Article Text
PDF
Abstract
Aim The aim of this review is to summarise issues surrounding the measurement of physical activity (PA) by self-report and accelerometry in youth (2–18 years old). Current levels and temporal trends in PA and sport participation and the effect of assessment method on data interpretation will be summarised.
Methods Relevant papers were extracted from a computerised literature search of MEDLINE and personal databases. Additional papers were extracted from reference lists of recently published reviews.
Results The criterion validity (direct comparison with an objective method) of self-reported instruments is low to moderate, with correlation coefficients generally between 0.3 and 0.4. Self-report instruments overestimate the intensity and duration of PA and sport participation. The interpretation of PA data from accelerometry is a challenge, and specific issues include the definition of intensity thresholds and the influence of age on intensity thresholds.
Recent data on self-reported PA in youth suggest that between 30% and 40% are sufficiently active. Prevalence values for sufficiently active youth measured by accelerometry range between 1% and 100%, depending on the intensity thresholds used. Sport participation is likely to contribute to higher levels of PA. The available evidence does not support the notion that PA levels and sport participation in youth have declined in recent decades.
Conclusion The number of youth meeting current PA guidelines varies by assessment method and the intensity thresholds used when PA is measured by accelerometry. The available evidence does not firmly support the notion that PA in young people has declined during the last decades. It is unlikely that any self-report method is sufficiently accurate for examining cross-cultural differences and temporal trends in young people's PA and sport participation over time. Surveillance systems therefore need to strive for an international standardisation using objective measurements of PA to complement existing self-report instruments.
Statistics from Altmetric.com
Introduction
Accurate and valid methods for measuring physical activity (PA) are necessary to determine dose–response associations between PA and health outcomes, to specify which aspect of PA is important for a specific health outcome, to monitor temporal trends in populations, to make cross-cultural comparisons and to determine the effect of interventions.1
Despite much progress in developing instruments for assessing PA, limitations for the accurate and feasible assessment of habitual PA, including sport participation, still exist. These limitations are amplified in youth, and some are unique to this age group as young people's PA patterns are different from those of adults due to cognitive, physiological and biomechanical changes that occur during natural growth and development and to the differences in the nature of their PA.2,–,6 Youth's PA tends to be intermittent and of more variable intensity than that of adults, generally consisting of less planned, time-bounded and organised activity. For example, it has been estimated that the majority of children's activity bouts last between 3 and 22 s,7 8 which has substantial implications for the measurement, processing and interpretation of PA data.
The aim of this review is to summarise the validity of self-report methods for assessing PA and issues surrounding the measurement of PA by self-report and accelerometry in young people aged 2–18 years. The effect of assessment method on data interpretation will be discussed, and current levels and temporal trend data will be summarised.
Self-report methods for assessing PA
Self-report PA questionnaires are the simplest method for assessing PA, but they have some unavoidable limitations. The accuracy of self-reports is influenced by the ability of the respondent to accurately recall all relevant activities retrospectively. Therefore, self-report methods are subject to recall bias.5 9 This can be intentional or accidental false recall; missed recall or differential reporting accuracy of different intensities, dimensions and domains of activity. The age of the respondent is another important variable in determining the method of administration. Accuracy of self-report data is particularly problematic in children.5 Therefore, parental and teacher-reported questionnaires or proxy reports are often used. However, recall of youth's PA is also difficult for adults, and unique limitations and errors are associated with this method.10
PA questionnaires
Most questionnaires developed for use in youth consider specific types of activity, such as duration and frequency of sport and exercise, as well as recreational PA,11,–,13 which may be more accurately recalled compared with habitual activity as the individual has made a conscious decision to carry out that activity in a defined period of time. However, the validity of specific types and domains of PA is difficult to assess as most self-report methods are scored in terms of duration of PA, moderate- and vigorous-intensity physical activity (MVPA) or estimates of energy expenditure.
Corder et al10 identified 18 studies that compared self-reported PA (21 different instruments) with data from objectively measured PA (ie, doubly labelled water, accelerometry, heart rate monitoring). Most questionnaires assessed time spent in MVPA, total PA or leisure time activity. No questionnaire was specifically developed for assessing frequency and duration of sport participation. The vast majority of instruments were developed to assess PA during the past 1–7 days, and the administration mode was self-report in approximately 50% of the studies. Criterion validity results were almost entirely reported as correlation coefficients, and absolute validity (ie, mean difference between the Physical Activity Questionnaire and the criterion measures reported in the same units (eg, minutes of MVPA per day)) was reported in only 5 of 18 studies. The magnitude of criterion validity correlation coefficients ranged from very small to large (median r=0.40, range=0.09–0.85). There was a tendency for higher correlation coefficients for interview-administered instruments, whereas the time frame had little influence on the association between self-reported and objectively measured PA.
Adamo et al11 systematically reviewed the literature to determine the extent of agreement between subjectively and objectively measured PA in youth (<19 years). The risk of bias (ie, internal and external validity and quality of reporting) was evaluated using a modified tool by Downs and Black.13 Eighty-three studies met inclusion criteria, and 59 of these reported Pearson or Spearman correlation coefficients with a wide range between −0.56 and 0.89. However, strong correlations (r>0.6) were usually reported as the highest value for a range of correlations within a study.14,–,18 The authors concluded that the correlations between subjective and objective methods for assessing PA in young people are low to moderate.
Twenty-four of the studies provided a directly comparable measure between subjective and objective instruments (ie, absolute validity). Overall, self-reported instruments overestimated objectively measured overall PA by 72%. Female participants were more likely to overestimate their activity levels compared with male participants (114% in male participants vs 584% in female participants). These substantial discrepancies between self-reported and objectively measured PA suggest that self-report methods may be useful for ranking individuals, whereas absolute validity for estimating overall PA and time in MVPA is poor.
Chinapaw et al12 performed a systematic review of the reliability, validity and responsiveness of PA questionnaires developed for use in youth (<18 years old). The authors used a standardised checklist to judge the quality of methodology and results.19 Fifty-four studies that examined 61 different versions of questionnaires met the quality control criteria for inclusion. Six questionnaires (all proxy reports) were developed for preschool children (<6 years), 25 questionnaires were developed for children between 6 and 11 years and 31 instruments were developed for adolescents between 12 and 18 years. Reliability results were reported for 35 different instruments, and the study quality was rated as positive if the study population was >50, with reliability quantified as intraclass correlations (Pearson correlations and κ statistics >0.8); the time interval between administrations was deemed adequate by the authors.12 Only 7 of the 35 reviewed questionnaires received a positive rating, 12 were rated negative and the remaining instruments either were rated indeterminate or received both a positive and a negative score (eg, one subcomponent of PA (eg, MVPA) was rated positive and another (eg, total PA) was rated negative) by the authors.12
Results for construct validity (ie, comparison with an objective measure of PA, another subjective measure of PA or an indirect measure of PA; eg, aerobic fitness) were reported for 7 questionnaires for preschool children, 25 questionnaires for children and 31 instruments for adolescents. The most commonly used criterion method was accelerometry (46 studies). In preschool children (<6 years old), the highest correlation was reported for the Children Physical Activity Questionnaire (r=0.4220) for total PA. In older children (6–11 years), similar correlations were reported for the PA Questionnaire for Parents and Teachers21 and the ACTIVITY Questionnaire.22 In adolescents (12–18 years), correlation coefficients were somewhat higher for the previous day physical activity recall (PDPAR) (r=0.7723) and the Swedish Adolescent Physical Activity Questionnaire (SAPAQ) (r=0.5124).
A few of the reviewed studies reported results for the absolute validity or the degree of agreement between questionnaires and the criterion method. Although some instruments appear to accurately assess PA energy expenditure or MVPA on a group level, the errors are usually large for estimates on an individual level.20 25
Issues associated with self-report
The reviews summarised above have highlighted a number of limitations associated with self-reported PA in young people.10,–,12 The results consistently suggest that the correlation between any self-report and an objective instrument is low to moderate, at best. This is not surprising given the limitations of self-report to assess the intermittent nature of young people's PA.7 8 Self-report instruments usually overestimate the intensity and duration of different types of sports. This is because many sports are intermittent in nature, and someone who (accurately) reports practicing or playing football, usually coded as vigorous intensity, for 1 h may only be physically active at vigorous intensity for a limited period within that hour.
Estimating energy expenditure from self-report is also problematic and associated with error. Usually, energy expenditure derived from self-report is expressed as metabolic equivalents (METs) minutes by multiplying the frequency, duration and intensity of all reported activities. This estimation generally uses adult-derived standard energy costs of specific activities.26 Further, there are differences between METs for the same activities in adults and youth.27 28 Nevertheless, self-report instruments are so far the only available method for assessing the setting and type of PA behaviour.
Assessing PA by accelerometry
Objective methods have substantially increased our ability to obtain accurate measurements of the volume, pattern, frequency, intensity and duration of youth's PA and sedentary behaviours. Further, when studies in youth have employed objective measures of PA, strong and consistent associations with health outcomes (eg, insulin sensitivity, blood pressure, adiposity) have been identified.29 Methods to objectively measure PA in youth have been extensively reviewed.30,–,34
Accelerometers are the most commonly used method to assess free-living PA.33 34 The instruments measure the acceleration of the body or different parts of the body in one, two or three dimensions for a specific, predefined time period (epoch). The Actigraph activity monitor (http://www.theactigraph.com) has been used to assess free-living PA in large-scale studies (n>1000) including the Avon Longitudinal Study of Parents and Children,35 the SPEEDY (Sport, Physical activity and Eating behaviour: Environmental Determinants in Young people) study,36 the National Health And Nutrition Examination Survey,37 the European Youth Heart Study (EYHS),38 the CHASE (Child Heart And health Study in England) study39 and Trial of Activity in Adolescent Girls.40
However, the interpretation of PA data from accelerometry is a challenge, and specific issues associated with data interpretation need consideration.
Intensity thresholds
Estimating the amount of time spent sedentary and in different intensity categories (ie, MVPA, VPA) requires interpretation of activity counts using intensity cut-off points from calibration studies. Intensity thresholds for the Actigraph accelerometer vary between 10041 and 1100 counts per minute (cpm)42 for time spent sedentary and between 61543 and 3600 cpm44 for time spent in MVPA.
Table 1 summarises examples of published intensity thresholds for time spent sedentary and in moderate-intensity activity in youth.
Examples of published intensity thresholds for time spent sedentary and moderate-intensity activity in youth
These thresholds are usually derived in the laboratory when body movement and energy expenditure are simultaneously measured by accelerometry and indirect calorimetry, respectively. The activities chosen for these calibration studies likely explain much of the variation in the slope and intercept when modelling energy expenditure as a function of activity counts and, therefore, influence the developed intensity cut points.45 There is no consensus on the most appropriate intensity cut points, and the interpretation of accelerometer data is therefore influenced by the cut points employed. Reilly et al32 demonstrated that time spent sedentary varied between 180 and 501 min/day in 4–7-year-old children by applying different cut points. In addition, time spent in MVPA varied between 28 and 266 min/day.32 It is unlikely that children spend more than 4 h in MVPA per day as indicated in studies where the lowest MVPA cut point was applied,43 with many studies suggesting that the intensity threshold for MVPA lies in the range between 3000 and 3600 cpm.41 44 46 47
It is also likely that a substantial amount of misclassification for time spent sedentary occurs when a ‘high’ (eg, 1100 cpm) sedentary cut point is applied. Activity counts from the Actigraph accelerometer while sitting are <100 cpm,48 49 and a conservative estimate for time spent sedentary (eg, <100 cpm) to avoid misclassification may therefore be preferred.49 Face validity for this threshold was demonstrated by data from the EYHS suggesting an independent positive association between time spent sedentary and clustered metabolic risk50 and insulin resistance51 in youth.
Future studies aimed at examining the influence of different sets of sedentary and MVPA cut points in relation to health outcomes are warranted. It is also unknown whether the association between amounts of time spent sedentary and metabolic risk indicators is independent of time spent at higher intensities (eg, MVPA).
Age-dependent intensity thresholds
Some investigators have suggested that accelerometry output is age dependent.52 53 However, an age-dependent association between accelerometer counts and energy expenditure is influenced by the expression of energy expenditure. When energy expenditure is expressed in MET units using the adult definition of 1 MET (ie, 3.5 ml O2/kg body weight), it is likely that age, in addition to activity counts, significantly contributes to the explained variance in METs. This is because mass-specific resting energy expenditure (ie, 1 MET) is higher in children than in adults.27
To systematically address whether age and body size influence the output from accelerometry, Reilly and coworkers32 observed children with a vide variation in age, body weight and height while the children wore an Actigraph accelerometer. There were no significant differences in activity counts for time spent sedentary and in MVPA between age groups (3–4, 5–8 and 9–10 years), and the authors concluded that the accelerometer output had little age- or size-related variation. This may suggest that the same cut points for intensity classification can be applied across age groups. However, further research is needed to confirm this observation.
Are children active according to public health recommendations?
Self-report
Most public health authorities (eg, Centers for Disease Control and Prevention, USA; Department of Health, UK; Public Health Agency of Canada/Canadian Society for Exercise Physiology) agree that young people should accumulate at least 60 min of MVPA on at least 5 days/week. In addition, the benefits of vigorous-intensity activity are acknowledged. Although the underlying evidence for these recommendations has been criticised,54 they provide a framework when evaluating population levels of PA in youth.
One of the most extensive data sets reporting on the prevalence of PA is the Health Behaviour in School-Aged Children (HBSC) study.55 The 2001/2002 survey included youth aged 11–15 years from 34 European countries/regions and the USA. On average, young people reported being physically active for 60 min or more for 3.8 days of the week and 34% of youth reported that they were physically active according to the 60-min recommendation for at least 5 days/week.55
Li and coworkers56 reported similar prevalence figures in US adolescents using data from the Youth Risk and Behaviour Surveillance System (YRBSS). In the 2007 survey, 34.7% of 15–18-year olds met the 60-min PA recommendation. However, approximately two thirds of the youth reported sufficient vigorous-intensity PA, defined as a minimum of 20 min of activity that made them breathe hard and sweat on three separate occasions per week.56
A recent study, from 34 developing countries participating in the WHO Global School-based Student Health Survey, including more than 72 000 youth aged 13–15 years, reported that, at best, 23.8% of boys and 15.4% of girls met the 60-min MVPA recommendation.57 Although these data should be interpreted with caution in relation to HBSC and YRBSS data due to differences in test methodology, it appears that self-reported activity levels in youth from developing countries are low.
Objective measurement
Data on the prevalence of sufficiently active (ie, accumulation of >60 min of MVPA per day) young people from studies using accelerometry vary considerably (from 1% to 100%). This variability is largely explained by the use of different intensity thresholds when defining MVPA. Using age-specific intensity thresholds (1000 and 1500 cpm for 9- and 15-year olds, respectively), Riddoch et al suggested from their EYHS data that more than 95% of 9-year olds and 82% and 62% of 15-year-old boys and girls were sufficiently active.38 Nader et al58 reported that ‘almost all’ 9- and 11-year-old US children accumulated at least 60 min of MVPA per day, whereas this figure dropped to 32% at age 15 years. The prevalence estimates for sufficiently active youth should be interpreted keeping in mind the age-dependent and remarkably low intensity thresholds used in these studies, although these data apparently confirm previous observations of gender and age differences in PA.
Using the same algorithm for defining intensity thresholds53 as that used by Nader et al,58 Troiano et al suggested from their National Health and Nutrition Examination Survey data that 42% of 6–11-year olds and <8% of 12–15-year olds accumulated at least 60 min of MVPA per day.37 Unfortunately, the results from the Nader et al58 and Troiano et al37 studies of US youth are not directly comparable, due to differences in the definition of moderate-intensity PA (three METs in Nader et al58 vs four METs in Troiano et al37). The pronounced steep decline in activity levels by age is partly the function of real age-related declines and differences in age-related intensity thresholds (which are up to 50% higher in adolescents than in children).
Riddoch et al35 analysed data from 5595 12-year-old UK boys and girls from the Avon Longitudinal Study of Parents and Children birth cohort. The intensity threshold for MVPA (>3600 cpm) used in this study was substantially higher than those used in the US studies discussed above, and the results suggested that ≤5% of UK boys and <1% of UK girls accumulated more than 60 min of MVPA per day. Another study (involving 1862 UK children aged 10 years) defined moderate intensity as 2000 cpm and observed that 69% of the children were sufficiently active.36 Similar results were observed in the CHASE study39 of UK children (who were of European, South Asian and African Caribbean descent) using the same intensity thresholds in which 76% of boys and 53% of girls accumulated more than 60 min of MVPA per day. Table 2 summarises data on the prevalence of sufficiently active youth measured by accelerometry in studies including more than 1000 participants.
Prevalence of youth accumulating 60 min or more of moderate- and vigorous-intensity activity per day measured by accelerometry in studies with more than 1000 participants
These results clearly highlight that the inconsistent use of intensity thresholds is a major issue when trying to quantify the prevalence of sufficiently active young people from accelerometry. Although it may be unlikely that a consensus can be reached on which intensity thresholds are the most appropriate to use, there is an urgent need to establish whether accelerometer-based intensity thresholds are age dependent. Furthermore, until consensus has been reached, researchers are encouraged to present prevalence data on sufficiently active youth using different intensity thresholds preferably in the range between 2000 and 3500 cpm. Reanalysing raw accelerometer data from multiple studies using the same inclusion criteria, data cleaning and reduction procedures and intensity thresholds should enhance our understanding of the prevalence of sufficiently active youth.
Time trends in PA and sport participation
The magnitude and direction of recent changes in PA levels and sport participation of youth are not well understood. Repeated cross-sectional surveys of population-representative samples of youth are needed in order to properly examine time changes.
Knuth and Hallal59 systematically reviewed the literature on temporal trends in PA. The authors identified nine studies (four from the USA, two from Australia and one each from Switzerland, Canada and Sweden) that met their inclusion criteria. Six studies reported time trend data on at least one component of PA. The results from these studies varied due to methodological differences, such as differences in self-report instruments and administration mode. For example, one study60 using data from the YRBSS reported that the prevalence of sufficiently active 9–12-year olds decreased from 84% to 82% between 1993 and 2003. Another study61 reported that participation in physical education classes and proportion of sufficiently active adolescents (grades 9–12) did not change between 1991 and 2003. Based on the available data, although limited, the authors suggested that PA in young people may have declined over time.59 Recent results from the HBSC, which examined time trends in the amount of vigorous-intensity PA between 1986 and 2002, suggested stability or small increases over time in both boys and girls.62 It is likely that the question on vigorous PA (‘How often do you usually exercise in your free time so much that you get out of breath or sweat’ (outside school hours)?) includes a substantial proportion of sport participation, although this was not specified in the questionnaire. Due to the nature of the question on vigorous-intensity PA (‘How often do you usually exercise in your free time so much that you get out of breath or sweat’ (outside school hours)?)
Li et al56 summarised time trend data on PA, participation in physical education and the time spent sedentary in US youth from seven previously published studies based on YRBSS data. The results suggested that the prevalence of having sufficient vigorous PA changed little between 1993 and 2005 (from 65.8% to 64.1%). Furthermore, the prevalence of youth attending daily physical education and exercising for more than 20 min during physical education increased by about 5% between 1995 and 2007.56 However, even if there was no clear evidence of a secular decline in PA, only one third of the US adolescents met the recommended levels of 60 min of daily MVPA.
Data from Iceland suggest that participation in vigorous-intensity PA and sports increased steadily from 1992, 1997, 2000, up until 2006 in 14–15-year olds.63 Participation in sports clubs increased consistently from 21.5% to 36.3% and from 12.8% to 27.3% in boys and girls, respectively.63
Few studies have examined time trends in PA by means of objective monitoring. A study in Swedish children examining daily step counts by pedometry in 7–9-year-old children reported significant increases in daily accumulated steps between 2000 and 2006.64 Ninety per cent of girls and 67% of boys met the preliminary weight control recommendations of accumulated steps65 in 2006 compared with 75% and 60% in 2000. Intriguingly, the prevalence of overweight and obesity increased from 16.5% to 21.8% and from 14.4 to 21.7% in boys and girls, respectively, between 2000 and 2006.64
A similar study in 8–10-year-old children from the Odense municipality area of Denmark measured habitual PA by accelerometry in 1997–1998 and again in 2003–2004. There was no difference over the 6-year period in weekday PA or weekend PA.66 Further, no changes in activity levels were observed across socioeconomic gradients. However, the authors suggested that these results should be interpreted cautiously due to the relatively small sample sizes, which may act to reduce the confidence in time trend estimates.
The evidence from the studies summarised above does not support the notion that PA levels in youth have declined, at least not during the last two decades. This is in contrast to a previous review suggesting that PA in defined contexts such as active transport, physical education and sport participation has typically declined.67 While the evidence of a secular decline in active transport appears consistent among children from different parts of the world,67 the pattern of change is less clear for sport participation. Much of the evidence cited in the review by Dollman et al67 was drawn from several Australian surveys, and data from other countries show conflicting results.68 Although the two studies65 67 that examined recent time trends in overall PA using objective measurements of PA were unable to demonstrate a decline in activity levels in youth, others69 have attempted to address whether young people's PA has declined by comparing children living a ‘traditional’ lifestyle (Old Order Amish and Old Order Mennonite) with those living a more ‘contemporary’ lifestyle (urban and rural Saskatchewan). Children living a traditional lifestyle were substantially more active than contemporary children (about 80 vs 55 min of MVPA per day), and the authors concluded that this difference was primarily due to higher habitual activity levels in those children living a traditional lifestyle, even despite not participating in organised sports and physical education.69 This begs the question whether sport participation contributes to overall activity levels in youth.
Some investigators have observed a higher amount of MVPA on sport participation days than non-sport participation days in 6–12-year-old boys.70 Leek et al71 measured PA by accelerometry in 200 7–14-year olds while training for sport. Overall, 24% of the youth met the 60-min PA recommendation during training sport.71 Results from another study directly examining time spent in MVPA by accelerometry during after-school programmes showed that children were significantly more active during free play than during structured activities.72 Nilsson et al73 examined correlates of objectively measured PA in more than 1300 9- and 15-year-old boys and girls from three European countries. In 9-year olds, playing outdoors after school was associated with higher percentage time in MVPA, while participation in sport clubs was associated with higher percentage time in MVPA in 15-year olds. It appears that sport participation contributes to higher levels of objectively measured time spent in MVPA. However, for the majority of youth who participate in sports, additional activities are needed to meet the current recommendations of 60 min of MVPA per day.
Conclusion
In light of the inherent limitations with self-reported PA in young people, the variability in age across studies and differences in recruitment, representativeness and sampling, and testing methodologies, it appears that approximately 30–40% of youth are sufficiently active based on current PA guidelines. The available evidence does not firmly support the notion that PA in young people has declined during the last decades. Surveillance systems need to strive for an international standardisation using objective measurements of PA complemented by existing self-report instruments. Consistent and standardised analyses from accelerometry data are urgently needed to enhance our understanding of the prevalence of sufficiently active youth.
References
Footnotes
-
Competing interests None.
-
Provenance and peer review Not commissioned; externally peer reviewed.