Teun Remmers

General discussion | 183 Recently, researchers have applied novel methodologies to estimate the context by combining accelerometers with self-reported diaries (12, 41), developing algorithms recognizing the context in triaxial accelerometer patterns (42-44), time-segmented analyses of accelerometer data (11, 45, 46), or even combining accelerometers with GPS (18, 19, 47). An important first step towards more detailed context-specific PA analyses is to perform time-segmented analyses of accelerometer data based on children's school schedule, for example as presented in chapter 7. Namely, as children's exposure to the environment during school hours is largely limited by regulations, schedules and geographic location of their school, conceptually-matched relationships between environmental attributes and PA are likely to be found in the afterschool period or during weekends (10, 11, 45, 48). In this thesis, chapters 7 and 8 addressed environmental determinants of afterschool PA in children. A second step towards context-specific PA lies in the combination of accelerometer and continuously logged GPS data. With the additional knowledge of the geographical location at which PA occurs, these data can be merged into Geographical Information Systems (GIS), which hold extensive data about characteristics of the built environment. In this way, the context of specific PA patterns (or sedentary behaviors) can be inferred from the contemporaneously-measured geographical location or travel speed (49-52). In this thesis, we have applied these recent methodological advances while investigating determinants of afterschool PA (chapter 8) and PA patterns in transition from primary to secondary school (chapter 9). These studies used the Personal Activity and Location Measurement System (PALMS), developed by the University of San Diego, to merge both measurements and to perform basic validation analyses (50). Several other studies developed their own methods for combining GPS and accelerometer data, which potentially hampers meaningful comparisons between studies (18, 19, 53-58). Although important steps have been taken to improve processing and analyses of the data (50-52, 59-61), and interpretation of the results (34, 49), GPS-methodologies are still developing. For example, future studies are encouraged to investigate the use of 1) accounting for measurement errors in the GPS location (59), 2) protocols for imputing missing GPS data points (61), and 3) handling accelerometer non-wear time. In addition, although analyses of both the behavioral- and spatial pattern add another layer of variability (and thus complexity) to the data, to date studies have based their wear time criteria on accelerometer-only criteria (33). When analysing combined GPS and accelerometer data, future studies may investigate the use of adapted criteria for daily wear time and the minimum number of days in which children should participate, to achieve sufficient intra- individual reliability. Future studies are encouraged to replicate methodologies in chapters 7, 8 and 9 while investigating the time-segmented association with contextually matched environmental determinants. In this way, relationships between objectively measured environmental determinants and PA may be further unravelled. However, when analyzing smaller time- segments, researchers should acknowledge its contribution to daily PA and potential health benefits in the future (i.e. 'clinical relevance'). Potential contributions to daily PA