It is becoming increasingly clear that variations in physical activity and sedentary behaviour are of enormous importance to the current and future health of children and adolescents.1 2 Sedentary behaviour is not simply the absence of physical activity, but involves purposeful engagement in activities that involve minimal movement and low energy expenditure.3 4 Increased awareness of the importance of physical activity and sedentary behaviour, combined with technological advances that have improved our ability to measure these variables in free-living children, have led to increased interest in paediatric physical activity and sedentary behaviour.

While subjective methods for measurement of physical activity and sedentary behaviour will continue to provide useful evidence on the context of these behaviours, and on the subjective perception of these behaviours by study participants, objective methods are now being regarded as optimal for quantification of the amount and intensity of physical activity and amount of sedentary behaviour. Recent reviews have concluded that accelerometry (motion sensing) provides an objective, practical, accurate and reliable means of quantifying the amount (“volume”) and intensity of habitual physical activity and the amount of sedentary behaviour in children.5^–10 These reviews also highlighted research needs, but did not consider the advantages of using accelerometry for quantifying the amount and intensity of physical activity compared with traditional alternative methods (such as questionnaires), and these advantages are not widely appreciated beyond the research community using accelerometry at present. In addition, the recent reviews on accelerometry have highlighted uncertainty over several fundamental practical issues concerning the optimum approach to using accelerometers in children, but did not provide new evidence to address this uncertainty or make evidence-based recommendations that might resolve these practical problems. This review, based on a synthesis of evidence cited in recent reviews of accelerometry,5^–10 combined with new empirical studies and secondary analysis, therefore aims to:

1. Make the case for objective measurement of physical activity in children, using examples from studies published recently.

2. Provide new evidence to address the uncertainty over practical issues in accelerometry.

IMPORTANCE OF OBJECTIVE MEASURING METHODS

Avoidance of reporting bias

Objective methods are unlikely to produce biased measures of the amount of physical activity or sedentary behaviour, in contrast to subjective self or proxy reports, which are the traditional methods of measuring physical activity in children and adolescents.

Children (or their families) involved in lifestyle modification-intervention studies have a tendency to over-report their physical activity, with the result that interventions may favour the intervention group; objective measures of physical activity made on the same children have suggested that such differences can be spurious.11 12 Biases of this kind should be expected — children and families involved in dietary interventions tend to report their intake in a biased manner, under-reporting in obesity-prevention studies, for example, while over-reporting intake of energy in trials aimed at increasing energy intake.12 13 Assessment of interventions aimed at physical activity and sedentary behaviour change should therefore use objective methods,11 both to confirm that apparent changes in physical activity are real and to quantify the magnitude of any change with confidence.

The problem of biased self-reporting of amount and intensity of physical activity using subjective methods – such as questionnaires — almost certainly extends to observational studies and surveys of physical activity in children carried out for surveillance purposes. In the United Kingdom, national surveillance of paediatric physical activity in health surveys still involves subjective (parental) reporting of physical activity and is associated with relatively high apparent levels of physical activity. In the Scottish Health Survey 2003, for example,14 >75% of 6–10 year olds were reported to exceed the public health target1 2 of an accumulated 60 min of moderate to vigorous physical activity (MVPA) per day every day, but recent UK studies that have measured MVPA by accelerometry suggest that <5% of children and adolescents meet this target.15^–17 Subjective measures appear to quantify the perception of physical activity, rather than physical activity per se, and current methods for national surveillance of the amount and intensity of habitual physical activity in the United Kingdom may provide a false sense of reassurance concerning population levels of physical activity.

Objective measurements of physical activity in children and adolescents have often produced counter-intuitive results, confirming the value of accelerometry. For example, interventions designed to promote physical activity such as walking to school and increasing the time allocated to school or pre-school physical education have been reported to be unrelated to total physical activity (as measured objectively by accelerometry18^–20). Promotion of active transport to school and increasing physical education may seem strategies that are so obviously effective at increasing physical activity as to not require evaluation, but the empirical evidence suggests that objective evaluation is essential.

Improved understanding of relationships between physical activity, sedentary behaviour and health

Accelerometry has the potential to improve our understanding of relationships between physical activity and health. For example, recent systematic reviews have concluded that relationships between childhood physical activity and obesity were unclear, largely because older studies (which used subjective methods) were unable to quantify physical activity adequately.21 Recent accelerometry studies have identified relationships between physical activity, sedentary behaviour, obesity and cardiovascular risk factors, in part because physical activity and sedentary behaviour have been measured with higher accuracy and precision using accelerometry.17 22 In addition, because accelerometers provide data on amount and intensity of activity, the methodology allows investigation of “dose–response” relationships between health and physical activity, providing important practical evidence that can be used to produce clinical or public-health recommendations with much greater confidence and which are quantitative.17

Discrepancies in findings from objective versus subjective methods

Accelerometry can provide insights that are not available from traditional self-reporting of physical activity in observational studies. For example, the influence of socio-economic status on child or adolescent physical activity is topical. Some recent UK studies that used self or parent reports23 24 found significant socio-economic differences in physical activity and/or sedentary behaviour. In contrast, recent UK accelerometry studies have found no significant differences between socio-economic groups.16 25^–27

AREAS OF UNCERTAINTY IN ACCELEROMETRY: EVIDENCE-BASED ANSWERS TO COMMON PRACTICAL QUESTIONS

Choice of accelerometer

The first practical issue facing users of accelerometry is which accelerometer to use. A recent systematic review found that the device most widely used, the MTI Actigraph (MTI, Florida), is also the device that has the greatest body of consistent and high-quality evidence to support its use:5 it is feasible, reliable and valid. In addition, there is a large body of evidence on “calibration” of the Actigraph. Accelerometers produce output in counts per unit time (epoch), but these counts have no biological meaning per se and must be converted to biological constructs such as MVPA or sedentary behaviour by empirical studies of their relationships to energy expenditure or direct observation of activity or some health outcome such as bone health in “calibration studies”.3 28 29 Supportive validation and calibration evidence is emerging for devices other than the Actigraph, and the field of accelerometry changes rapidly. Table 1 summarises the accelerometers used in paediatric studies and provides access to manufacturer’s websites for further information, including technical details relating to the devices and references to primary literature.

A variety of important methodological issues confront users of accelerometry: these are summarised in table 2. The evidence base that addresses some of these issues is limited and contentious, and we address three issues below by providing new empirical evidence. The practical approaches to accelerometry taken should be fully described in future studies, a practice that a previous review noted has not been common.30

Choice of epoch

Older accelerometry studies have tended to use 1-minute sampling intervals (epochs). It is widely believed that shorter epochs would be more appropriate in children because of the perception that children’s patterns of physical activity are highly intermittent, based on a single study31 in which 15 6-to-10 year olds were observed over 3 days. More recent studies using direct observation and heart-rate monitoring to measure patterns of physical activity in children suggest a much more sedentary pattern of behaviour with limited physical activity and patterns of physical activity much more like adults.32 33 If children do undertake high-intensity activity only in very short bouts, “long” epochs of around 1 min might mis-classify high-intensity activity as being of lower intensity, by averaging with bouts of lower intensity activity within the same epoch.

Only two empirical studies appear to have addressed this issue. Rowlands et al34 compared apparent intensities of activity measured with 1 sec and 60 sec epochs with the RT3 accelerometer in 25 7-to-11 year olds over 1 h. Rowlands et al reported that differences between the two epochs were minimal, affecting only “very hard” intensity activity (mis-classified as “hard”).34 Nilsson et al35 found no significant effect of epoch on amounts of light and moderate intensity physical activity with the Actigraph in 16 7 year olds, although vigorous intensity activity was misclassified as moderate intensity to some extent in the longer epochs. One practical solution to this potential problem when using longer epochs is to classify moderate and vigorous activity together, as MVPA:15 this is also biologically and clinically meaningful because public-health targets for physical activity in children and adolescents are currently expressed in terms of MVPA.1 2

To add to the evidence base on the question of epoch using the Actigraph, we undertook a secondary analysis of existing 7–10 day accelerometry data from 32 free-living children (age 5 and 6 years) using the same methods adopted by Nilssen et al:34 data from a previous study, originally saved in 15 sec epochs but reported using 60 sec epochs,15 were reintegrated in 15, 30, 45 and 60 sec epochs. We then expressed MVPA using the cut-points of Puyau et al.29 and sedentary behaviour using the cut-point derived from our previous calibration study3 for data summarised in the four epochs. The results are shown in figure 1. We found that the differences between the epochs for sedentary behaviour were not statistically significant. For MVPA, the differences were significant statistically, but the differences were small, consistent with the other two studies to address this question summarised above. The biological significance of the differences in MVPA observed is unclear.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 (A) Minutes per day in sedentary behaviour in 32 5 and 6 year olds with data expressed in 15, 30, 45 and 60 sec epochs. No significant difference between epochs (Kruskal Wallis, p = 0.907). (B) Minutes per day in moderate to vigorous physical activity (MVPA) in 32 5 and 6 year olds, with data expressed in 15, 30, 45 and 60 sec epochs. Differences between categories are statistically significant (Kruskal Wallis test followed by paired Mann–Whitney tests, p<0.01 in all cases), ranging from median of 28 min/day (interquartile range 19 to 42) to 17 minutes per day (inter-quartile range 10 to 28) at 15 and 60 sec epochs, respectively.

In summary, despite a widespread perception that shorter epochs are essential to measure physical activity in children, the empirical evidence on the topic is limited and does not support the notion that “short” epochs are essential. One exception to this conclusion might be in circumstances where the outcome of interest is vigorous intensity physical activity.

Effect of different accelerometry cut-points on apparent levels of physical activity and sedentary behaviour

To measure the amount of sedentary behaviour (no trunk movement, largely consisting of time spent seated3) and the amount of time in activities of moderate to vigorous intensity (equivalent to energy expenditures above around three times their energy expenditure at rest), accelerometry counts are interpreted using cut-points derived from calibration studies for the reasons noted above. There is currently enormous variation in practice between researchers in the use of cut points, and widespread use of cut-points derived from adult studies, from the manufacturers of accelerometers (with provenance unknown), from calibration studies (which set out to derive cut-points) and observational studies (which simply report mean cut-points during particular activities).

The first practical issue to consider is whether meaningful differences in the amount of measured MVPA and sedentary behaviour arise from the use of the different cut-points. This question has not been examined systematically to date. For the present review we have re-analysed previously published data15 from CSA Actigraph accelerometry in 72 children (31 boys: 41 girls: mean age 5.8, SD 0.5 years) studied over 7 days (mean 10.5 h/day: SD 1.1). We took the data set and applied three commonly used cut-points for MVPA: Puyau et al29 from a calibration study based on free-living energy expenditure in 26 6–14 year olds; the Trost/Freedson cut-point36, apparently based on extrapolation from adult treadmill data, and age-dependent (for our sample MVPA was defined as a cut-point of 630 cpm); Treuth et al37 from a calibration study based on free-living energy expenditure in 74 13–14 year old girls, cut-off 3000 cpm. The effect of the three different cut-points on min/day and % of daily time in MVPA are shown table 3. For sedentary behaviour we took the same approach, comparing the three most popular cut-points from calibration studies in the literature: 800 cpm from the study of Puyau et al;29 1100 cpm from the study of Reilly et al3 from a study in which accelerometry was calibrated to sedentary behaviour (measured by direct observation); Treuth et al,37 100 cpm, based on an energy-expenditure study in 64 13–14 year-old girls. Table 3 shows statistically and biologically significant differences in amounts of sedentary behaviour and MVPA when the various cut-points were applied to the same data.

These new findings illustrate the extent to which engagement in MVPA and sedentary behaviour is dependent on the cut-point applied to the data and provide evidence as to the magnitude of the differences that can be expected. This leads to the question of which cut-points are most appropriate. Several lines of evidence are relevant to this question. First, biological plausibility — is it plausible that children engage in >4 h/day MVPA, a common observation in studies which use low cut-points to define MVPA (table 3 summarises data using the Trost/Freedson cut-point36)? This seems implausible to us given secular trends of increased fatness of children, even among non-obese children,11 evidence from accelerometry carried out simultaneously with total energy expenditure measured using doubly labelled water,15 (where both methods independently suggested that MVPA was low), and other evidence suggesting that MVPA is low in children from studies using direct observation, heart rate monitoring and pedometry (step counting).32 33 Second, what is the nature and quality of the evidence on paediatric cut-points? We make a marked distinction between calibration versus observational studies. Calibration studies aim specifically to determine the most appropriate cut-points by relating accelerometry output to energy expenditure and/or direct observation of movement, and with the most appropriate statistical analysis used to calculate the “diagnostic accuracy” of various cut-points. By contrast, observational studies simply describe typical accelerometry output for a given activity, and typical output may not represent the optimal diagnostic cut-point to identify that activity when the child is free-living, particularly given the marked variation in accelerometry output that exists between individuals for the same activity.3 38 39 A hierarchy of calibration studies exists; the calibration evidence that is most applicable to free-living activity in children will come from paediatric studies where children participate in a range of usual activities. Cut-points on the basis of adult data, or extrapolated from adult or treadmill data, should be viewed with caution: biomechanics of movement differ between treadmill and non-treadmill movement and cut-points differ markedly between treadmill and non-treadmill-based calibration studies.40 Cut-points should also be based on published studies so that their provenance can be considered critically. Finally, the mass and consistency of evidence is important: confidence in cut-points requires a mass of high quality and consistent evidence from published paediatric calibration studies. Current evidence from high-quality calibration studies in children and adolescents is fairly consistent in suggesting that the most appropriate cut-point when using the Actigraph, with 1 min epochs, lies in the range 3000–3600 counts/min.29 37 41 42

Effect of age on accelerometry output

A concern among users of accelerometry in children and adolescents is the possibility that accelerometry output may vary systematically with age, as a result of age-related changes in height or weight, or biomechanics of movement.7 This important practical issue has not been studied systematically in children. If accelerometry output was fairly independent of body size or age then the practical utility of the method would be enhanced, since there would be no need to adjust output for age/size. To address this gap in the literature, we conducted a study in which we recruited 108 children in three distinct age groups (3–4 years n = 35; mean height 1.08 m, mean weight 18.8 kg), 5–8 years (n = 42; mean height 1.20 m, weight 34.1 kg), 9–10 years (n = 31; mean height 1.40 m, weight 37.1 kg) who were participating in 45–55 min physical activity classes at the University of Glasgow. The work had ethics approval from the Yorkhill Hospitals Research Ethics Committee, and informed written consent was obtained. We observed children while they undertook a wide range of activities during the classes and quantified the intensity of movement using direct observation with the “Children’s Physical Activity Form” (CPAF).3 During the classes participants wore the Actigraph accelerometer at the right hip and accelerometry output data were collected in 1 min epochs. We synchronised our 1 min epochs from the CPAF with the Actigraph accelerometer by video recording the classes and setting the camera and accelerometer time to the PC clock. We then extracted epochs, on the basis of direct observation, which were entirely sedentary (CPAF category 1, stationary with no trunk movement) such as watching TV or moderate intensity (CPAF category 3, movement of the trunk at low-moderate speeds) such as ball games. Even with three groups of children who differed so markedly and significantly in height, weight and age, we found no evidence of systematic variation in accelerometry output across the three groups during either sedentary behaviour or MVPA. Mean (SD) accelerometer output was 666 cpm (418), 716 (694) and 607 (515) during sedentary behaviour in the groups from youngest to oldest respectively. During CPAF-classified moderate intensity physical activity, mean (SD) accelerometry output was 2650 (841), 2524 (688) and 2688 (773) from youngest to oldest groups respectively.

The present study therefore suggests that accelerometry output (at least with the Actigraph) has little age- or size-related systematic variation for the same behavioural input across a wide age/size range (3–10 years in the present study). This observation should increase the practical utility of the methodology by simplifying data interpretation.

CONCLUSIONS

Accelerometry now provides a practical, reliable and valid means of quantifying the amount and intensity of physical activity, and amount of sedentary behaviour, in children. For most applications where the amount and intensity of physical activity is of interest, accelerometry will offer a marked improvement over more traditional methods. Use of accelerometry should avoid bias in physical activity measurement and should improve our understanding of relationships between physical activity, sedentary behaviour and health. While a degree of uncertainty remains over certain practical issues, evidence-based accelerometry measurement protocols are now available.