Research design and participants
The present study is an integrated part of the KIDSCAPE II project that explores the co-beneficial impact of the outdoor environment on Swedish school children’s health at compulsory school. A repeated measurement study design was used to elucidate the characteristics of school environment that trigger spontaneous physical activity during fall, late winter, and late spring. The study was approved by the Regional Ethics Committee of Stockholm.
The sample was drawn from available 2nd, 5th and 8th graders (aged 7–14 years) at four municipal schools in mid-southern Sweden. The schools were selected considering differences in the overall layout of the outdoor environment, taking into consideration the total size of the school yard, its topography, surfaces with woodland, trees and bushes, and the presence of ball play areas, play equipment, and scheduled outdoor education. Further, the schools were selected to reflect the socio-economic composition of the municipalities of the whole country, with a majority of the population living in medium-sized cities outside or on the outskirts of a metropolitan area (Statistics Sweden). The ISCO code was applied for socioeconomic classification (European socio-economic classification, ISCO, 1988).
The parents of 259 pupils attending the selected schools were asked to let their children participate in the study. Both parents and their participating children signed a written consent to participate in the study. After obtaining permission, the school management, the pupils, teachers and parents received relevant and detailed information at an early stage and were given the opportunity to discuss the study. A sample of 196 (76%) agreed to participate.
Procedures, data collection
In selecting school outdoor environments we considered all typical behavior settings and combinations of behavior settings that are representative for Northern and central European school outdoor environments. For a varied sample in terms of design and overall layout the school yards were selected in collaboration with a landscape researcher. To provide an estimate of the total school yard areas (m
2), ball areas (soccer fields, basket fields, markings e.g.), green areas (woodland, grass, trees and bushes) and areas varying in use depending on season (used play area), the Google™ Earth Pro (GEP) software using aerial pictures of the playgrounds and the polygon measurement tool, described by Ridgers et al. 2010, were applied at each one of the four schools [
26]. Topography (sum of hills and slopes), vegetation (amount of trees, bushes and grass), fixed equipment (swings, slides, climbing frames) and playground markings were inspected and counted by ocular inspection. The computed amount of space that the pupils used during the various seasons was based on the maps on which all the pupil’s positions during recess were marked. This patterned use of space was made up by different behavior settings (hereafter “used play space”). The percentage of used play space was obtained by dividing used play space by available school yard area. To calculate the variations between school, grade and seasons of the average play space (m
2) used per child, the measured used play spaces were divided by the number of pupils for each grade.
Weather conditions were observed and ranked at AM and PM every day of fieldwork according to the following weather index, 1 = clear sky, 2 = partly cloudy, 3 = white cloudiness, 4 = grey cloudiness, 5 = precipitation. Temperature recordings were obtained from the Swedish Metrological and Hydrological Institute (SMHI).
Prior to data collection a questionnaire was filled in by the parents of the 2nd graders and by the 5th and 8th graders themselves about family conditions (members of adults in the household, parents’ education and occupation, immigrant and/or indigenous, diseases etc.).
Data collection was carried out during five consecutive school days on three occasions during one school year, fall-September 2012, late winter-March 2013, and spring-May 2013. During each measurement period the pupils filled in a diary about leisure-time PA, considering other possible confounders (e.g. feeling unwell, sports etc. outside school). The pupils were weighed (scale: Beurer GS 27 Happystripes, CE Utrecht) and waists and heights measured using a measuring tape and a stadiometer (Seca 217, UK Birmingham) (Table
1).
Table 1
Descriptive data of the participants by school, grade and sex
2:nd grade (n = 74) | Height (cm) | 136,5 | ± 7,5 | 131,6 | ± 7,8 | 132,7 | ± 4,8 | 135,2 | ± 5,7 | 133,9 | ± 6,7 | 133,6 | ± 6,6 |
Weight (kg) | 32,8 | ± 6,7 | 31,6 | ± 6,2 | 29,0 | ± 4,2 | 30,9 | ± 3,9 | 31,5 | ± 6,3 | 30,0 | ± 4,0 |
Waist (cm) | 65,2 | ± 8,3 | 64,6 | ± 6,4 | 61,0 | ± 3,8 | 60,2 | ± 5,6 | 63,4 | ± 7,0 | 61,6 | ± 5,0 |
BMI (kg/m2) | 17,5 | ± 2,1 | 18,1 | ± 2,1 | 16,4 | ± 1,6 | 16,9 | ± 1,5 | 17,4 | ± 2,2 | 16,8 | ± 1,4 |
5:th grade (n = 85) | Height (cm) | 153,3 | ± 7,1 | 151,0 | ± 7,6 | 155,6 | ± 9,8 | 150,1 | ± 6,7 | 151,8 | ± 6,4 | 152,9 | ± 9,4 |
Weight (kg) | 46,3 | ± 10,4 | 45,5 | ± 12,0 | 45,0 | ± 11,6 | 43,6 | ± 7,2 | 44,3 | ± 8,4 | 45,8 | ±12,2 |
Waist (cm) | 70,3 | ± 11,1 | 70,3 | ± 10,0 | 67,3 | ± 8,5 | 69,5 | ± 6,8 | 69,8 | ± 8,8 | 69,0 | ± 9,6 |
BMI (kg/m2) | 19,6 | ± 3,7 | 19,8 | ± 3,5 | 18,4 | ± 3,2 | 19,3 | ± 2,5 | 19,1 | ± 2,8 | 19,4 | ± 3,6 |
8:th grade (n = 25) | Height (cm) | 167,4 | ± 4,7 | 169,2 | ± 7,8 | 171,0 | ± 5,5 | 187,0 | ± 10,5 | 175,5 | ± 8,6 | 165,7 | ± 4,9 |
Weight (kg) | 64,4 | ± 9,0 | 69,2 | ± 15,9 | 59,4 | ± 7,1 | 74,4 | ± 7,7 | 65,7 | ± 12,2 | 63,7 | ±11,0 |
Waist (cm) | 82,5 | ± 6,8 | 81,7 | ± 16,2 | 73,4 | ± 6,2 | 73,0 | ± 7,8 | 75,1 | ± 10,5 | 80,7 | ± 9,7 |
| BMI (kg/m2) | 23,0 | ± 3,6 | 24,0 | ± 5,1 | 20,3 | ± 1,8 | 21,2 | ± 0,3 | 21,3 | ± 3,6 | 23,1 | ± 3,2 |
PA was measured using hip-mounted accelerometers (Actigraph GT3X+ Activity monitors, US Pensacola) which enable time-stamped analysis of duration, intensity and location of activity in terms of indoor or outdoor activity (in this case important for the elimination of indoor stay). The accelerometers were activated on the first day of fieldwork and analyzed after the five consecutive days of each data collection period. Epochs were set at 10 seconds for detailed PA data [
35,
36]. We applied previously validated (Evenson et al., 2008) and recommended cut-points set to <17 counts/10s for sedentary, 17 – 382 counts/10s for light PA, 383 – 682 counts/10s for moderate PA and >682 counts/10s for vigorous PA to estimate time spent in sedentary, light-, moderate-, and vigorous-intense PA in pupils aged 7–14 years [
37]. Additionally, the built-in light sensor (Actilux) of the accelerometers for the registration of ambient light supplemented the separation of outdoor from indoor time registered by ocular observation. The sensitivity of the light sensor is 74% and specificity 86% [
38], and valid for separating outdoor time in free living children 3–5 years [
39]. Ambient light data were sampled and stored to memory at a 1 Hz rate. The downloaded data file was converted into an accumulated *.agd format with epoch lengths of 10 seconds, and average lux values for each epoch. In the Actigraph GT3X+ manual recommended values for indoor light are 1 – 500 lux (1 Hz) and for outdoor light >100 lux (1Hz). The cut point for lux value explaining outdoor time was set to >130 lux for average data of 10 second epochs, after comparing with observed values for in- and outdoor data, which were fairly consistent with newly presented validations of the GT3X+ light sensor [
38,
39]. The accelerometers, attached to elastic belts, were put on upon arrival and removed at departure from school. The observer, one for each class, made sure that the accelerometers by were correctly mounted, worn outside and tightly strapped to the clothes. Each observer tracked the pupils of the designated class, marking the locations and activities for girls and boys separately on the school yard during recess. Sufficient data were obtained for 189 (88 girls) pupils, after the first, for 184 (85 girls) pupils after the second, and for 180 (86 girls) pupils after the third and last measurement period.
Evaluation and statistical analysis
Data were analyzed using SPSS for Windows (22.0). School day accelerometer counts were condensed into time spent in MPVA during outdoor stay (observed and measured outdoor time) and divided by total outdoor time to provide percent of MPVA (%MVPA) of outdoor stay time. The mean numbers of measured schooldays per child were 4.3, 4.7 and 4.2 in September, March and in May respectively. A one-way ANOVA test showed no differences between measured days (1–5) and daily minutes in MVPA, p > 0.05, thus all pupils with at least one day of data were included in the analysis and the pupils’ average daily outdoor %MPVA used as the outcome variable. After scheduled lessons, the 2nd graders spent the rest of the day at the after-school center situated at the school premises until their parents or older siblings fetched them. As differences in their MVPA between recess during the scheduled part of the day and outdoor stay at the after-school center were non-significant during all seasons (paired t-test, p > 0.05), and the children were confined to exactly the same outdoor environment, MVPA data were collapsed for the whole day.
Independent t- tests were applied to identify differences in physical activity means. Paired t-tests were used to evaluate the differences between seasons and outdoor physical activity. Intra-class correlation coefficients were calculated to estimate the influence from factors at school level upon %MVPA during outdoor stay. For bivariate analyses of confounder variables’ association vs. %MVPA the t-test, Spearman’s rho correlation coefficient and Pearson’s correlation coefficient were applied depending on which test was appropriate. Significantly associated variables were jointly tested in linear mixed model analysis, i.e. entered into the equation and sequentially removed (criterion for removal p ≥ 0.05) by highest p-value. Potential confounders related to PA and socio-economic status were non-significant and thus removed.
The choice of modeling was a linear regression model with random effects allowing for the clustering of measures within subjects, along seasons. With only three repeated measures within one school year, seasonality could play an important role in the autocorrelation patterns of the within-subject measures. A larger number of repeated measures per subject would have been necessary in order to correctly estimate the outcome pattern of autocorrelation. The matrix of variance-covariance for the correlated measures was therefore regarded as unstructured. The model specification regarded indicator variables for gender, for season when the measure was taken as well as for the school and grade the subject attended.