Assessing the effect of up-grading playgrounds on children’s physical activity

Playgrounds are facilities that may increase children’s physical activity. Territorial authorities provide playgrounds in public parks. The link between children’s needs for physical activity and subsequent provision of play equipment was acknowledged in the Dunedin City Council’s (DCC) 2006 Play Strategy. Implementation of the Strategy through planned up-grades in a spatially-defined area provided an opportunity to test whether children’s physical activity increased with improved play facilities. Two playgrounds were up-graded by the DCC in early 2008 providing the opportunity to conduct a natural experiment study known as LOCATE (Location of Children's Activity in Their Environment) set in Dunedin, New Zealand (NZ). The Location Of Children’s Activity in Their Environment Study (LOCATE) was a natural experiment, set in Dunedin, New Zealand (NZ). At baseline (October - December 2007), 184 children, aged five to ten years old, were recruited from participating schools in intervention and control communities with comparable demographic and spatial characteristics. There were 156 participants at follow-up (15% attrition). There were two phases: a baseline assessment and the follow-up assessment, with 12 months between each assessment. To obtain parents/guardians perceptions of their neighbourhood and other information associated with physical activity, a self-administered questionnaire was developed and tested for reliability in a pilot study. It was mailed to each participant’s home when they began each assessment phase. Participants’ heights and weights were measured, and age- and sex-standardised BMIs determined. Mean Total Daily Physical Activity (TDPA), measured by an Actigraph GT1M accelerometer, was matched with location data collected at one minute intervals using a GlobalSat DG-100 Global Positioning System Data Logger. A regression model was used to compare the log-transformed mean TDPA at follow-up, with baseline scores included as a covariate. All models controlled for repeated measures and clustering by schools by using random effects. Univariate screening was carried out using p