Unveiling the Obesogenic Neighborhood Food Environment Factors and Typologies in Tianjin, China: An Integrative Analysis of Perceived and Objective Measures

Yue Sun¹Wei Lu^1*Jinyuan Gu¹Yishu Yao^1*Tianyue Wan²

• ¹Dalian University of Technology, Dalian, China
• ²Fujian Agriculture and Forestry University School of Agriculture, Fuzhou, China

Introduction: Assessing and intervening in food environments constitutes a critical strategy for addressing the obesity epidemics. However, existing assessments predominantly focus on either objective or perceived dimensions, with limited attention to developing countries. This study investigates the impact of neighborhood-level food environments on resident obesity in a national central city of China and establishes a typology of obesogenic community profiles. Methods: We developed an integrative tool that harmonizes objective geospatial data with subjective perceptual metrics. Leveraging stratified sampling survey data on neighborhood food environments (N=405) and multiscale geospatial datasets from Tianjin, China (2023), we establish a comprehensive indicator repository for neighborhood food environments. Dimensionality reduction via principal component analysis (PCA) was applied to all measured indicators, followed by an ordinal multinomial regression model to identify significant obesogenic determinants at the neighborhood level. Finally, the K-means clustering algorithm was subsequently implemented to delineate prototypical obesogenic neighborhood typologies. Results: Among 10 principal components derived from PCA, four obesogenic factors were identified, ranked by effect magnitude: FAC_8(Perceived Community Food Accessibility Index, β=-0.382, P=0.001, OR=0.68), FAC_4(Food Availability and Diversity within 500-1000m, β=0.225, P=0.061, OR=1.25), FAC_6 (Unhealthy Dietary Behavior, β=-0.191, P=0.066, OR=0.68), and FAC_3(Retail Food Environment Index within 500m, β=-0.184, P=0.078, OR=0.83). K-means clustering delineated three obesogenic neighborhood types: Objective Deprived (N=10, 6.1%), Objective Overloaded (N=37, 22.56%), and Objective Overloaded-Dietary Behavior Integrated (N=117, 71.34%).