Development of a measurement approach to assess time children participate in organized sport, active travel, outdoor active play, and curriculum-based physical activity

Background Children participate in four main types of physical activity: organized sport, active travel, outdoor active play, and curriculum-based physical activity. The objective of this study was to develop a valid approach that can be used to concurrently measure time spent in each of these types of physical activity. Methods Two samples (sample 1: n = 50; sample 2: n = 83) of children aged 10–13 wore an accelerometer and a GPS watch continuously over 7 days. They also completed a log where they recorded the start and end times of organized sport sessions. Sample 1 also completed an outdoor time log where they recorded the times they went outdoors and a description of the outdoor activity. Sample 2 also completed a curriculum log where they recorded times they participated in physical activity (e.g., physical education) during class time. Results We describe the development of a measurement approach that can be used to concurrently assess the time children spend participating in specific types of physical activity. The approach uses a combination of data from accelerometers, GPS, and activity logs and relies on merging and then processing these data using several manual (e.g., data checks and cleaning) and automated (e.g., algorithms) procedures. In the new measurement approach time spent in organized sport is estimated using the activity log. Time spent in active travel is estimated using an existing algorithm that uses GPS data. Time spent in outdoor active play is estimated using an algorithm (with a sensitivity and specificity of 85%) that was developed using data collected in sample 1 and which uses all of the data sources. Time spent in curriculum-based physical activity is estimated using an algorithm (with a sensitivity of 78% and specificity of 92%) that was developed using data collected in sample 2 and which uses accelerometer data collected during class time. There was evidence of excellent intra- and inter-rater reliability of the estimates for all of these types of physical activity when the manual steps were duplicated. Conclusions This novel measurement approach can be used to estimate the time that children participate in different types of physical activity. Electronic supplementary material The online version of this article (10.1186/s12889-018-5268-1) contains supplementary material, which is available to authorized users.

This code is meant to serve as a starting point for researchers to replicate two algorithms that we describe in the above paper -one for deriving estimates of outdoor active play and the other for deriving estimates of curriculum-based physical activity. To use this code you will need to have an intermediate understanding of SAS syntax and be familiar with methods that are commonly used when collecting, managing, and analyzing accelerometer and GPS data in children.
2.3 Identify time spent indoors and outdoors. Obtain GIS data from the geographic region that you collected accelerometer and GPS data. Combine the cleaned, merged accelerometer and GPS data with building footprint information to determine whether a GPS point is indoors or outdoors using ArcGIS or similar GIS software. You may want to develop your own approach to clean this indoor/outdoor variable for GPS jitter and GPS drift. An alternative approach for this step would be to use a satellite signal-to-noise ratio built into the GPS device to determine if each of the GPS data collection points occurred while the participant was indoors or outdoors.
2.4 Import and concatenate all of these files into SAS. Flag all the relevant times that participants recorded on their activity logs -including time spent sleeping, participating in organized sports and school schedule information (day of the week, start and end times of school and recess).
2.5 Identify accelerometer non-wear time and categorize movement into intensity categories. This will be specific to the device used, wear location, epoch length, and population being studied.
The algorithm provided below is meant as a starting point for researchers to develop their own unique approach to predicting time spent in curriculumbased physical activity. It is provided as a guide, and by no means will it provide a valid estimate of time spent in curriculum-based physical activity in other studies/samples. 6.1 Start with dataset derived in step 2.6. Remove all data except for epochs that occur during school hours, but not during recess time (referred to here as data1_PE). Develop and apply an algorithm to predict time spent in curriculum-based physical activity. The algorithm developed for our paper is below.
*Set-up an initial prediction variable Physed_pred = initial prediction variable. Is equal to 0 if curriculum-based physical activity is not likely to have occurred at this time and is set to 1 if curriculum-based physical activity is probable during this time.