10-year assessment of predictive relationships between different dimensions of urban environments and incidents cardio-metabolic risk in South Australia

Abstract

Evidence on the attributes of urban areas related to health is limited by cross-sectional studies focusing on few dimensions at a time. This study simultaneously assessed several dimensions of environments in relation to incident cardiometabolic risk (CMR) over 10 years. Initially disease-free adults in a biomedical cohort in Adelaide, Australia provided address and clinical data over 3 waves of follow up (n=2555; 138 suburbs). CMR was defined as the count of 5 clinical risk factors. A GIS was used to derive environmental measures for a 1600m network buffer around each residential location, as follows. Food environment - fast-food outlet count, density, retail food environment index; Natural environment - public open space (POS) area, count, normalised difference vegetation index (NDVI); Built environment – proximity, connectivity to services; Locational property wealth - relative location factor (RLF) from a hedonic model of sales data blind to location. Poisson growth models accounting for within suburb clustering, age, gender and education were used to estimate associations between environmental measures and change in CMR across 10 years. CMR counts increased over time. Fitted linear trajectories had statistically significant mean values of intercepts and slopes ((RR)=1.04, 95% CI: 1.02-1.05, per 5 years). CMR trajectories were positively associated with individual age, male gender and low education. Higher RLF, walkability and POS area were protective against CMR but POS count and NDVI did not differ from the null. Only RLF (RR=0.74, 95% CI: 0.65-0.84) and POS area (RR=0.53, 95% CI: 0.34-0.82, per 10 m2) remained associated with CMR after adding food environment measures. This study found no support for a role of food or built environments. POS area and locational property wealth were strongly inversely related to CMR. Results support a health rationale for larger POS in urban areas, and the utility of a locational measure of property wealth for predicting CMR.