Assessing the Validity of OpenStreetMap for Food Environment Research

Abstract

This study assessed agreement between food environment measures derived from OpenStreetMap (OSM) data, a commercial dataset, and an administrative dataset (the Canadian Food Environment Dataset, Can-FED) to better understand the suitability of OSM food-related data for food environment research. We calculated Spearman's correlations between continuous retail food environment measures in Can-FED and those derived from OSM and DMTI Spatial. Additionally, using Can-FED as the reference, we assessed the accuracy of categorical food environment variables derived from OSM and DMTI data. OSM consistently reported fewer food retailers than Can-FED, but correlations between density and proportion measures from OSM, DMTI, and Can-FED were moderate to very strong. OSM and DMTI reliably identified areas with low proportions of healthier food retailers and fast-food outlets, though accuracy was lower in areas with higher proportions. In metropolitan areas, where categorized variables from OSM differed from Can-FED, proportions of healthier retailers and fast-food outlets were often underestimated. This study highlights OSM's limitations, such as missing data and error in accurately classifying neighborhood food environments, yet suggests that OSM may be useful for capturing general trends or measuring food environments in low-density areas when higher quality administrative data is not accessible.