Background

The physical and psychological health benefits of being physically active are well known [1]; however secular declines in children's physical activity are evident [2]. In order to develop effective strategies to arrest these declines, understanding factors that influence children's activity is important. Ecological models posit that there are multiple levels of influence on physical activity, including intrapersonal, social and physical environmental factors [3]. Intrapersonal factors have been widely investigated, and although measures are well conceptualised and established, these factors appear only to explain part of the variance in children's activity [4].

The home and the local neighbourhood are two important settings for children's physical activity. Studies among adults have commonly defined the physical environment in terms of accessibility and availability of destinations and facilities, aesthetics, and safety [5]. Few studies have attempted to define, operationalise or measure these among children, and none have examined the psychometric properties of such measures. Children's social environment has typically been defined in terms of the proximal social influences (e.g. parental support), and measures are well established [6]. However, with few exceptions [7], almost no studies have assessed children's broader social environment (e.g. the local neighbourhood).

The aim of this study was to develop and test the reliability and acceptability of an instrument to assess a broad range of environmental perceptions that might predict physical activity among children, particularly the physical and social environments at home and in the neighbourhood.

Participants and methods

A convenience sample of Grade 5 and 6 children attending a metropolitan Melbourne elementary school participated (21 boys, 18 girls, aged 11.1 ± 0.7 years).

Measures

Children were asked to report their perceptions of the physical and social environments at home and in the neighbourhood. All items are shown in Table 1.

Table 1 The psychometric properties and the number of missing responses to items examining children's perceptions of the physical activity environment

Items examining the size of the yard (1 item) and physical activity opportunities at home (16 items) were adapted from a published parent proxy-report survey examining children's leisure activities [8]. Responses were either dichotomous, or on a seven-point scale (later dichotomised), indicating whether the child did or did not have these items at home. A summed score indicating the number of opportunities for physical activity at home was created (range 0–16).

Ten items assessed the home social environment. Six items on family physical activity were adapted from a parent proxy-report survey [8]. Responses ranged from 'never' to 'daily', and were collapsed into no (never) or yes (all other options) response options. Four items on encouragement and support for activity were derived from previous study of psychosocial influences on children's physical activity [6]. Response options ranged from 'never' to 'daily', which were collapsed into yes/no categories. A scale evaluating the overall home social environment was generated by summing the positive (yes) responses to each of the 10 items (range 0–10), with a higher score indicating a more positive perceived home social environment.

Children's perceptions of the neighbourhood physical environment (access to destinations, aesthetic, and safety characteristics) were assessed with 29 items. Access (by active transport) to 15 neighbourhood destinations was assessed using items from a study examining Australian children's leisure activities [8]. Responses were on a seven-point scale with options ranging from never to daily as well as 'it's not within walking/cycling distance'. Based on this last option, responses were collapsed into two categories; 'can't access' and 'can access'. A score indicating the total number of neighbourhood destinations accessible by active transport was created by summing the positive responses to each item (range 0–15).

Fourteen items were used to assess children's perceptions of the neighbourhood aesthetic and safety characteristics, adapted from questions used in a study among adults [9]. Each response was dichotomised ('yes', or 'no/don't know'). Overall aesthetic and safety scales were developed by reverse coding negatively-worded items and summing the 'yes' responses. A higher score on these scales indicated more positive perceptions of neighbourhood aesthetics (range 0–5) and safety characteristics (range 0–9).

Questions assessing children's perceptions of the neighbourhood social environment were developed specifically for use in this study. Seven items (dichotomised to 'yes' or 'no/don't know' response options) assessed children's perceptions of the social connections in their neighbourhood. An overall neighbourhood social environment scale was generated by summing the positive responses, with a higher score indicating more positive perceptions.

Survey administration

All questions were completed on two occasions (Times 1 and 2) up to nine days apart, during class time with researchers and the classroom teacher present.

Statistical analyses

The survey was assessed for acceptability by calculating the proportion of missing responses to each item. Intra-class correlations (ICC's) were used to examine similarities between responses for continuous variables [10], while the Kappa statistic (κ) and percent agreement between responses (the proportion of participants grouped within the same response category for test and retest), were used to determine the repeatability of the categorical variables. Adequate test-retest reliability was defined as ICC ≥ 0.75 for continuous variables [11], and the strength of agreement between responses to categorical variables using κ was defined as poor to fair (0.00–0.40), moderate (0.41–0.60), substantial (0.61–0.80) and almost perfect (0.81–1.0) [12]. Significance (p-value) was set at 0.05, which indicated if the strength of agreement between responses was statistically significant according to κ. Percent agreement was also calculated since the stability of κ is dependent on the prevalence of responses to a question and when there is a low prevalence of one response, κ may be low [13]. Percent agreement values greater than 66% were classified as fair [14].

Internal reliability analyses (Cronbach's α) were performed on scaled responses to multiple items (e.g. the neighbourhood safety scale). Internal reliability was deemed acceptable if Cronbach's α was greater than 0.6 [11].

Results

The survey appeared acceptable for children to complete, as they asked few questions, did not report any difficulties understanding the questions and only 11% of responses were missing (Table 1). The items with the most missing responses were those asking about physical activity with family members, particularly with grandparents (13/39 missing), suggesting these items may be difficult for children to complete (possibly reflecting the fact that some children may not have grandparents or siblings, or recall difficulties). On balance, the survey appears to be feasible and acceptable for children.

Results from test-retest reliability analyses are shown in Table 1. The home physical environment items (17 items) showed at least moderate reliability with the exception of two items (having a covered area outdoors and having active toys). All items showed at least fair agreement between test and retest. The physical activity opportunities at home score showed acceptable repeatability (ICC = 0.80, CI = 0.55–0.91). Of the 10 home social environment items, one (receiving offers for physical activity from family members) showed poor reliability according to κ (0.16); however all items assessing the home social environment showed at least fair agreement (69% or better) from test to re-test. The home social environment scale also showed adequate test-retest reliability (ICC = 0.84, CI = 0.56–0.94) and good internal reliability (Cronbach's α = 0.73).

The 29 neighbourhood physical environment items showed at least moderate reliability, with the exception of two access to destinations items (can access the post box and public transport), one aesthetic item (there is lots of litter and rubbish) and one safety item (worried about older kids hanging around), which showed poor κ values. All items showed at least fair agreement (greater than 75%) between test and retest. The total number of accessible destinations in the neighbourhood score showed excellent test-retest reliability (ICC = 0.84, CI = 0.66–0.93) and the aesthetics and safety scales showed acceptable test-retest reliability (ICC = 0.72, CI = 0.45–0.86; ICC = 0.88, CI = 0.76–0.94 respectively). Internal reliability analyses showed that while the safety scale was acceptable (Cronbach's α = 0.65), the aesthetics scale was not (Cronbach's α = 0.43).

Of the seven items examining the neighbourhood social environment, only one showed less than moderate test-retest reliability (having lots of other children around to play with). All items showed at least fair agreement (68% or better) from test to retest. The neighbourhood social environment scale showed adequate test-retest reliability (ICC = 0.92, CI = 0.84–0.96) and good internal reliability (Cronbach's α = 0.77).

Although several items showed poor to moderate test-retest reliability according to κ, all items in the questionnaire showed at least fair agreement (greater than 68%) between test and retest.

Conclusion

This study is the first to assess the test-retest reliability of a broad range of questions examining children's perceptions of their physical and social environment at home and in the neighbourhood. The questions were mostly found to be acceptable and appropriate for this age group. Although several items did have poor κ values, all items showed at least adequate agreement between test and retest. Further research may be required to design a questionnaire using language that may be more appropriate for children, to better define and operationalise the neighbourhood aesthetic environment for children, and to validate such measures of the environment. The results of this study provide some support for the use of these items when examining environmental determinants of physical activity among children.