1. Introduction
Physical activity (PA) is associated with numerous health benefits in school-aged youth [
1,
2]. Results from observational studies generally show dose–response relationships; that is, any increment in PA, irrespective of the type, frequency or duration, is related to increasing cardiovascular, musculoskeletal, cognitive and metabolic benefits for children’s general health and well-being [
2,
3,
4,
5]. In addition, experimental studies have shown that PA programs can achieve promising results on the anthropometrical and cardiovascular risk profiles in high-risk youth [
6]. As physical inactivity tends to track from youth to adulthood [
7], promoting PA of children is a crucial component of strategies that combat the associated health consequences of physically inactive lifestyles.
The development of objective measurements (e.g., accelerometry) allows researchers to continuously monitor children’s daily PA behavior, and to investigate separate time periods that are promising for interventions by filtering data outputs based on time segments. Acknowledgement of these specific time segments is essential, as children’s PA patterns fluctuate during the day, and can be highly context-specific [
8,
9]. The afterschool period is such a context-specific time segment that is often referred to as “critical hours” for PA promotion, because it contributes up to half of the daily amount of moderate to vigorous PA (MVPA) [
10], and afterschool PA declines as children reach adolescence [
11]. During the afterschool period, children have more discretion over the activities in which they engage [
11,
12]. Because of this, afterschool behaviors may reflect more autonomously regulated behavior, which may be more likely to persist into habits and routines [
13,
14]. Hence, afterschool PA has been considered highly predictive of overall sustained PA patterns [
15], which makes it a primary time segment for PA interventions [
16].
Since 2010, the number of studies that used objective measurements of PA to investigate afterschool PA rapidly increased [
17]. However, interpretation of results of the studies to date is limited by several challenges. First, there are inconsistent definitions of the afterschool time segment [
11]. For example, while some studies used generic start time thresholds such as 3 PM [
12,
18,
19,
20] or 3:30 PM [
21,
22,
23], other studies used reported schools’ schedules [
10,
11,
24,
25,
26,
27] or excluded activity during school hours [
28]. Second, during the afterschool period children are exposed to diverse attributes of the physical environment which may have an influence on children’s PA [
29,
30]. Hence, various studies have investigated attributes that foster afterschool PA. To date, six studies have investigated relationships between afterschool PA and the physical environment using objective measures of PA [
22,
25,
28,
31,
32,
33]. Four studies solely investigated environmental features around the residential neighborhood [
22,
25,
31,
33]. However, not only the residential environment, but also other environments such as the children’s school environment and the daily transport route between home and their school are of interest when investigating environmental attributes related to afterschool PA, as especially during the afterschool period, children spend significant parts of their time outside their residential neighborhood [
34]. Consequently, the second challenge is to analyze children’s PA in relation to so-called daily activity spaces (e.g., including school, residence and daily transport route as spatial anchor points) may be especially suited for investigating associations with PA during the afterschool time period [
35,
36]. Third, although investigating afterschool PA separately from total daily PA is an important first step in examining context-specific determinants of PA [
32], the afterschool period still consists of multiple distinct “activity types” (e.g., active transport, organized sports participation and leisure time PA) [
24]. The ability to differentiate between several activity types is essential, as the influence of potential (environmental) determinants depend on these activity types [
9]. Only very few studies to date, however, have succeeded to measure these activity types and attributes of the physical environment using objective measurements [
30]. Thus, the third challenge is to differentiate between activity types in the afterschool time period, in order to understand which attributes of the physical environment influences specific types of objectively measured PA [
9].
Combined accelerometer and Global Positioning Systems (GPS) methodologies may help to overcome the challenges presented above. First, afterschool time segments may now be individually validated by the actual presence of a child on its school parcel, rather than based on generic start-time thresholds or reported school schedules. Second, when integrated with registries of the physical environment such as Geographic Information Systems (GIS), researchers can measure underlying characteristics of the environment based on a participant’s unique GPS-based mobility pattern. In this way, children’s environmental exposure can be measured in a broader perspective than solely their residential environment, as operationalized in the majority of previous studies [
29,
30,
37]. Third, when combining the parameters activity, location and time, combined accelerometer and GPS methodologies can provide essential information on the relative importance of afterschool activity types and the context in which these occur. This may increase our understanding of the influence of attributes of the physical environment in fostering these afterschool activity types.
To date, three studies have investigated the influence of the physical environments on afterschool PA, using accelerometers and GPS methodologies [
21,
24,
28]. These studies have all used a so-called contemporaneous momentary design, in which accelerometer-measured activities and GPS-derived geographical location (and its attributes of the physical environment) are collected at the exact same moment in time. In other words, these designs investigate relationships between behavior and environment within simultaneously occurring pairs of (1) PA or sedentary behavior and (2) attributes of the physical environment. However, these designs may be vulnerable to selective daily mobility bias [
35,
37]. Selective daily mobility occurs when researchers investigate simultaneously occurring PA-environment pairs, and include locations that children purposively visit to perform the desired behavior (e.g., visiting sports grounds for performing the planned PA) [
35,
38,
39]. In this case, individual preferences for performing PA behavior in specific settings confounds the causal relationship between environmental exposure and PA behaviors [
37]. Although these studies can reveal relevant short-term insights into where PA takes place [
28,
40,
41,
42,
43,
44], it remains unclear whether long-term PA patterns are actually influenced by attributes of the physical environment [
38,
39].
One possible way to overcome potential selective daily mobility bias is to reconsider our view of environmental exposure towards more non simultaneous or long-term exposure of the environment. Based on work in space–time geography, transportation research and environmental psychology, Perchoux and colleagues conceptualized a theoretical framework of environmental exposure [
35]. This theoretical framework defines environmental exposure as a dynamic individualized activity space based on space and time [
45]. These individualized spaces are shaped by the location of usual visited places (e.g., home or school) [
46]. Furthermore, individual spatial freedom is delimited by various constraints (e.g., transportation mode or regulations concerning home or school hours) [
47]. Moreover, these individual activity spaces are also formed by the social interactions that are often fixed by location or time, and may over time even evolve to a sense of belonging to a certain neighborhood [
48]. Relative accessibility of PA opportunities within a child’s daily activity space may influence its afterschool PA behavior. The aim of the present study was thus to improve our understanding of the relationship between characteristics of children’s daily physical environment and their afterschool PA behavior, by using combined accelerometer and GPS methodologies, and framing analyses within the activity space framework [
35].
4. Discussion
This study examined relationships between features of the physical environment and specific afterschool PA behaviors (i.e., afterschool leisure time PA behavior and afterschool active transport). Our first, more methodological aim, was to investigate context-specific afterschool leisure time and active transport by filtering these contexts from other afterschool contexts, such as organized sports participation. This may be important since previous studies suggested that this may be a confounding factor in the relationship between PA and the physical environment [
37,
72]. We showed that GPS devices provide additional descriptive information about the context of daily PA and mobility patterns, which enables more context-specific analyses of the relations between leisure PA and its environmental determinants [
9].
4.1. Empirical Findings
This study showed that greenery density (i.e., lawns and shrubs) was associated with more afterschool leisure time PA and walking, but we found no association with the density of general vegetation. Systematic reviews, including studies until 2010, reported a mixed association between environmental greenspaces and PA [
29,
73]. However, studies from 2010 onwards using objective PA and GPS-determined environmental exposure consistently suggest that children are more active in greenspace environments such as parks [
28,
41,
74,
75]. Findings from the present study not only support the suggestion that shrubs and lawns (often found in public green spaces) may be important facilitators for children’s PA, but also show that children with a higher density of shrubs and lawns in their activity space, generally perform more afterschool leisure time PA than children with a lower density of these environments (irrespective of whether PA is actually performed around shrubs or lawns).
From the six studies that investigated relationships between objective afterschool PA and the physical environment [
22,
25,
28,
31,
32,
33], two studies focused on public features of the environment [
25,
32]. These two studies suggest that not only the public open space closest to the children’s residence is associated with afterschool PA, but also the larger home–school environment or parts thereof. In contrast to an earlier study [
32], we found no evidence that a higher density of public playgrounds was associated with more afterschool leisure time PA or active transport. This may be explained by the fact that GIS data in the present study did not enable us to look at quality, maintenance status or age-appropriateness of these playgrounds. In addition, some unstructured PA at playgrounds may have been lost when excluding behavior performed at sports facilities. Additional use of activity diaries may help in identifying planned PA behavior at specific locations. Similarly, we found that a higher density of buildings was associated with more minutes of leisure time MVPA. Results may be comparable with results from Rodriguez et al., who found increased MVPA in environments with a higher population density.
We found that the distance between the children’s school and home was an important determinant of leisure time PA, cycling and walking. More specifically, greater home–school distances were related to more cycling, but less leisure time MVPA and walking. As results from the electronic questionnaire showed that the vast majority of our participants used active transport to get to and from school, greater distances may be related to more cycling as a replacement of walking during the home–school commute (and vice versa). In addition, greater distances may also reflect the subgroup of participants living in neighborhoods that may be somewhat further away from facilities (e.g., supermarkets or afterschool activities) [
75,
76]. This was supported by our finding that a lower density of buildings and pedestrian paths (typical for less urbanized areas) was also associated with more cycling. Moreover, we found that a higher density of main roads was associated with more cycling. As a higher density of main roads may relate to lower connectiveness for cyclists, this relationship may be explained by an increased likelihood of detour trips. For example, in order to visit the desired location, children are forced to take detours to safely cross main roads, which in turn may result in increased daily minutes of cycling. The same explanation may be valid for the unexpected positive relationship between pedestrian areas and cycling. As cycling is usually prohibited in pedestrian areas, children may be forced to take detours, which in turn may increase daily minutes of cycling.
4.2. Methodological Considerations
Systematic reviews investigating relationships between the environment and PA urged for objective measurements of both PA and the physical environment [
29,
73]. Consequently, there has been an increase in studies combining accelerometer and GPS measurements, to investigate relationships between an environment and behavior using contemporaneous momentary designs. In this design, objective data on characteristics of the environment (based on the GPS location) and PA intensity are analyzed in simultaneously occurring PA–environment pairs [
28,
34,
41,
44,
74,
75,
77,
78,
79,
80]. However, as PA is a complex interplay between spontaneous and planned behavior, involving memory of PA facilities, time and capacity constraints, social interactions and compensation mechanisms [
81,
82,
83], children’s behavioral response may not occur simultaneously with exposure to PA-supportive attributes in their direct physical environment. Furthermore, studies using contemporaneous designs are vulnerable to selective daily mobility bias [
35,
37].
Quality and specificity of GIS data may depend from one data source to another. As variability in the quality and data structure of GIS data hampers between-study comparisons, researchers are encouraged to provide insight into the various levels of spatial detail underlying their spatial analyses. The present study’s GIS data consisted of high-quality, fine-grained polygons that were routinely validated, but presented a rather rudimentary categorization of the physical environment. Increasing the specificity of GIS-based categorization may pinpoint more precise environmental attributes, but on the other hand may be more vulnerable to erroneous classifications.
Children’s daily activity spaces aggregate to the same neighborhood environment if children live closer to school (see
Figure 2). The percentage of “shared” neighborhood environment (and thus the added value of using individual activity spaces instead of regular residential or school environments) depends on the distance children reside from their school. This means that future studies are encouraged to make informed decisions about children’s daily exposure or accessibility to their environment, based on distances between school and homes, and knowledge of potential other frequently visited anchor points. For example, some studies incorporated the participant’s own perceptions of their daily mobility environments (i.e., neighborhoods) [
39].
In the present study, we have reconsidered children’s environmental exposure in the afterschool time period, following the theoretical framework of Perchoux et al. We created individual daily activity spaces for each child by assigning residential and school environments, in combination with the shortest route between these locations. Multiple subjective experiences of the same attributes of the physical environment may lead to increasing knowledge about the environmental qualities within their activity space [
28,
35,
37], and a sense of belonging to a certain neighborhood [
48]. Environmental PA opportunities within a child’s “belonging neighborhood” may have more influence on PA behavior than more distant PA opportunities [
32]. In addition, social networks also influence daily activity spaces [
48]. For children, the residence and school are prominent locations from which these social networks may arise in the afterschool time period. Therefore, we theorized that afterschool daily mobility due to social opportunities (or constraints) may center around the children’s residence and their school. In conclusion, we propose this operationalization of environmental exposure area to investigate meaningful relationships between environmental attributes that are located within a child’s daily area of influence, and children’s afterschool PA behavior. In our view, this may be an important step forward in understanding causal mechanisms between relative accessibility or opportunity to various characteristics of the objective physical environment and afterschool PA in primary-school children.
4.3. Strengths and Weaknesses
Study strengths are the utilization of GPS devices in order to investigate associations between the objectively assessed built environment features (using GIS) and domain-specific PA, adjusting for meteorological differences and the nested structure of measurement days within children, and children within schools.
We however acknowledge that in some instances GPS data may not indicate the true behavioral context of the location that was visited (e.g., sports grounds visited for sports participation or as a spectator). Although it may be possible to improve validity of the GPS signal by (random) confirmation of the participant, this comes with the cost of participant burden. In addition, a small proportion of the afterschool records (0.04%) were located outside the municipality, of which we had no GIS data available. If these records may have consisted of regular PA participation, our results may have been biased. However, considering the small percentage of records outside the municipality, we expect the magnitude of this potential bias to be small. This study also had some other weaknesses. Although children were instructed to wear the devices during organized sports programs, children sometimes indicated that they removed the devices because they perceived them as uncomfortable. Although fast developing innovations facilitate application of smaller and thus more comfortable devices combining accelerometry and GPS loggers (e.g., smartphone applications), extensive studies are warranted to validate their performance both in PA and location assessment.
Although we have found some interesting and statistically significant associations that were consistent with previous literature, the direct influence attributes of the objectively measured environment on actual minutes PA or active transport was relatively small in our study. Also, investigating PA behavior in a school setting requires advanced statistical multilevel methodologies. In addition, children’s context-specific PA tends to deviate from normal distributions and non-parametric or transformed models are needed. In our study, log-transformations and standardization of independent variables may have hampered straightforward interpretation of these influences. In addition, based on the relatively low ICC for light PA, modeling random variability of the school-level may not have been absolutely necessary.
This study focused on improving the understanding of relationships between children’s daily physical environment and their afterschool PA behavior. However, according to the socioecological model, the influence of the physical environment interacts with other types of the environment such as parental rules regarding children’s PA, or the social influence of peers regarding PA initiatives. In addition, child characteristics such as attitudes or habits also play a major role in determining children’s afterschool PA participation. With the improved understanding about how attributes of the physical environment influence PA behavior, future studies or policy makers are warranted to implement interventions that contain aspects from multiple types and at different levels of the environment [
84].
4.4. External Validity
Although we only investigated data within the municipality border from which GIS data was available, findings of this study may be generalizable to other environments with comparable meteorological circumstances, afterschool time segments, population density and residential density. As the Netherlands also have facilities that support cycling (i.e., separate cycling paths), our results may have limited generalizability to environments with less favorable infrastructure for active transport. In addition, findings of this study may be generalizable to samples with comparable distances from children’s residences to their primary schools, and similar motives for active transport. In addition, our sample was considered highly educated; this could imply that child or parental motives regarding leisure time PA may be different from other samples. In our multivariate analyses, we tried to adjust for social economic status using the neighborhood SES-score that was based on relatively small administrative ZIP-code units (average number of the total households in 2016 was 3371 (SD = 1112.10) per area. As in our sample children did not automatically attend the primary school closest to their residence, there was only a small relationship between children attending the same school and their neighborhood SES-score. In addition, time constraints due to competing activities such as organized sports participation or homework may also be considered in comparing results with other studies. However, as we separately investigated cycling, walking and leisure time, results of leisure time PA may be generalizable to less cycle-friendly environments.