This study applied the SLOTH theoretical model  to examine GPS and accelerometer measured PA in the leisure, school, transport and home domain, and further expanded the model to encompass 11 subdomains . MVPA was detected in all domains and subdomains, with almost 90% of all week day MVPA being objectively assigned to domains and subdomains. This is a significant improvement when compared with a recent study which used a similar methodology (only 60% of MVPA was assigned to locations or transport ).
There was large variation in how often subdomains were used for MVPA. In active transport, almost all participants accumulated some MVPA, while less than half of participants accumulated MVPA at playgrounds. There are a number of possible explanations for these findings; a) a large number of participants did not visit some subdomains; b) participants did not accumulate any MVPA whilst visiting some subdomains; or c) the employed data collection period was not suited to capture the use of certain subdomains (i.e., subdomains such as playgrounds that are used infrequently (i.e., less than weekly) but may be important places for accumulating MVPA). In support of our findings Qiugg and colleagues found that the overall amount of PA occurring at playgrounds was quite low . However, in the current study, the proportion of time spent in MVPA when visiting a playground was relatively high (21-35%). Research focusing on the use of, and PA accumulated in subdomains such as playgrounds, sports facilities or urban green space, however, may need to employ a longer data collection period to capture behaviour in these subdomains.
In support of previous research we found that school and leisure are important domains in regard to the total amount of MVPA accumulated [10, 12, 15, 34, 35]. The proportion of time spent in MVPA when at a specific location may warrant attention as this can highlight domains or subdomains where PA promotion may be most efficient. The proportion of time spent in MVPA was found to be particularly high in active transport, playgrounds, sports facilities and urban green space. Other studies have found that active transport to school can contribute to children’s overall PA levels  and that the proportion of time spent in MVPA in active transport to school can be as high as 50% . The current study found that active transport overall accounts for almost 20% of all MVPA and the average proportion of time spent in MVPA in active transport is between 30-35%. Rainham and colleagues  found similar results for their suburban population when investigating locations for youth PA. But contrary to our findings, girls achieved more MVPA than boys . These differences may be due to GPS processing methods, or that the Rainham study assessed total amount of MVPA across valid days, without adjusting for differences in accumulated time or number of valid days between groups.
It is well established that in general, boys are more physically active than girls [6, 15], however, gender differences in context specific PA behaviours has seldom been investigated. Our findings suggest that gender differences in PA may be partly explained by a generally lower level of PA among girls throughout the day and not correlated to specific behaviours, environments or contexts (home being the exception as no differences were observed). A recent study using accelerometer and GPS, found similar results to our study however the participant age range was younger (i.e., 6–11 year old children) . Overall, these findings suggest that intrapersonal factors and social, cultural and perceived environments may be more influential in shaping girls PA behaviour than physical environment features of specific domains or subdomains.
Compared with adolescents, children accumulated more overall MVPA, and this was primarily explained by differences in MVPA accumulated during school, in particularly during recess whereas no age difference was detected during PE. In Denmark, children from grade 7 are permitted to leave the school grounds which may result in more sedentary activities (such as sitting in a café). Further, as children age they may prefer to socialize with peers as opposed to “playing”, which may also induce more sedentary behaviours. Thus, interventions aimed at reducing the decline in PA from childhood to adolescence within the school setting may be important; for example providing active programs for recess and keeping students on campus. Very few studies have examined correlates for adolescents PA during recess  and no intervention studies targeting adolescents were found in a recent review investigating the effect of school recess interventions on PA . Future research in school settings should investigate adolescent school behaviour, in particular correlates for recess PA behaviour as well as effects of recess interventions. The lack of an age difference in MVPA during PE may be explained by PE being mandatory. Interventions aiming to increase the number of PE lessons as children become adolescents may thus help prevent or reduce the age related decline in PA. However, a recent prospective study concluded that the importance of the after school period for accumulating MVPA increases with age  and a strategy focusing on increasing PA during both the school and leisure domain may be warranted. Future prospective studies should also consider using a domain approach to track the relative and cumulative importance of domains and subdomains of MVPA from the early years through to adolescence.
Furthermore children’s MVPA patterns over the weekend were collected because different patterns of PA have been observed between week and weekend days . In supplementary analyses of weekend data (not shown) it was found that all groups (boys, girls, children and adolescent) had fewer minutes of MVPA compared with week days (p < 0.05), that the gender pattern prevailed (boys compared to girls were significantly more active in all domains except the home domain, p < 0.05), but that no age differences were present (p > 0.1).
Strengths and limitations
The current study pre-defined potential domains and subdomains relevant for children’s MVPA based on a theoretical model and published literature, and used a systematic approach to objectively identify and assess locations for PA. Considering the relatively novel application of accelerometer and GPS data to health research, the strength of this study was its large sample of detailed and matched accelerometer and GPS data. The use of accelerometers and GPS has limitations. The PALMS algorithms used in this study currently provides 80% accuracy in detecting travel modes. Also the GPS signal inaccuracy varies with the environment (e.g., urban canyons can reduce accuracy). As this study is conducted in an inner city environment, some locations and travel modes are likely to have been misclassified, and further studies are needed to determine the impact of different environments on GPS inaccuracy and further improve the PALMS algorithms. Accelerometer measurements also have limitations predominantly associated with data processing decisions and the choice of a threshold to determine the level of MVPA. As the aim of this study was to describe patterns between domains and subdomains by age and gender, the actual threshold value chosen is unlikely to impact the results. The subdomains included in this study were also derived to reflect places where children and adolescents accumulate MVPA. However, future research may also consider the inclusion of context specific domains for other activity intensities such as sedentary, light or vigorous activity, and the inclusion of counts per minute as an outcome variable.
The study sample was from a deprived area, which is rare in studies of this nature. The included schools are urban schools situated in one of the densest areas in Denmark, with a high proportion of ethnic minorities  and a high unemployment rate . As such the results may differ for suburban and rural schools, or schools situated in more affluent areas. Future research should apply the PA behaviour measurement methods developed as part of this study to children living in a wide range of socio-economic areas and countries. This study did not aim to provide population estimates of domain specific PA among Danish children and adolescents as it was based on a sample from deprived area. As such the generalizability of this study is limited to children from schools in deprived neighbourhoods with a high unemployment rate and a high proportion of immigrants and descendants. However, the methodology developed as part of this study provides a significant advancement in the measurement of children’s PA behaviour which will advance the promotion of PA and inform new PA interventions. Focusing on a population in need, however, may have generated findings that would not have been seen elsewhere and helps demonstrate the utility of the current methods. This study included outcome measures for two variables, each consisting of four domains and 11 subdomains. Thus Type I error (i.e., false significance) may have been an issue due to multiple testing. Moreover, data was only collected during early fall and late spring, where daylight and overall weather conditions were quite similar, and therefore it is unknown if the results are valid during the winter months .