Background
For people of all ages, being physically active is a well-established precondition to health [
1‐
4]. Especially for adolescents, an active lifestyle is necessary for growing up healthy and has been associated with numerous health benefits [
5,
6]. For example, being regularly physically active increases cardiorespiratory fitness and high-density lipoprotein cholesterol and enhances psychological health [
7]. International activity guidelines for adolescents require a minimum level of moderate-to-vigorous physical activity (MVPA) of 60 minutes per day [
8,
9]. Yet, there is scientific consensus that many adolescents do not engage in a sufficient level of physical activity (PA) [
10,
11].
Socio-ecological and socio-cognitive models explain health behaviors such as PA and have also been applied to children and adolescents [
12]. These models postulate an interaction of individual and environmental factors that influence human health behaviors [
13‐
15]. It has been suggested [
16] that environment-changing interventions are important in the field of public health because they may reach a large proportion of the population and achieve sustainable effects. Especially in adolescence, health promotion strategies should consider social and physical environmental aspects [
14].
The concept of social support plays an important role in socio-ecological and socio-cognitive theories and is one of the most important social determinants on health behavior. This concept differs conceptually from social norms, modeling, social influence and social networks [
17]. In adolescents, the main critical distinction in the social support construct has been made between the different sources or providers of support, respectively. Adolescence is the period of the second separation-individualization process where relationships are getting refined and renegotiated and the social environment changes [
18]. While peers may take on a more prominent socially influential role, parents—who were the most important reference persons during childhood—continue to exert considerable influence on adolescents’ behavior. Other reference persons such as grandparents or teachers are less frequent sources of support as children are getting older [
19]. As the most important reference persons, parents and peers are the main providers of supportive behavior for adolescents [
18]. In addition, the findings of previous studies summarized by various reviews suggest positive effects of parental and peer support on adolescent PA [
20‐
24]. A meta-analysis confirms a significant moderate positive effect of parental behavior on children’s and adolescents’ PA behavior [
22].
Furthermore, physical environments provide potential opportunities and barriers to engage in physically active lifestyles [
25]. While various models postulate influences of physical environmental aspects on health behaviors, no comprehensive definition of the physical environment has been established. The behavior specific model, the ecological model of active living [
15] describes the influences of the environment on active living on multiple levels. The perceived physical environment as one predictive dimension for active living includes the aspects safety, attractiveness, comfort, accessibility and convenience [
15].
In the last decade public health research has focused on physical environmental influences in adolescents’ PA [
26]. The state of research has been summarized in several reviews [
27‐
29] which have mostly confirmed the impact of the physical environmental dimensions postulated in the ecological model of active living. The most recent review [
29] found associations between PA and some variables of the perceived physical environment including walking/biking facilities, traffic speed/volume, unspecified traffic safety and access to recreation facilities. Limstrand [
27] identified significant correlations between environmental aspects such as short distance, nearby parks, playgrounds, sidewalks, perceived safety and young people’s use of sports facilities. A meta-analysis on the perceived environment and PA in adults revealed that having shops and services within walking distance explained the greatest amount of variance in PA [
25].
While results of previous studies emphasize the importance of social support and physical environmental factors for increasing PA among adolescents, there is a lack of longitudinal and European studies, respectively [
21,
28]. Appropriate methods for assessing social support and physical environment in adolescents are necessary to bridge this gap. During the last decade, research efforts have increasingly focused on the development of new instruments on this topic [
30‐
37]. However, some of these instruments are very detailed and extensive measurement methods [
32,
33,
36], and most existing questionnaires are in English language and have been developed for use in North-America or Australia [
31,
32,
36]. Because of differences in sports culture (e.g. sports clubs vs. institutionalized sports) and land-use culture (e.g. urban construction), cross-cultural invariance of the instruments can not be expected implicitly [
35]. Thus, applying a questionnaire that accounts for the structural and social conditions of the country where a study will take place is critical [
35].
Therefore, the development of appropriate, valid and reliable instruments on social support and physical environmental correlates of adolescent’s PA is necessary for longitudinal and European studies. The aim of this study was to assess the psychometric properties of a new German short self-report instrument for measuring social support and physical environmental determinants of PA in adolescents. Specifically, we assessed the construct validity, internal consistency, test-retest reliability and predictive validity.
Discussion
The aim of this study was to assess the psychometric properties of a new German short self-report instrument for measuring social support and physical environmental determinants of PA in adolescents. The longitudinal design of Study I allowed for examination of test-retest reliability and predictive validity of the social support and physical environment scales. In addition, the internal consistency and construct validity were assessed.
Social support
According to theoretical considerations [
18], the scale comprised the two factors that characterize the two main providers of social support in adolescents:
parental support and
peer support. The two-factorial structure that was identified in Study I using PCA was confirmed using cross-validation technique by performing CFA with the independent nationwide cross-sectional sample of Study II. The CFA showed good model fit for the two factor model which suggests adequate construct validity of this new social support scale. The fit indices of this study were comparable to those reported by Dishman and colleagues [
30] for a two-factor model of social support that includes parental and friend support.
The factors
parental support and
peer support showed sufficient internal consistency. In comparison with other instruments on social support, the current scale featured comparable internal consistency and test-retest reliability. The alpha estimates were slightly higher than those reported for some other instruments of social support for children [
32,
34] and adolescents [
33,
35]. However, Pirasteh and colleagues [
37] found a higher internal consistency of a five-item friend support scale in a validation study of a psychosocial determinants questionnaire in Iranian adolescent girls (α = 0.77). Satisfying internal consistency of the brief factor scales suggests that there was a good balance between scale length and reliability. Hence, this scale might be suitable in large-scale health surveys that require short scales.
In this study, the test-retest reliability of the latent factors was moderate to good indicating test-retest-reliability during an interval of one week. In addition, test-retest coefficients were comparable to other instruments that were developed for children and adolescents [
30‐
32,
37]. The correlation coefficients were higher for
parental support than for
peer support. This finding is consistent with the results of a study by Dishman and colleagues [
30] that test-retest reliability for parental support was higher than that for friends support over a two-week interval. This result indicates that either the reliability of the scale for parental support is higher than that of the scale for peer support or that the construct of parental support is more stable over time than the construct of peer support. This discrepancy in reliability between factors may be caused by the fact that the items on
parental support partly relate to more general perceptions (e.g. “How important is it for your parents that you do sport?”) than the potentially more variable aspects of items on
peer support (e.g. “How often do you do sport with your friends?”).
Both factors of the social support scale correlated positively with all PA measures (objective MVPA, self-reported recent MVPA and self-reported habitual MVPA) suggesting predictive validity of the instrument. Adolescents who received greater social support from their parents and their peers were more likely to be physically active (measured objectively as well as self-reported) than those who received less social support. The correlation coefficients in this study are comparable to those reported in other studies on the relationship between social support and adolescent PA [
22,
60]. The highest longitudinal correlations of both parental support and peer support were observed with self-reported habitual MVPA (r
parental support
= 0.30 and r
peer support
= 0.31, respectively). The results of this study suggest that social support of parents and peers predicts rather habitual MVPA than recent MVPA (measured objectively as well as self-reported). Because the social support scale contains habitual variables of supportive behavior (e.g. “How often do your parents watch you doing sport?” – “never” to “always”), the higher correlation coefficients may have been caused by the conceptual conformance of the measures of habitual MVPA and habitual social support.
Physical environment
The CFA showed a good fit for the four-factor model and provides evidence for the construct validity of the physical environment scale. Fit indices were comparable with the indices identified by Ommundsen and colleagues [
35] for a model of physical environment containing the three factors opportunity, facility and license. The variability of factor structures across physical environmental questionnaires may be related to the diversity of the construct and its definitions and the lack of behavior-specific theories on physical environmental influences on PA behavior [
15,
61].
The current scale on the physical environment has a four-factor structure with the latent factors convenience, public recreation facilities, safety and private sport providers. The items that were developed for measuring accessibility of recreation facilities built upon two separate factors that distinguish between private sport providers and public recreation facilities. These factors seem to describe different aspects of the physical environment that may offer different opportunities for adolescents for being physically active. It is possible that these items load on two separate factors because the incidence of these facilities is traceable to the population density in the residential area. In Germany, private sports providers are more likely to be located in the center of cities whereas public recreation facilities such as playgrounds or soccer fields may be located in neighborhoods with lower population density and outside of city centers.
The internal consistency of two of the physical environment factors was low which may be related to the brevity of the factors as only two-item measures. However, other validation studies of physical environment instruments on different target populations also showed low internal consistency on the factor level [
32,
34,
35,
62], which may be related to the heterogeneity among the physical environment items. Some characteristics of the physical environmental that are relevant for adolescent’s PA may be independent of each other. For example, there is no association between the accessibility of different types of recreation facilities such as soccer fields or playgrounds. Therefore, the physical environment construct may not be appropriately assessed by internal consistency as a measure of reliability [
63].
The results of this study showed satisfying one-week test-retest reliability that was comparable to that of other physical environment instruments for children or adolescents [
31,
32,
36]. Nevertheless, it should be noted that most of the dimensions of the physical environment measured here and in the instruments described above are relatively stable local conditions (e.g. access, convenience). Consequently, the limitations of test-retest reliability are likely related to changes in response behavior of participants. Adolescents seem to have varying perceptions of the physical environment or are not able to describe aspects of their physical environment appropriately. Rephrasing items and providing a description of the area of interest (e.g. neighborhood) might improve the participants’ understanding of the questions and consequently increase reliability.
The predictive validity of the physical environment factors could not be confirmed. The correlation coefficients of relationships between the physical environment factors and overall PA indices were in most cases not statistically significant. Only the factor
convenience positively correlated with self-reported habitual MVPA in adolescents. This result supports findings of Rosenberg and colleagues [
36] who reported correlations between land use mix-access (which is a scale of 6 items on access of stores, places and transit stops, parking opportunities, and walking barriers) and routinely walking to shops and routinely walking to parks in adolescents. There is inconclusive evidence on predictive effects of physical environment measures on adolescents’ PA [
28] and the concept that the environment has direct effects on behavior is contentious [
64]. Hence, the physical environment may be a prerequisite and a moderator of the relationship between psychosocial correlates such as social support and PA behavior in adolescents without a direct effect [
65]. In addition, the influence of environmental factors on behavior may develop slowly [
64]. The timespan between the measurement points of physical environmental variables and PA behavior in the current study was one week. This period may be too short to detect physical environmental effects. Hence, future research should focus on longitudinal data with longer between-test intervals. Finally, the physical environment should be investigated as a moderator of associations between other PA correlates and adolescents’ PA behavior [
65].
Strength/ limitations
The strength of this study was that the questionnaire was developed based on theoretical considerations and built upon previous instruments. It was cross-validated with two independent study samples of 9 to 17 years old girls and boys. Nevertheless, this study had some limitations. For instance, our considerations were based on two studies with suboptimal sample size. The results of test-retest correlations should be carefully considered because these results are based only on the sample of the Study I with 196 participants. In addition, Study I included a sample of adolescents from only one public school, and hence these students may have had a somewhat similar physical environment. The variability in physical environment variables may have been higher if participants had originated from different areas.
In addition, the instrument has not been validated in a comparison with a 'gold standard' because to date a 'gold standard' for measuring social support or physical environment in adolescents is lacking [
34]. Different methods are available for objectively examining the physical environment. However, objectively measured and subjectively measured physical environments do not equally correlate to PA in adolescents [
29] and differ by definition, because the objectively measured environment is the actual environment while the subjectively measured environment is the person’s perception of environmental attributes. Hence, future studies could validate the physical environment scale by comparison with objective measurement methods and the social support scale by comparison with proxy-report methods of social support (e.g. parent-report).
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
AKR was responsible for the overall conception and design of this manuscript, statistical analysis and interpretation of data. DJ designed Study I, contributed to the design of Study II, collected the data of Study I, contributed to the statistical analysis and provided edits to the manuscript. FM and NM revised the manuscript. AW designed Studies I and II. All authors read and approved the final manuscript.