Introduction

Physical Activity (PA) and its measurement have quickly become a focus for many researchers who target the childhood obesity crisis that faces this nation. Psychometrically sound measures are critical for research efforts designed to evaluate the effects of interventions. Experts have identified three categories of activity measures: self-reported measures (e.g., Physical Activity Questionnaire - Child [PAQ-C]; [1]; direct observation methods (e.g., Child Activity Rating System [CARS]; [2]; and motion monitors (e.g., pedometers, accelerometers).

Identifying valid and reliable measures of PA with young children is a complicated task because children younger than 10 years old are unable to accurately recall and report this information; [3,4] as a result parents are often called upon to report the child’s activity. Just as self-reported data may be biased, the accuracy of second-hand reports or parent reports is suspect [5]. Moreover, direct observational methods of children’s PA are costly and time-intensive and thus not optimal to use with moderate to large samples in appropriately powered full-scale intervention studies. Thus, researchers may rule out the use of parent reports and direct observation and opt to use motion monitors. The valid use of motion monitors in young children should be initiated by trained experts because of the physiologic differences that exist in children as compared with adults which directly affect which monitors are used and how data is interpreted.

The body size and leg length of young children (3-4 years of age) may affect the function of the internal mechanisms of pedometers, rendering data with questionable reliability and validity [6]. The use of accelerometers, on the other hand, has increased dramatically as a way to assess the PA of young children largely because they provide a feasible, valid and reliable objective measure of habitual activity (activity over time [7]. Unlike parent report and direct observation, they measure the frequency of PA in addition to the intensity [8]. Software programs have significantly reduced the time and cost involved in the process of data reduction, necessary step when using accelerometers, thereby facilitating their use. This paper describes the variation in outcomes researchers can anticipate using different approaches to the data reduction process.

Background

Accelerometers function by measuring change in velocity or acceleration, identified as an activity count, over a specified time interval (epoch), thus enabling intensity of PA to be quantified [9]. The ability of researchers to determine the epoch length when using accelerometers with young children is critical because young children frequently have short bouts of high-level activity (i.e., Moderate-To-Vigorous Physical Activity [MVPA]) which should not be averaged into long periods of time with less activity. If one calculates PA with epochs greater than 15 seconds the result may be a finding of low cumulative amounts of time spent in MVPA intensity and large amounts of time spent in lower levels of PA (i.e., sedentary and light activity). A Metabolic Equivalent of Task (MET) is a standard physiological measure expressing the energy cost of PA (1 MET = 1 kcal/kg/hour; [10]. The activity count/epoch that results from accelerometer use can be translated into energy expenditure (METs) using the following calculation: 1 MET = 1 kcal/kg of body weight * height [11], making this a helpful tool for obesity investigators.

Several researchers have conducted calibration and validation studies specifically with young children using accelerometry to determine PA intensity levels (i.e., sedentary, light, or MVPA [12-15]. These studies have resulted in thresholds of activity counts/epoch that equate to child-specific intensities of PA [16-25] and thus energy expenditure. These differences may be attributable to the use of different brands of accelerometers; however, controversy remains over what the thresholds or cut points should be and the best methodology to establish the validity of those threshold levels [12,13,22,26,27].

Sirard and colleagues [22] conducted a study with 281 preschool aged children in which accelerometers were placed on the children, and the CARS was the criterion PA measure [2]. The children were asked to complete activities representative of the PA intensity categories of CARS for specific time intervals; then the data from the accelerometers were temporally matched with the activities performed. This validation study yielded threshold levels for sedentary, light and MVPA for the children (Table 1). Shortly thereafter, Pate et al. [17] completed an accelerometer validation study with 29 preschool-aged children. The researchers asked the children to wear Actigraph^® accelerometers and complete some structured and unstructured activities; in this study, however, the children’s energy expenditure was concurrently assessed by a metabolic criterion measure, Vo2 or oxygen consumption, an indirect calorimetry assessment regarded as a more valid assessment of energy expenditure than direct observation (e.g., the CARS). The findings of this study resulted in very different PA cut points than those identified by Sirard et al. (Table 1) [22] but many experts agree that this method of criterion validity testing was more rigorous than that of Sirard et al. [22].

Some data reduction software products use the Pate, et al. [17] reference to designate PA thresholds for different activity intensity levels, including light activity, even though Pate and colleagues did not publish a validated threshold distinguishing light and sedentary activity using the metabolic criterion method mentioned previously. The light threshold estimation from the Pate 2006 research was based on visual inspection of the data rather than indirect calorimetry [29]. A later study published by a group of researchers [28] identified a different threshold that distinguished sedentary from light activity differently, but they did not indicate the validation process used to identify this new threshold value for the Actigraph accelerometer^®. Thus, researchers have derived activity-level thresholds that distinguish different activity levels that have been developed from various validation methods. These discrepancies have resulted in differences in reported amounts of cumulative daily time children spend in light and sedentary PA levels. The application of different cut points may significantly affect the study PA outcomes of research investigations.

Objective

The purpose of this report is to bring attention to the critical phase of data reduction (i.e., applying PA cut points when summing the number of minutes spent in different intensities of PA) when using accelerometers, with a focus on the Actigraph^® accelerometer, and the effect of data reduction on reports of sample characteristics and PA outcome variables of interest, specifically time spent in sedentary and light PA. This secondary data analysis of accelerometer data has been conducted to understand the differential effect of the data reduction process on the outcome variable: minutes of sedentary and light PA of overweight/obese children, 4-8 years of age, an age group for whom the Actigraph^® accelerometer is validated for use.

Analysis

Data were analyzed using SPSS (Version 20; [30]. Significance was set at p = .05. Descriptive characteristics of the parents and children are presented as means and standard deviations. Paired samples t-tests were conducted to compare the means of PA levels of two cut points within subjects [17,28].

Methods

Methods of the Primary Study

The primary study from which baseline data for this study was extracted was an obesity treatment intervention Randomized Controlled Trial (RCT). We recruited parents of 60 overweight or obese (> 85^th BMI percentile) children, 4-8 years, from 14 primary care provider offices in the Southwestern United States [31]. Following IRB approval and completion of the formal consenting procedure, we undertook several baseline assessments and then randomly assigned the parent-child dyads to treatment or control conditions. For this secondary data analysis, we extracted raw baseline Actigraph^® accelerometry data collected in 15-second epochs over two days from the original dataset.

Methods of this Secondary Data Analysis

To examine the effect of applying different cut points to the raw, extracted accelerometry data, we used the same process to reduce the data using a commercially available software program, [32] but we sequentially applied two different sets of cut points to the data, those identified by Pate, et al. [17] and those of Pfeiffer et al. [28]. We were most interested in appreciating the different cut points found in the literature that distinguish light from sedentary levels of PA intensity.

To complete the data reduction, we used the raw accelerometer data from participant children (N = 53; demographic data were missing from 2 dyads) who completed up to 2 days of activity with at least 4 hours of recorded time each day but not more than 18 hours of recorded active time each day, which we defined as a “Valid” day of data collection. To assure measurement of actual active time identified, we assessed “Nonwear Time” as no activity counts recorded on the accelerometer for 60 or more continuous minutes [7], and we removed these data. We defined “Spurious Data” (i.e., abnormally high activity counts and perhaps erroneous activity) as activity that exceeded 10,000 counts per 15-second epoch or more than 9 standard deviations from the median [33]. We also removed these data from the final dataset.

The first set of cut points we applied to the data was published in a 2006 [17] Actigraph^® accelerometer validation study, in which indirect calorimetry (Vo2), a gold standard, was used as the criterion measure that distinguished light PA from MVPA using the threshold of 420 activity counts/15-second epoch. This group of researchers specified the threshold level of 200 activity counts/15-second epoch that separated sedentary and light PA intensities based upon visual inspection of the data, a less rigorous method of validation. Using threshold values determined by Pfeiffer, et al. [28], a research team that also using the Actigraph^® accelerometer, we then applied the cut point of 420 activity counts/15-second epoch as the threshold that distinguishes light from MVPA. Pfeiffer’s research team identified the threshold of 37.5 activity counts/15-second epoch to differentiate sedentary from light PA; although the validation method for that threshold level is unclear. This threshold was subsequently validated through the work of Fischer and Yıldırım [16] and Trost et al. [15], both using direct observation as the method of validation. After we applied the two sets of thresholds, we summed the time spent in each activity level.

Sample

Sixty participants were enrolled in the primary study [31]; of those, 7 did not have baseline accelerometry data, and of the remaining 53 participants 2 did not have fully completed demographic information. We extracted and used all participant data when baseline accelerometry data were present. Demographic information is depicted in (Table 2).

Results

To examine the difference in PA levels between the 2006 and 2009 cut points, we computed paired samples t tests to compare the average minutes of sedentary, light, and MVPA in preschool-aged children across different cut points (Table 3). The difference in average minutes of sedentary activity (M_diff = 257.14 minutes or 4 hours and 17.12 minutes) was significant, with lower levels of sedentary activity found when using the 2009 cut point than when using the 2006 cut point (t = 16.71, p = .00); the average minutes of light activity was significantly higher when using the 2009 cut point than when using the 2006 cut point (t = -16.71, p = .00). The average percentage time spent in sedentary activity was significantly lower using the 2009 cut point compared to the 2006 cut point (t = 44.25, p = .00). The average minutes of MVPA was unchanged across the 2006 and 2009 cut points because the threshold values were the same in both studies (Figure 1).

Discussion

We identified the Pate research team’s [17] validation study that distinguished different levels of physical activity intensity using Actigraph^® accelerometers with young children to be a high-level criterion validation study due to the use of indirect calorimetry (Vo2). However, upon closer review we identified the threshold demarcating light PA from sedentary time (200 activity counts/15 second epoch) by the Pate team [17] was determined by visual inspection of the data. To identify the best threshold that distinguishes light PA from sedentary time we continued to review validation studies that included Actigraph accelerometers^® and young children. Pfeiffer’s research team [28] identified that threshold as 37.5 activity counts/15 second epoch although it was unclear how this threshold was determined this was validated by the work of Fischer and Yıldırım [16] and Trost, et al. [15], both using direct observation as the method of validation.

Our research team had specific interest in the sedentary and light physical activity of the overweight/obese child participants, especially in light of the limited effects of previously conducted intensive PA interventions on children [34-38]. Additionally, light activity has recently been reported as a distinct category for intensity of activity [22,28,39], and its positive contribution to health has been recognized in the research literature over the last 25 years [40-45]. We believe that 37.5 activity counts/15 second epoch is the most valid threshold to apply to Actigraph^® accelerometry data gained from young children given our investigation to date. However, we acknowledge that differences in child age may confound our findings similar to the findings of other researchers [46-48].

The change in the percentage of reported sedentary time that resulted with the different cut points, a change from 74% to 53%, was also important. Several researchers have commented about the significant amount of time that preschool children spend in sedentary activities: estimates range from 53% to 86% of total activity time [31,49-51]; however, if those research teams reevaluated their data and applied the different threshold values, their reported findings would likely be significantly different. Moreover, if researchers apply different cut points during the data reduction phase, accelerometer findings will not be comparable across studies, and sample characteristics and outcomes may appear unique or unusual (e.g., samples are reported to be more active than those in other studies), suggesting that data from that study may have limited generalizability, when in fact the external validity of the study has not been compromised.

Identifying and applying cut points during the data reduction process are critical steps that have been facilitated by new software products, allowing researchers to select cut points by author or manuscript title or by specifying select thresholds for each intensity level. The predetermined threshold values offered may not be based on the most current research (e.g., validation studies using the best methods). The thresholds applied during the data reduction process should be taken from the most rigorously conducted validation studies (e.g., criterion validation studies). Furthermore, researchers should carefully review and appraise the original validation studies to be certain that each threshold value was validated similarly. It is critical, therefore, that researchers appreciate the importance of the application of thresholds in the data reduction step, continue to stay up-to-date with current research (i.e., validation studies) and remain closely involved with all data decisions.

Conclusions

Accelerometer use requires data reduction, during which the minutes of different PA intensities are accumulated and represented as aggregate daily group means. The critical decisions that researchers make with this increasingly used measure affect the data reduction process and likely have direct effects on reported sample characteristics and study findings; therefore, researchers’ decisions need to be intentional, justified, and clearly presented when the study findings are disseminated [33]. Individuals evaluating research that has used accelerometer measurement should be knowledgeable about the data reduction process and application of cut points. This knowledge will enhance the ability to critically analyze and synthesize results across studies and develop practice-related decisions.

Acknowledgements

We are grateful for the funding provided by the National Institute of Diabetes, Digestive, and Kidney Diseases (DK074428-01A1) and the Aetna Foundation for Healthy Communities. We are deeply appreciative to all of the pediatric providers and staff who recruited participants and provided office space for the multiple parent visits.

Declaration of Conflicting Interests

The authors declare no potential conflicts of interest with respect to authorship and/or publication of this manuscript.

Figure 1: Shift in Daily Average Minutes of Physical Activity with the Application of Different Cut Points.

1.       Sallis JF (1991) Self-report measures of children's physical activity. The Journal of School Health 61: 215-219.

2.       Puhl J, Greaves K, Hoyt M, Baranowski T (1990) Children's activity rating scale (CARS): Description and calibration. Research Quarterly for Exercise and Sport 61: 26-36.

3.       Sallis JF, Buono MJ, Roby JJ, Micale FG, Nelson JA (1993) Seven-day recall and other physical activity self-reports in children and adolescents.Medicine and Science in Sports and Exercise 25: 99-108.

4.       Trost SG, Ward DS, McGraw B, Pate RR (1999) Validity of the previous day physcial activity recall (PDPAR) in fifth-grade children. Pediatric Exercise Science 11: 341-348.

5.       Small L, Bonds-McClain D, Gannon A (2013) Physical activity of young overweight and obese children: Parent-reports of child activity level compared with objective measures. Western Journal of Nursing Research 35: 638-654.

6.       Bellows L, Silvernail S, Caldwell L, Bryant A, Kennedy C, et al. (2011) Parental perception on the efficacy of a physical activity program for preschoolers. Journal of Community Health 36: 231-237.

7.       Cain KL, Sallis JF, Conway TL, Van Dyck D, Calhoon L (2013) Using accelerometers in youth physical activity studies: A review of methods. Journal of Physical Activity & Health 10: 437-450.

8.       Pate RR, O'Neill JR, Mitchell J (2010) Measurement of physical activity in preschool children. Medicine and Science in Sports and Exercise 42: 508-512.

9.       Sirard JR, Pate RR (2001) Physical activity assessment in children and adolescents. Sports Medicine 31: 439-454.

10.    Jette M, Sidney K, Blumchen G (1990) Metabolic equivalents (METS) in exercise testing, exercise prescription, and evaluation of functional capacity. Clinical Cardiology 13: 555-565.

11.    Roemmich JN, Clark PA, Walter K, Patrie J, Weltman A, et al. (2000) Pubertal alterations in growth and body composition. V. energy expenditure, adiposity, and fat distribution. American Journal of Physiology.Endocrinology and Metabolism 279: E1426-36.

12.    Bornstein DB, Beets MW, Byun W, McIver K (2011) Accelerometer-derived physical activity levels of preschoolers: A meta-analysis. Journal of Science and Medicine in Sport / Sports Medicine Australia 14: 504-511.

13.    Cliff DP, Reilly JJ, Okely AD (2009) Methodological considerations in using accelerometers to assess habitual physical activity in children aged 0-5 years.Journal of Science and Medicine in Sport / Sports Medicine Australia 12: 557-567.

14.    Sirard JR, Trost SG, Pfeiffer KA, Dowda M, Pate RR (2005b) Calibration and evaluation of an objective measure of physical activity in preschool children. Journal of Physical Activity & Health 3: 345-357.

15.    Trost SG, Fees BS, Haar SJ, Murray AD, Crowe LK (2012) Identification and validity of accelerometer cut-points for toddlers. Obesity 20: 2317-2319.

16.    Fischer C, Yildirim M, Salmon J, Chinapaw MJM (2012) Comparing different accelerometer cut-points for sedentary time in children. Pediatric Exercise Science 24: 220-228.

17.    Pate RR, Almeida MJ, McIver KL, Pfeiffer KA, Dowda M (2006) Validation and calibration of an accelerometer in preschool children. Obesity 14: 2000-2006.

18.    Pfeiffer KA, Dowda M, McIver KL, Pate RR (2009a) Factors related to objectively measured physical activity in preschool children. Pediatric Exercise Science 21: 196-208.

19.    Puyau MR, Adolph AL, Vohra FA, Butte NF (2002a) Validation and calibration of physical activity monitors in children. Obesity Research 10: 150-157.

20.    Reilly JJ, Coyle J, Kelly L, Burke G, Grant S, et al. (2003) An objective method for measurement of sedentary behavior in 3- to 4-year olds. Obesity Research 11: 1155-1158.

21.    Robertson W, Stewart-Brown S, Wilcock E, Oldfield M, Thorogood M (2011) Utility of accelerometers to measure physical activity in children attending an obesity treatment intervention. Journal of Obesity 2011: 398919.

22.    Sirard JR, Trost SG, Pfeiffer KA, Dowda M, Pate RR (2005a) Calibration and evaluation of an objective measure of physical activity in preschool children. Journal of Physical Activity & Health 3: 345-357.

23.    Stone MR, Rowlands AV, Eston RG (2009) Relationships between accelerometer-assessed physical activity and health in children: Impact of the activity-intensity classification method. Journal of Sports Science & Medicine 8: 136-143.

24.    Treuth MS, Schmitz K, Catellier DJ, McMurray RG, Murray DM, et al. (2004) Defining accelerometer thresholds for activity intensities in adolescent girls. Medicine and Science in Sports and Exercise 36: 1259-1266.

25.    van Cauwenberghe E, Labarque V, Trost SG, de Bourdeaudhuij I, Cardon G (2011) Calibration and comparison of accelerometer cut points in preschool children. International Journal of Pediatric Obesity : IJPO : An Official Journal of the International Association for the Study of Obesity 6: e582-589.

26.    Alhassan S, Robinson TN (2010) Defining accelerometer thresholds for physical activity in girls using ROC analysis. Journal of Physical Activity & Health 7: 45-53.

27.    Kim Y, Beets MW, Welk GJ (2012) Everything you wanted to know about selecting the "right" actigraph accelerometer cut-points for youth, but...: A systematic review. Journal of Science and Medicine in Sport / Sports Medicine Australia 15: 311-321.

28.    Pfeiffer KA, Dowda M, McIver KL, Pate RR (2009b) Factors related to objectively measured physical activity in preschool children. Pediatric Exercise Science 21: 196-208.

29.    ActiGraph Support (2012) What's the difference among the cut points available in ActiLife?

30.    IBM Corp (2011) IBM SPSS statistics for WIndows (Version 20.0 Edition). Armonk NY: IBM Corp.

31.    Small L, Bonds-McClain D, Vaughan L, Melnyk BM, Gannon A (2013) The preliminary effects of a primary care based randomized treatment trial with overweight and obese young children and their parents. Journal of Pediatric Health Care 28:198-207.

32.    ActiGraph (2013) ActiLife (Version 6.7.0 Edition). Pensacola, FL: ActiGraph.

33.    Masse LC, Fuemmeler BF, Anderson CB, Matthews CE, Trost SG, et al. (2005) Accelerometer data reduction: A comparison of four reduction algorithms on select outcome variables. Medicine and Science in Sports and Exercise 37: S544-S554.

34.    Brown H, Hume C, Pearson N, Salmon J (2013) A systematic review of intervention effects on potential mediators of children's physical activity. BMC Public Health 13: 165.

35.    Dobbins M, De Corby K, Robeson P, Husson H, Tirilis D (2009) School-based physical activity programs for promoting physical activity and fitness in children and adolescents aged 6-18. The Cochrane Database of Systematic Reviews 1: CD007651.

36.    Mehtala MA, Saakslahti AK, Inkinen ME, Poskiparta ME (2014) A socio-ecological approach to physical activity interventions in childcare: A systematic review. The International Journal of Behavioral Nutrition and Physical Activity 11: 22.

37.    Reilly JJ, Kelly L, Montgomery C, Williamson A, Fisher A, et al. (2006) Physical activity to prevent obesity in young children: Cluster randomised controlled trial. BMJ 333: 1041.

38.    van Sluijs EM, Kriemler S, McMinn AM (2011) The effect of community and family interventions on young people's physical activity levels: A review of reviews and updated systematic review. British Journal of Sports Medicine 45: 914-922.

39.    Puyau MR, Adolph AL, Vohra FA, Butte NF (2002b) Validation and calibration of physical activity monitors in children. Obesity Research 10: 150-157.

40.    Healy GN, Wijndaele K, Dunstan DW, Shaw JE, Salmon J, et al. (2008) Objectively measured sedentary time, physical activity, and metabolic risk: The australian diabetes, obesity and lifestyle study (AusDiab). Diabetes Care 31: 369-371.

41.    Leon AS, Connett J, Jacobs DR Jr, Rauramaa R (1987) Leisure-time physical activity levels and risk of coronary heart disease and death. the multiple risk factor intervention trial. JAMA : The Journal of the American Medical Association 258: 2388-2395.

42.    Powell KE, Paluch AE, Blair SN (2011) Physical activity for health: What kind? how much? how intense? on top of what? Annual Review of Public Health 32: 349-365.

43.    Saunders TJ, Tremblay MS, Mathieu ME, Henderson M, O'Loughlin J, et al. (2013) Associations of sedentary behavior, sedentary bouts and breaks in sedentary time with cardiometabolic risk in children with a family history of obesity. PloS One 8: e79143.

44.    van der Niet AG, Smith J, Scherder EJ, Oosterlaan J, Hartman E, et al. (2014) Associations between daily physical activity and executive functioning in primary school-aged children. Journal of Science and Medicine in Sport / Sports Medicine Australia 18: 673-677.

45.    Woodcock J, Franco OH, Orsini N, Roberts I (2011) Non-vigorous physical activity and all-cause mortality: Systematic review and meta-analysis of cohort studies. International Journal of Epidemiology 40: 121-138.

46.    Cliff DP, Okely AD (2007) Comparison of two sets of accelerometer cut-off points for calculating moderate-to-vigorous physical activity in young children.Journal of Physical Activity & Health 4: 509-513.

47.    Janssen X, Cliff DP, Reilly JJ, Hinkley T, Jones RA, et al. (2013) Predictive validity and classification accuracy of ActiGraph energy expenditure equations and cut-points in young children. PloS One 8: e79124.

48.    Morrison JA, Barton BA, Biro FM, Daniels SR, Sprecher DL (1999) Overweight, fat patterning, and cardiovascular disease risk factors in black and white boys. J Pediatr 135: 451-457.

49.    Butte NF, Puyau MR, Adolph AL, Vohra FA, Zakeri I (2007) Physical activity in nonoverweight and overweight hispanic children and adolescents.Medicine and Science in Sports and Exercise 39: 1257-1266.

50.    Spittaels H, Van Cauwenberghe E, Verbestel V, De Meester F, Van Dyck D, et al. (2012) Objectively measured sedentary time and physical activity time across the lifespan: A cross-sectional study in four age groups. The International Journal of Behavioral Nutrition and Physical Activity 9: 149.

51.    Trost SG, Kerr LM, Ward DS, Pate RR (2001) Physical activity and determinants of physical activity in obese and non-obese children. International Journal of Obesity and Related Metabolic Disorders : Journal of the International Association for the Study of Obesity 25: 822-829.