Pediatric Exercise Science, 2006, 18, 483-491
© 2006 Human Kinetics, Inc.
Monitoring of Physical Activity
in Young Children:
How Much Is Enough?
Victoria Penpraze, John J. Reilly, Christina M. MacLean,
Colette Montgomery, Louise A. Kelly, James Y. Paton,
Thomas Aitchison, and Stan Grant
There is limited evidence on how much and on which days accelerometry monitoring should be performed to obtain a representative measurement of physical
activity (PA) in young children. We measured 76 children (40 M and 36 F, mean
age 5.6 years ([SD ± 0.4]) on 7 days using Actigraph accelerometers. Mean daily
PA was expressed in counts per min (cpm). Reliability increased as the number
of days and hours of monitoring increased, but only to 10 hr per day. At 7 days
of monitoring for 10 hr per day, reliability was 80% (95% CI [70%, 86%]). The
number of days was more important to reliability than the number of hours. The
inclusion or exclusion of weekend days made relatively little difference. A monitoring period of 7 days for 10 hr per day produced the highest reliability. Surprisingly
short monitoring periods may provide adequate reliability in young children.
Regular physical activity (PA) has been recognized as providing protection against
many chronic diseases of adult life (4). There is substantial evidence that physical
activity also provides important health benefits to children (2,30) and establishes
behavior patterns that may continue into adulthood (12, 21).
The measurement of physical activity in children has its own particular difficulties. These include the sporadic nature of the childrenʼs activity patterns and
their lack of cognitive ability to accurately recall the amount and intensity of activity (16). Accelerometry has become widely used for characterizing two important
dimensions of physical activity: total amount (volume) of physical activity and
intensity of activity (time spent at or above a specified intensity level). The Actigraph accelerometer (formerly known as MTI Actigraph or Computer Science
Applications [CSA] 7164, Fort Walton Beach, FL) is one of the most widely used
accelerometers and has been shown to be a valid and reliable tool for measuring
physical activity volume and intensity in young children (1,5,15,27,29).
Penpraze and Grant are with the University of Glasgow, Institute of Diet, Exercise and Lifestyle, Faculty
of Biological and Life Sciences, West Medical Building, Glasgow, Scotland, UK; Reilly, Montgomery,
Kelly, and Paton are with the University of Glasgow Division of Developmental Medicine, Yorkhill
Hospitals, Glasgow, G3 8SJ, Scotland, UK; MacLean and Aitchison are with the University of Glasgow
Department of Statistics, 15 University Gardens, Glasgow, Scotland, UK.
483
484
Penpraze et al.
Some outstanding methodological issues exist in the use of accelerometers to
assess physical activity in children. One of the most important is how long monitoring should be to allow a representative estimate of physical activity to be determined.
At present, there is limited evidence on the recommended measurement periods in
free-living environments. In the literature, examples of monitoring periods range
from as few as 2 days, frequently 3 or 4 days including 1 weekend day, up to 2
weeks (3,6,7,9,20). There appears to be little theoretical or empirical justification
for the duration of monitoring. Trost et al. (26) investigated age-related differences
in the reliability of physical activity measurements in children aged 6 to 16 years.
They suggested that, depending on age, between 4 and 7 days of monitoring,
including a weekend day, are required in order to obtain a reliable measurement
of childrenʼs physical activity levels. Trost and associates demonstrated that the
monitoring period required to produce a representative measurement of physical
activity varies with age. However, to our knowledge, there are no studies using
accelerometers, which have been considered the most suitable monitoring period
in children younger than 6 years.
The primary aim of this study was to investigate the number of days and hours
of monitoring required to obtain representative measures of physical activity of
younger children. We also wished to determine whether monitoring should include
weekend days.
Methods
Participants
Accelerometry data were collected from young children, participating in the Study
of Preschool Children Activity, Lifestyle, and Energetics (SPARKLE) in Glasgow,
Scotland (8,18,19). This was a mixed longitudinal study of physical activity and
sedentary behavior in a socioeconomically representative sample of children. The
SPARKLE study has been described in detail elsewhere (8,19). The Research Ethics
Committee of the Royal Hospital for Sick Children at Yorkhill, Glasgow, approved
the study, and written informed consent was obtained from all parents.
Initial screening identified participants with acceptable compliance with the
measurement protocol as those participants with a minimum of 3 days of monitoring
of at least 6 hr per day (8, 13). In practice, adherence to the measurement protocol
was good and duration of monitoring was generally much longer than the minimum
stipulated time. To ensure uniformity of monitoring periods in this study, those
participants from the original study with missing data were excluded. This present
study was a secondary analysis of data collected from a subset of 76 children (40
boys and 36 girls) participating in the SPARKLE study who had provided at least
7 consecutive days of accelerometry data (mean age 5.6 years, SD ± 0.4).
Protocol
Accelerometry
Physical activity data were collected using Actigraph activity monitors. The Actigraph is a small (51 × 38 × 15 mm) and lightweight (43 g, including the weight of
Monitoring Physical Activity in Children
485
the belt) lithium-battery-powered monitoring device that measures accelerations in
a vertical axis. The uniaxial accelerometer measures movements within the range
0.5 to 2 G and frequencies of 0.25 and 2.5 Hz. Movement produces a voltage signal
that is proportional to acceleration. A digital count is produced by converting this
signal using an 8-bit A/D converter and the counts are summed over a user-defined
period of time (epoch) (28).
Participants’ Accelerometer Use
Participantsʼ parents were given an Actigraph and accelerometer belt. They were
asked to complete a daily diary of what time the belt was attached and removed
each day. The accelerometers and diaries were collected by the researchers 7 to
10 days later. Parents and children were informed that the child should wear the
accelerometer on the right hip underneath clothing (5,14,25), and they were shown
how to secure the accelerometer belt according to the manufacturerʼs guidelines.
Participants were asked to wear the device from the time they woke in the morning
until bedtime in order to capture all their waking physical activity for each day. The
only exception was during showering or bathing when the parents were instructed
to remove the accelerometer, which is not waterproof, in order to prevent damage
to the device. Parents were asked to note daily in the diary provided the time the
child began wearing the device, when it was removed at the end of the day, and
any time the device was removed and reattached during the day.
Data Reduction and Analyses
Data were uploaded according to the manufacturerʼs guidelines for analysis. All
participantsʼ monitoring periods included the waking hours of the day; start and
finish times were determined manually in conjunction with the parental record of
times from the diary. Monitoring start times for each participant on each day were
identified as the beginning of the third complete minute after the appearance of
counts above zero and not before the parental record indicated that the activity
monitor had been attached. Finish times were identified as the end of the third last
minute before the beginning of the overnight period of extended zero counts and
not after the time the device was worn as recorded in the diary. The exclusion of
the first and last 2 min of monitoring ensured the counts produced from movement
associated with attaching and removing the accelerometer were not included in the
analysis. The median start and finish times for monitoring period were 8:36 a.m.
(interquartile range [IQ] 8:05 to 9:50 a.m.) and 8:13 p.m. (IQ range 7:17 p.m. to
9:08 p.m.), respectively. Mean number of hours of monitoring per day was 10.9
(SD ± 1.9). Mean total number of hours of monitoring per participant was 84.5 hr
(SD ± 16.4).
For each participant, data were expressed in counts per minute. All accelerometer records were reviewed to identify periods of repeated zeros. These periods of
repeated zeros were seen once on 1 day of monitoring in 9 participants and once
on each of 2 days of monitoring in 1 out of 76 participants. Such periods were
checked against parental diaries to determine whether the Actigraph was being
worn. Those periods of repeated zero counts corresponding to parental records were
excluded. Other, essentially identical, periods of repeated zeros where the parental
486
Penpraze et al.
record gave no information were also excluded. Mean daily physical activity was
calculated in accelerometer counts per minute (cpm) by dividing the total counts
per minute by the number of minutes of monitoring per day.
Statistics
Data were summarized using standard descriptive statistics. Repeated measures
analysis of variance was performed to determine whether there was a difference in
activity levels between girls and boys on the weekend and weekdays.
Logarithms of the mean daily counts per minute were calculated, because
of the heteroscedasticity of counts at increasing activity levels, to ensure withinparticipant variability remained constant. Before each analysis an assessment of
data normality was determined on the logarithmically transformed data. Reliability coefficients and 95% confidence intervals (95% CI) were calculated using
the Spearman Brown prophecy formula (S-B; 22,23,24). This statistical method
is used to determine the internal consistency of data (26). The effect on reliability of lengthening or shortening a monitoring period can be computed using the
Spearman-Brown prophecy formula. The S-B states the following:
R = σB2 / [σB2 + (σW2 / Number of Days of Measurements)] × 100
where: R = reliability, σB2 = between-participant variance, σW2 = within-participant
variance.
Results
Total Volume of Physical Activity
The mean (±SD) daily physical activity was significantly higher for males (870
± 187 cpm) than females (771 ± 161 cpm) by 99 cpm (p = .011). Both males and
females were significantly (p = .02) more active on weekends (900 ± 245 cpm and
833 ± 219 cpm, respectively) than on weekdays (858 ± 183 cpm and 747 ± 164
cpm, respectively).
Influence of Monitoring Period on Reliability
Reliability coefficients are shown for various combinations of days and hours per day of activity monitoring in a matrix (Table 1).
As indicated by the S-B prophecy formula, the most reliable measure of physical
activity was a monitoring period of 7 days and 10 hrs per day (R = 80%; 95% CI
70%, 86%) (Table 1). These data also showed that the reliability of the physical
activity measurement clearly depended on both the number of hours and the number
of days. The evidence from Table 1 suggests that the reliability of the measurement
of physical activity in our sample and setting remained relatively stable from as
little as 3 hr per day up to and no more than 10 hr per day of monitoring. With
constant or even reduced daily periods of monitoring, higher reliabilities are possible if the number of days of monitoring is increased. Perhaps surprisingly, in the
present study the reliability was lower if participants were monitored for 11 hr or
Number of Days and Hours.
Monitoring Physical Activity in Children
487
Table 1 Reliability (%) of Monitoring Period for Mean Daily
Physical Activity Based on 76 Participants With Complete Data
Sets of 7 Days
Hours
13
12
11
10
9
8
7
6
5
4
3
Day 1
Day 2
Day 3
Day 4
Day 5
Day 6
Day 7
n
13
25
33
36
35
33
34
34
35
35
35
24
41
49
53
52
50
51
51
51
51
52
32
51
59
62
62
60
61
61
61
61
62
38
58
66
69
68
67
67
67
68
68
69
44
63
71
73
73
71
72
72
72
73
73
48
67
74
77
76
75
76
76
76
76
77
52
71
77
80
79
78
78
78
79
79
79
33
58
73
75
76
76
76
76
76
76
76
more per day (e.g., the estimated reliability for 7 consecutive days of measurement
for 13 hr per day was 52% compared with 80% for 10 hr of monitoring per day;
see Table 1).
Table 2 shows reliability coefficients
and 95% confidence intervals from Spearman-Brown analysis based on data sets
with 4 days of monitoring for periods including weekdays only, and with 4 days of
data including 1 weekend day. This period was chosen on the basis of the analysis
as it is frequently quoted as inclusion criteria for the minimal required duration
of monitoring (9,11). Reliability coefficients and 95% confidence intervals show
that the inclusion of weekend days in the monitoring period either had a negligible
effect on reliability or slightly decreased the reliability of the estimation of total
physical activity in this sample of young children (i.e., 4-day monitoring period
including 3 weekdays and 1 weekend day rather than 4 weekdays reduced the reliability slightly from 84% to 82%).
Inclusion of Weekdays and Weekend Days.
Discussion
In the present study, reliability of measurement of physical activity became higher
as the number of days of monitoring increased up to peak reliability of 80% (95%
CI 71%, 86%) for 7 days. Whether the monitoring period included a weekend day
did not significantly alter the reliability.
There was a high degree of consistency between the estimated reliabilities
and their confidence intervals in the present study and those reported by Trost et
al. (26) for older children. In both studies, 7 days of monitoring for >8 hr per day
produced the highest reliability for young children. However, there are implications
to both cost and participant compliance with such a long monitoring period. In our
488
Penpraze et al.
Table 2 Reliability (%) and 95% Confidence Intervals of Mean Daily
Physical Activity Measurements
2 days
3 days
4 days
Weekdays only
73%
(55%, 83%)
80%
(64%, 88%)
84%
(71%, 91%)
Including 1 weekend day
70%
(54%, 80%)
78%
(64%, 86%)
82%
(70%, 89%)
Note. Based on data including 4 days of measurement with and without inclusion of a weekend day.
experience, accelerometers used in studies are sometimes returned with fewer than
7 days of monitoring and the monitoring may not always include a weekend day.
The results of the present study (Table 1) should allow researchers to determine
the likely level of reliability for shorter periods of monitoring in young children.
In the present study, reliability exceeded 70% (widely regarded as acceptable, e.g.,
Trost et al.; [26]) for even relatively short days of monitoring (e.g., 3-4 hr/day in
Table 1) so long as at least 5 days were monitored. In the present study the number
of days of monitoring had more marked effects on reliability than number of hours
per day (Table 1).
It has been reported that participation in MVPA in 6- to 8-year-old children
occurs in two distinct and independent time components (i.e., on weekdays 7:00
a.m. to 10:59 a.m. and 11:00 a.m. to 8:59 p.m.; [26]). Trost et al. proposed that
participation in MVPA was consistent within each component though different
between time periods; thus selecting a monitoring period with fewer than 10 hr
may potentially bias the resultant physical activity estimation (26). The present
study suggests that this influence is not evident in the younger children in our
setting, who do not exhibit the same structure of a school day as the volunteers in
Trostʼs study. The results of the present study may not be generalizable across all
age populations, and researchers may need empirical data for other populations
with different lifestyles.
Anecdotally it is thought that the longer the monitoring period, the more
representative the estimation of physical activity. However, the findings of the
present study show this may be true only up to a point, in young children, since
the reliability showed a marked decrease if the monitoring period extended to
11 or more hours per day of measurement, irrespective of the number of days of
monitoring (Table 1).
The percentage of time spent in activity categories of sedentary, light, and
moderate to vigorous, determined using published pediatric accelerometer cut
points (17,18), is an important construct of physical activity. The conclusions of
this current study were very similar irrespective of whether accelerometry data were
expressed as counts per minute or as percentage of time in these activity categories.
For instance, R = 80% (95% CI 71%, 81%) for 4 days of monitoring when activity
is expressed as percentage time spent in each of the activity categories.
Monitoring Physical Activity in Children
489
A number of studies using objective assessment methods have reported higher
levels of physical activity in children during the weekend (e.g., 10,26), and for this
reason it has been considered important to include a weekend day in the monitoring
period (26). In the present study, both boys and girls were statistically significantly
more active on weekends. However, the magnitude of the difference between weekend and weekday may not be biologically meaningful (boys were more active on
weekends by just 42 cpm and girls by only 86 cpm). Further consideration of the
width of published pediatric cut points delineating activity levels (e.g., light activity
ranging from 801 to 3,199 cpm and moderate activity ranging from 3,200 to 8,200
cpm; [17]) would also suggest that the relatively small differences in accelerometry
counts per min in the present study reflect very small differences in activity levels.
With the analyses in the present study, for young children, reliability did not differ
significantly with the inclusion or exclusion of weekend days.
Conclusions
The results of the present study suggest that a monitoring period of 7 days and
10 hr per day produced the highest level of reliability in measurements of physical activity in young children. However, surprisingly short monitoring periods
may provide an appropriate level of reliability in this age group. These shorter
monitoring durations need not include a weekend day. The evidence presented
here should allow researchers to design practical research protocols at a specified
level of reliability.
Acknowledgments
We wish to thank the parents, children, nursery staff, and Glasgow City Council
for their involvement and cooperation in this research. The SPARKLE study was
funded by SPARKS 00GLW/01.
References
1. Basset, D.R. Validity and reliability issues in objective monitoring of physical activity.
Res. Q. Exerc. Sport. 71:30, 2002.
2. Caspersen, C.J., P.A. Nixon, and R.H. Du Rant. Physical activity epidemiology applied
to children and adolescents. Exerc. Sport Sci. Rev. 26:314-403, 1998.
3. Cooper, A.R., A.S. Page, L.J. Foster, and D. Qahwaji. Commuting to school: are children
who walk more physically active? Am. J. Prev. Med. 25:273-276, 2003.
4. Dietz, W.H. The role of lifestyle in health: the epidemiology and consequences of
inactivity. Proc. Nutr. Soc. 55:829-840, 1996.
5. Fairweather, S.C., J.J. Reilly, S. Grant, A. Whittaker, and J.Y. Paton, Using the Computer Science and Applications (CSA) activity monitor in preschool children. Pediatr.
Exerc. Sci. 11:413-420, 1999.
6. Finn, K., N. Johannsen, and B. Specker. Factors associated with physical activity in
preschool children. J. Pediatr. 140:81-85, 2002.
7. Hoos, M.B., G. Plasqui, W.J.M. Gerver, and K.R. Westerterp. Physical activity level
measured by doubly labelled water and accelerometry in children. Eur. J. App. Phys.
89:624-626, 2003.
490
Penpraze et al.
8. Jackson, D.M., J.J. Reilly, L.A. Kelly, C. Montgomery, S. Grant, and J.Y. Paton. Objectively measured physical activity in a representative sample of 3-4 year-old children.
Obes. Res. 11:420-425, 2003.
9. Janz, K.F., S.M. Levy, T.L. Burns, J.C. Torner, M.C. Willing, and J.J. Warren. Fatness,
physical activity and television viewing in children during the adiposity rebound period:
the Iowa Bone Development Study. Prev. Med. 35:563:571, 2002.
10. Janz, K.F., J. Witt, and L.T. Mahoney. Validation of the CSA accelerometer for assessing childrenʼs physical activity. Med. Sci. Sports Exerc. 26:369-375, 1995.
11. Kalakanis, L.E., G.S. Goldfield, R.A. Paluch, and L.H. Epstein. Parental activity as a
determinant of physical activity level and pattern of activity in obese children. Res. Q.
Exerc. Sport. 72:202-209, 2001.
12. Kraut, A., D. Melamed, D. Gofer, and P. Froom. Effects of school age sports on leisure
time physical activity in adults: the CORDIS Study. Med. Sci. Sports Exerc. 35:20382042, 2003.
13. Montgomery, C., J.J. Reilly, D.M. Jackson, L.A. Kelly, C. Slater, J.Y. Paton, and S.
Grant. Relation between physical activity and energy expenditure in a representative
sample of young children. Am. J. Clin. Nutr. 80:591-596, 2004.
14. Nilsson, A., U. Ekelund, A. Yngve, and M. Sjostrom. Assessing physical activity among
children with accelerometers using different time sampling intervals and placements.
Pediatr. Exerc. Sci. 14:87-96, 2002.
15. Ott, A.E., R.R. Pate, S.G. Trost, D.S. Ward, and R. Saunders. The use of uniaxial and
triaxial accelerometers to measure childrenʼs “free-play” physical activity. Pediatr.
Exerc. Sci. 12:360-370, 2000.
16. Pate, R.R. Physical activity assessment in children and adolescents. Crit. Rev. Food
Sci. Nutr. 33:321-326, 1993.
17. Puyau, M.R., A.L. Adolph, F.A. Vohra, and N.F. Butte. Validation and calibration of
activity monitors in children. Obes. Res. 10:150-157, 2002.
18. Reilly, J.J., J. Coyle, L.A. Kelly, G. Burke, S. Grant, and J.Y. Paton. An objective
method for measurement of sedentary behavior in 3-4 year olds. Obes. Res. 11:11551158, 2003.
19. Reilly, J.J., D.M. Jackson, C. Montgomery, L.A. Kelly, C. Slater, S. Grant, and J.Y.
Paton. Total energy expenditure and physical activity in young Scottish children: mixed
longitudinal study. Lancet. 363:211-212, 2004.
20. Riddoch, C.J., L.B. Anderson, N. Wedderkopp, et al. Physical activity levels and patterns of 9- and 15-yr-old European children. Med. Sci. Sports Exerc. 36:86-92, 2004.
21. Serdula, M.K., D. Ivery, R.J. Coates, D.S. Freedman, D.F. Williamson, and T. Byers.
Do obese children become obese adults? A review of the literature. Prev. Med. 22:167177, 1993.
22. Spearman, C.E. ʻGeneral Intelligenceʼ objectively determined and measured. Am. J.
Psych. 5:201-293, 1904.
23. Spearman, C.E. Proof and measurement of association between two things. Am. J.
Psych. 5:72-101, 1904.
24. Stanley, J.C. Reliability. In: Educational Measurement, R.L. Thorndike (Ed.). Washington, DC: American Council of Education, 1971, pp. 395.
25. Swartz, A.M., S.J. Strath, D.R. Bassett Jr., W.L. OʼBrien, G.A. King, and B.E. Ainsworth. Estimation of energy expenditure using CSA accelerometers at hip and wrist
sites. Med. Sci. Sports Exerc. 32(9 Supp.):S450-S456, 2000.
26. Trost, S.G., R.R. Pate, P.S. Freedson, J.F. Sallis, and W.C. Taylor. Using objective
physical activity measures with youth: how many days of monitoring are needed? Med.
Sci. Sports Exerc. 32:426-431, 2000.
27. Trost, S.G., D.S. Ward, S.M. Moorehead, P.D. Watson, W. Riner, and J.R. Burke. Validity of the computer science and applications (CSA) activity monitor in children. Med.
Sci. Sports Exerc. 30:629-633, 1998.
Monitoring Physical Activity in Children
491
28. Tyron, W.W., and R. Williams. Fully proportional actigraphy: A new instrument. Behav.
Methods. Instruments Comput., 28(3):392-403, 1996.
29. Welk, G.J., J.A. Schaben, and J.R. Morrow Jr. Reliability of accelerometry-based activity monitors: a generalizability study. Med. Sci. Sports Exerc. 36:1637-1645, 2004.
30. Williams, C.L., L.L. Hayman, S.R. Daniels, T.N. Robinson, J. Steinberger, S. Paridon, and T. Bazzarre. Cardiovascular health in childhood. Circulation. 106:143-160,
2002.