Teun Remmers

Afterschool PA and the built environment using GPS, GIS and accelerometers | 129 We exported PALMS datasets separately for each school, and integrated these separate datasets into a PostgreSQL database (http://www.postgresql.com) . We first deleted data from pre-selected participants with insufficient accelerometer wear time. Subsequently, based on reports from the school principal, we performed queries to identify relevant time-segments based on individual school's schedules, and these separate datasets were subsequently merged into one time-segmented dataset (i.e. afterschool and the last 60 minutes in-school datasets). Time-segmented datasets, containing both accelerometry and GPS data, were integrated into ArcGIS version 10.4.1 (ESRI, Redlands, California). Additionally, we overlaid GIS-data from the municipality of 's-Hertogenbosch, which contained detailed geospatial information regarding for example buildings, roads, land-use, and facilities. These time- segmented datasets consisted of millions of records. We applied basic validation rules by deleting records (i.e. 10-second epoch measurements) with incidental missing accelerometer data, or records that were located outside the municipality study-area. In accordance with previous studies investigating afterschool time-segments (5, 15, 23), we ensured reliability of the time-segment by only selecting records from days with > 4 hours of valid wear time in the afterschool time-segment (see Figure 1). Spatial analyses to validate afterschool time-segments Subsequent spatial analyses were conducted in order to 1) filter afterschool leisure time context from other afterschool contexts, and 2) define children's exposure to features of their multi-place environment (i.e. school, residence, and daily transport environment). We geo-located the residential and school buildings and accompanied geo-referenced parcels for each participant's residence and school, using ArcGIS geocode-functionality and the municipality's address-database. Subsequently, we validated children's precise afterschool time-segments by calculating the daily percentage of GPS points that were within a distance of 10 meters to their school's geographic parcel during the last hour of school time. A buffer of 10 meters was chosen to account for potential imprecision of the GPS logger (18, 49, 50). Time-segments from children with < 80% of their GPS points within the school-parcel during this last hour of school time, were carefully inspected in individual ArcGIS maps, and the majority of the accompanied days were deleted from the afterschool dataset (see Figure 1). Nevertheless we retained such data in some instances because they represented short walking trips outside the school-parcel in order to participate in physical education classes at neighbouring sports halls or fields, or slight erroneous deviations of the GPS-signal due to urban canyoning (50).