the calculated signal features to discriminate between the tested activities
based on data from the developmental group. The original adult thresholds determined by Skotte et
al. (2014) are presented in all figures (black vertical line) for comparison. The optimal threshold to be
used with the preschool children, children and adolescents was made by the visual inspection of the
distribution in combination with the threshold known from the adults. Different analytical methods
were explored (Receiver Operating Characteristics, Naïve Bayesian classification). However, the
thresholds determined from these methods were very different from those of the adults, which
indicated that an analytical method is not optimal. The SDx distribution for the stationary and
dynamical movements activities is presented in Figure 2. The 0.1G SDx threshold determined with
adults is similarly applicable to preschool children, children and adolescents.
Figure 2. The pooled distribution of the SDMAX signal for moving (green curve) and non-moving (red
curve) activities. The vertical line represents the original adult 0.1 g SDMAX threshold (Skotte et al.
2014).
The angle distribution for biking and locomotion activities is presented in Figure 3, with only
children and adolescents in left plot and all the age groups included in the right plot. From the right
plot it is clear that the proportion of angle below 25 degrees is substantially increased by adding the
preschool children. However, the optimal angle threshold for discriminating biking from locomotive
activities is 22.5 degrees, which is slightly lower than the 24 degrees used with adults. Setting the
detection of biking at this threshold seems to suggest an increased misclassification with biking in
the preschool group as compared to the children and adolescents’ groups. This is most likely caused
by the use of running bikes with the preschool children and is further addressed in the discussion.
0
20
40
0.0 0.1 0.2 0.3 0.4
SDMAX
Density
Moving
No
Yes
Figure 2.
The pooled distribution of the SD
MAX
signal for moving (green curve) and non-moving (red
curve) activities. The vertical line represents the original adult 0.1 g SDMAX threshold (Skotte et al. 2014).
Children 2020,7, 72 8 of 15
Figure 3.
The pooled distribution of the angle signal feature for locomotion (red curve) and biking
(green curve) only including the children and adolescents group in the left plot and all groups included
in the right plot. The vertical line represents the original adult 24-degree threshold (Skotte et al. 2014).
Children 2020, 7, x FOR PEER REVIEW 8 of 15
Figure 3. The pooled distribution of the angle signal feature for locomotion (red curve) and biking
(green curve) only including the children and adolescents group in the left plot and all groups
included in the right plot. The vertical line represents the original adult 24-degree threshold (Skotte
et al. 2014).
The SDx distribution for walking and running is presented in Figure 4. The SDx threshold for
discriminating walking from running is 0.65G, which is lower than the 0.72G determined with adults.
Figure 4. The pooled distribution of the SDX signal feature for walking (red curve) and running (green
curve). The vertical line represents the original adult 0.72 g SDX threshold (Skotte et al. 2014).
0.00
0.02
0.04
0.06
−25 0 25 50 75
Angle
Density
Status
Locomotion
Bike
0.00
0.02
0.04
0.06
−25 0 25 50 75
Angle
Density
Status
Locomotion
Bike
0
1
2
3
0.0 0.5 1.0 1.5
SDX
Density
Status
Walk
Run
Figure 4.
The pooled distribution of the SD
X
signal feature for walking (red curve) and running (green
curve). The vertical line represents the original adult 0.72 g SDXthreshold (Skotte et al. 2014).
Children 2020,7, 72 9 of 15
Children 2020, 7, x FOR PEER REVIEW 9 of 15
The inclination angle distribution for the sitting and standing activities is presented in Figure 5.
The optimal inclination angle discriminating sitting from standing is 47.5 degrees, which is slightly
higher than the 45 degrees used with adults.
Figure 5. The pooled distribution of the inclination angle signal feature for sitting (red curve) and
standing (green curve) activities. The vertical line represents the original adult 45.0 degrees threshold
(Skotte et al. 2014).
The SDMAX distribution for standing still and locomotion activities is presented in Figure 6. The
SDMAX threshold for discriminating standing still from moving/shuffling is 0.13G, which is slightly
higher than the 0.1 found with adults.
3.3. Algorithm Validation
Applying the identified thresholds in the decision tree with the validation group resulted in a
sensitivity above 64.8% and specificity above 95% for all activities. The individual activity related
sensitivity and specificity for each age group are presented in Table 2.
The sensitivity and specificity are above 99% for sitting and standing activities across all age
groups, and the sensitivity and specificity for the remaining activities biking, walking and running is
above 85.8% or close to 100% for children and adolescents. The sensitivity for biking is only 64.8%
with the pre-school children, which is lower than with children and adolescents. The basketball
activity performed by the children and adolescents was identified as walking (62%) with only 28.7%
of time identified as running. The playground activity performed by nine school children was mainly
identified as walking (83.5%), with only 8.5% identified as running. A minor part of the basket and
playground activities (<1%) were identified as moving and standing. The playground activities
performed by the preschool children were identified as walking (40.9%), moving (28.0%), biking
(12.8%), standing, (9.8%), running (8.2%), and sitting (0.3%). The swing activities were identified as
biking (50.0%), sitting (38.7%), standing (7.9%), and moving (3.4%).
0.00
0.02
0.04
0.06
0 40 80 120
Incl
Density
Status
Sitting
Stand
Figure 5.
The pooled distribution of the inclination angle signal feature for sitting (red curve) and
standing (green curve) activities. The vertical line represents the original adult 45.0 degrees threshold
(Skotte et al. 2014).
Children 2020, 7, x FOR PEER REVIEW 10 of 15
Figure 6. The pooled distribution of the SDMAX signal feature for standing (red curve) and walking
(green curve) activities. The vertical line represents the original adult 0.1 g SDMAX threshold (Skotte et
al. 2014).
Table 2. Activity type identification accuracy for sitting, standing, biking, walking and running for
the three age groups—preschool, children and adolescents—using the validation data.
Pre-School Children Adolescents
Sensitivity Specificity Sensitivity Specificity Sensitivity Specificity
Sitting 100.0 100.0 99.8 99.7 100.0 99.3
Standing 100.0 99.8 100.0 99.8 100.0 100.0
Biking 64.8 100.0 85.8 100.0 94.9 100.0
Walking 82.6 98.1 93.3 100.0 100.0 99.9
Running 92.4 95.0 99.9 97.3 99.6 99.9
4. Discussion
In this study, we evaluated a method for identifying sitting, standing, moving, walking, biking
and running using a single Axivity AX3 (Axivity, Newcastle UK) accelerometer worn on the thigh
with preschoolers, children and adolescents. The results demonstrate that the proposed method
provides an excellent sensitivity and specificity with all the proposed activities, with the only
exception for biking on running bikes with pre-school children. The accuracy found in this study is
comparable to the original study conducted on adults using the same method for the identification
[16].
Other studies have evaluated techniques to identify children’s time spent in different activity
types from accelerometer data. Trost et al. [19] used logistic regression to identify seven activity types
(lying, sitting, standing, walking, running, basketball, and dancing) in 52 children and adolescents
using accelerometers worn either on the hip or wrist. The hip- and wrist-based models achieved
91.0% and 88.4% accuracy, respectively. In a study by Stewart et al. [17] a dual-accelerometry system
was evaluated with both children and adults for the identification of six activity types. One device
was worn at the thigh and another device worn on the lower back. A random forest algorithm and a
total of 142 different signal features generated from the raw acceleration were used in the
identification of activity type. The random forest algorithm is an ensemble learner and performs the
0
20
40
60
0.0 0.1 0.2 0.3 0.4 0.5
SDMAX
Density
Status
Stand
Walk
Figure 6.
The pooled distribution of the SD
MAX
signal feature for standing (red curve) and walking (green
curve) activities. The vertical line represents the original adult 0.1 g SD
MAX
threshold (Skotte et al. 2014).
Children 2020,7, 72 10 of 15
3.3. Algorithm Validation
Applying the identified thresholds in the decision tree with the validation group resulted in a
sensitivity above 64.8% and specificity above 95% for all activities. The individual activity related
sensitivity and specificity for each age group are presented in Table 2.
Table 2.
Activity type identification accuracy for sitting, standing, biking, walking and running for the
three age groups—preschool, children and adolescents—using the validation data.
Pre-School Children Adolescents
Sensitivity Specificity Sensitivity Specificity Sensitivity Specificity
Sitting 100.0 100.0 99.8 99.7 100.0 99.3
Standing 100.0 99.8 100.0 99.8 100.0 100.0
Biking 64.8 100.0 85.8 100.0 94.9 100.0
Walking 82.6 98.1 93.3 100.0 100.0 99.9
Running 92.4 95.0 99.9 97.3 99.6 99.9
The sensitivity and specificity are above 99% for sitting and standing activities across all age
groups, and the sensitivity and specificity for the remaining activities biking, walking and running is
above 85.8% or close to 100% for children and adolescents. The sensitivity for biking is only 64.8% with
the pre-school children, which is lower than with children and adolescents. The basketball activity
performed by the children and adolescents was identified as walking (62%) with only 28.7% of time
identified as running. The playground activity performed by nine school children was mainly identified
as walking (83.5%), with only 8.5% identified as running. A minor part of the basket and playground
activities (<1%) were identified as moving and standing. The playground activities performed by
the preschool children were identified as walking (40.9%), moving (28.0%), biking (12.8%), standing,
(9.8%), running (8.2%), and sitting (0.3%). The swing activities were identified as biking (50.0%), sitting
(38.7%), standing (7.9%), and moving (3.4%).
4. Discussion
In this study, we evaluated a method for identifying sitting, standing, moving, walking, biking
and running using a single Axivity AX3 (Axivity, Newcastle UK) accelerometer worn on the thigh with
preschoolers, children and adolescents. The results demonstrate that the proposed method provides
an excellent sensitivity and specificity with all the proposed activities, with the only exception for
biking on running bikes with pre-school children. The accuracy found in this study is comparable to
the original study conducted on adults using the same method for the identification [16].
Other studies have evaluated techniques to identify children’s time spent in dierent activity
types from accelerometer data. Trost et al. [
19
] used logistic regression to identify seven activity types
(lying, sitting, standing, walking, running, basketball, and dancing) in 52 children and adolescents
using accelerometers worn either on the hip or wrist. The hip- and wrist-based models achieved 91.0%
and 88.4% accuracy, respectively. In a study by Stewart et al. [
17
] a dual-accelerometry system was
evaluated with both children and adults for the identification of six activity types. One device was
worn at the thigh and another device worn on the lower back. A random forest algorithm and a total
of 142 dierent signal features generated from the raw acceleration were used in the identification of
activity type. The random forest algorithm is an ensemble learner and performs the identification of
activity types from acceleration using multiple individual decision trees [
20
]. The method used in the
present study only uses a single decision tree and just six dierent signal features. Despite the large
dierence in algorithm complexity and number of features, the sensitivity and specificity are above
98% for most activities in both studies. Thus, using a more complex algorithm or additional features
does not seem to improve the identification accuracy per se, suggesting that wear location and optimal
selection of signal features might be more important than the algorithm and number of signal features.
Children 2020,7, 72 11 of 15
Including biking in the identification of activity type is primarily due to the known health benefits
of this activity, but also due to the important assessment of transportation mode which is great
importance in many countries. The lower sensitivity and specificity demonstrated for the identification
of biking with preschool children is clearly caused by the use of running bikes rather than actual bikes
with pedals. Running bikes do not have pedals as normal bikes do, and driving the running bike
forward is carried out by either “running”, or double pushing both legs on the ground, or simply
by having a break resting the legs on the footrests, although the children were encouraged not to
do so. The accuracy for identifying the running bike as biking could be increased by decreasing the
forward/backward angle threshold in the identification of biking from locomotion. Decreasing the angle
threshold, however, would also potentially increase the misclassification of some running activities
as biking, which will decrease the sensitivity of identifying locomotion in real data. The acceleration
not identified as biking with the preschoolers using a running bike is primarily identified as running,
but also as walking. Considering the actual movement with the running bike, it is more correct to the
actual movement performed rather than as an indented biking activity.
The activities included in the present study and the study by Stewart et al. (2018) are performed
in controlled environments, which simplifies the data processing and analysis. However, the frequent
transitions between activities—which are an important element of children’s common movement
behavior during free-living—are not included, suggesting that the sensitivity and specificity estimated
by Stewart et al. (2018) and in the present study might not accurately reflect the performance of the
algorithms in a free-living environment. In an attempt to address this, we included a basketball activity
with the children and adolescents, and a playground activity with the children and preschoolers.
The basketball and playground activities included movements such as standing still, moving, walking,
running and jumping, varying in both duration and organization. We did not assess the amount of
time spent with the dierent activity categories with the basketball and playground activities, but the
estimated duration of the individual activity types seems to reflect the overall nature to be expected
of these activities. The method proposed by Stewart et al. (2018) was evaluated in a free-living
environment, and the results provide excellent sensitivity and specificity with most activities [
21
].
However, transitions were excluded from the analysis and the results also further indicate that the
identification of dynamic standing and movement is challenged with children. Increasing the number
of available features and the complexity of the algorithm will increase the risk for overfitting and
thus misclassification of some activity types in a free-living environment. The complex and sporadic
movement behaviors, especially of younger children, seem to suggest that a robust identification
(balance between sensitivity and specificity) of common activity types with children is most optimally
approached using a limited number of signal features. The children enrolled in the free-living
evaluation by Steward et al. (2018) were at the age of 10 years, and further studies seem to be
required to evaluate the classification accuracy with even younger children in a free-living environment.
Moreover, the implementation of the proposed method described by Stewart et al. (2018) is not publicly
available, which makes it dicult for other researchers to use, replicate and modify.
The identification of stair walking was included in the original method proposed by Skotte et al.
(2014). The identification of stair walking was implemented using an individual defined threshold,
determined using the median value for the forward/backward angle (
Θ
) below 5 degrees, and adding 4.5
degrees (
Θd
=
Θm
+4.5). The threshold for the identification of stair walking is therefore <9.5 degrees,
and thus sensitive to the misclassification of walking and running as stair walking. Running and
walking movements with children are commonly performed in a complex environment and, combined
with children’s short legs, it seems to increase the risk for the children to generate backward/forward
angles that resemble stair walking. The question of whether to include or exclude stair walking is
a balance between the misclassification of stair walking as normal walking or the misclassification
of walking and running as stair walking. Most children do perform stair walking, but considering
the nature of movement in children, we might introduce a systematic bias by the misclassification of
walking and running as stair walking rather than obtaining an accurate estimate of children’s stair
Children 2020,7, 72 12 of 15
walking. In some environments, stair walking might be an important element of children’s movement
behavior, which requires the accurate quantification of stair walking. However, only including the
forward/backward angle in the identification of stair walking seems inadequate for obtaining a robust
identification with younger children. This could argue for using additional features, as with the study
by Stewart et al. (2018) for the identification of stair walking. However, as previously mentioned,
increasing the number of signal features increases the risk of overfitting, and thus the misclassification
of the actual activity performed. Currently, there is no single robust signal feature available for the
accurate identification of stair walking in children, and further investigation of the biomechanical
properties of children’s stair walking in relation to the acceleration measured at the thigh seems to
be required.
All signal features generated in the method proposed by Stewart et al. (2018) are determined
using a 5-s non-overlapping window, whereas only a 2-s 50% over-lapping window (providing second
by second resolution) is used in the method proposed by Skotte et al. (2014) and the present study.
The need to use a longer time window with the method proposed by Stewart et al. (2018) is most likely
to provide sucient resolution with the frequency-related signal features, in order to accurately detect
the cyclic or non-cyclic nature of some activities. The dominant frequency is commonly included in
many machine learning algorithms [
17
,
22
], and estimated using the Fast Fourier transformation (FFT).
The resolution and minimal detectable frequency is coupled with the total number samples used to
estimate the feature. The sporadic nature of children’s movement behavior in a free-living environment
seems to suggest a general increased misclassification, with longer time windows as compared to
shorter time windows. Window size or epoch length has previously been investigated extensively with
intensity-derived measures from acceleration, and it is clear that the intermittent nature of children’s
activity has to be analyzed with short-duration epochs [
23
,
24
]. Another interesting dierence between
the present study and the study conducted by Steward et al. (2018) is the use of non-overlapping
and overlapping windows. Non-overlapping windows determine the activity type for each window
independently, whereas over-lapping windows consider a smoother transition between windows.
Preschoolers and children often perform a transition between dierent activity types lasting less than 5
s, and most likely not in synchronization with the window by window classification [
25
]. This will
cause the rate of misclassification to follow the number of activity transitions performed by the subject.
However, using over-lapping windows might also be challenged with identifying the actual onset and
oset between activities. The optimal selection of window size and overlapping windows seems to be
an important aspect of the accurate classification of activity types in children, which seems to require
further investigation in the future.
The lying posture is commonly interpreted as an indirect measure of sleep and the identification
of lying using accelerometry has been approached using both single and multiple devices worn on
both wrist, thigh, and hip [
17
,
21
,
26
,
27
]. For many children and adolescents, it is not uncommon in the
late afternoon or evening, during weekdays or generally during weekend days, to lie on the couch or
in bed watching TV, using a tablet or their cell phone. This seems to suggest that with the indirect
measure of sleep from the identification of lying, as validated with laboratory conducted experiments,
there is a substantial risk of misclassifying sedentary behavior as sleep. The lying posture allocation
associated with evening sedentary behavior is clearly not sleep and should not be included as such.
The accurate identification of lying and sleep with 24-h free-living recordings is challenging and
only including a laboratory lying activity, even though it is in dierent positions, does not provide
sucient information for distinguishing lying as sedentary behavior from time in bed and thus sleep.
If accelerometry is to be used to provide an indirect assessment of time in bed and potentially sleep,
it is of utmost importance to distinguish between these behaviors. The accurate identification of lying
and time in bed/asleep has to be performed using measurements conducted with subjects during
their free-living behavior similar to in previous studies [
28
32
], and thus including the important
temporal information regarding circadian rhythm and essentially sleep behaviors. Currently, there is
only one study investigating the assessment of sleep using free-living recordings and a thigh-worn
Children 2020,7, 72 13 of 15
device. However, this study relies on the proprietary ActivPAL device and the validity was assessed
with adults [
27
]. Future studies should investigate the accurate identification of children’s lying from
time in bed with thigh-worn accelerometers using free-living recordings, rather than only relying on
laboratory data or standardized protocols.
Strength and Limitations
A major strength of this study was that all activities were performed in the subject’s natural
environment, as well as the inclusion of subjects across multiple age groups. Although the subjects
followed a strict protocol during the field validation, we allowed some natural movement adjustment
during most activities—for example, adjusting sitting position when sitting. The use of short and
overlapping windows in the generation of the signal features is a strength of the method, and provides
a resolution which is required with the sporadic nature of this population. The method implementation
is made publicly available and open source.
A major limitation with the present study that a free-living validation is not included. Conducting
a true free-living validation is challenging, and the sensitivity and specificity estimated in this study do
not reflect the true sensitivity and specificity with the method using real recordings. In the discussion,
we addressed the possible limitations of the current method in comparison to more advanced methods
and, considering the complexity of human behavior and movement, it seems to suggest that selecting
a simpler and more robust method might perform better with real recordings. It is a minor limitation
to use running bikes to perform biking for preschool children. However, the number of preschool
children capable of using a real bike with pedals is likely to be very small, suggesting that this activity
is not common with real recordings. It is a limitation that the method is implemented in the commercial
and costly software Matlab
®
. However, GNU Octave (https://www.gnu.org/software/octave/) is a
free and open-source alternative to Matlab. We implemented the described method using standard
functions which are also available with Octave. Moreover, using a standard function also provides
an easy replication of the method with common statistical software such as R or Python, which are
freely available.
5. Conclusions
The identification of six common activity types with preschoolers, children and adolescents with
a single accelerometer worn at the thigh has been presented and evaluated. The cross validation
demonstrated an excellent sensitivity and specificity, suggesting that the proposed method is valid for
determining the time spent in dierent activity types in preschoolers, children and adolescents in a
free-living environment.
Author Contributions:
Conceptualization, J.C.B., A.G., L.B.A., D.A. and L.G.O.; Data curation, J.C.B. and L.G.O.;
Formal analysis, J.C.B. and L.G.O.; Funding acquisition, A.G. and L.B.A.; Investigation, J.C.B., D.A. and L.G.O.;
Methodology, J.C.B., D.A. and L.G.O.; Project administration, L.B.A., D.A. and L.G.O.; Resources, D.A.; Software,
J.C.B.; Supervision, A.G., L.B.A. and D.A.; Validation, J.C.B., A.G. and L.G.O.; Visualization, J.C.B., D.A. and
L.G.O.; Writing—original draft, J.C.B.; Writing—review and editing, J.C.B., A.G., L.B.A., D.A. and L.G.O. All
authors have read and agreed to the published version of the manuscript.
Funding:
Anders Grøntved and Line Grønholt Olesen were supported by a European Research Council Grant
(grant number 716657) and Trygfonden (grant number 114536).
Acknowledgments:
The authors are very grateful to the participating children, their parents and the school- and
preschool stawho devoted their time and made this study possible. The preschool study was supported by the
Danish TrygFonden.
Conflicts of Interest: The authors declare no conflicts of interest.
References
1.
Lou, D. Sedentary behaviors and youth: Current trends and the impact on health. In Proceedings of the
Active Living Research, San Diego, CA, USA, 2014.
Children 2020,7, 72 14 of 15
2.
Panahi, S.; Tremblay, A. Sedentariness and Health: Is Sedentary Behavior More Than Just Physical Inactivity?
Front. Public Health 2018,6, 258. [CrossRef] [PubMed]
3.
Atkin, A.J.; Foley, L.; Corder, K.; Ekelund, U.; van Sluijs, E.M. Determinants of Three-Year Change in
Children’s Objectively Measured Sedentary Time. PLoS ONE 2016,11, e0167826. [CrossRef] [PubMed]
4.
Sievanen, H.; Kujala, U.M. Accelerometry-Simple, but challenging. Scand. J. Med. Sci. Sports
2017
,27,
574–578. [CrossRef] [PubMed]
5.
Sirard, J.R.; Pate, R.R. Physical activity assessment in children and adolescents. Sports Med.
2001
,31, 439–454.
[CrossRef]
6.
Trost, S.G.; Loprinzi, P.D.; Moore, R.; Pfeier, K.A. Comparison of accelerometer cut points for predicting
activity intensity in youth. Med. Sci. Sports Exerc. 2011,43, 1360–1368. [CrossRef]
7.
Rowlands, A.V.; Cli, D.P.; Fairclough, S.J.; Boddy, L.M.; Olds, T.S.; Parfitt, G.; Noonan, R.J.; Downs, S.J.;
Knowles, Z.R.; Beets, M.W. Moving Forward with Backward Compatibility: Translating Wrist Accelerometer
Data. Med. Sci. Sports Exerc. 2016,48, 2142–2149. [CrossRef]
8.
van Loo, C.M.; Okely, A.D.; Batterham, M.J.; Hinkley, T.; Ekelund, U.; Brage, S.; Reilly, J.J.; Trost, S.G.;
Jones, R.A.; Janssen, X.; et al. Wrist Accelerometer Cut-points for Classifying Sedentary Behavior in Children.
Med. Sci. Sports Exerc. 2016. [CrossRef]
9.
Hildebrand, M.; Hansen, B.H.; van Hees, V.T.; Ekelund, U. Evaluation of raw acceleration sedentary
thresholds in children and adults. Scand. J. Med. Sci. Sports 2016. [CrossRef]
10.
Fairclough, S.J.; Noonan, R.; Rowlands, A.V.; Van Hees, V.; Knowles, Z.; Boddy, L.M. Wear Compliance and
Activity in Children Wearing Wrist- and Hip-Mounted Accelerometers. Med. Sci. Sports Exerc.
2016
,48,
245–253. [CrossRef]
11.
Schneller, M.B.; Bentsen, P.; Nielsen, G.; Brond, J.C.; Ried-Larsen, M.; Mygind, E.; Schipperijn, J. Measuring
Children’s Physical Activity: Compliance Using Skin-taped Accelerometers. Med. Sci. Sports Exerc.
2017
.
[CrossRef]
12.
Duncan, S.; Stewart, T.; Mackay, L.; Neville, J.; Narayanan, A.; Walker, C.; Berry, S.; Morton, S. Wear-Time
Compliance with a Dual-Accelerometer System for Capturing 24-h Behavioural Profiles in Children and
Adults. Int. J. Environ. Res. Public Health 2018,15, 1296. [CrossRef] [PubMed]
13.
Aminian, S.; Hinckson, E.A. Examining the validity of the ActivPAL monitor in measuring posture and
ambulatory movement in children. Int. J. Behav. Nutr. Phys. Act. 2012,9, 119. [CrossRef]
14.
Ridley, K.; Ridgers, N.D.; Salmon, J. Criterion validity of the activPAL and ActiGraph for assessing children’s
sitting and standing time in a school classroom setting. Int. J. Behav. Nutr. Phys. Act.
2016
,13, 75. [CrossRef]
[PubMed]
15.
van Loo, C.M.; Okely, A.D.; Batterham, M.J.; Hinkley, T.; Ekelund, U.; Brage, S.; Reilly, J.J.; Jones, R.A.;
Janssen, X.; Cli, D.P. Validation of thigh-based accelerometer estimates of postural allocation in 5-12 year-olds.
J. Sci. Med. Sport 2017,20, 273–277. [CrossRef] [PubMed]
16.
Skotte, J.; Korshøj, M.; Kristiansen, J.; Hanisch, C.; Holtermann, A. Detection of physical activity types using
triaxial accelerometers. J. Phys. Act. Health 2014,11, 76–84. [CrossRef]
17.
Stewart, T.; Narayanan, A.; Hedayatrad, L.; Neville, J.; Mackay, L.; Duncan, S. A Dual-Accelerometer System
for Classifying Physical Activity in Children and Adults. Med. Sci. Sports Exerc.
2018
,50, 2595–2602.
[CrossRef] [PubMed]
18.
Jackson, D.G. OMGui Software. Available online: https://github.com/digitalinteraction/openmovement/wiki/
AX3-GUI (accessed on 1 April 2020).
19.
Trost, S.G.; Zheng, Y.; Wong, W.K. Machine learning for activity recognition: Hip versus wrist data.
Physiol. Meas. 2014,35, 2183–2189. [CrossRef]
20. Breiman, L. Random Forests. Mach. Learn. 2001,45, 5–32. [CrossRef]
21.
Narayanan, A.; Stewart, T.; Mackay, L. A Dual-Accelerometer System for Detecting Human Movement in a
Free-living Environment. Med. Sci. Sports Exerc. 2020,52, 252–258. [CrossRef]
22.
Staudenmayer, J.; Pober, D.; Crouter, S.; Bassett, D.; Freedson, P. An artificial neural network to estimate
physical activity energy expenditure and identify physical activity type from an accelerometer. J. Appl.
Physiol. (1985) 2009,107, 1300–1307. [CrossRef]
23.
Colley, R.C.; Harvey, A.; Grattan, K.P.; Adamo, K.B. Impact of accelerometer epoch length on physical activity
and sedentary behaviour outcomes for preschool-aged children. Health Rep. 2014,25, 3–9. [PubMed]
Children 2020,7, 72 15 of 15
24.
Nettlefold, L.; Naylor, P.J.; Warburton, D.E.; Bredin, S.S.; Race, D.; McKay, H.A. The Influence of Epoch
Length on Physical Activity Patterns Varies by Child’s Activity Level. Res. Q. Exerc. Sport
2016
,87, 110–123.
[CrossRef] [PubMed]
25.
Bailey, R.C.; Olson, J.; Pepper, S.L.; Porszasz, J.; Barstow, T.J.; Cooper, D.M. The level and tempo of children’s
physical activities: An observational study. Med. Sci. Sports Exerc.
1995
,27, 1033–1041. [CrossRef] [PubMed]
26.
van Hees, V.T.; Sabia, S.; Jones, S.E.; Wood, A.R.; Anderson, K.N.; Kivimaki, M.; Frayling, T.M.; Pack, A.I.;
Bucan, M.; Trenell, M.I.; et al. Estimating sleep parameters using an accelerometer without sleep diary.
Sci. Rep. 2018,8, 12975. [CrossRef]
27.
Lyden, K.; John, D.; Dall, P.; Granat, M.H. Dierentiating Sitting and Lying Using a Thigh-Worn Accelerometer.
Med. Sci. Sports Exerc. 2016,48, 742–747. [CrossRef]
28.
Barreira, T.V.; Redmond, J.G.; Brutsaert, T.D.; Schuna, J.M., Jr.; Mire, E.F.; Katzmarzyk, P.T.; Tudor-Locke, C.
Can an automated sleep detection algorithm for waist-worn accelerometry replace sleep logs? Appl. Physiol.
Nutr. Metab. 2018,43, 1027–1032. [CrossRef]
29.
Tudor-Locke, C.; Barreira, T.V.; Schuna, J.M., Jr.; Mire, E.F.; Katzmarzyk, P.T. Fully automated waist-worn
accelerometer algorithm for detecting children’s sleep-period time separate from 24-h physical activity or
sedentary behaviors. Appl. Physiol. Nutr. Metab. 2014,39, 53–57. [CrossRef]
30.
Bellettiere, J.; Zhang, Y.; Berardi, V.; Full, K.M.; Kerr, J.; LaMonte, M.J.; Evenson, K.R.; Hovell, M.; LaCroix, A.Z.;
Di, C. Parameterizing and validating existing algorithms for identifying out-of-bed time using hip-worn
accelerometer data from older women. Physiol. Meas. 2019,40, 075008. [CrossRef]
31.
Full, K.M.; Kerr, J.; Grandner, M.A.; Malhotra, A.; Moran, K.; Godoble, S.; Natarajan, L.; Soler, X. Validation
of a physical activity accelerometer device worn on the hip and wrist against polysomnography. Sleep Health
2018,4, 209–216. [CrossRef]
32.
Tracy, J.D.; Acra, S.; Chen, K.Y.; Buchowski, M.S. Identifying bedrest using 24-h waist or wrist accelerometry
in adults. PLoS ONE 2018,13, e0194461. [CrossRef]
©
2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).
... However, validation studies of postures and movements based on thigh-worn accelerometry among children are scarce [19]. Five studies have assessed the accuracy of a thigh-worn accelerometer for estimating postures and movements among children [20][21][22][23][24], of which only three have examined multiple postures and movements [20,21,24]. The studies by Stewart et al. and Brønd et al. used the Axivity AX3 (Axivity, Newcastle, UK) accelerometer and customised software (i.e., machine learning or decision tree algorithms). ...
... However, validation studies of postures and movements based on thigh-worn accelerometry among children are scarce [19]. Five studies have assessed the accuracy of a thigh-worn accelerometer for estimating postures and movements among children [20][21][22][23][24], of which only three have examined multiple postures and movements [20,21,24]. The studies by Stewart et al. and Brønd et al. used the Axivity AX3 (Axivity, Newcastle, UK) accelerometer and customised software (i.e., machine learning or decision tree algorithms). ...
... We found comparable levels of balanced accuracy with what has previously been reported by validation studies using thigh-worn accelerometry and custom-made software to measure standing, walking, and running among children [20][21][22][23]. The studies by Stewart et al. and Lendt et al. found slightly higher levels of balanced accuracy for standing, walking, and running (>86%) than the current study [21,24]. ...
Article
Full-text available
Background: ActiMotus, a thigh-accelerometer-based software used for the classification of postures and movements (PaMs), has shown high accuracy among adults and school-aged children; however, its accuracy among younger children and potential differences between sexes are unknown. This study aimed to evaluate the accuracy of ActiMotus to measure PaMs among children between 3 and 14 years and to assess if this was influenced by the sex or age of children. Method: Forty-eight children attended a structured ~1-hour data collection session at a laboratory. Thigh acceleration was measured using a SENS accelerometer, which was classified into nine PaMs using the ActiMotus software. Human-coded video recordings of the session provided the ground truth. Results: Based on both F1 scores and balanced accuracy, the highest levels of accuracy were found for lying, sitting, and standing (63.2-88.2%). For walking and running, accuracy measures ranged from 48.0 to 85.8%. The lowest accuracy was observed for classifying stair climbing. We found a higher accuracy for stair climbing among girls compared to boys and for older compared to younger age groups for walking, running, and stair climbing. Conclusions: ActiMotus could accurately detect lying, sitting, and standing among children. The software could be improved for classifying walking, running, and stair climbing, particularly among younger children.
... Four studies have assessed the accuracy of a thigh-worn accelerometer for estimating postures and movements among children [20][21][22][23] of which only two have examined multiple postures and movements 20,21 . Both of these used the Axivity AX3 (Axivity, Newcastle, UK) accelerometer and customised software (i.e. machine learning or decision tree algorithms). ...
... Four studies have assessed the accuracy of a thigh-worn accelerometer for estimating postures and movements among children [20][21][22][23] of which only two have examined multiple postures and movements 20,21 . Both of these used the Axivity AX3 (Axivity, Newcastle, UK) accelerometer and customised software (i.e. machine learning or decision tree algorithms). ...
... We found comparable levels of balanced accuracy with what has previously been reported by validation studies using thigh-worn accelerometry and custom-made software to measure standing, walking and running among children [20][21][22][23] 20 than what we observed. In contrast, two studies using a thigh-worn activPAL accelerometer and the activPAL software to measure standing and walking among children aged 4-6 years found lower or comparable levels of balanced accuracy for classifying standing (76-89%) and walking (74-88%) 22,23 . ...
Preprint
Full-text available
Background Robust measurements of children’s postures and movements are required to understand their impact on health and wellbeing. Recent advances in wearable sensor technology may enable the development of accurate measurements. Motus, a wearable sensor based system for surveillance of postures and movements, has shown high accuracy among adults. However, its accuracy to measure postures and movements among children is unknown. This study aimed to evaluate the criterion validity of Motus to measure common postures and movements among children between 3–14 years old in a laboratory setting. We further assessed if the sex or age of children impacted accuracy. Method Data were collected on 48 children who attended a structured ~ 1-hour data collection session at a Curtin University laboratory with their caregivers. The session was video recorded and thigh acceleration was measured using a SENS accelerometer. Data from the accelerometer were processed and classified into nine postures and movements using the Motus software. Human-coded video provided the ground truth to calculate sensitivity, specificity, precision, F1-scores, and balanced accuracy. Results We observed good to very good overall accuracy (F1-score = 61.9, balanced accuracy = 81.1%) and for classifying lying, sitting, standing (ranging between 63.2–85.3%). Walking and running were classified with moderate to very good accuracy. The lowest accuracy was observed for classifying stair climbing. We found a higher accuracy for stair climbing among girls compared to boys and for older compared to younger age-groups for walking, running and stair climbing. Conclusion Motus showed moderate to very good accuracy for detecting lying, sitting, standing, and running among children. The system could be improved for classifying the more dynamic postures and movements (i.e. walking, running and stair climbing), particularly among younger children and developed further to measure more child-specific postures and movements.
... Participants were included in the analysis if they had at least one valid day of accelerometer data, BMJ Public Health defined as a minimum of 10 hours of awake wear time per day. Physical behaviours were classified second-bysecond using algorithms by Brønd et al. 13 Physical behaviours classified from the accelerometer data included sitting or lying; standing still or with minor movement; and walking or running. The intensity of physical activity was categorised into sedentary, light and moderate-tovigorous intensity derived from ActiGraph counts in 10 s epochs. ...
... We used a thigh-worn accelerometer, allowing for accurate second-by-second classification of common physical behaviours, in addition to assessing time spent on physical activities of varying intensities. [13][14][15] Furthermore, by collecting school time schedules from each participant, we were able to accurately classify accelerometer data to the specific periods of recess. These methodological strengths, combined with the consistent associations observed across different age groups and genders, suggest that our findings have wider generalisability to the broader adolescent population. ...
Article
Full-text available
Background The widespread ownership of digital devices among adolescents presents challenges and opportunities in schools, particularly during recess. This study aimed to examine the association of screen use during recess with physical activity behaviours. Methods The study was based on a population-based cross-sectional study including 1347 adolescents from 28 schools (2017–2018). Physical activity behaviours were assessed using thigh-attached accelerometers, worn 24/7 for up to 6 weekdays and 2 weekend days. We examined accelerometer data for all recess periods, aligned with each adolescent’s school schedule, along with leisure-time activity from the same days. Physical behaviours were classified during recess and leisure-time behaviour (negative control). Frequency of screen use during recess was based on self-report using a 5-point ordinal scale. Results Greater frequency of screen use during recess was associated with less time engaged in physically active behaviours and more time spent sitting, consistent with dose-dependent associations. In multivariable-adjusted analysis, adolescents with no screen use during recess spent an absolute 11.1% (95% CI 5.4 to 6.8) more of their recess time being physically active compared to those with frequent use. Based on the mean recess duration observed in the sample, adolescents not using screens during recess engaged in physical activities for an average of 44.9 min per day (95% CI 42.3 to 47.6), compared with 35.1 min (95% CI 26.0 to 44.3) for frequent screen users. Leisure time activities, used as a negative control, showed no link to screen use during recess. Conclusion Increased screen use during recess was associated with lower physical activity levels. These findings suggest that regulating digital device use during recess could enhance physical activity among adolescents.
... All periods identified as not worn were marked as missing data and not excluded from the subsequent analysis. PA intensity was determined using the algorithm proposed by Brønd et al. 30 . Moderate and vigorous PA time was estimated using 10-s epochs accounting for the elevated post oxygen consumption during intermittent PA using the second-by-second epoch data to improve the assessment of vigorous activity 31 . ...
... The PA types were determined in 1-s epochs from the raw acceleration using a simple decision-tree algorithm 30 . This method has been validated with a standardised protocol and provide a sensitivity and specificity > 99.3% for sitting and standing and > 85.8% for walking and running activities with children in the same age range as included in this study. ...
Article
Full-text available
Physical inactivity increases risks of cardiovascular disease, poor mental health, and morbidity. School-based physical activity (PA) promotion can reach children from differing backgrounds in a large proportion of their waking hours. Education outside the classroom (EOtC) is an PA-integrating pedagogical approach that aligns PA with the primary goals of schools and is therefore thought to present an acceptable, feasible, and efficacious model for school-based PA. This registered report used a cluster randomised wait list design to evaluate the efficacy of an EOtC intervention that aims to increase adolescents’ (ages 10–16 years) school-based (school week) and overall (full week) PA by providing a course on the pedagogical and didactic methods of EOtC to teachers across 30 Danish schools, and subsequently the teachers implementing EOtC in their classes > 5 h weekly over the course of one school year. Across 20 schools, 503 pupils aged 9–14 years were enrolled. The intervention group classes delivered a weekly mean of 238 (± 50) minutes of EOtC per-protocol and engaged in 20.4% moderate-to-vigorous school-based PA above the control group, however with no effect on overall PA. Enhancing effectiveness on overall PA, supplementing the intervention with additional PA activities, or adjusting its delivery and implementation is necessary. Protocol registration The stage 1 protocol for this Registered Report was accepted in principle on 22/08/22. The protocol, as accepted by the journal, can be found at: https://doi.org/10.17605/OSF.IO/NTM9K.
... This study deployed Axivity AX3 accelerometers (Axivity, Newcastle upon Tyne, UK), which have demonstrated high reliability in measuring sedentary and PA levels in children [37,38]. The device is a small, lightweight triaxial accelerometer that measures acceleration along three axes (vertical, anterior-posterior, and medio-lateral) at a sampling frequency of 100 Hz with a range of +8 g to −8 g. ...
... Indeed, AX3 models can differentiate between six distinct activity classes with exceptionally high accuracy in children (97.3%) [39]. Moreover, the device is valid for calculating walking steps [38,40]. The AX3 was used to assess PA levels and daily steps at pre-and post-intervention. ...
Article
Full-text available
This study examines how the 8-week Hoosier Sport program impacts cardiovascular disease (CVD) risks by promoting physical activity (PA) among rural, low-income children. Using a human-centered participatory co-design approach, the program aimed to increase PA levels (e.g., total PA, daily steps) in at-risk children. The present study explored the feasibility of the intervention as well as physiological and psychological changes across the intervention using a hybrid type 2 design (a model that evaluates both the effectiveness of an intervention and its implementation in real-world settings). Favorable feasibility indicators like attendance, acceptability, and compliance, with a 23.3% recruitment rate and 94.3% retention rate, were observed. Moreover, participants attended over 80% of sessions across the 8 weeks. Accelerometers (AX3) tracked daily steps and total PA for 7 days before and after the intervention, revealing increased PA levels throughout. At post-intervention, notable improvements were observed in psychological factors such as autonomy, social competence, and global self-worth. This study highlights the importance of tailored PA interventions in schools, emphasizing their potential to improve PA levels among rural, low-income children.
... The acceleration was subsequently resampled to 30 Hz. A sampling frequency of 30 Hz is sufficient to capture enough detail of the fast accelerations performed by children [28]. Data preprocessing was done in Matlab R2021b Version 9.11 (The Mathworks Inc., Natick, MA, USA) and prediction model training and evaluation were done in Python 3.9. ...
... Table 1. The activities performed by the participants during the structured protocol [28]. Playground activity was only performed by nine participants. ...
Article
Full-text available
The accurate estimation of energy expenditure from simple objective accelerometry measurements provides a valuable method for investigating the effect of physical activity (PA) interventions or population surveillance. Methods have been evaluated previously, but none utilize the temporal aspects of the accelerometry data. In this study, we investigated the energy expenditure prediction from acceleration measured at the subjects’ hip, wrist, thigh, and back using recurrent neural networks utilizing temporal elements of the data. The acceleration was measured in children (N = 33) performing a standardized activity protocol in their natural environment. The energy expenditure was modelled using Multiple Linear Regression (MLR), stacked long short-term memory (LSTM) networks, and combined convolutional neural networks (CNN) and LSTM. The correlation and mean absolute percentage error (MAPE) were 0.76 and 19.9% for the MLR, 0.882 and 0.879 and 14.22% for the LSTM, and, with the combined LSTM-CNN, the best performance of 0.883 and 13.9% was achieved. The prediction error for vigorous intensities was significantly different (p < 0.01) from those of the other intensity domains: sedentary, light, and moderate. Utilizing the temporal elements of movement significantly improves energy expenditure prediction accuracy compared to other conventional approaches, but the prediction error for vigorous intensities requires further investigation.
... The activity types standing, standing with minor movement, walking, running, and cycling were considered physically active behaviors. Algorithms were used to determine activity types with high accuracy in children and adults [21,22]. Details on the identi cation of non-wear periods are described in the study protocol [18]. ...
Preprint
Full-text available
Engagement in shared activities between parents and children is potentially compromised by the pervasive use of digital screens in familial contexts. In this randomized trial in 326 parent-child dyads nested in 87 families, we investigated the effects of limiting screen use in parents and children on the amount of synchrony in physical behaviors and family cohesion. Families were randomly assigned to wither undergo an extensive screen media reduction intervention or to control. For seven days at baseline and follow-up, parents and children each wore two accelerometers, positioned on the thigh and trunk, 24 hours/day, enabling the second-by-second classification of their physical behaviors. Time-series sequence analysis of physical behavior revealed significant enhancements in dyadic synchrony for the screen reduction group. In shared leisure time, the between-group mean difference in change favored the screen reduction group, with a -0.18 point (95%CI -0.27 to -0.10) decrease in time-warp edit distance dissimilarity score and a 32.9 min/day (95%CI 16.0 to 49.9) of more direct matched activity. Additionally, parents in the screen reduction group reported enhanced family communication, more collaborative tasks, and engagement in new shared activities. Our findings highlight the potential benefits of reducing screen time for improving parent-child behavioral synchrony and familial cohesion.
Article
Background: The ActiPASS software was developed from the open-source Acti4 activity classification algorithm for thigh-worn accelerometry. However, the original algorithm has not been validated in children or compared with a child-specific set of algorithm thresholds. This study aims to evaluate the accuracy of ActiPASS in classifying activity types in children using 2 published sets of Acti4 thresholds. Methods: Laboratory and free-living data from 2 previous studies were used. The laboratory condition included 41 school-aged children (11.0 [4.8] y; 46.5% male), and the free-living condition included 15 children (10.0 [2.6] y; 66.6% male). Participants wore a single accelerometer on the dominant thigh, and annotated video recordings were used as a reference. Postures and activity types were classified with ActiPASS using the original adult thresholds and a child-specific set of thresholds. Results: Using the original adult thresholds, the mean balanced accuracy (95% CI) for the laboratory condition ranged from 0.62 (0.56-0.67) for lying to 0.97 (0.94-0.99) for running. For the free-living condition, accuracy ranged from 0.61 (0.48-0.75) for lying to 0.96 (0.92-0.99) for cycling. Mean balanced accuracy for overall sedentary behavior (sitting and lying) was ≥0.97 (0.95-0.99) across all thresholds and conditions. No meaningful differences were found between the 2 sets of thresholds, except for superior balanced accuracy of the adult thresholds for walking under laboratory conditions. Conclusions: The results indicate that ActiPASS can accurately classify different basic types of physical activity and sedentary behavior in children using thigh-worn accelerometer data.
Article
There are currently several methods available to generate summary measures from acceleration, while ActiGraph (AG) counts as the first method to be used at large scale. The recent disclosure of the AG counts method exposes its intrinsic properties, which has not been accessible before. The intrinsic properties are the raw acceleration processing elements like filtering, rectification, or dead-band elimination, which are used to estimate physical activity intensity. The aim of this technical note is to compare the intrinsic properties of AG counts method with five alternatives (Euclidean Norm Minus One, mean average deviation, Activity Index, Rate of Change Accelerometry Movement, and Monitor-Independent Movement Summary) and how rescaling of AG counts and Monitor-Independent Movement Summary/minute into the International System of Units can be used to harmonize all summary measures and facilitate direct comparison. A total of 12 intrinsic properties are compared, and the overview demonstrates that there is large diversity regarding the specific intrinsic property elements being included, and with Monitor-Independent Movement Summary to be the only summary measure, which has been developed considering all elements. The harmonized output generated from all summary methods is highly comparable within common activities, but to obtain a robust summary measure recorded in subjects during free-living conditions, more research is warranted to evaluate the effect of the different intrinsic properties.
Article
Purpose: To study the effectiveness of a preschool staff-delivered motor skills intervention on body composition and physical activity over a 2.5-year time frame. Methods: In this pragmatic parallel cluster randomized controlled trial (16 preschools), outcome data were collected after 6 (body composition only), 18, and 30 months of intervention. The main physical activity outcomes were accelerometer behavior measures summarizing the total percentage of child daily movement (walk, run, cycle, and standing that included minor movements) and preschool movement during preschool attendance. To estimate between-group mean differences in outcomes, mixed-linear regression analyses including baseline value of the selected outcome and a treatment × time interaction term as a fixed effect were applied. In addition, the baseline preschool and child were included as a random effect. Results: For body mass index, a total of 437 children (90%) had at least one valid baseline and one follow-up assessment. The corresponding numbers for preschool movement and daily movement were 163 (55%) and 146 (49%), respectively. No significant between-group mean difference was identified for body mass index, waist-to-height ratio, or any physical activity outcomes. Conclusion: Overall, this preschool motor skills intervention had no effect on either child anthropometry or physical activity, consistent with previous studies.
Article
Full-text available
Sedentary behavior refers to certain activities in a reclining, seated, or lying position requiring very low energy expenditure. It has been suggested to be distinct from physical inactivity and an independent predictor of metabolic risk even if an individual meets current physical activity guidelines. Over the past decades, a shift in the activity profile of individuals has been observed with vigorous physical activity and sleep being partly replaced by cognitive work, a potential neurogenic stress component considering its hormonal and neurophysiological effects, leading to various impacts on health. Mental work, for instance, may significantly increase glycemic instability leading to an increase in the desire to eat and thus, higher energy intakes. Furthermore, screen-based leisure activities (e.g., television watching) and screen-based work activities (e.g., computer use for work purposes) have often been considered together while they may not trigger the same stress response and/or use of substrate. Thus, the problems of sedentariness may not only be attributed to a lack of movement, but also to the stimulation provided by replacing activities. The objective of this review is to discuss the (1) recent evidence and current state of knowledge regarding the health impact of sedentary behaviors on health; (2) potential neurogenic effects of cognitive work as a sedentary behavior; (3) link between sedentary behaviors and the diet; (4) resemblance between sedentary behaviors and the inadequate sleeper; and (5) potential solutions to reduce sedentary behaviors and increase physical activity.
Article
Full-text available
Abstract Wrist worn raw-data accelerometers are used increasingly in large-scale population research. We examined whether sleep parameters can be estimated from these data in the absence of sleep diaries. Our heuristic algorithm uses the variance in estimated z-axis angle and makes basic assumptions about sleep interruptions. Detected sleep period time window (SPT-window) was compared against sleep diary in 3752 participants (range = 60–82 years) and polysomnography in sleep clinic patients (N = 28) and in healthy good sleepers (N = 22). The SPT-window derived from the algorithm was 10.9 and 2.9 minutes longer compared with sleep diary in men and women, respectively. Mean C-statistic to detect the SPT-window compared to polysomnography was 0.86 and 0.83 in clinic-based and healthy sleepers, respectively. We demonstrated the accuracy of our algorithm to detect the SPT-window. The value of this algorithm lies in studies such as UK Biobank where a sleep diary was not used.
Article
Full-text available
To advance the field of time-use epidemiology, a tool capable of monitoring 24 h movement behaviours including sleep, physical activity, and sedentary behaviour is needed. This study explores compliance with a novel dual-accelerometer system for capturing 24 h movement patterns in two free-living samples of children and adults. A total of 103 children aged 8 years and 83 adults aged 20-60 years were recruited. Using a combination of medical dressing and purpose-built foam pouches, participants were fitted with two Axivity AX3 accelerometers—one to the thigh and the other to the lower back—for seven 24 h periods. AX3 accelerometers contain an inbuilt skin temperature sensor that facilitates wear time estimation. The median (IQR) wear time in children was 160 (67) h and 165 (79) h (out of a maximum of 168 h) for back and thigh placement, respectively. Wear time was significantly higher and less variable in adults, with a median (IQR) for back and thigh placement of 168 (1) and 168 (0) h. A greater proportion of adults (71.6%) achieved the maximum number of complete days when compared to children (41.7%). We conclude that a dual-accelerometer protocol using skin attachment methods holds considerable promise for monitoring 24-h movement behaviours in both children and adults.
Article
Full-text available
The purpose of this study was to test whether estimates of bedtime, wake time, and sleep period time (SPT) were comparable between an automated algorithm (ALG) applied to waist-worn accelerometry data and a sleep log (LOG) in an adult sample. A total of 104 participants were asked to log evening bedtime and morning wake time and wear an ActiGraph wGT3X-BT accelerometer at their waist for 24 h/day for 7 consecutive days. Mean difference and mean absolute difference (MAD) were computed. Pearson correlations and dependent-sample t tests were used to compare LOG-based and ALG-based sleep variables. Effect sizes were calculated for variables with significant mean differences. A total of 84 participants provided 2+ days of valid accelerometer and LOG data for a total of 368 days. There was no mean difference (p = 0.47) between LOG 472 ± 59 min and ALG SPT 475 ± 66 min (MAD = 31 ± 23 min, r = 0.81). There was no significant mean difference between bedtime (2348 h and 2353 h for LOG and ALG, respectively; p = 0.14, MAD = 25 ± 21 min, r = 0.92). However, there was a significant mean difference between LOG (0741 h) and ALG (0749 h) wake times (p = 0.01, d = 0.11, MAD = 24 ± 21 min, r = 0.92). The LOG and ALG data were highly correlated and relatively small differences were present. The significant mean difference in wake time might not be practically meaningful (Cohen’s d = 0.11), making the ALG useful for sample estimates. MAD, which gives a better estimate of the expected differences at the individual level, also demonstrated good evidence supporting ALG individual estimates.
Article
Full-text available
OBJECTIVES:To adapt and refine a previously-developed youth-specific algorithm to identify bedrest for use in adults. The algorithm is based on using an automated decision tree (DT) analysis of accelerometry data. DESIGN:Healthy adults (n = 141, 85 females, 19-69 years-old) wore accelerometers on the waist, with a subset also wearing accelerometers on the dominant wrist (n = 45). Participants spent ≈24-h in a whole-room indirect calorimeter equipped with a force-platform floor to detect movement. METHODS:Minute-by-minute data from recordings of waist-worn or wrist-worn accelerometers were used to identify bedrest and wake periods. Participants were randomly allocated to development (n = 69 and 23) and validation (n = 72 and 22) groups for waist-worn and wrist-worn accelerometers, respectively. The optimized DT algorithm parameters were block length, threshold, bedrest-start trigger, and bedrest-end trigger. Differences between DT classification and synchronized objective classification by the room calorimeter to bedrest or wake were assessed for sensitivity, specificity, and accuracy using a Receiver Operating Characteristic (ROC) procedure applied to 1-min epochs (n = 92,543 waist; n = 30,653 wrist). RESULTS:The optimal algorithm parameter values for block length were 60 and 45 min, thresholds 12.5 and 400 counts/min, bedrest-start trigger 120 and 400 counts/min, and bedrest-end trigger 1,200 and 1,500 counts/min, for the waist and wrist-worn accelerometers, respectively. Bedrest was identified correctly in the validation group with sensitivities of 0.819 and 0.912, specificities of 0.966 and 0.923, and accuracies of 0.755 and 0.859 by the waist and wrist-worn accelerometer, respectively. The DT algorithm identified bedrest/sleep with greater accuracy than a commonly used automated algorithm (Cole-Kripke) for wrist-worn accelerometers (p
Article
Full-text available
Study purpose The integration of methods to assess daytime physical activity (PA) and sedentary behavior (SB) and nighttime sleep would allow the evaluation of 24-hour daily activity using a single device. Accelerometer devices used to assess daytime PA have not been substantially validated to evaluate sleep. The objective of this study was to use polysomnography (PSG) to validate a commonly used PA accelerometer worn on both wrists and the hip. Methods Seventeen participants (50-75 years) completed a single-night in-home PSG recording while concurrently wearing 3 PA accelerometers. Accelerometer devices were worn on each wrist and the hip. Total sleep time (TST), sleep efficiency (SE), and wake after sleep onset (WASO) were compared for each device against PSG. Correlation coefficients estimated measurement agreement. Paired t tests and Bland-Altman plots assessed measurement differences. Results Between PSG and devices, mean TST ranged from 361.6 to 403.2 minutes. Mean SE estimates ranged from 86.9% to 96.9%. Mean WASO estimates ranged from 12 to 51.2 minutes. For TST, SE, and WASO hip estimates differed significantly from PSG estimates (paired t tests, TST: P = .03, SE: P < .001, WASO: P< .001). No significant differences were found between wrist accelerometers and PSG estimates of TST, SE, or WASO. Conclusions PA accelerometer devices worn on either wrist provide valid estimates of TST, WASO, and SE when compared with PSG. Further studies are needed to investigate methods to improve assessment of sleep parameters by PA accelerometer devices to advance device integration and assessment 24-hour activity in populations.
Article
Purpose: Accurate measurement of various human movement behaviors is essential in developing 24-h movement profiles. A dual-accelerometer system recently showed promising results for accurately classifying a broad range of behaviors in a controlled laboratory environment. As a progressive step, the aim of this study is to validate the same dual-accelerometer system in semi free-living conditions in children and adults. The efficacy of several placement sites (e.g., wrist, thigh, back) was evaluated for comparison. Methods: Thirty participants (15 children) wore three Axivity AX3 accelerometers alongside an automated clip camera (clipped to the lapel) that recorded video of their free-living environment (ground truth criterion measure of physical activity). Participants were encouraged to complete a range of daily-living activities within a 2-h timeframe. A random forest machine-learning classifier was trained using features generated from the raw accelerometer data. Three different placement combinations were examined (thigh-back, thigh-wrist, back-wrist), and their performance was evaluated using leave-one-out cross-validation for the child and adult samples separately. Results: Machine learning models developed using the thigh-back accelerometer combination performed the best in distinguishing seven distinct activity classes with an overall accuracy of 95.6% in the adult sample, and eight activity classes with an overall accuracy of 92.0% in the child sample. There was a drop in accuracy (at least 11.0%) when other placement combinations were evaluated. Conclusions: This validation study demonstrated that a dual-accelerometer system previously validated in a laboratory setting also performs well in semi free-living conditions. Although these results are promising and progressive, further work is needed to expand the scope of this measurement system to detect other components of behavior (e.g., activity intensity and sleep) that are related to health.
Article
Objective: To parameterize and validate two existing algorithms for identifying out-of-bed time using 24 h hip-worn accelerometer data from older women. Approach: Overall, 628 women (80 ± 6 years old) wore ActiGraph GT3X+ accelerometers 24 h d-1 for up to 7 d and concurrently completed sleep-logs. Trained staff used a validated visual analysis protocol to measure in-bed periods on accelerometer tracings (criterion). The Tracy and McVeigh algorithms were adapted for optimal use in older adults. A training set of 314 women was used to choose two key thresholds by maximizing the sum of sensitivity and specificity for each algorithm and data (vertical axis, VA, and vector magnitude [VM]) combination. Data from the remaining 314 women were then used to test agreement in waking wear time (i.e. out-of-bed time while wearing the accelerometer) by computing sensitivity, specificity, and kappa comparing the algorithm output with the criterion. Waking wear time-adjusted means of sedentary time, light-intensity physical activity (light PA) and moderate-to-vigorous-intensity physical activity (MVPA) were then estimated and compared. Main results: Waking wear time agreement with the criterion was high for Tracy_VA, Tracy_VM, McVeigh_VA, and highest for McVeigh_VM. Compared to the criterion, McVeigh_VM had mean sensitivity = 0.92, specificity = 0.87, kappa = 0.80, and overall mean difference (±SD) of -0.04 ± 2.5 h d-1. Minutes of sedentary time, light PA, and MVPA adjusted for waking wear time using the criterion measure and McVeigh_VM were not statistically different (p > 0.43|all). Significance: The McVeigh algorithm with optimal parameters using VM performed best compared to criterion sleep-log assisted visual analysis and is suitable for automated identification of waking wear time in older women when visual analysis is not feasible.
Article
Introduction: Accurately monitoring 24-h movement behaviors is a vital step for progressing the time-use epidemiology field. Past accelerometer-based measurement protocols are either hindered by lack of wear time compliance, or the inability to accurately discern activities and postures. Recent work has indicated that skin-attached dual-accelerometers exhibit excellent 24-h uninterrupted wear time compliance. This study extends this work by validating this system for classifying various physical activities and sedentary behaviors in children and adults. Methods: Seventy-five participants (42 children) were equipped with two Axivity AX3 accelerometers; one attached to their thigh, and one to their lower back. Ten activity trials (e.g., sitting, standing, lying, walking, running) were performed while under direct observation in a lab setting. Various time- and frequency-domain features were computed from raw accelerometer data, which were then used to train a random forest machine learning classifier. Model performance was evaluated using leave-one-out cross-validation. The efficacy of the dual-sensor protocol (relative to single sensors) was evaluated by repeating the modeling process with each sensor individually. Results: Machine learning models were able to differentiate between six distinct activity classes with exceptionally high accuracy in both adults (99.1%) and children (97.3%). When a single thigh or back accelerometer was used, there was a pronounced drop in accuracy for nonambulatory activities (up to a 26.4% decline). When examining the features used for model training, those that took the orientation of both sensors into account concurrently were more important predictors. Conclusions: When previous wear time compliance results are taken together with our findings, it represents a promising step forward for monitoring and understanding 24-h time-use behaviors. The next step will be to examine the generalizability of these findings in a free-living setting.
Article
Purpose: A Youth Compendium of Physical Activities (Youth Compendium) was developed to estimate the energy costs of physical activities using data on youth only. Methods: Based on a literature search and pooled data of energy expenditure measurements in youth, the energy costs of 196 activities were compiled in 16 activity categories to form a Youth Compendium of Physical Activities. To estimate the intensity of each activity, measured oxygen consumption (V[Combining Dot Above]O2) was divided by basal metabolic rate (Schofield age-, sex- and mass-specific equations) to produce a youth MET (METy). A mixed linear model was developed for each activity category to impute missing values for age ranges with no observations for a specific activity. Results: This Youth Compendium consists of METy values for 196 specific activities classified into 16 major categories for four age groups, 6-9, 10-12, 13-15, and 16-18 years. METy values in this Youth Compendium were measured (51%) or imputed (49%) from youth data. Conclusion: This Youth Compendium of Physical Activities uses pediatric data exclusively, addresses the age-dependency of METy and imputes missing METy values and thus represents advancement in the physical activity research and practice. This Youth Compendium will be a valuable resource for stakeholders interested in evaluating interventions, programs, and policies designed to assess and encourage physical activity in youth.This is an open-access article distributed under the terms of the Creative Commons Attribution-Non Commercial-No Derivatives License 4.0 (CCBY-NC-ND), where it is permissible to download and share the work provided it is properly cited. The work cannot be changed in any way or used commercially without permission from the journal.