Background:
Increasingly the combination of GPS, GIS and accelerometry is being employed in research on physical activity and the environment. GPS has potential as an objective and accurate method of collecting data on journeys and places used that does not suffer some of the drawbacks of alternative methods. Several studies have used this combined technology to research travel to school or work in relation to physical activity. None have reported yet on its use in researching greenspace and physical activity among adolescents.
Objectives:
This paper reports on methodological issues encountered in using GPS-enabled mobile phones, accelerometers and greenspace map data provided by a Local Authority in a study with Scottish adolescents. This was part of a wider research project aimed at exploring the role of greenspace in promotion of mental well-being and physical activity in Scottish adolescents.
Methods:
In May/June 2010 a sample of 35 13 and 15 year old Scottish adolescents from one Local Authority were loaned a GPS-enabled Blackberry (8900) and an AM7164 uniaxial accelerometer for 6 days. They were asked to wear the accelerometer from when they got up to when they went to bed and to take the Blackberry with them wherever they went.
GPS points were recorded initially only via “live” tracking. However, problems were identified with this system and so an additional GPS data collection method was used. An open source piece of software called “GPSlogger” [emacberry.com] was downloaded to each phone which enabled logging of data onto the phones themselves. GPS and accelerometer readings were set to record every 30 seconds.
Data handling and Analysis
GPS data was downloaded as a .csv file. It was processed using Microsoft excel, PAWS and ArcGIS. Data was trimmed to retain only timestamp, longitude and latitude data. Local Authority greenspace data was used to assign a land use code to the GPS data using ArcGIS. GPS and accelerometer data were merged by matching the time and date by 30 second epochs and then trimmed to retain “leisure time/school run”. The merged and trimmed data was then returned to ArcGIS for checking of spurious and missing GPS data before further analyses.
Results:
A number of methodological issues were identified. It was not possible to identify missing data using the live tracking GPS system. Using GPSlogger resolved this problem. When merged with the accelerometer data and using ArcGIS it was clear when and where missing data had occurred.
Unexpectedly several participants managed to accidentally switch off their GPSloggers within a few hours of receiving the phones. Initially the folder containing the software was hidden but this failed to prevent the problem. An application was then found that enabled password protection of GPSlogger.
Phones needed charging every night which may have been expected to reduce compliance, however, in general this was very good and overall acceptability of the methods was high. The use of Blackberrys as the GPS monitor was motivated by concerns over recruitment and the need to use something that was attractive to teenagers and likely to maintain compliance. This appeared to work well.
The memory capacity of the phone memory cards was good (1GB). The number of logged points ranged from around 6500 to over 13500 points.
Approximately 25% of trimmed GPS data was missing and 4% of GPS data was lost through duplication from rounding the time values to match the accelerometer format. GPS locations adjacent to the missing data were tagged to enable visualization of missing data in ArcGIS so that the pattern of missing data could be analysed. It is suspected that most will occur when participants were in buildings. Duplicated data will also be checked to see if deleting them affects greenspace analyses.
Data handling was indeed complex and, although several steps were common to other research projects using similar methods, certain processes were developed for analysis specific to greenspace and physical activity.
Conclusions:
Some of the issues associated with the use of GPS have been previously highlighted such as signal reception problems leading to inaccurate points or missing data; capacity of units to store recorded information; battery life of units leading to increased participant burden or compliance problems and complexity in data cleaning and handling. Despite some of the drawbacks the combination of accelerometry, GPS and GIS offers great potential in public health research but there is still some way to go in refining protocols. The more researchers that engage in this type of work then the more refined and established these techniques can become.
Support:
This research has been funded through an ESRC and Scottish Government PhD studentship grant.
