Examining non-linear built environment effects on elderly’s walking: A random forest approach

Abstract

Previous studies often assume a pre-specified relationship between walking and the built environment. Using data from 702 older adults in Nanjing (China), this study employs a random forest method to scrutinise the refined associations between the built environment and walking time among older adults, while controlling for socio-demographics. Results show that all the analysed built environment attributes tend to impose salient non-linear and threshold effects on walking time. For example, population density and land use mix only increase older adults’ walking at certain levels. Ultra-densely populated areas and excessively mixed land uses may even lead to a decline in walking. Built environment interventions would consequently only be effective in certain ranges. By quantifying the relative importance of built environment attributes, we further indicate the hierarchy of intervention priorities. These results offer nuanced and appropriate guidance for building age-friendly neighbourhoods.