Creation of synthetic homogeneous neighbourhoods using zone design algorithms to explore relationships between asthma and deprivation in Strasbourg, France

The concept of ‘neighbourhood’ as a unit of analysis has received considerable research attention over the last decade. Many of these studies raise the question of the influence of local characteristics on variations in health and more recently, researchers have sought to understand how the neighbourhood can influence individual health through individual behaviour. Relatively few studies discuss the question of the borders and definition of a neighbourhood but we know that the results from health or population datasets are very sensitive to how zones are constructed – part of the Modifiable Areal Unit Problem (MAUP). In reality, we know that neighbourhoods are not constrained by artificial statistical boundaries, but rather exist as complex multi-dimensional living communities. This paper tries to better represent the reality on the ground of these communities to better inform studies of health. In this work, we have developed an experimental approach for the automated design of neighbourhoods using a small tessellated cell as a basic building block. Using the software AZTool, we considered population, shape and homogeneity constraints to develop a highly innovative approach to zone construction. The paper reports the challenges and compromises involved in building these new synthetic neighbourhoods. We provide a fully worked example of how our new synthetic homogeneous zones perform using data from Strasbourg, France. We examine data on Asthma reported through calls to the emergency services, and compare these rates with an index of multiple deprivation (NDI) which we have constructed and reported elsewhere. Higher correlations between Asthma and NDI were found using our newly constructed synthetic zones than using the existing French census areas of similar size. The significance of our work is that we show that careful construction of neighbourhoods – which we claim are more realistic than census areas – can greatly aid unpacking our understanding of neighbourhood relationships between health and the social and physical environments.