Aim: While studies using global positioning systems (GPS) have the potential to refine measures of exposure to the neighbourhood environment in health research, one limitation is that they do not typically identify time spent undertaking journeys in motorised vehicles when contact with the environment is reduced. This paper presents and tests a novel methodology to explore the impact of this concern. Methods: Using a case study of exposure assessment to food environments, an unsupervised computational algorithm is employed in order to infer two travel modes: motorised and non-motorised, on the basis of which trips were extracted. Additional criteria are imposed in order to improve robustness of the algorithm. Results: After removing noise in the GPS data and motorised vehicle journeys, 82.43% of the initial GPS points remained. In addition, after comparing a sub-sample of trips classified visually of motorised, non-motorised and mixed mode trips with the algorithm classifications, it was found that there was an agreement of 88%. The measures of exposure to the food environment calculated before and after algorithm classification were strongly correlated. Conclusion: Identifying non-motorised exposures to the food environment makes little difference to exposure estimates in urban children but might be important for adults or rural populations who spend more time in motorised vehicles.
|Journal||International Journal of Geographical Information Science|
|Early online date||24 Feb 2016|
|Publication status||Published - 2016|
- food environments
- Global positioning systems
- travel mode
- unsupervised algorithm