How the physical inactivity is affected by social-, economic- and physical-environmental factors: an exploratory study using the machine learning approach

Lee, Kangjae; Wang, Jue; Heo, Joon

doi:10.1080/17538947.2023.2230944

Cited by 3 publications

(1 citation statement)

References 61 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…On one hand, clustering analysis has the advantage of providing insight into real-world scenarios and holding a high scalability to uncover hidden patterns. On the other hand, tree-based machine learning can be applied as a pluralistic analysis platform to synthesize evidence between a range of urban physical exposures and physical activity 58,64,65 . Comparing to conventional analyses, the XGBoost model enhances our assessments with several advantages: 1) unraveling nonlinear relationships through visualization, 2) disentangling complex interactions among multiple exposures, and 3) offering robust computation for multi-inference approaches 66 .…”

Section: Discussionmentioning

confidence: 99%

The urban physical exposome and leisure-time physical activity in early midlife: a FinnTwin12 study

Wang,

Aaltonen,

Teeuwen

et al. 2024

Preprint

View full text Add to dashboard Cite

Leisure-time physical activity is beneficial for health and is associated with various urban characteristics. Using the exposome framework, the totality of the environment, this study investigated how urban physical environments were associated with leisure-time physical activity during early midlife. A total of 394 participants (mean age: 37, range 34-40) were included from the FinnTwin12 cohort residing in five major Finnish cities in 2020. We comprehensively curated 145 urban physical exposures at residential addresses of participants and measured three leisure-time physical activity measures: (1) total leisure-time physical activity (total LTPA) and its sub-domains (2) leisure-time physical activity without commuting activity (LTPA) and (3) commuting activity. Using K-prototypes cluster analysis, we identified three urban clusters: “original city center,” “new city center,” and “suburban”. Results from adjusted linear regression models showed that participants in the “suburban” cluster had lower levels of total LTPA (beta: -0.13, 95% CI: -0.23, -0.03) and LTPA (beta: -0.17, 95% CI: -0.28, -0.05), compared to those in the “original city center” cluster. The eXtreme Gradient Boosting models ranked exposures related to greenspaces, pocket parks, and road junctions as the top important factors influencing outcomes, and their relationships with outcomes were largely non-linear. More road junctions and more pocket parks correlated with higher total LTPA and LTPA. When the all-year normalized difference vegetation index within a 500 m buffer fell below 0.4, it correlated with higher levels of total LTPA, whereas above 0.4, it correlated with lower levels. To conclude, our findings revealed a positive correlation between urbanicity and physical activity in Finnish cities and decomposed this complexity into crucial determinants. Importance rankings and nonlinear patterns offer valuable insights for future policies and projects targeting physical inactivity.

show abstract

Section: Discussionmentioning

confidence: 99%