Non-linear Relationships Between the Built Environment and Walking Frequency Among Older Adults in Zhongshan, China

Abstract

Aim: Promoting walking activity is an effective way to improve the health of older adults. Walking frequency is a critical component of walking behavior and an essential determinant of daily walking levels. To decipher the association between the built environment and walking frequency among older adults, this study's aims are as follows: (1) to empirically test whether non-linear relationships between the two exist, and (2) to identify the thresholds of the built environment characteristics that promote walking.Methods: The walking frequency of old adults was derived from the Zhongshan Household Travel Survey (ZHTS) in 2012. The sample size of old adults aged 60 or over was 4784 from 274 urban and rural neighborhoods. A semi-parametric generalized additive model (GAMM) is used to analyze the non-linear or non-monotonic relationships between the built environment and the walking frequency among older adults.Results: We found that non-linear relationships exist among five out of the six built environment characteristics. Within certain thresholds, the population density, sidewalk density, bus stop density, land use mixture, and the percentage of green space are positively related to older adults' walking trips. Furthermore, the land use mixture and the percentage of green space show an inverse “V”-shaped relationship.Conclusions: Built environment features can either support or hinder the walking frequency among older adults. The findings in the current study contribute to effective land use and transport policies for promoting active travel among older adults.

Nonlinear Associations of the Built Environment with Cycling Frequency among Older Adults in Zhongshan, China

“…The results imply that the population density of 5000 persons/km 2 is sufficient to promote cycling among old adults. That echoes the results of a recent study in Zhongshan on the nonlinearity of the built environment on walking [25]. When the population density is around 5000 persons/km 2 , the thresholds appear both in walking and cycling frequency among older adults.…”

“…The commonly used methods include Poisson regression [17], negative binomial regression [18], and multilevel logistic regression [19]. However, recent studies have displayed nonlinear relationships between the built environment and travel behavior [20][21][22][23][24][25][26]. A complex nonlinear association may also exist between the built environment and cycling activity among older adults.…”

Non-Linear Effects of the Built Environment and Social Environment on Bus Use among Older Adults in China: An Application of the XGBoost Model

“…Compared with traditional linear models, the non-linear methods offer higher prediction precision and perform better in interpreting complex relationships. These studies support that the pre-specified relationship between the independent variables and dependent variable in the linear regression is flawed [30], and renders the non-linear relationship unobservable [31].…”

“…However, the results are not able to determine how small the proportion should be. The assumption of linearity in variables in Zhang, He, Wu, and Li 's study cannot capture the non-linear relationships between variables [31]. Compared with the previous study, the current study not only indicates whether the independent variable has a promotion or an inhibitory effect, but also depicts the changing curves and threshold effects of bus use frequency when the independent variable changes.…”

“…When the proportion of older adults in a neighborhood exceeds the threshold of 15%, the over-aggregation of the older population may even have an inhibitory effect. Planners may consider developing neighborhoods with proper dwelling unit density and relatively balanced age structures [31]. Second, it is important to ensure an adequate green space land use coverage rate, but within an optimal range (preferably around 10%).…”