Objectives: Geographical Information Systems (GIS) are widely used in environmental epidemiology studies to locate study population by geocoding addresses and to evaluate exposures and relationship with health outcomes. Despite this, Italian environmental epidemiologists poorly discuss quality of address geocoding results.
Design: two case-studies have been carried out in Tuscany Region (Central Italy): one in the mountain area in the Municipality of Piancastagnaio (Siena Province) and one in the urban area around the airport of Florence. Three geocoding systems have been compared: the geographical database produced by Tuscany Region and two commercial systems (Google and Bing-Microsoft); 1,549 addresses in Piancastagnaio and 2,946 addresses in Florence have been tested.
Results: Tuscan geographical database showed better performance than the two commercial systems, with bigger differences in Piancastagnaio. In this area, mean difference between regional system and Google service is more than 300 mt, with peaks of 7-8 km. Bing- Microsoft system does not provide any information on addresses in Piancastagnaio: all input addresses were geocoded in the centroid of the municipality or in the centre of a few principal streets. Lowest differences among the three methods were observed in the urban area of Florence: mean difference between Tuscany and Goggle systems was 150 mt, with less than 2 km peaks; between Tuscany and Bing-Microsoft mean difference was 100 mt with 3 km peaks. In both case-studies, but especially in Piancastagnaio area, these differences gave rise to great misclassification in the evaluation of individual exposure and health outcome.
Conclusion: the study highlighted the impacts of address geocoding process in exposure assessment in environmental health research and pointed out the need of specifically evaluate the quality of cartographic data.