Objective measures of physical activity, specifically uniaxial hip-worn accelerometers, were introduced into national surveys in the United States (National Health and Nutrition Examination Survey [NHANES]) in 2003 (29), Canada (Canada Health Measures Survey) in 2007 (7,8), and the UK (Health Survey for England) in 2008 (17). Also in 2008, the International Children’s Accelerometry Database (ICAD) was initiated: a compilation of accelerometer-derived estimates of children’s physical activity from a wide range of studies, settings, and countries (28). The accelerometers used in these surveys and studies converted accelerations into proprietary counts stored in 5- to 60-s epochs and time accumulated in moderate-to-vigorous physical activity (MVPA) was subsequently estimated.
Over the past decade, there have been rapid developments in accelerometry resulting in the commercial availability of triaxial microelectromechanical (MEMS) accelerometers that continuously sample and store raw accelerations at up to 100 Hz, such as the ActiGraph GT3X+ and the GENEActiv. There has also been a move to 24-h wear protocols with wrist wear to maximize compliance (2,9,14) and facilitate measurement of the full spectrum of physical behaviors (physical activity, sedentary behavior, and sleep) (6,25). As a result, since 2011, wrist-worn ActiGraph GT3X+ monitors that collect and store raw accelerations at 100 Hz have been used in NHANES (30). Other large-scale adult (2,9,21) and children’s (9,10,20,34) studies are also using 24 h wrist-worn accelerometer protocols using the GENEActiv.
As the ActiGraph GT3X+ and the GENEActiv store raw accelerations rather than proprietary counts, their data should, theoretically, be comparable. Outputs from the GENEActiv and the ActiGraph GT3X+, when processed and calibrated identically using the open source package GGIR (32,33) in R [http://cran.r-project.org], have high agreement for acceleration magnitudes >50–80 mg, indicative of light activity and MVPA, although not for lower acceleration magnitudes indicative of sedentary time (27).
Advances in measurement methods (e.g., self-report to objective measurement) and/or measurement technologies (e.g., proprietary count uniaxial accelerometers to raw acceleration triaxial accelerometers) bring reduced bias, improved precision, and enhanced measurement opportunities (30), but at a cost of limited comparability to past data. There is a wealth of MVPA data on children estimated from uniaxial hip-worn ActiGraphs (28,29), and it is desirable to use these data to contextualize future estimates of MVPA, to map trends in physical activity, to compare effectiveness of past and present interventions, and to understand the clinical significance of intervention changes in PA, by contextualizing current data with the extant historical evidence on the impact of physical activity on health. To complicate comparisons further, hip-worn ActiGraph data have been analyzed using an extensive range of cut points leading to widely varying estimates of MVPA even for the same data set (4,5,15).
The purpose of this study is to provide a means for quickly and simply comparing raw acceleration data from wrist-worn accelerometers at a group level to past estimates of children’s MVPA from a range of cut points applied to hip-worn ActiGraph data. To do this, we used data from three studies that have concurrent 7-d accelerometer wear at the wrist (GENEActiv) and hip (ActiGraph) to determine and cross validate the acceleration magnitudes most closely associated with established MVPA cut points. As the GENEActiv and ActiGraph GT3X+ have high agreement for accelerations indicative of light activity and MVPA (27), the results will be applicable to studies measuring raw triaxial accelerations at the wrist in children with either the ActiGraph GT3X+ or the GENEActiv.
This is a secondary data analysis using data from three studies: 1) 58 children, age 10–12 yr, recruited from primary schools in South Australia (26); 2) 129 children, age 9–10 yr, recruited from primary schools in Liverpool, UK (12); and 3) 81 children, age 9–11 yr, recruited from one primary school in Liverpool, UK. The appropriate university research ethics committee approved each study. Written informed consent and assent were obtained from the parents/guardians and children, respectively. Height was measured to the nearest 0.1 cm and body mass to the nearest 0.1 kg.
Assessment of activity
Free-living physical activity was measured by concurrent wear of the GENEActiv on the nondominant wrist and the ActiGraph GT3X+ positioned above the right hip, on an elasticated belt worn around the waist, for seven consecutive days. In study 1, children were requested to wear both monitors day and night, removing the hip-worn ActiGraph for water-based activities only. In studies 2 and 3, children were requested to wear both monitors at all times except when sleeping or during water-based activities.
The GENEActiv is a triaxial accelerometry-based activity monitor with a dynamic range of ±8g (Gravity Estimator of Normal Everyday Activity; ActivInsights Ltd., Cambridgeshire, UK). The ActiGraph GT3X+ is a triaxial accelerometry-based activity monitor with a dynamic range of ±6g (ActiGraph LLC, Pensacola, FL). Study 1: the GENEActivs were initialized to collect data at 87.5 Hz and data uploaded using GENEActiv PC software version 2.2. The ActiGraphs were initialized to collect data at 80 Hz, and data were uploaded using Actilife version 6.5.3. Data were collected between April and December 2012. Studies 2 and 3: The GENEActivs and ActiGraphs, were both initialized to collect data at 100 Hz, and data were uploaded using GENEActiv PC software version 2.2 and Actilife version 6.11.4, respectively. Study 2 data were collected between January and May 2014, and study 3 data were collected in January and February 2015.
Wrist-worn GENEActiv (raw acceleration) GENEActiv.bin files were analyzed with R-package GGIR version 1.2-0 (http://cran.r-project.org) (32,33). Signal processing in GGIR includes the following steps: 1) autocalibration using local gravity as a reference (32), 2) detection of sustained abnormally high values, 3) detection of nonwear, and 4) calculation of the average magnitude of dynamic acceleration, i.e., the vector magnitude of acceleration corrected for gravity (Euclidean Norm minus 1g, ENMO) over user-defined epochs:
with negative values set to zero. In study 1, ENMO was averaged over 5-s epochs; in studies 2 and 3, ENMO was averaged over 1-s epochs. As studies applying GGIR to wrist accelerometer data have used both 1-s (12) and 5-s epochs (9), inclusion of both epochs increases the generalizability of the findings.
Files were excluded from all analyses if postcalibration error was greater than 0.02g (9), and individual days were classified as invalid and excluded if wear time was insufficient (16 h for the 24-h protocol in study 1, 10 h for the waking wear protocol in studies 2 and 3). The detection of nonwear has been described in detail previously (see “Procedure for non-wear detection” in supplementary document to van Hees et al. ). In brief, nonwear is estimated based on the standard deviation and value range of each axis, calculated for 60 min windows with 15-min moving increments. If for at least two of the three axes the SD is less than 13 mg or the value range is less than 50 mg, the time window is classified as nonwear. The default nonwear setting was used, i.e., invalid data were imputed by the average at similar time points on different days of the week.
The distribution of time spent across ENMO levels in 50-mg resolution (0–50 mg, 50–100 mg, …, ≥400 mg) was calculated using the argument “ilevels” from the GGIR package. The time spent above thresholds of 100, 150, 200, 250, 300, 350, and 400 mg was calculated for comparison to widely used hip-worn ActiGraph MVPA cut points.
Hip-worn ActiGraph (counts)
Data were analyzed using Actilife version 6.13.0. The raw.gt3x files were summarized into uniaxial (vertical) proprietary counts in 1-, 5-, 15-, and 60-s epochs, resulting in four ActiGraph files for analysis per participant. Nonwear was defined as 60 min of consecutive zero counts, with an allowance for 1–2 min of counts between 0 and 100 (29). Individual days were classified as invalid and excluded if wear time was insufficient (16 h for the 24-h protocol in study 1, 10 h for the waking wear protocol in studies 2 and 3).
Each file was analyzed with four widely used MVPA cut points: very low (1100 counts per minute, approximately equivalent to the cut point for an 11-yr-old (3 METs) using the age-specific criteria of the Freedson group, published by Trost et al. ), low (1680 counts per minute, Pate et al. ), medium (2296 counts per minute, Evenson et al. ), and high (3200 counts per minute, Puyau et al. ). This resulted in 16 outputs per participant: MVPA classified using very low, low, medium, and high cut points, with each cut point applied to data integrated into 1-, 5-, 15-, and 60-s epochs.
For each participant, days were only included if classified as valid for both the wrist-worn GENEActiv and hip-worn ActiGraph; therefore, to be included, a participant needed a minimum of 1 d where both the ActiGraph and the GENEActiv recorded sufficient wear time. The daily means for all output variables were taken for each participant. For data from study 1, GENEActiv 5-s epoch outputs were compared with the ActiGraph 5-, 15-, and 60-s epoch outputs. For data from studies 2 and 3, the GENEActiv 1-s epoch files were compared with the ActiGraph 1-, 15-, and 60-s epochs. The 5-s data from study 1 and the 1-s data from studies 2 and 3 were designated a ≤5-s epoch.
Descriptive statistics (mean ± SD) were calculated for all variables. Data from studies 1 and 2 (approximately 70% of the total sample) were analyzed with data from study 3 reserved for cross validation. The wrist-worn GENEActiv ENMO thresholds (100+, 150 +, 200+, 250+, 300+, 350+, and 400+ mg), which most closely approximated time accumulated in each of the hip-worn ActiGraph MVPA cut points (very low, low, medium, and high) for each epoch length (≤5 s, 15 s, 60 s), were examined with a series of limits of agreement (LoA) analyses (3) and intraclass correlation (ICC; single measures, absolute agreement) with 95% confidence interval.
For each hip-worn ActiGraph MVPA cut point/epoch combination, the wrist-worn ENMO threshold with the closest agreement was selected and the agreement between these optimal pairings tested in the independent cross-validation sample. The distributions for each of the optimal pairings were illustrated on kernel density plots (bandwidth = 10) for the total sample (data from studies 1, 2, and 3 combined).
Demographic data, by study, are presented in Table 1. The final sample size was 238 (test sample n = 159, cross-validation sample n = 79) with 30 participants excluded because of no days of concurrent valid wear for both monitors. Figure 1 shows the time recorded in each of the intensity categories by the hip-worn ActiGraph (very low, low, medium, and high MVPA cut points) and the wrist-worn GENEActiv (100+, 150+, 200+, 250+, 300+, 350+, and 400+ mg ENMO thresholds) by epoch (ActiGraph ≤5 s, 15 s, 60 s; GENEActiv ≤5 s) for the total sample.
The agreement between each wrist-worn GENEActiv ENMO threshold and each hip-worn ActiGraph MVPA cut point is shown for each epoch length in Table 2. The ENMO threshold with the highest agreement for each ActiGraph MVPA cut point/epoch combination is highlighted in bold in Table 2. The optimal wrist-worn ENMO thresholds for comparison to hip-worn ActiGraph MVPA cut points were as follows:
- Very low MVPA ActiGraph cut points (1100 counts per minute, Trost et al. )
- ○ ENMO 150+ mg, irrespective of the ActiGraph epoch (ICC ≥ 0.65, mean bias [ENMO – ActiGraph] = −2.9 to −18.0 min [−2.7% to −14.9% of mean MVPA])
- Low MVPA ActiGraph cut points (1680 counts per minute, Pate et al. )
- ○ ENMO 200+ mg, irrespective of the ActiGraph epoch (ICC ≥ 0.67, mean bias = −4.1 to −10.7 min [−5.4% to −13.0% of mean MVPA])
- Medium MVPA cut points (2296 counts per minute, Evenson et al. )
- ○ ENMO 250+ mg for ≤5- and 15-s epochs (ICC ≥ 0.69, mean bias = −3.0 to −7.3 min [−5.4% to −12.0% of mean MVPA])
- ○ ENMO 300+ mg for 60-s epochs (ICC = 0.73, mean bias = −5.0 min [−10.6% of mean MVPA])
- High MVPA cut points (3200 counts per minute, Puyau et al. )
- ○ ENMO 300+ mg for ≤5-s epochs (ICC = 0.73, mean bias = +1.8 min [+4.7% of mean MVPA])
- ○ ENMO 350+ mg for 15-s epochs (ICC = 0.73, mean bias = +2.7 min [+8.7% of mean MVPA])
- ○ ENMO 400+ mg for 60-s epochs (ICC = 0.65, mean bias = +6.5 min [+28.6% of mean MVPA]).
The agreement of each of these optimal pairings of wrist-worn ENMO threshold and hip-worn ActiGraph MVPA cut point was tested in the cross-validation sample (Table 3, Fig. 2). Agreement was robust with ICC’s similar to the test sample for 15-s epochs (very low, low, and medium MVPA cut points, mean bias = ∣4.9∣% ± 0.9% of mean MVPA) and ≈ 0.01–0.11 lower than the test sample (0.61 to 0.71, mean bias = ∣8.9∣% ± 4.8% of mean MVPA) for other MVPA cut point/epoch combinations, except for the high MVPA cut point/60-s epoch where the ICC was considerably reduced (0.42). The mean biases and 95% LoA were also similar in magnitude to the test sample. However, the values of the mean bias for specific pairings were not consistent between the test sample and the cross-validation sample.
The distribution of the ActiGraph and ENMO data for each of the optimal pairings is shown on kernel density plots for the total sample (Fig. 3). The columns represent cut points (left to right: very low, low, medium, high), and the rows represent ActiGraph epochs (top to bottom: ≤5 s, 15 s, 60 s). The agreement statistics for the total sample are shown in Supplemental Digital Content 1 (see Table, Supplemental Digital Content 1, Agreement between each hip-worn ActiGraph cut point and each wrist-worn GENEActiv ENMO threshold by epoch length in the total sample, http://links.lww.com/MSS/A716).
Rapid progress in accelerometer technology has led to changes in the data collected and study protocols followed, with a shift from uniaxial proprietary count outcomes collected using accelerometers worn at the hip to triaxial raw accelerations measured using wrist-worn accelerometers (30). We have developed a quick and simple method to facilitate the comparison of group level estimates of children’s MVPA from uniaxial hip-worn count-based ActiGraphs to triaxial raw acceleration data measured at the wrist processed using the open source R-package, GGIR (32,33). The method was developed using the GENEActiv wrist-worn accelerometer, but evidence suggests it will also be applicable to raw acceleration measured at the wrist using the ActiGraph and processed in GGIR (27).
Mean biases for optimal pairings of ENMO thresholds and ActiGraph MVPA cut points were relatively low (test sample: mean bias = ∣9.4∣% ± 4.2% of mean MVPA; cross-validation sample: mean bias = ∣7.8∣% ± 4.9% of mean MVPA), indicating good group level agreement, excluding high ActiGraph MVPA cut points assessed using a 60-s epoch where mean bias was high relative to the low means (29% in the test sample, 60% in the cross-validation sample). Similarly, the ICC values for optimal pairings were all between 0.61 and 0.76 in the test and cross-validation sample, indicating good agreement (13), except the high ActiGraph MVPA cut points assessed using a 60-s epoch in the cross-validation sample (ICC = 0.42). The 95% LoA were moderate to large, indicating that individual-level comparisons are not advised. The MVPA recorded in the cross-validation sample was lower than the test samples, in particular when applying high cut points with a 60-s epoch (Figs. 1 and 2); this may have contributed to the lower robustness for the high cut point/60-s epoch combination. Hildebrand et al. (16) developed an MVPA threshold of approximately 200 mg for use with wrist-worn ActiGraph and GENEActiv accelerometers. On the basis of the current findings, MVPA determined by applying the 200 mg threshold to wrist-worn accelerometer data should compare best to MVPA determined from low cut points (23) applied to hip-worn ActiGraph data, irrespective of epoch. Overall, the cross validation suggests that agreement may be closest when comparing ENMO 150+, 200+, and 250+ thresholds to MVPA estimated from ActiGraph 15-s epoch data processed using very low, low, and medium cut points, respectively.
The potential for application of these comparisons is extensive. By 2010, more than 46,000 physical activity data sets from hip-worn ActiGraphs had been collated in the ICAD, approximately 19,000 from children age 9–12 yr (28). More recently, the International Study of Childhood Obesity, Lifestyle, and the Environment collected data on 6000 children, age 9–11 yr from 12 countries across five diverse regions of the world using hip-worn ActiGraphs (18). The latter study collected triaxial raw acceleration data using ActiGraph GT3X+ and has developed novel analytical tools for application to the raw acceleration data, e.g., to determine sleep duration (1), but as the hip was the measurement site, these data have also been summarized in proprietary counts and analyzed using count cut points (19). Since NHANES moved to assessing physical activity using triaxial raw acceleration data measured at the wrist for the NHANES cycles 2011–2012 and 2013–2014 (30), many other large studies have also used wrist-worn accelerometers. For example, data have already been collected in ≈4000 children, age 9–11 yr, in the Child Health Checkpoint (Melbourne, Australia ); ≈1800 girls, age 11–14 yr, in Girls Active (Leicester, UK ); ≈1000 children, age 8–11 yr, in the Cork Children’s Lifestyle Study (Ireland ); and ≈ 4000 children, age 7 yr, in the Pelotas Birth cohort (Brazil ). The comparisons presented will facilitate interpretation of these data in relation to past estimates of children’s MVPA, e.g., from NHANES, ICAD, and International Study of Childhood Obesity, Lifestyle, and the Environment.
The data collated for this study came from three different sources and were collected using two differing protocols. Study 1 took place in South Australia, used a 24-h wear protocol, and summarized the GENEActiv ENMO data in 5-s epochs. Studies 2 and 3 took place in the UK, used a waking time only protocol, and summarized the ENMO data in 1-s epochs. While the results were similar across studies and the cross validation (study 3 data) showed the agreement statistics were robust, these differences limit the internal validity of the study. However, the external validity is enhanced, as results are applicable to ENMO data collected in 1- and 5-s epochs using either a waking or 24 h protocol. Given the outcome of interest was MVPA, it is not surprising that the use of a waking or 24-h protocol did not impact on the results.
ActiGraph epochs of ≤5-, 15-, and 60 s were considered, whereas ENMO data were only summarized into ≤5-s epochs. The use of longer epochs in the past was due to the memory limitations of accelerometers (30). Accelerations were integrated on board the accelerometer and stored in epochs, normally 60-s epochs, to ensure 1 wk of data could be stored before downloading the data. Because of technological progress, onboard memory is no longer a problem and raw acceleration data collected at 100 Hz can be stored for 1 wk. Therefore, it is unlikely that epochs longer than the default 5-s epoch in GGIR will be used, particularly when assessing children’s activity where the typical sporadic activity patterns are best captured using short epochs (22). It should be noted that the participants in this study were from a relatively narrow age range, and the results cannot be generalized beyond the 9- to 12-yr age-group tested.
In summary, this study indicates that, in 9–12 yr old children, time accumulated above the appropriate incremental ENMO threshold has good agreement at a group level with a range of widely used very low to high ActiGraph MVPA cut points. It is important to note this is a simple pooled data comparison study that enables group-level comparisons, but individual-level comparisons are not advised. We recommend that when processing triaxial raw acceleration wrist accelerometer data using GGIR, the times accumulated above ENMO thresholds ranging from ≥100 to ≥400 mg, or in incremental acceleration bins (e.g., 9), are presented. In addition to providing an activity profile, this will enable the reader to quickly and simply compare the findings to past estimates of children’s MVPA from hip-worn ActiGraph data across a range of widely used cut points.
Study 1 was funded by the University of South Australia, and Studies 2 and 3 were funded by Liverpool John Moores University. AR is with the National Institute for Health Research (NIHR) Diet, Lifestyle and Physical Activity Biomedical Research Unit based at University Hospitals of Leicester and Loughborough University, the National Institute for Health Research Collaboration for Leadership in Applied Health Research and Care–East Midlands (NIHR CLAHRC–EM), and the Leicester Clinical Trials Unit. The views expressed are those of the authors and not necessarily those of the NHS, the NIHR, or the Department of Health. DPC is funded by an Australian Research Council (ARC) Discovery Early Career Researcher Award (DE140101588). The results of the present study do not constitute endorsement by the authors or the American College of Sports Medicine of the products described in this article.
The results of the study are presented clearly, honestly, and without fabrication, falsification, or inappropriate data manipulation. There are no conflicts of interest.
1. Barreira TV, Schuna JM Jr, Mire EF, et al. Identifying children’s nocturnal sleep using 24-h waist accelerometry. Med Sci Sports Exerc
2. Bell JA, Hamer M, van Hees VT, Singh-Manoux A, Kivimäki M, Sabia S. Healthy obesity and objective physical activity. Am J Clin Nutr
3. Bland JM, Altman GA. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet
4. Bornstein DB, Beets MW, Byun W, et al. Equating accelerometer estimates of moderate-to-vigorous physical activity: in search of the Rosetta Stone. J Sci Med Sport
5. Brazendale K, Beets MW, Bornstein DB, et al. Equating accelerometer estimates among youth: the Rosetta Stone 2. J Sci Med Sport
6. Buman MP, Hu F, Newman E, Smeaton AF, Epstein DR. Behavioral periodicity detection from 24 h wrist accelerometry and associations with cardiometabolic risk and health-related quality of life. Biomed Res Int
7. Colley RC, Garriguet D, Janssen I, Craig CL, Clarke J, Tremblay MS. Physical activity of Canadian adults: accelerometer results from the 2007 to 2009 Canadian Health Measures Survey. Health Rep
8. Colley RC, Garriguet D, Janssen I, Craig CL, Clarke J, Tremblay MS. Physical activity of Canadian children and youth: accelerometer results from the 2007 to 2009 Canadian Health Measures Survey. Health Rep
9. da Silva IC, van Hees VT, Ramires VV, et al. Physical activity levels in three Brazilian birth cohorts as assessed with raw triaxial wrist accelerometry. Int J Epidemiol
10. Edwardson CL, Harrington DM, Yates T, et al. A cluster randomised controlled trial to investigate the effectiveness and cost effectiveness of the ‘Girls Active’ intervention: a study protocol. BMC Public Health
11. Evenson KR, Catellier DJ, Gill K, Ondrak KS, McMurray RG. Calibration of two objective measures of physical activity for children. J Sports Sci
12. Fairclough SJ, Noonan R, Rowlands AV, van Hees V, Knowles Z, Boddy LM. Wear compliance and activity in children wearing wrist and hip mounted accelerometers. Med Sci Sports Exerc
13. Fleiss J. The Design and Analysis of Clinical Experiments
. New York: John Wiley & Sons; 1986. pp. 1–31.
14. Freedson PS, John D. Comment on “estimating activity and sedentary behavior from an accelerometer on the hip and wrist. Med Sci Sports Exerc
15. Guinhouya BC, Samouda H, de Beaufort C. Level of physical activity among children and adolescents in Europe: a review of physical activity assessed objectively by accelerometry. Public Health
16. Hildebrand M, Van Hees VT, Hansen BH, Ekelund U. Age group comparability of raw accelerometer output from wrist- and hip-worn monitors. Med Sci Sport Exerc
17. Joint Health Surveys Unit, National Centre for Social Research and University College London Research Department of Epidemiology and Public Health. The Health Survey for England 2008. Volume 1: Physical Activity and Fitness
. Leeds (UK): NHS Information Centre for Health and Social Care; 2009. Available from: http://www.hscic.gov.uk/catalogue/PUB00430/heal-surv-phys-acti-fitn-eng-2008-rep-v2.pdf
. Accessed March 31, 2016.
18. Katzmarzyk PT, Barreira TV, Broyles ST, et al. The International Study of Childhood Obesity, Lifestyle and the Environment (ISCOLE): design and methods. BMC Public Health
19. Katzmarzyk PT, Barreira TV, Broyles ST, et al. Relationship between lifestyle behaviors and obesity in children ages 9–11: results from a 12-country study. Obesity
20. Keane E, Kearney PM, Perry IJ, Browne GM, Harrington JM. Diet, Physical activity, lifestyle behaviors, and prevalence of childhood obesity in Irish children: the Cork Children’s Lifestyle Study Protocol. JMIR Res Protoc
21. Kearney PM, Harrington JM, McCarthy VJC, Fitzgerald AP, Perry I. Cohort profile: the Cork and Kerry Diabetes and Heart Disease Study. Int J Epidemiol
22. Nilsson A, Ekelund U, Yngve A, Sjostrom M. Assessing physical activity among children with accelerometers using different time sampling intervals and placements. Ped Exerc Sci
23. Pate RR, Almeida MJ, McIver KL, et al. Validation and calibration of an accelerometer in preschool children. Obesity
24. Puyau MR, Adolph AL, Vohra FA, Butte NF. Validation and calibration of physical activity monitors in children. Obes Res
25. Rowlands AV, Olds TS, Hillsdon M, et al. Assessing sedentary behavior with the GENEActiv: introducing the sedentary sphere. Med Sci Sports Exerc
26. Rowlands AV, Rennie K, Kozarski R, et al. Children’s physical activity assessed with wrist- and hip-worn accelerometers. Med Sci Sports Exerc
27. Rowlands AV, Yates T, Davies M, Khunti K, Edwardson CL. Raw accelerometer data analysis with GGIR R-package: does accelerometer brand matter? Med Sci Sport Exerc
28. Sherar LB, Griew P, Esliger DW, et al. International children’s accelerometry database (ICAD): design and methods. BMC Public Health
29. Troiano RP, Berrigan D, Dodd KW, Mâsse LC, Tilert T, McDowell M. Physical activity in the United States measured by accelerometer. Med Sci Sports Exerc
30. Troiano RP, McClain JJ, Brychtta RJ, Chen KY. Evolution of accelerometer methods for physical activity research. Br J Sports Med
31. Trost SG, Pate RR, Sallis JF, et al. Age and gender differences in objectively measured physical activity in youth. Med Sci Sports Exerc
32. van Hees VT, Fang Z, Langford J, et al. Autocalibration of accelerometer data for free-living physical activity assessment using local gravity and temperature: an evaluation on four continents. J Appl Physiol (1985)
33. van Hees VT, Gorzelniak L, Dean León EC, et al. Separating movement and gravity components in an acceleration signal and implications for the assessment of human daily physical activity. PLoS One
34. Wake M, Clifford S, York E, et al. Introducing growing up in Australia’s Child Health CheckPoint: a physical health and biomarkers module for the Longitudinal Study of Australian Children. Fam Matters