Physical activity has long been accepted as providing important health benefits in adulthood (26), and emerging evidence suggests that even in young children, immediate and long-term health and other benefits are evident (8,15,19). However, protocols for the measurement of physical activity in preschool children (3 to 5 yr) are not well established and are hampered by several practical and methodological issues.
Preschool children lack the cognitive ability to self-report their behaviors (18). Because preschool children’s physical activity is highly sporadic in nature (2,5,11), measurement instruments that can accurately capture short bursts of intermittent behavior are required. Accelerometers capture acceleration and deceleration in real time and therefore allow for interpretation of the data to assess frequency, intensity, and duration of physical activity. Accelerometry has been widely used in the preschool population to measure physical activity and is a well-established method of reliably and validly capturing habitual physical activity in that population (5,11). However, there is lack of agreement about appropriate cut points for thresholds between physical activity intensities (several exist, including Refs. (12,16,22,28)) and a lack of evidence about the volume of data (i.e., number of days, number of hours within days, weekdays, and/or weekend days) required to reliably estimate physical activity in the preschool population.
Although several studies have now investigated where cut points should lie to identify thresholds between various intensities of physical activity (16,22,28), there is limited evidence about the volume of accelerometry data required to reliably estimate physical activity in preschool children. Published studies of preschool children’s physical activity assessed by accelerometry have used various minimum criteria to determine inclusion of a given case in analyses, and those criteria vary greatly from one or more hours on each of three or more days (13) to a minimum of 8 h·d−1 for 7 d (20). There is little consistency regarding the inclusion of both weekdays and weekend days, and studies largely fail to report reliability estimates of their accelerometry data inclusion criteria.
Only one study has reported on the volume of accelerometry data required to reliably estimate physical activity in preschool children (14). However, that study (14) used a small sample of 76 participants, a less sensitive sampling epoch of 1 min (17), and expressed data in counts per minute (cpm) rather than determining time in physical activity as has been done with studies determining reliability in older children and adolescents (25). That study found that 10 h of monitoring on each of 7 d was required to achieve a reliability of 0.8, and that 5 d with three or more hours of data on each day were required to achieve a reliability of 0.7. With only 4 d of monitoring, commonly used in studies with school-age children and likely to be much more achievable for participants, reliability was estimated at between 0.38 and 0.69, depending on the number of hours included (14). That study concluded that it was not necessary to include data from weekend days to reliably estimate habitual physical activity, despite lower estimates of reliability potentially suggesting differences between days (14). Given recent evidence that preschool children’s physical activity behavior varies between weekdays and weekend days (4), it may be necessary to determine whether or not data from both weekdays and weekend days are necessary to reliably estimate physical activity in the preschool population. Furthermore, differences in reliability estimates may be evident when comparing outcomes from cpm, as previously done (14), and actual time in physical activity, now commonly used as a primary outcome variable. New generations of ActiGraph accelerometers are now also commonly used, and potential small differences between those and older models, such as the 7164 used by Penpraze et al. (14), may also affect reliability estimates in the preschool population.
The aims of the study were to investigate the amount of physical activity data required to reliably estimate the percentage of time preschool children spend in physical activity per day and whether or not data should be included from both weekday and weekend day monitoring.
Recruitment and Participants
Data were drawn from the baseline period of the Healthy Active Preschool Years study, a cohort study established to investigate correlates of physical activity and sedentary behavior and to track changes in those behaviors in young children over time. Children and their families were recruited from randomly selected preschools and childcare centers in two randomly selected local government areas in each of low, medium, and high socioeconomic position areas of metropolitan Melbourne, Australia. Recruitment and data collection occurred in two phases: July to November 2008 and May to October 2009. In total, 156 childcare centers and 137 preschools were approached. The final sample consisted of 71 (46%) childcare centers and 65 (47%) preschools. From those childcare centers and preschools, 1032 preschool children and their parents consented to participate, an overall response rate of 10.5%. Of the 1032 participants, 485 were recruited through childcare centers (9.0% response) and 547 through preschools (12.4% response); four children were older than 5 yr and 24 withdrew before data collection, leaving a final sample of 1004. Deakin University Human Research Ethics Committee and the Victorian Department of Education and Early Childhood Development provided ethical approval for data collection. All participating center managers and parents provided informed written consent.
Measures and Data Management
Parents completed a survey that sought information on demographic details including the child’s sex and date of birth. Child’s date of birth was used to determine child’s age at the time of measurement. Data from surveys were entered by a commercial data entry company and then transferred into SPSS version 17.0 (23) (SPSS, Chicago, IL) for cleaning.
Children were fitted with an ActiGraph Model GT1M accelerometer on an elasticized belt. ActiGraph 7164 accelerometers are the only accelerometers that have established utility, validity, and reliability in children age 3 to 5 yr (5), and the GT1M model superseded the 7164. Children and parents (via written information) were instructed that the accelerometer was to be worn on the right hip from waking for the entire day, removed only for sleeping and aquatic activities (swimming and bathing). Data were collected in 15-s epochs to maximize opportunities to more accurately capture the sporadic nature of young children’s physical activity (5,22). Children wore accelerometers for an 8-d period, including the day of fitting and removal; however, because of some monitors not being returned on time, the actual wear time for some children was up to 10 d. Each monitor was initialized to commence recording at 9 a.m. on the day of fitting, and data management was adjusted for actual wear time. Data were downloaded after return of the monitor, and raw data files were then managed by a specially developed series of macros in Excel, followed by specially developed code in Stata version 11.0 (StataCorp LP, College Station, TX).
Data Reduction and Analyses
Monitoring start times for each participant on each day were identified as the beginning of the fourth complete minute of the appearance of counts above zero after 4 a.m., with a tolerance of four epochs (1 min) of zero counts. Non-wear time was determined as 10 min or more of consecutive zero counts, indicating the accelerometer recorded no movement at all for that period. In hourly increments, days with data ranging from 3 or more hours per day to 10 or more hours per day were entered into analyses. That is, days with 3 or more hours of data were tested for reliability, followed by assessment of days with 4 or more hours of data, then 5 or more hours of data, and so on, for each hourly increment until 10 or more hours of data were included in analyses. This approach was adopted to account for the age of the sample, where many children were still having daytime naps, thereby limiting the number of available hours during which the accelerometer could be worn. Days with 18 h or more of recorded data were excluded as being improbable.
Age-specific cut points (22) were applied to the data to identify physical activity. In this study, physical activity was operationalized as total physical activity, that is, light–moderate–vigorous activity, in accordance with interpretation of the National Association of Sport and Physical Education (1) physical activity guidelines in recent studies (3,4,27) and the Australian (6) and UK (7) physical activity recommendations for children from birth to 5 yr.
The percentage of time in total physical activity was determined for each day by dividing the number of minutes spent in total physical activity by the total wear time for that day and multiplying by 100. This approach was taken to account for potential differences in wear time within and between participants. The mean percentage of time in physical activity for weekdays was determined for those children who had three or more weekdays of data and for weekend days for those children who had one or more weekend days of data, at each of the hourly increments (i.e., 3 h or more to 10 h or more of data). Stata 11.0 was used to assess differences between weekday and weekend day physical activity and between estimates of mean percentage of time in physical activity at different inclusion criteria (3–10 h·d−1) using t-tests, which controlled for age and clustering by center of recruitment.
SPSS 17.0 was used to assess reliability using intraclass correlation coefficients (ICCs). The Spearman–Brown prophecy formula (24) was used to determine the number of days required at each of the estimated ICC values to meet reliabilities of 0.7, 0.8, and 0.9 (25). For each participant, 4 d from across the entire monitoring period (including week and potentially weekend days) were randomly selected and entered into analyses. Similarly, to ensure weekend days were included in reliability analyses, each of three weekdays and one weekend day were randomly selected and entered into analyses. A total of 4 d was chosen because this number of days is commonly used in studies in this population (4,9,14). Once the number of days to achieve each of the chosen reliabilities had been determined from the Spearman–Brown prophecy formula, the number of children achieving the criteria (no. of days, no. of hours per day) was calculated. Where the number of days required was a decimal, it was rounded up to the next whole number of days (i.e., 5.2 and 5.7 were both rounded up to 6) to ensure the minimum reliability criteria were met. For reliabilities for weekdays and weekend days, possible combinations of days including at least one weekend day were used (i.e., 5 d could be four weekdays and one weekend day or three weekdays and two weekend days) and the possible range of participants meeting the criteria are reported.
In total, sufficient accelerometry data for between 493 and 799 children was available for reliability analyses, depending on the minimum number of hours per day used as the inclusion criteria. The mean percentage of time in total physical activity is presented for boys and girls, and by weekdays and weekend days, in Table 1. Results show that children spent between 15.4% and 16.6% of their time on weekdays and between 17.2% and 17.4% of their time on weekend days in total physical activity. t-tests revealed that children spent a significantly higher mean percentage of their time in physical activity on weekend days compared with weekdays, regardless of the daily volume of data used (P < 0.001 for all). Estimates of the percentage of time in physical activity varied significantly between each of the minimum number of hours required for analyses (3–10 h) for both sexes on both weekdays and weekend days.
Table 2 presents the results for the ICCs and Spearman–Brown estimates of the number of days required to achieve the reliability estimates of 0.7, 0.8, and 0.9. In total, 570 children had four or more days of 10 or more hours of data and 493 had at least three weekdays and one weekend day each with 10 or more hours of data. The number of children having sufficient data to be entered into analyses increased as the number of hours of data collected per day decreased, and when three or more hours of data were used, 799 children had four or more days, and 744 children had at least three weekdays and one weekend day. Overall, the reliability derived from a single day of monitoring was moderate, ranging from 0.30 to 0.47. Higher ICC values were obtained for days with a greater minimum volume of data. Spearman–Brown analyses revealed that 2.7–5.4 d were required to achieve a reliability of 0.7, 4.6–9.3 d were required to achieve a reliability of 0.8, and 10.3–21.0 d were required to achieve a reliability of 0.9. Differences between reliability estimates for any 4 d of data compared with three weekdays and one weekend day of data were between 0.1 and 1.2 d.
Table 2 also shows the number (and percentage) of participating children who achieved the required number of days. When optimizing the sample size (10) for reliability of 0.7 (i.e., maximizing the sample size to achieve the predetermined reliability level), 781 children (78%) achieved four or more days with five or more hours of data per day, and 723 children (72%) achieved three weekdays and one weekend day with six or more hours of data. To optimize the sample size for inclusion in analyses at reliability of 0.8, 631 children (63%) had 6 d with seven or more hours of data each day, whereas only 567 children (56%) had four weekdays and two weekend days. Between 67% and 78% of children met the reliability criteria when the required ICC was set at 0.7, whereas no more than 63% met the criteria for an ICC of 0.8. Overall, fewer children met the reliability criteria when weekend days were necessarily included than when any days could be included in analyses. The “any 4 d” criteria may include participants with only weekday data and no weekend days. Because weekdays and weekend days are significantly different, if only weekdays are included, the days would be more similar to each other, requiring fewer days to achieve the same level of reliability, than if weekend days were necessarily included. No children achieved the 0.9 criteria for any of the monitoring periods.
This article has described the methods used to determine the amount of accelerometry data required to reliably estimate preschool children’s physical activity and results of those analyses. Because measurement of preschool children’s physical activity is still an emerging field, there is little evidence available to inform decisions regarding minimum inclusion criteria for accelerometry data. This study, therefore, adds to the evidence base informing such decisions for future studies. Reliability estimates varied depending on the minimum amount of wear time each day and whether or not valid data for weekend days were required for a child to be entered into analyses. The estimated number of days required to achieve any of the predetermined reliability estimates increased as the minimum amount of wear time on each day decreased. The substantially lower numbers of children with data for increasingly stringent criteria (i.e., a greater reliability level or a higher minimum number of hours per day) clearly show the trade-off between the more stringent criteria and retaining an adequate sample size for analyses. The percentage of time estimated to be spent in total physical activity also differed with the amount of data entered into analyses, by as much as 1.3%. Such differences may affect prevalence estimates and researchers, and public health practitioners should therefore be aware of the effect of data-related inclusion criteria used in those cases.
These results are comparable with published findings in older children and adolescents where 10 h of data per day on 4 to 5 d and 8 to 9 d of monitoring, respectively, are required (25). The results also suggest that a slightly lower volume of data may be required in the preschool population than previously published (14) when percentage of time in physical activity is used as the outcome rather than cpm and 15-s rather than 1-min epochs. Although the express criteria of including at least one weekend day made little difference to the estimated number of days required to achieve reliability, it did appear to affect the available sample meeting the criteria, with fewer children achieving the minimum numbers of hours required on weekend days. Differences between physical activity on weekdays and weekend days were significant, suggesting that it would be prudent to include weekend days in analyses.
Only one study has previously reported on reliability of physical activity data in preschool children (14). That study included only 76 participants and used cpm collected in 1-min epochs as the outcome. Because cpm includes all intensities of physical activity, including sedentary time, that study did not, therefore, investigate reliability of physical activity per se, when defined as light, moderate, and vigorous intensity physical activity. The current study included a much larger sample, 15-s epochs, and used percentage of time in total physical activity (i.e., percentage of time in light, moderate, and vigorous intensity physical activity) as the outcome. This study consequently found that fewer days of data were required to reliably estimate physical activity than previously suggested (14). Differences between findings may be due to differences in sample size, different inclusion criterion (6 h for 7 d compared with varying numbers of hours on each of the included days), epoch duration (1 min compared with 15 s), or differences in physical activity outcome variables (cpm compared with total physical activity).
An additional consideration in determining reliability of accelerometry data is the time frame during the day when monitoring takes place (i.e., morning, afternoon, etc.). In older children and adolescents (25), physical activity has been shown to systematically vary depending on the time of monitoring, and distinct periods are evident. However, in preschool children, who are subject to less structure in their day than school-age children, the same variability has not been evident (14). Therefore, in this study, specific periods of the day were not considered. Nonetheless, this study does show that estimates of percentage of time in total physical activity are significantly associated with the volume of daily data used in analyses.
In this sample and setting, 56% of the sample could be included in analyses if a 0.8 criterion was required for including weekend days (four weekdays and two weekend days, each with 7 h of data). Clearly, researchers need to maintain adequate sample sizes to ensure analyses are meaningful while being considerate of the estimated reliability of their chosen amount of data. An alternative may be to consider accepting lower reliability levels from accelerometry data, say 0.7, thereby requiring fewer days (three to four). With the inclusion of weekend days, this saw 20% more of the sample meeting the criteria than when a 0.8 reliability level was sought, equating to almost three quarters of the sample in total (72% had three weekdays and one weekend day each with 6 h of data, which meets the 0.7 reliability criteria). Implications of these findings include ensuring large initial sample sizes and implementing protocols to encourage and support participants wearing accelerometers for the required period. Suggestions for encouraging accelerometer wear in adolescents have previously been published (21,29); however, no known articles address this issue in the preschool population.
It is important to note that these analyses only used data collected from contiguous days because participants were instructed to wear the accelerometer daily over the 8 d of data collection. Contiguous days may not be representative of habitual physical activity for a variety of reasons, including seasonal influences, unusual social engagements (birthdays, other celebrations, and holidays), or health reasons. Data from noncontiguous days may provide different results.
Despite minimal differences in reliability estimates when weekend days were specifically included, physical activity differed significantly between weekdays and weekend days. Thus, it would be prudent for future studies to ensure data were included from both weekdays and weekend days to accurately represent preschool children’s physical activity across the entire week.
This study has shown that weekday and weekend day physical activities were significantly different. Therefore, data from both types of days should be included in analyses to accurately represent habitual physical activity. In addition, estimates of the mean percentage of time children spent being active varied with the minimum number of hours required for inclusion in analyses. Researchers should be aware of this when reporting estimates of prevalence of physical activity and compliance with recommendations. Results clearly show the tradeoff between using more stringent inclusion criteria for accelerometry data and retaining an adequate sample size for analyses. Further studies are required to determine the variability of physical activity during the day for the preschool population.
Trina Hinkley was supported by a Deakin University APA PhD Scholarship during the first half of data collection. Jo Salmon is supported by a National Heart Foundation of Australia Career Development Award and sanofi-aventis. Kylie Hesketh is supported by a National Heart Foundation of Australia Career Development Award. David Crawford is supported by a Victorian Health Promotion Foundation Senior Research Fellowship. The project was funded by Deakin University. The authors declare no conflict of interest. The results of the present study do not constitute endorsement by the American College of Sports Medicine.
1. American Alliance for Health P.E., Recreation and Dance (AAHPERD). NASPE releases first ever physical activity
guidelines for infants and toddlers. Illinois J Health PE Rec Dance
. 2002; 50: 31–2.
2. Bailey R, Olson J, Pepper S, Porszasz J, Barstow T, Cooper D. The level and tempo of children’s physical activities: an observational study. Med Sci Sports Exerc
. 1995; 27 (7): 1033–41.
3. Beets MW, Bornstein D, Dowda M, Pate RR. Compliance with national guidelines for physical activity
in U.S. preschoolers: measurement and interpretation. Pediatrics
. 2011; 127 (4): 658–64.
4. Cardon G, De Bourdeaudhuij I. Are preschool
children active enough? Objectively measured physical activity
levels. Res Q Exerc Sport
. 2008; 79 (3): 326–32.
5. Cliff DP, Reilly JJ, Okely AD. Methodological considerations in using accelerometers to assess habitual physical activity
in children aged 0–5 years. J Sci Med Sport
. 2009; 12 (5): 557–67.
6. Department of Health and Aging (DoHA). Get Up and Grow: Healthy Eating and Physical Activity for Early Childhood
. 2009 [cited 2009 28 October]; Physical Activity
Guidelines for 0–5 year olds. Available from: http://www.health.gov.au/internet/main/publishing.nsf/Content/EDFEDB588460BCE3CA25762B00232A13/$File/gug-directorscoord-2.pdf
7. Department of Health, Physical Activity
, Health Improvement and Protection. Start Active, Stay Active: A Report on Physical Activity from the Four Home Countries’ Chief Medical Officers
. London: Department of Health, Physical Activity
, Health Improvement and Protection; 2011. p. 1–62.
8. Janz KF, Burns TL, Torner JC, et al.. Physical activity
and bone measures in young children: the Iowa Bone Development Study. Pediatrics
. 2001; 107 (6): 1387.
9. Janz KF, Levy SM, Burns TL, Torner JC, Willing MC, Warren JJ. Fatness, physical activity
, and television viewing in children during the adiposity rebound period: the Iowa Bone Development Study. Prev Med
. 2002; 35 (6): 563–71.
10. Matthews CE, Hagstromer M, Pober DM, Bowles HR. Best practices for using physical activity
monitors in population-based research. Med Sci Sports Exerc
. 2012; 44 (suppl 1): S68–76.
11. Oliver M, Schofield G, Kolt G. Physical activity
in preschoolers: understanding prevalence and measurement issues. Sports Med
. 2007; 37 (12): 1045–70.
12. Pate RR, Almeida M, McIver K, Pfeiffer K, Dowda M. Validation and calibration of an accelerometer in preschool
. 2006; 14 (11): 2000–6.
13. Pate RR, Pfeiffer KA, Trost SG, Ziegler P, Dowda M. Physical activity
among children attending preschools. Pediatrics
. 2004; 114 (5): 1258–63.
14. Penpraze V, Reilly JJ, MacLean CM, et al.. Monitoring of physical activity
in young children: how much is enough? Pediatr Exerc Sci
. 2006; 18 (4): 483–91.
15. Reilly JJ. Physical activity
, sedentary behaviour and energy balance in the preschool
child: opportunities for early obesity prevention. Proc Nutr Soc
. 2008; 67 (3): 317–25.
16. Reilly JJ, Coyle J, Kelly L, Burke G, Grant S, Paton J. An objective method for measurement of sedentary behavior in 3- to 4-year olds. Obes Res
. 2003; 11 (10): 1155–8.
17. Reilly JJ, Kelly LA, Montgomery C, et al.. Validation of ActiGraph accelerometer estimates of total energy expenditure in young children. Int J Pediatr Obes
. 2006; 1 (3): 161–7.
18. Sallis JF. Self-report measures of children’s physical activity
. J Sch Health
. 1991; 61 (5): 215–9.
19. Shea S, Basch CE, Gutin B, et al.. The rate of increase in blood pressure in children 5 years of age is related to changes in aerobic fitness and body mass index. Pediatrics
. 1994; 94 (4 Pt 1): 465–70.
20. Sigmund E, De Ste Croix M, Miklankova L, Fromel K. Physical activity
patterns of kindergarten children in comparison to teenagers and young adults. Eur J Public Health
. 2007; 17 (6): 646–51.
21. Sirard JR, Slater ME. Compliance with wearing physical activity
accelerometers in high school students. J Phys Act Health
. 2009; 6 (suppl 1): S148–55.
22. Sirard JR, Trost SG, Pfeiffer KA, Dowda M, Pate RR. Calibration and evaluation of an objective measure of physical activity
children. J Phys Act Health
. 2005; 2 (3): 345.
23. SPSS Inc. SPSS Statistics 17.0
. Release 17.0.0. Chicago: SPSS Inc; 2008.
24. Stanley JC. Reliability, in Educational Measurement
. In: Thorndike RL, editor. Washington (DC): American Council of Education; 1971. p. 395.
25. Trost SG, Pate RR, Freedson PS, Sallis JF, Taylor WC. Using objective physical activity
measures with youth: how many days of monitoring are needed? Med Sci Sports Exerc
. 2000; 32 (2): 426–31.
26. US Department of Health and Human Services. Physical Activity and Health: A Report of the Surgeon General
. Atlanta (GA): US Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion; 1996.
27. Vale S, Silva P, Santos R, Soares-Miranda L, Mota J. Compliance with physical activity
guidelines in preschool
children. J Sports Sci
. 2010; 28 (6): 603–8.
28. van Cauwenberghe E, Labarque V, Trost SG, de Bourdeaudhuij I, Cardon G. Calibration and comparison of accelerometer cut points in preschool
children. Int J Pediatr Obes
. 2010; 6 (2–2): e582–9.
29. Van Coevering P, Harnack L, Schmitz K, Fulton JE, Galuska DA, Gao S. Feasibility of using accelerometers to measure physical activity
in young adolescents. Med Sci Sports Exerc
. 2005; 37 (5): 867–71.