Long-term exposure to objective and perceived residential greenness and diabetes mortality: A census-based cohort study.

BACKGROUND: Residing close to green spaces might reduce diabetes mellitus (DM) risk; however, evidence for diabetes mortality is limited. Moreover, individual and neighbourhood social factors may determine DM risk. Exposure to green spaces may also depend on socioeconomic position (SEP). This study examined the associations between residential greenness and diabetes-related mortality, and the role of the social environment in these associations. METHODS: We used the 2001 Belgian census linked to mortality register data for the period 2001-2014. We included individuals aged 40-79 years old and residing in the five largest Belgian urban areas at baseline. Exposure to residential greenness was assessed with surrounding greenness using the Normalized Difference Vegetation Index (NDVI) within 500-m of residence (objective indicator), and perceived neighbourhood greenness (subjective indicator). We conducted mixed-effects Cox proportional hazards models to obtain hazard ratios (HR) for diabetes-related mortality per interquartile range (IQR) increments of residential greenness. We assessed effect modification by social factors through stratification. RESULTS: From 2,309,236 individuals included at baseline, 1.2% died from DM during follow-up. Both residential greenness indicators were inversely associated with diabetes-related mortality after adjustment for individual social factors. After controlling for neighbourhood SEP, the beneficial association with surrounding greenness disappeared [HR 1.02 (95%CI:0.99,1.06)], but persisted with perceived neighbourhood greenness [HR 0.93 (95%CI:0.91,0.95)]. After stratification the inverse associations with perceived neighbourhood greenness were strongest for women, the lowest educated, and individuals residing in least deprived neighbourhoods. CONCLUSIONS: Our findings suggest that an overall positive perception of neighbourhood green spaces reduces independently the risk of diabetes-related mortality, regardless of the neighbourhood social environment. Nevertheless, neighbourhood SEP may be a strong confounder in the associations between diabetes-related mortality and greenness indicators derived from satellite images. Perception factors not captured by objective measurements of green spaces are potentially relevant in the association with DM, especially among disadvantaged groups.