A protective association was observed between levels and variability of neighborhood greenness and coronary heart disease or stroke. The odds of hospitalization was 37% lower, and the odds of self-reported heart disease or stroke was 16% lower, among adults with highly variable greenness around their home, compared to those in neighborhoods with low variability in greenness. These effects were independent of the absolute level of greenness in the neighborhood. We observed weaker evidence for association between the cardiovascular outcomes and mean neighborhood greenness. The odds of self-reported heart disease or stroke decreased by 7% per unit (interquartile range) increase in the level of greenness.
The lower prevalence of heart disease may be attributable to higher levels of physical activity, such as neighborhood walking which is positively influenced by the natural and built environment. Previous studies have reported that adults with access to a large high-quality park within walking distance (also 1600
m) from home have elevated levels of walking [9
] and longevity [18
]. Our results indicate that these results might be explained by neighborhood variability in greenness, which is more strongly associated with a lower risk of coronary heart disease or stroke. This suggests that in terms of cardiovascular health, the mix of greenness is more relevant than the extent of greenness. That is, both green and non-green areas are necessary within walking distance, rather than vegetation alone. Neighborhood attributes that may contribute to a high variability in greenness might include prevalence of tree lined streets/cycleways/footpaths, presence of parks with parking, or green reserves with good road connectivity. The coexistence of both aesthetically pleasing natural vegetation to entice people out of their homes and destinations within walking distance would also contribute to variability in neighborhood greenness. A review of the environmental influences on walking concluded that aesthetic neighborhood attributes (which included ratings of natural features) were found to be associated with walking [7
]. Furthermore, a cross-sectional study conducted in Seattle, US, reported that the most frequently walked non-green destinations were grocery stores, restaurants, libraries, coffee shops, and post offices[19
]. Further studies are needed to identify the specific attributes of neighborhoods with a high degree of variability in greenness, such as the co-location of green and non-green areas (e.g. tree-lined paths, parks surrounded by well-connected streets). Future cohort studies of coronary heart disease and stroke should include measurement of variability in neighborhood greenness.
A limitation of this study was that the mediating effects on physical activity were not directly examined because we did not have a measure of physical activity undertaken in the neighborhood. For example, neighborhood walking might be more appropriate to specifically test the hypothesis. The results of this cross-sectional study should be interpreted cautiously, as they might be explained by self-selection of healthier individuals into neighborhoods with high levels of variability in greenness.
Although this study was limited to a cross-sectional design, temporal relevance of hospitalizations was improved by ensuring that cases were defined if the hospitalization occurred within a 3
year window of the year of the health survey and calculation of neighborhood greenness. Adjustment for a wide range of risk factors, such as nutrition, was an advantage of this study. However, adjustment was not made for known risk factors for cardiovascular disease such as heritability, which were not available from the health survey. The possibility of residual confounding by individual level socio-economic status cannot be dismissed. However, adjustment was also made for multiple correlates of socioeconomic position, including education and household income. Adjustment was also made for biological and behavioral factors that also exhibit gradients in socioeconomic status.
A further limitation is that imagery obtained from remote sensing is accurate to a specified level of resolution. It is possible that the spatial variability in greenness could have been deflated for neighborhoods with smaller block sizes as an artifact of the spatial resolution of the imagery. However, smaller block sizes in our study area were typically located closer to the city with greater land-use mix. Therefore, we suspect that the deflation in spatial variability in greenness due to the effect of the resolution of the imagery was small relative to effect of actual variability in greenness and land-use mix. Finally, spatial autocorrelation in cardiovascular outcomes was not directly modeled in the analyses. However, such autocorrelation would be limited to that not already accounted by adjustment for socioeconomic factors (e.g. education and household income) which also cluster spatially and are strong predictors for cardiovascular disease. Perhaps more importantly, it is impossible to completely rule out the chance of spatial autocorrelation in the cardiovascular outcomes due to an unknown factor.