BackgroundBuilt environment can influence physical conditions of older adults (e.g. osteoporosis). However, traditional methods using 2-dimensional circular buffer as a spatial structure to measure neighbourhood effect may create bias in health estimation, especially for the hilly and compact environment across low-income neighbourhoods (e.g. public housing estates). MethodsWe evaluated the environmental influences on self-reported osteoporosis among “old residents” (age≥65) in Hong Kong (n = 2077). Twelve public housing estates across hilly neighbourhoods in Hong Kong were selected as study sites. A cross-validated approach was developed to evaluate four spatial structures (2D circular, 2D service area, 3D circular, 3D service area). To determine problems of spatial uncertainty, we compared odds ratios (OR) and differences in effect sizes from models using different spatial structures. When all adjusted models achieve significant results based on 95% confidence intervals (CI) and with all positive/negative ORs, this study reported to have reached “a result with consistency”. Results from the 3D service area were then used to explain the environment-health relationship. ResultsDifferent spatial structures can yield different results. Particularly, circular buffers overestimated environmental effects on self-reported osteoporosis. Overestimated measures were related to walkability and accessibility but not greenery. Specifically, results from the 3D service area showed that more public space and health facilities within a walkable distance (500 m) from a location of subject's residence were negatively associated with self-reported osteoporosis (adjusted ORs: 0.44 [0.29, 0.66]; 0.94 [0.90, 0.99]). However, more major transport facilities at the immediate distance from residence (200 m) was positively associated with self-reported osteoporosis (adjusted OR: 1.11 [1.01, 1.23]). ConclusionsPhysical conditions (e.g. osteoporosis) of older adults living in a hilly neighbourhood could be driven by walking behaviours. It is necessary to include local terrain and road network to define a walkable neighbourhood for environment-health estimations to minimize spatial bias.
Read full abstract