With an expected increase in urban population and the burden of healthcare costs, the role of the environment for human health has become a topical issue in the scientific and political debate. Numerous studies have investigated the benefits of green spaces on specific morbidities and mortalities while governments are looking to green environments as an alternative to improve citizens' wellbeing and decrease healthcare expenditures. This thesis first explores the definitions and measures of "green" through an analysis of the literature and empirical tests conducted on four landuse data sources in Belgium. It shows that the choice of the measure of green, as well as its level of aggregation, can lead to contradictory conclusions concerning the associations between green and health. Then, using medication reimbursement data recently made available in Belgium, this thesis investigates spatial associations of five groups of medication (a priori associated with the environment) with green spaces. At different scales and at different aggregation levels, medication reimbursement data appear not to be a proxy of individual health but of healthcare practices. Spatial variations of reimbursement are rather linked to exogenous factors such as education of the practitioners, or pharmaceutical and commercial practices. This thesis emphasizes the importance of a clear definition and interpretation of environmental and health data to correctly inform decision makers, particularly when data in use are diverted from their original aim and include spatial information. There is a clear risk of misinterpretation when using new types of data without a multidisciplinary perspective. ; (SC - Sciences) -- UCL, 2020