Next Article in Journal
Transforming Public Space with Nature-Based Solutions: Lessons from Participatory Regeneration in Lorca, Spain
Previous Article in Journal
Combining Resilience and Sustainability in Urban Mobility: A Scoping Review and Thematic Analysis
Previous Article in Special Issue
Indigenous Knowledge, Gender and Agriculture: A Scoping Review of Gendered Roles for Food Sustainability in Tonga, Samoa, Solomon Islands and Fiji
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Land
Volume 14
Issue 10
10.3390/land14102064
land-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Nonlinear Associations Between Built Environment and Overweight: Gender and Marital Status Differences in Urban China

by
Xiaohua Zhong
1,
Yang Xiao
2 and
Yihui Huang
2,*
1
Institute of Sociology, Shanghai Academy of Social Sciences, No. 7, Lane 622, Middle Huaihai Road, Shanghai 200020, China
2
College of Architecture and Urban Planning, Tongji University, 1239 Siping Road, Shanghai 200092, China
*
Author to whom correspondence should be addressed.
Land 2025, 14(10), 2064; https://doi.org/10.3390/land14102064
Submission received: 25 August 2025 / Revised: 1 October 2025 / Accepted: 14 October 2025 / Published: 16 October 2025
(This article belongs to the Special Issue Land, Ageing, Gender and Environment: Problems and Challenges from Different Disciplines II)
Browse Figures
Versions Notes

Abstract

Overweight has become a major public health concern in China’s rapidly urbanizing cities. Patterns of environmental exposure differ notably between men and women, both before and after marriage. This study examines how built environment characteristics influence the risk of overweight, with particular attention to nonlinear associations as well as variations by marital status and gender. Drawing on survey data from 2634 Shanghai residents, we applied extreme gradient boosting to model complex environment–health relationships. The results indicate that greenness, park accessibility, population density, and transit conditions are associated with overweight through nonlinear pathways, with threshold and plateau effects suggesting that benefits taper off, or risks escalate, beyond certain levels. These optimal ranges differ across gender–marriage groups: moderate density and green exposure were generally protective, but the effective ranges were narrower for women and unmarried individuals. Married men benefited more consistently, likely supported by healthier routines reinforced through spousal support, whereas married women showed weaker or even adverse effects, potentially due to greater family responsibilities. Overall, the findings reveal that overweight is shaped by socially differentiated nonlinearities in environmental exposures. Urban planning and public health policies should therefore optimize built environment attributes within effective ranges and tailor interventions to diverse demographic groups.

1. Introduction

Overweight has become a pressing public health concern worldwide, substantially increasing the risk of cardiovascular disease, diabetes, and other chronic conditions. According to WHO data, approximately 2.5 billion adults, 43% of the global adult population, were overweight in 2022, underscoring its growing burden on healthcare systems [1]. In China, the prevalence of overweight has risen sharply in recent decades, driven by rapid urbanization and lifestyle changes [2]. This trend highlights the urgency of identifying the determinants of overweight to guide effective prevention strategies and public health interventions.
Among these determinants, the built environment has attracted increasing attention [3,4]. A well-designed urban form characterized by adequate density, diversity, and accessibility can promote physical activity and healthy dietary habits. By contrast, car-dependent environments with limited green space or abundant fast-food outlets tend to foster sedentary behaviors and poor nutrition, thereby elevating overweight risk [5,6]. Growing evidence suggests that features such as green space, population density, and public transport accessibility influence energy balance through both physical activity and dietary pathways, positioning the built environment as a key upstream determinant of overweight [7,8,9].
Behavioral factors are also central to understanding overweight [10]. Insufficient physical activity, prolonged sedentary time, irregular eating patterns, and frequent consumption of energy-dense foods are well-documented risk factors [11,12]. Additional lifestyle routines, including smoking, alcohol consumption, and sleep quality, further affect weight regulation [13,14]. Moreover, subjective life satisfaction has been shown to shape health-related behaviors: individuals with higher satisfaction are more likely to maintain healthier routines, whereas those with lower satisfaction may adopt sedentary or stress-related behaviors [15,16]. These findings emphasize that overweight cannot be explained solely as a matter of individual choice, but rather emerges from the interplay between behaviors, psychological states, and environmental conditions.
Despite notable progress, relatively few studies have examined population heterogeneity in these relationships [17,18]. The Social Determinants of Health framework highlights that demographic and social attributes shape how individuals perceive and interact with their environments [19]. Gender and marital status are particularly relevant, as they influence role expectations, lifestyle routines, and access to social support. Women often face structural and cultural barriers, such as household responsibilities and concerns about safety in public spaces, that restrict opportunities for exercise and limit their ability to benefit from green spaces and recreational facilities [20,21]. By contrast, men are typically more active in outdoor and recreational environments, although their higher rates of smoking and alcohol consumption may counteract some of these health benefits [22,23]. Marital status further complicates these dynamics: while marriage can provide protective resources through spousal support [24], it may also impose additional constraints, especially on women, due to greater family and caregiving responsibilities [25]. However, empirical evidence on how overweight is differentially linked to built environments and behavioral habits across gender–marriage subgroups remains limited. Addressing this heterogeneity is crucial for identifying vulnerable populations and designing more tailored interventions.
A further issue concerns the nonlinearity of environment–behavior–health linkages. Emerging evidence suggests that the effects of population density, green space, and transit accessibility on overweight often follow nonlinear patterns, with thresholds or optimal ranges beyond which benefits diminish or risks increase [26,27]. However, little is known about whether these nonlinearities apply uniformly across demographic subgroups. Incorporating gender and marital status into this discussion is particularly important, as these attributes not only shape behavioral practices but also condition how individuals access and benefit from environmental resources.
To address these gaps, we propose three hypotheses:
Hypothesis 1 (H1):
Nonlinear associations between the built environment and overweight vary across social groups, as lifestyle routines influence how individuals can access and benefit from environmental resources.
Hypothesis 2 (H2):
Marriage provides spousal support that enables married men to adopt healthier lifestyles [24], thereby lowering their risk of overweight compared with unmarried men.
Hypothesis 3 (H3):
Married women, constrained by limited opportunities for physical activity and greater family responsibilities imposed by social expectations [25], are more likely to face an elevated risk of overweight.
This study drew on survey data from 2634 residents in Shanghai and applied XGBoost to examine how built environment, lifestyle, and socio-demographic factors are associated with overweight.

2. Literature Review

2.1. Built Environment and Overweight

The built environment is a key determinant of overweight, shaping both opportunities and constraints for healthy living. Core elements such as green space, population density, and transport accessibility influence physical activity, dietary behaviors, and psychosocial conditions, thereby affecting weight outcomes. Green space has consistently been shown to reduce overweight risk by providing venues for exercise, alleviating stress, and fostering social interaction [7,8,9]. Importantly, not only the presence but also the accessibility and quality of green areas matter, with some studies identifying threshold effects beyond which health benefits diminish [28,29]. Population density exerts mixed effects: moderate levels promote walkability and access to services, whereas very high density can reduce satisfaction, restrict recreational space, and increase sedentary time, thereby elevating overweight risk [30]. Transport accessibility also has dual implications. Proximity to transit supports active commuting and incidental walking [31], but long commuting times may reinforce sedentary behavior and reduce opportunities for exercise or healthy eating [32]. Taken together, these findings underscore that the built environment shapes overweight through both physical access and behavioral incentives, highlighting its importance as a modifiable factor in weight prevention.

2.2. Gender and Marital Status as Moderators of Overweight

Existing research highlights the important role of marital status and gender in shaping health outcomes. Marriage is frequently linked to protective effects, whereas divorce or widowhood elevates risks of morbidity and mortality, particularly among men [33,34]. In parallel, urban and public health studies emphasize the influence of the built environment on physical activity and obesity. Features such as neighborhood walkability, recreational facilities, and safety are positively associated with healthier lifestyles, though these associations often differ by gender [35].
While the built environment provides opportunities and constraints for health, its effects on overweight are mediated by behaviors and conditioned by gender and marital status. Behavioral factors, including physical activity, sedentary lifestyle, diet, sleep, smoking, alcohol use, and life satisfaction, directly shape overweight risk [36,37,38,39]. Regular exercise reduces risk by improving energy balance, whereas inactivity, poor diet, and inadequate sleep promote weight gain [40,41,42]. Smoking is sometimes associated with lower body weight but harms overall health, while alcohol consistently contributes excess calories [43,44]. In addition, life satisfaction influences weight regulation through stress, emotional eating, and motivation to maintain healthy routines [15,45].
These behaviors do not develop in isolation but are strongly shaped by socio-economic contexts [46,47]. Gender and marital status, in particular, influence both health behaviors and the use of the built environment, thereby reshaping daily routines and health outcomes [48,49]. Women often face structural and cultural constraints, such as safety concerns in public spaces or greater household responsibilities, that limit opportunities for physical activity and reduce their ability to benefit from urban green spaces, recreational facilities, or walkable neighborhoods [20,50]. Men, by contrast, tend to report higher levels of physical activity but also greater engagement in risky behaviors such as smoking and alcohol consumption [22], which may offset the advantages of their more active use of parks, sports facilities, and other urban amenities [22]. These differences suggest that identical environmental or behavioral exposures may yield divergent outcomes by gender. Marital status adds another layer of complexity. Marriage is frequently associated with weight gain due to shared meals, reduced activity, and lifestyle adjustments, a phenomenon sometimes described as the “marriage market hypothesis” [25]. At the same time, married individuals may benefit from spousal support that fosters healthier routines, particularly among men [24]. Single, divorced, or widowed individuals often display different behavioral patterns shaped by social networks, stress, and economic stability [51]. When combined, gender and marital status create intersecting subgroups with distinct risk profiles. Married women, for example, may experience greater time constraints from household responsibilities, leaving them less able to take advantage of nearby parks, green spaces, or recreational facilities, even when such resources are physically accessible [50]. By contrast, men, especially single men, may be more vulnerable to unhealthy dietary patterns [24]. They are also more likely to use outdoor and recreational spaces actively while simultaneously frequenting fast-food outlets or bars, linking physical activity with riskier lifestyle habits [23]. These dynamics underscore that overweight is not simply the outcome of individual choices or environmental exposures but rather of social positions that mediate how people interact with and respond to them.

2.3. Nonlinear Effects and Subgroup Heterogeneity in Overweight

Despite a growing body of literature on overweight [22,46,47], conventional approaches often fail to capture the complex dynamics of its determinants. Traditional regression models typically assume linear and additive relationships between variables [31,52], thereby overlooking the threshold effects and nonlinear patterns that characterize the influence of the built environment on health. For example, the benefits of green space may plateau beyond a certain level, while the risks associated with population density may only emerge once critical thresholds are exceeded [30]. Similarly, lifestyle factors such as physical activity, sleep, alcohol consumption, and smoking have been shown to follow nonlinear trajectories, with optimal ranges rather than monotonic effects [53,54]. These lifestyle routines also vary substantially across demographic subgroups [43,55], meaning that the same built environment may produce divergent health outcomes depending on how it is used. Ignoring such nonlinearities and subgroup interactions risks obscuring vulnerable populations and generating oversimplified policy recommendations. Furthermore, comparative analyses across demographic subgroups remain limited. Recent advances in machine learning offer tools well-suited to address these complexities [27,56], providing more precise and context-sensitive insights into overweight risks.

3. Data and Methodology

3.1. Data Source

The data for this study were drawn from the project Spatial Structure, Social Structure, and Residential Satisfaction of Neighborhoods in Shanghai, jointly initiated by the Center for the Study of Social Transformation at Fudan University and the Center for Urban and Social Research at Tongji University. Neighborhood surveys were conducted between 2015 and 2017, and resident questionnaires between 2019 and 2021. Two structured questionnaires were employed. The neighborhood survey collected information on locational features, spatial structure, social composition, and governance, while the resident questionnaire covered demographics, occupation and life course, social capital, values and attitudes, housing, and self-rated health.
A five-stage probability-proportional-to-size sampling design was used to ensure representativeness. First, 30 subdistricts/towns within the Shanghai Outer Ring were selected based on location and the latest population data. From each, 4 neighborhood committees were chosen (120 total). Within these committees, 300–400 residential communities were sampled. In each community, approximately 60 households were contacted to obtain 20 completed interviews, yielding 3648 interviews in total (a 30% success rate). Finally, one individual aged 18–70 was randomly selected from each household using a Kish table. Neighborhood-level data were obtained from community committees and property managers, while resident-level data were collected through face-to-face household interviews, supplemented by field observations of greenery, facilities, and community activities. In total, 399 neighborhoods and 3630 residents were surveyed. After excluding incomplete or erroneous cases, 2634 valid samples were retained (Figure 1). In Figure 1, red squares denote the sampled neighborhoods, with their size corresponding to the boundary extent of each community.

3.2. Variable Description

The dependent variable in this study is the presence of overweight, defined as a body mass index (BMI) of 24 kg/m2 or higher [57]. BMI is widely used in large-scale epidemiological research because it is simple to calculate, enables cross-study comparability, and is available in most survey datasets [58]. However, it reflects weight relative to height and does not capture body fat distribution or composition [59]. Alternative measures such as waist circumference, waist-to-hip ratio, or body fat percentage provide more direct assessments of central adiposity and metabolic risk, but these indicators were not available in the present dataset. Guided by the framework of health determinants, we examine both built environment attributes and individual behavioral habits as key explanatory factors of overweight. Built environment characteristics include the normalized difference vegetation index (NDVI) within a 1000 m buffer, distance to the nearest park, green view index (GVI) within a 1000 m buffer, total park area, population density, community green space ratio, and public transit travel time. Together, these variables capture greenery exposure, spatial accessibility, and mobility conditions, and are widely used in the literature to represent built environment features. Behavioral habit variables include monthly exercise frequency, alcohol consumption frequency, smoking frequency, sleep quality, and life satisfaction. These indicators are commonly employed to reflect individual lifestyle patterns and subjective health perceptions. In addition, several sociodemographic covariates were included: gender (male/female), age squared (continuous), marital status (married/unmarried), education level (categorical), and log-transformed personal income (continuous, in CNY).

3.3. Model and Analytical Approach

We employed extreme gradient boosting (XGBoost) to model associations between built environment and lifestyle attributes and overweight status. XGBoost is an advanced implementation of gradient boosting decision trees that integrates regularization, shrinkage, feature subsampling, and second-order optimization, thereby reducing overfitting and enhancing computational efficiency [60,61]. These properties make it particularly well suited for modeling complex, nonlinear relationships between environmental and behavioral factors and health outcomes in urban contexts [26,56].
Compared with traditional linear regression, XGBoost can flexibly capture nonlinear and interactive effects without assuming linearity, handle multicollinearity, and accommodate missing data [62]. It also performs robustly with moderate sample sizes and class imbalance, delivering strong out-of-sample accuracy. These strengths are especially relevant for built environment–health research, where predictors often interact across spatial, behavioral, and socioeconomic domains [63,64]. In this study, XGBoost allowed more accurate estimation of predictor importance and effect shapes than conventional regression, thereby improving interpretability and predictive power in multidimensional datasets.
Analyses were conducted in Python 3.11 using the XGBoost library. Data preprocessing involved variable cleaning, one-hot encoding of categorical predictors, and a stratified 75/25 train–test split. Class imbalance was addressed by setting the scale_pos_weight parameter according to the prevalence of overweight. The final model was configured with a binary logistic objective, 800 boosting iterations, a maximum tree depth of 3, a learning rate of 0.05, a subsample ratio of 0.8, and a column sampling ratio per tree of 0.8. On the test set, the model achieved an AUC of 0.529, accuracy of 0.530, sensitivity of 0.436, specificity of 0.599, and an F1-score of 0.440. Four subgroup models (female/male × unmarried/married) were fitted using identical parameters to assess heterogeneity in predictor importance and partial dependence curves across demographic strata.
We reported gain-based relative importance (RI) scores to quantify each predictor’s contribution to loss reduction. Partial dependence plots (PDPs) were generated on quantile grids (5th–95th percentiles), with background data restricted to the same range to avoid extrapolation. Curves were smoothed using PCHIP for continuous features and LOWESS for ordinal behavioral variables, applying milder smoothing to built environment indicators and stronger smoothing to behavioral variables. Subgroup PDPs included 95% bootstrap confidence bands (B = 50) to capture uncertainty, enabling nuanced interpretation of nonlinear and threshold effects across domains [65].

4. Results

4.1. Descriptive Statistics

Table 1 summarizes the dependent variable and individual-level covariates. In our sample, 42% of respondents were classified as overweight, a proportion somewhat higher than the national average of approximately 34–35% reported in recent Chinese surveys [2]. Males represented 37% of participants, and 80% were married. On a four-point categorical scale of education level (1 = primary school or below, 4 = postgraduate or above), the sample mean was 2.35, suggesting that the typical respondent had completed education between the secondary and undergraduate levels. The mean value of log-transformed personal income was 10.20, corresponding to an annual income of approximately CNY 74,352. This figure is slightly above the 2019 average annual disposable income of Shanghai residents (CNY 69,442; Shanghai Municipal Bureau of Statistics, 2020). The higher average income likely reflects the inclusion of employed residents living in built-up districts, resulting in a sample with somewhat higher socioeconomic status than the general Shanghai population.
Table 2 presents the characteristics of lifestyle attributes and built environment variables in the study area. Regarding lifestyle, nearly half of respondents (48%) reported exercising daily, while smaller proportions exercised several times per week (18%), monthly (8%), or less frequently (8%); 18% reported no exercise at all. Most participants consumed alcohol less than once per week or not at all (81%), and 70% were non-smokers. Sleep quality was generally stable, with 86% reporting no deterioration, and life satisfaction was high, with 81% expressing satisfaction or strong satisfaction.
With respect to the built environment, within a 1000 m buffer of participants’ residences, the mean NDVI was 0.16, the mean GVI was 0.22, and the average park area was 78,102 m2. The average distance to the nearest park was 651 m, while the mean green space ratio was 26.89%. Population density averaged 37,745 persons/km2, consistent with the high-density urban form of the study area. Public transport accessibility was strong, with an average walking time of just over four minutes. Taken together, these indicators suggest that respondents lived in highly urbanized neighborhoods characterized by dense populations, moderate vegetation cover, convenient access to green spaces, and abundant public transportation options.

4.2. Feature Importance Comparison

Table 3 reports the RI of predictors for overweight in the total sample and across gender–marital subgroups. In the total sample, built environment factors contributed the largest share of predictive power (37.79%), followed by sociodemographic characteristics (35.57%) and lifestyle attributes (26.64%). At the individual level, gender (12.23%), education level (7.30%), and drinking frequency (5.75%) emerged as the strongest predictors, alongside built environment indicators such as distance to the nearest park (5.65%) and population density (5.54%).
Gender comparisons revealed that men’s overweight risk was more strongly shaped by built environment factors (47.41–49.05%) than women’s (47.00–48.24%), with population density, public transport accessibility, and green space ratio exerting greater influence in male models. Women’s outcomes, by contrast, were more affected by sociodemographic and lifestyle factors, particularly education and income. Marital status also moderated predictor importance. Unmarried respondents, both male and female, showed greater reliance on built environment characteristics (male–unmarried: 49.05%; female–unmarried: 48.24%) than their married counterparts (male–married: 47.41%; female–married: 47.00%). Married groups gave greater weight to lifestyle factors such as drinking and exercise frequency, while unmarried groups were more sensitive to public transport accessibility, green space ratio, and education.
In the four-way subgroup breakdown, male–married participants were most influenced by built environment factors (47.41%), especially population density and NDVI, with drinking frequency as the leading lifestyle contributor. Male–unmarried participants exhibited the highest overall built environment share (49.05%), dominated by public transport accessibility and green space ratio, while drinking frequency ranked first across domains. Female–married participants were similarly shaped by built environment factors (47.00%), with the GVI as the top predictor, complemented by exercise frequency and sleep quality. Female–unmarried participants showed the strongest effect of education across all subgroups, alongside significant contributions from population density and park area within 1000 m.

4.3. Nonlinear Relationship Analysis

Full-sample partial dependence analysis revealed distinct and often nonlinear associations between overweight risk and built environment, lifestyle, and socioeconomic predictors (Figure 2). Both the green space ratio and GVI were positively associated with overweight, with risk increasing steadily and peaking at approximately 70–80% and 0.30, respectively. Park area within 1000 m and distance to the nearest park followed inverted U-shaped patterns: overweight probability peaked at ~100,000–175,000 m2 for park area and ~700–900 m for park distance, with lower risks observed when park area exceeded 180,000 m2 or distance reached 1100–1200 m. NDVI, population density, and public transit time displayed wave-like curves with clear inflection points. NDVI shifted from protective to risk-enhancing at ~0.18; population density minimized risk between 20,000 and 60,000 persons/km2 but increased beyond 60,000 persons/km2; and public transit time showed the lowest risk at ~12–14 min, whereas shorter times (6–8 min) were associated with higher risk. Lifestyle behaviors exhibited similarly complex associations. Smoking frequency and sleep quality both reduced risk up to moderate levels (around three to four times per week for smoking), after which the effect attenuated or reversed. Drinking frequency followed an inverted U-shape, with risk peaking at ~three times per week before declining at higher levels. Exercise frequency was consistently protective, with risk declining sharply until two to three sessions per week, after which the effect plateaued. Life satisfaction showed a positive association, with overweight risk rising sharply up to a score of three and stabilizing thereafter. Socioeconomic attributes produced comparatively stable associations. Gender was positively linked to overweight, with men at higher risk than women. Marital status exerted a protective effect, as married individuals consistently exhibited lower risk compared with their unmarried counterparts. Education level was inversely associated with overweight, suggesting that higher educational attainment reduces risk. By contrast, personal income displayed fluctuations without a clear monotonic trend, while age (squared) showed a nonlinear trajectory with oscillating risks across age groups rather than a consistent directional effect.
Subgroup heterogeneity was also evident in lifestyle behaviors (Figure 3). For drinking frequency, female subgroups displayed relatively flat risk curves, while male–married individuals experienced a decline at the level of former drinker but an increase at three to four drinks per week. Male–unmarried individuals showed little variation across drinking levels. Smoking frequency revealed more distinct thresholds. Among men, an inflection point emerged at non-smoker with passive smoke exposure, after which overweight risk declined consistently. Among women, divergence appeared at the level of former smoker: female–married individuals exhibited increasing risk, whereas female–unmarried individuals showed decreasing risk. Sleep quality also highlighted subgroup-specific thresholds. Male–married individuals maintained relatively stable risk up to about the same as usual, followed by a modest increase, whereas male–unmarried individuals displayed a steady upward trend across the entire scale. In contrast, both female subgroups faced higher risk once sleep quality reached as good as usual, with the increase steeper for female–married than for female–unmarried individuals. Exercise frequency produced contrasting effects by marital status. Married subgroups demonstrated a consistent decline in overweight risk with greater exercise frequency, while unmarried subgroups displayed an inflection at daily exercise, beyond which risk rose. Life satisfaction revealed further divergence. Male–married individuals experienced reduced risk once satisfaction reached satisfied, with male–unmarried individuals showing a similar decline only at very satisfied. By contrast, female–married individuals exhibited a continuous rise in risk with increasing satisfaction, whereas female–unmarried individuals showed a marked decline between neutral and satisfied.

4.4. Robustness Check

To assess robustness, we compared feature importance rankings from XGBoost with SHAP explanations (Figure 4). Both methods consistently identified gender, age, and personal income as the most influential predictors, underscoring their stability in explaining overweight risk. NDVI also ranked highly across approaches, confirming the importance of greenness exposure. However, the two methods differed in emphasis: SHAP attributed relatively greater importance to behavioral and green space attributes, such as exercise frequency, sleep quality, green space ratio, and GVI, reflecting their nonlinear contributions at the individual level. In contrast, XGBoost highlighted variables such as park area and population density, which serve as frequent splitting criteria in tree construction.
The SHAP value distributions further validated the direction of the main effects (Figure 5). Male gender was consistently associated with higher overweight risk, whereas higher income, more frequent exercise, and better sleep quality were protective. Smoking frequency also showed a negative association, though this should be interpreted cautiously as it reflects body weight outcomes only and does not imply broader health benefits. Several predictors displayed nonlinear or threshold-like patterns, including NDVI, green space ratio, GVI, park area, and distance to the nearest park, as well as behavioral factors such as drinking frequency, exercise frequency, and sleep quality. By contrast, life satisfaction and marital status exerted only weak and unstable effects. Taken together, the convergence of key determinants across methods and the presence of theoretically plausible nonlinearities strengthen confidence in the robustness of the results.

5. Discussion and Limitation

First, our findings support Hypothesis 1, which posits that built environment attributes, including greenness (NDVI, GVI), park accessibility, population density, and public transport proximity, exhibit nonlinear associations with overweight risk, and that the optimal ranges vary across subgroups. NDVI was protective up to ~0.18 before reversing to risk-enhancing, consistent with diminishing returns of greenness [66]. Population density followed a U-shaped pattern, with the lowest risk at ~20,000–60,000 persons/km2 and increased risk beyond 60,000, suggesting that density, like greenness, operates within an optimal range [30]. Subgroup analyses revealed distinct thresholds: park accessibility benefits diminished earlier for unmarried men, whose risk was more strongly linked to alcohol consumption, while GVI remained protective up to ~0.15 for unmarried men and ~0.17 for married women. These patterns demonstrate that “optimal ranges” are socially differentiated, shaped by gendered lifestyles and marital status [53,54,55]. Evidence from green exercise research further supports this interpretation, showing that brief exposures to natural environments produce the largest gains in self-esteem and mood, with benefits that diminish but persist under longer or more intensive engagement [67]. Overall, our findings highlight that built environment effects operate through threshold-dependent mechanisms, underscoring the need to target effective ranges rather than assuming that “more is always better.”
Second, our findings support Hypothesis 2, which posits that married men are more likely to maintain healthier lifestyles and therefore face a lower risk of overweight compared with unmarried men. Overall, men exhibited a higher probability of overweight than women, consistent with prior research [68,69], as shown in Figure 2 (full sample) and Figure 3 (subgroups). Within male subgroups, however, marriage played a protective role: married men maintained healthier routines and thus experienced lower overweight risk relative to their unmarried counterparts. Lifestyle factors contributed most strongly among married men (33.71%), the highest of all subgroups, with drinking frequency ranked first (RI = 8.46), followed by exercise and sleep quality. Although unmarried men were even more sensitive to alcohol (RI = 9.75), their overall lifestyle contribution (32.84%) was slightly lower. In contrast, women displayed smaller lifestyle contributions (31.49% for married, 29.13% for unmarried), indicating stronger influence from socio-demographic and environmental conditions. These patterns suggest that marriage reinforces healthier behaviors and amplifies the benefits of favorable environments, consistent with prior evidence of spousal support shaping men’s diets and health practices [24,70]. Importantly, such protective effects also extended to the built environment, where married men responded more stably and beneficially to greenness and park accessibility [25].
Third, our findings support Hypothesis 3, showing that married women face a higher risk of overweight. Subgroup analyses indicated that unmarried women consistently displayed the lowest risk. For instance, their overweight risk declined with higher NDVI and longer public transport walking time, whereas risk in other groups either plateaued or increased. By contrast, married women exhibited weaker protective effects from exercise and sleep quality, and in some cases even experienced rebound risks at higher levels of NDVI. These outcomes reflect how gendered social roles condition health behaviors: unmarried women often have more discretionary time for outdoor activity and may engage in healthier routines to manage body shape [21], while married women face role overload and reduced autonomy in lifestyle choices, limiting the benefits of protective behaviors [71]. These findings emphasize the need for urban planning and public health interventions that alleviate time constraints and provide safe, accessible environments tailored to women with family responsibilities.
This study has several limitations. First, the cross-sectional design precludes causal inference between built environment attributes and overweight. Second, BMI was partly self-reported, and the dataset lacked detailed measures of diet, food environments, and occupational activity, which may introduce residual confounding. Third, the sample was drawn from urban neighborhoods in Shanghai, so the findings may not be generalizable to rural settings or cities with different social and environmental contexts. Despite these limitations, the study provides novel evidence on nonlinear and socially differentiated associations between built environments and overweight.

6. Conclusions

This study drew on survey data from 2634 residents in Shanghai and applied XGBoost to examine how built environment, lifestyle, and socio-demographic factors are associated with overweight. The analysis revealed nonlinear relationships characterized by thresholds and plateaus rather than simple linear trends. Key built environment attributes, greenness, park accessibility, population density, and transit proximity, all displayed “optimal ranges,” though in distinct nonlinear forms: population density and NDVI followed U-shaped patterns with clear minimum-risk points; park accessibility showed an inverted U-shaped association with a peak protective interval; and transit proximity exhibited oscillating patterns with a specific low-risk range.
The central contribution of this study is to demonstrate that these nonlinear associations are socially differentiated. Gender and marital status condition daily lifestyles and mediate how built environment features influence overweight risk. Subgroup-specific thresholds and turning points highlight the need for contextualized health planning. Characteristics of green space in Shanghai, such as fragmentation, small scale, and limited accessibility [72], may partly explain subgroup differences, while East Asian cultural norms and gendered household roles [73] further magnify the interaction between marriage, lifestyle, and health.
Overall, the findings underscore the importance of socially differentiated strategies in overweight prevention. In high-density cities like Shanghai, planning should maintain greenness and density within effective ranges and improve access to fragmented, small-scale green spaces. Public health measures should also address subgroup vulnerabilities, such as alcohol-related risks among men and time constraints faced by women with caregiving responsibilities. By identifying subgroup-specific thresholds and turning points, this study provides empirical evidence to inform more precise and equitable urban health planning, showing how cultural context, gender, and marital status shape the benefits of environmental exposures.

Author Contributions

Conceptualization, Y.H. and Y.X.; methodology, X.Z.; software, Y.H.; validation, X.Z., Y.X., and Y.H.; formal analysis, X.Z. and Y.X.; investigation, X.Z. and Y.H.; resources, Y.X.; data curation, Y.H.; writing—original draft preparation, X.Z., Y.X., and Y.H.; writing—review and editing, Y.H. and Y.X.; visualization, X.Z.; supervision, Y.H.; project administration, X.Z. and Y.X.; funding acquisition, X.Z. All authors have read and agreed to the published version of the manuscript.

Funding

This work was supported by the Discipline Construction Fund of the College of Architecture and Urban Planning, Tongji University; the Industry-Academia Collaborative Education Project of Tongji University-TJUPDI.

Informed Consent Statement

Informed consent was obtained from all subjects involved in the study. The ethical approval and informed consent documents are attached as Non-published Material.

Data Availability Statement

Data are available from the authors upon reasonable request.

Conflicts of Interest

The authors declare no conflicts of interest.

References

  1. WHO. Obesity and Overweight. Available online: https://www.who.int/news-room/fact-sheets/detail/obesity-and-overweight (accessed on 23 July 2025).
  2. Pan, X.-F.; Wang, L.; Pan, A. Epidemiology and determinants of obesity in China. Lancet Diabetes Endocrinol. 2021, 9, 373–392. Available online: https://www.thelancet.com/diabetes-endocrinology (accessed on 28 July 2025). [CrossRef]
  3. An, R.; Shen, J.; Yang, Q.; Yang, Y. Impact of built environment on physical activity and obesity among children and adolescents in China: A narrative systematic review. J. Sport Health Sci. 2019, 8, 153–169. [Google Scholar] [CrossRef]
  4. Sun, B.; Yan, H.; Zhang, T. Built environmental impacts on individual mode choice and BMI: Evidence from China. J. Transp. Geogr. 2017, 63, 11–21. [Google Scholar] [CrossRef]
  5. Papas, M.A.; Alberg, A.J.; Ewing, R.; Helzlsouer, K.J.; Gary-Webb, T.L.; Klassen, A.C. The built environment and obesity. Epidemiol. Rev. 2007, 29, 129–143. [Google Scholar] [CrossRef]
  6. Day, K. Built environmental correlates of physical activity in China: A review. Prev. Med. Rep. 2016, 3, 303–316. [Google Scholar] [CrossRef] [PubMed]
  7. Cen, L.; Xiao, Y. Assessing Urban Park Accessibility via Population Projections: Planning for Green Equity in Shanghai. Land 2025, 14, 1580. [Google Scholar] [CrossRef]
  8. Klompmaker, J.O.; Hoek, G.; Bloemsma, L.D.; Gehring, U.; Strak, M.; Wijga, A.H.; van den Brink, C.; Brunekreef, B.; Lebret, E.; Janssen, N.A.H. Green space definition affects associations of green space with overweight and physical activity. Environ. Res. 2018, 160, 531–540. [Google Scholar] [CrossRef]
  9. Lopez, R.P.; Hynes, H.P. Obesity, physical activity, and the urban environment: Public health research needs. Environ. Health 2006, 5, 25. [Google Scholar] [CrossRef] [PubMed]
  10. Matheson, E.M.; King, D.E.; Everett, C.J. Healthy lifestyle habits and mortality in overweight and obese individuals. J. Am. Board Fam. Med. 2012, 25, 9–15. [Google Scholar] [CrossRef] [PubMed]
  11. Gubbels, J.S.; van Assema, P.; Kremers, S.P.J. Physical activity, sedentary behavior, and dietary patterns among children. Curr. Nutr. Rep. 2013, 2, 105–112. [Google Scholar] [CrossRef]
  12. Iannotti, R.J.; Wang, J. Patterns of physical activity, sedentary behavior, and diet in U.S. adolescents. J. Adolesc. Health 2013, 53, 280–286. [Google Scholar] [CrossRef]
  13. Buchvold, H.V.; Pallesen, S.; Øyane, N.M.F.; Bjorvatn, B. Associations between night work and BMI, alcohol, smoking, caffeine and exercise—A cross-sectional study. BMC Public Health 2015, 15, 1112. [Google Scholar] [CrossRef]
  14. Hayes, J.F.; Balantekin, K.N.; Altman, M.; Wilfley, D.E.; Taylor, C.B.; Williams, J. sleep patterns and quality are associated with severity of obesity and weight-related behaviors in adolescents with overweight and obesity. Child. Obes. 2018, 14, 11–17. [Google Scholar] [CrossRef]
  15. Grant, N.; Wardle, J.; Steptoe, A. The relationship between life satisfaction and health behavior: A cross-cultural analysis of young adults. Int. J. Behav. Med. 2009, 16, 259–268. [Google Scholar] [CrossRef] [PubMed]
  16. Kuroki, M. Life satisfaction, overweightness and obesity. Int. J. Wellbeing 2016, 6, 93–110. [Google Scholar] [CrossRef]
  17. Lissner, L.; Wijnhoven, T.M.A.; Mehlig, K.; Sjöberg, A.; Kunesova, M.; Yngve, A.; Petrauskiene, A.; Duleva, V.; Rito, A.I.; Breda, J. Socioeconomic inequalities in childhood overweight: Heterogeneity across five countries in the WHO European Childhood Obesity Surveillance Initiative (COSI–2008). Int. J. Obes. 2016, 40, 796–802. [Google Scholar] [CrossRef]
  18. Tremblay, M.S.; Pérez, C.E.; Ardern, C.I.; Bryan, S.N.; Katzmarzyk, P.T. Obesity, overweight and ethnicity. Health Rep. 2005, 16, 23–34. [Google Scholar] [PubMed]
  19. Marmot, M.; Wilkinson, R. Social Determinants of Health; Oup Oxford: Oxford, UK, 2005. [Google Scholar]
  20. Navarrete-Hernandez, P.; Vetro, A.; Concha, P. Building safer public spaces: Exploring gender difference in the perception of safety in public space through urban design interventions. Landsc. Urban Plan. 2021, 214, 104180. [Google Scholar] [CrossRef]
  21. Wood, W.; Eagly, A.H. A cross-cultural analysis of the behavior of women and men: Implications for the origins of sex differences. Psychol. Bull. 2002, 128, 699–727. [Google Scholar] [CrossRef]
  22. Haenle, M.M.; Brockmann, S.O.; Kron, M.; Bertling, U.; Mason, R.A.; Steinbach, G.; Boehm, B.O.; Koenig, W.; Kern, P.; Piechotowski, I.; et al. Overweight, physical activity, tobacco and alcohol consumption in a cross-sectional random sample of German adults. BMC Public Health 2006, 6, 233. [Google Scholar] [CrossRef]
  23. Leasure, J.L.; Neighbors, C.; Henderson, C.E.; Young, C.M. Exercise and alcohol consumption: What we know, what we need to know, and why it is important. Front. Psychiatry 2015, 6, 156. [Google Scholar] [CrossRef] [PubMed]
  24. Lillard, L.A.; Panis, C.W.A. Marital status and mortality: The role of health. Demography 1996, 33, 313–327. [Google Scholar] [CrossRef]
  25. Averett, S.L.; Sikora, A.; Argys, L.M. For better or worse: Relationship status and body mass index. Econ. Hum. Biol. 2008, 6, 330–349. [Google Scholar] [CrossRef]
  26. Cheng, L.; De Vos, J.; Zhao, P.; Yang, M.; Witlox, F. Examining non-linear built environment effects on elderly’s walking: A random forest approach. Transp. Res. Part D Transp. Environ. 2020, 88, 102552. [Google Scholar] [CrossRef]
  27. Yin, C.; Cao, J.; Sun, B. Examining non-linear associations between population density and waist-hip ratio: An application of gradient boosting decision trees. Cities 2020, 107, 102899. [Google Scholar] [CrossRef]
  28. Shanahan, D.F.; Fuller, R.A.; Bush, R.; Lin, B.B.; Gaston, K.J. The health benefits of urban nature: How much do we need? BioScience 2015, 65, 476–485. [Google Scholar] [CrossRef]
  29. Xiao, Y.; Cen, L. Do new urban parks really improve green equity? A longitudinal analysis of Shanghai (2000–2020). Landsc. Urban Plan. 2026, 265, 105519. [Google Scholar] [CrossRef]
  30. Miao, S.; Xiao, Y. Exploring the spatial trade-off effects of green space on older people’s physical inactivity: Evidence from Shanghai. Landsc. Urban Plan. 2024, 251, 105155. [Google Scholar] [CrossRef]
  31. Barr, A.; Bentley, R.; Simpson, J.A.; Scheurer, J.; Owen, N.; Dunstan, D.; Thornton, L.; Krnjacki, L.; Kavanagh, A. Associations of public transport accessibility with walking, obesity, metabolic syndrome and diabetes. J. Transp. Health 2016, 3, 141–153. [Google Scholar] [CrossRef]
  32. Raza, A.; Pulakka, A.; Hanson, L.L.M.; Westerlund, H.; Halonen, J.I. Commuting distance and behavior-related health: A longitudinal study. Prev. Med. 2021, 150, 106665. [Google Scholar] [CrossRef] [PubMed]
  33. Williams, K.; Umberson, D. Marital Status, Marital Transitions, and Health: A Gendered Life Course Perspective. J. Health Soc. Behav. 2004, 45, 81–98. [Google Scholar] [CrossRef] [PubMed]
  34. Wang, Y.; Jiao, Y.; Nie, J.; O’nEil, A.; Huang, W.; Zhang, L.; Han, J.; Liu, H.; Zhu, Y.; Yu, C.; et al. Sex differences in the association between marital status and the risk of cardiovascular, cancer, and all-cause mortality: A systematic review and meta-analysis of 7,881,040 individuals. Glob. Health Res. Policy 2020, 5, 4. [Google Scholar] [CrossRef]
  35. Valson, J.S.; Kutty, V.R. Gender differences in the relationship between built environment and non-communicable diseases: A systematic review. J. Public Health Res. 2018, 7, 43–49. [Google Scholar] [CrossRef]
  36. Alzamil, H.A.; Alhakbany, M.A.; Alfadda, N.A.; Almusallam, S.M.; Al-Hazzaa, H.M. A profile of physical activity, sedentary behaviors, sleep, and dietary habits of Saudi college female students. J. Fam. Community Med. 2019, 26, 1–8. [Google Scholar] [CrossRef]
  37. Janssen, I.; Katzmarzyk, P.T.; Boyce, W.F.; King, M.A.; Pickett, W. Overweight and obesity in Canadian adolescents and their associations with dietary habits and physical activity patterns. J. Adolesc. Health 2004, 35, 360–367. [Google Scholar] [CrossRef]
  38. Martínez-De-Quel, Ó.; Suárez-Iglesias, D.; López-Flores, M.; Pérez, C.A. Physical activity, dietary habits and sleep quality before and during COVID-19 lockdown: A longitudinal study. Appetite 2021, 158, 105019. [Google Scholar] [CrossRef]
  39. Moschonis, G.; Trakman, G.L. Overweight and obesity: The interplay of eating habits and physical activity. Nutrients 2023, 15, 2896. [Google Scholar] [CrossRef] [PubMed]
  40. Strasser, B. Physical activity in obesity and metabolic syndrome. Ann. N. Y. Acad. Sci. 2013, 1281, 141–159. [Google Scholar] [CrossRef] [PubMed]
  41. Vilela, D.L.; Fonseca, P.G.; Pinto, S.L.; Bressan, J. Influence of dietary patterns on the metabolically healthy obesity phenotype: A systematic review. Nutr. Metab. Cardiovasc. Dis. 2021, 31, 2779–2791. [Google Scholar] [CrossRef]
  42. Gonnissen, H.K.; Adam, T.C.; Hursel, R.; Rutters, F.; Verhoef, S.P.; Westerterp-Plantenga, M.S. Sleep duration, sleep quality and body weight: Parallel developments. Physiol. Behav. 2013, 121, 112–116. [Google Scholar] [CrossRef]
  43. Audrain-McGovern, J.; Benowitz, N.L. Cigarette smoking, nicotine, and body weight. Clin. Pharmacol. Ther. 2011, 90, 164–168. [Google Scholar] [CrossRef]
  44. Suter, P.M.; Tremblay, A. Is alcohol consumption a risk factor for weight gain and obesity? Crit. Rev. Clin. Lab. Sci. 2005, 42, 197–227. [Google Scholar] [CrossRef]
  45. Vieira, P.N.; Silva, M.N.; Mata, J.; Coutinho, S.R.; Santos, T.C.; Sardinha, L.B.; Teixeira, P.J. Correlates of health-related quality of life, psychological well-being, and eating self-regulation after successful weight loss maintenance. J. Behav. Med. 2013, 36, 601–610. [Google Scholar] [CrossRef] [PubMed]
  46. Proper, K.; Cerin, E.; Brown, W.; Owen, N. Sitting time and socio-economic differences in overweight and obesity. Int. J. Obes. 2007, 31, 169–176. [Google Scholar] [CrossRef]
  47. Wake, M.; Hardy, P.; Canterford, L.; Sawyer, M.; Carlin, J.B. Overweight, obesity and girth of Australian preschoolers: Prevalence and socio-economic correlates. Int. J. Obes. 2007, 31, 1044–1051. [Google Scholar] [CrossRef]
  48. Dinour, L.; Leung, M.M.; Tripicchio, G.; Khan, S.; Yeh, M.-C. The association between marital transitions, body mass index, and weight: A review of the literature. J. Obes. 2012, 2012, 294974. [Google Scholar] [CrossRef]
  49. García-Álvarez, A.; Serra-Majem, L.; Ribas-Barba, L.; Castell, C.; Foz, M.; Uauy, R.; Plasencia, A.; Salleras, L. Obesity and overweight trends in Catalonia, Spain (1992–2003): Gender and socio-economic determinants. Public Health Nutr. 2007, 10, 1368–1378. [Google Scholar] [CrossRef]
  50. Bove, C.F.; Olson, C.M. Obesity in low-income rural women: Qualitative insights about physical activity and eating patterns. Women Health 2006, 44, 57–78. [Google Scholar] [CrossRef] [PubMed]
  51. Mata, J.; Richter, D.; Schneider, T.; Hertwig, R. How cohabitation, marriage, separation, and divorce influence BMI: A prospective panel study. Health Psychol. 2018, 37, 948–958. [Google Scholar] [CrossRef] [PubMed]
  52. Liu, G.C.; Wilson, J.S.; Qi, R.; Ying, J. Green neighborhoods, food retail and childhood overweight: Differences by population density. Am. J. Health Promot. 2007, 21 (Suppl. S4), 317–325. [Google Scholar] [CrossRef]
  53. Paraschiv-Ionescu, A.; Buchser, E.; Rutschmann, B.; Aminian, K. Nonlinear analysis of human physical activity patterns in health and disease. Phys. Rev. E Stat. Nonlinear Soft Matter Phys. 2008, 77, 021913. [Google Scholar] [CrossRef]
  54. Wang, F.; Sun, Z.; Lin, F.; Xu, Y.; Wu, E.; Sun, X.; Zhou, X.; Wu, Y. Nonlinear relationships between sleep duration, mental health, and quality of life: The dangers of less sleep versus more sleep. Sleep Med. 2024, 119, 565–573. [Google Scholar] [CrossRef]
  55. Visontay, R.; Sunderland, M.; Slade, T.; Wilson, J.; Mewton, L. Are there non-linear relationships between alcohol consumption and long-term health? A systematic review of observational studies employing approaches to improve causal inference. BMC Med Res. Methodol. 2022, 22, 16. [Google Scholar] [CrossRef]
  56. Yin, C.; Zhang, Z.; Dai, S.; Chen, Y. Built environments and obesity: A framework considering residences, commute routes, and workplaces. Cities 2025, 160, 105842. [Google Scholar] [CrossRef]
  57. Chinese Center for Disease Control and Prevention. Say No to Obesity. Available online: https://en.chinacdc.cn/in_focus/202208/t20220823_260921.html (accessed on 1 August 2025).
  58. Daniels, S.R. The use of BMI in the clinical setting. Pediatrics 2009, 124 (Suppl. S1), S35–S41. [Google Scholar] [CrossRef] [PubMed]
  59. Gosse, M.A. How accurate is self-reported BMI? Nutr. Bull. 2014, 39, 105–114. [Google Scholar] [CrossRef]
  60. Chen, T.; Guestrin, C. Xgboost: A scalable tree boosting system. In Proceedings of the 22nd Acm Sigkdd International Conference on Knowledge Discovery and Data Mining, San Francisco, CA, USA, 17 August 2016; pp. 785–794. Available online: https://hdl.handle.net/10419/181026 (accessed on 1 August 2025).
  61. Friedman, J.H. Greedy function approximation: A gradient boosting machine. Ann. Stat. 2001, 29, 1189–1232. [Google Scholar] [CrossRef]
  62. Zhang, X.; Yan, C.; Gao, C.; Malin, B.A.; Chen, Y. Predicting missing values in medical data via XGBoost regression. J. Health Inform. Res. 2020, 4, 383–394. [Google Scholar] [CrossRef] [PubMed]
  63. Chen, Y.; Zhang, X.; Grekousis, G.; Huang, Y.; Hua, F.; Pan, Z.; Liu, Y. Examining the importance of built and natural environment factors in predicting self-rated health in older adults: An extreme gradient boosting (XGBoost) approach. J. Clean. Prod. 2023, 413, 137432. [Google Scholar] [CrossRef]
  64. Ma, J.; Ding, Y.; Cheng, J.C.P.; Tan, Y.; Gan, V.J.L.; Zhang, J. Analyzing the leading causes of traffic fatalities using XGBoost and grid-based analysis: A city management perspective. IEEE Access 2019, 7, 148059–148072. [Google Scholar] [CrossRef]
  65. Molnar, C. Interpretable Machine Learning; Lulu. com: Morrisville, NC, USA, 2020. [Google Scholar]
  66. Barton, J.; Pretty, J. What is the Best Dose of Nature and Green Exercise for Improving Mental Health? A Multi-Study Analysis. Environ. Sci. Technol. 2010, 44, 3947–3955. [Google Scholar] [CrossRef] [PubMed]
  67. Scully, D.; Kremer, J.; Meade, M.M.; Graham, R.; Dudgeon, K. Physical exercise and psychological well being: A critical review. Br. J. Sports Med. 1998, 32, 111–120. [Google Scholar] [CrossRef] [PubMed]
  68. Kamel, E.G.; McNeill, G. Men are less aware of being overweight than women. Obes. Res. 2000, 8, 604. [Google Scholar] [CrossRef] [PubMed]
  69. Tsai, S.A.; Lv, N.; Xiao, L.; Ma, J. Gender Differences in Weight-Related Attitudes and Behaviors Among Overweight and Obese Adults in the United States. Am. J. Men’s Health 2016, 10, 389–398. [Google Scholar] [CrossRef]
  70. Kemmer, D.; Anderson, A.; Marshall, D. Living together and eating together: Changes in food choice and eating habits during the transition from single to married/cohabiting. Sociol. Rev. 1998, 46, 48–72. [Google Scholar] [CrossRef]
  71. Artazcoz, L.; Borrell, C.; Cortàs, I.; Escribà-Agüir, V.; Cascant, L. Occupational epidemiology and work related inequalities in health: A gender perspective for two complementary approaches to work and health research. J. Epidemiol. Community Health 2007, 61, ii39–ii45. [Google Scholar] [CrossRef]
  72. Nguyen, P.-Y.; Astell-Burt, T.; Rahimi-Ardabili, H.; Feng, X. Green Space Quality and Health: A Systematic Review. Int. J. Environ. Res. Public Health 2021, 18, 11028. [Google Scholar] [CrossRef]
  73. Yang, W.-S.; Yen, P.-C. A Comparative Study of Marital Dissolution in East Asian Societies: Gender Attitudes and Social Expectations towards Marriage in Taiwan, Korea and Japan. Asian J. Soc. Sci. 2011, 39, 751–775. Available online: http://www.jstor.org/stable/43498087 (accessed on 4 August 2025). [CrossRef]
Land 14 02064 g001
Figure 1. Sample distributions.
Figure 1. Sample distributions.
Land 14 02064 g001
Land 14 02064 g002
Figure 2. Nonlinear Relationship Analysis of All Samples.
Figure 2. Nonlinear Relationship Analysis of All Samples.
Land 14 02064 g002
Land 14 02064 g003
Figure 3. Nonlinear Relationship Analysis of Subgroup Sample.
Figure 3. Nonlinear Relationship Analysis of Subgroup Sample.
Land 14 02064 g003
Land 14 02064 g004
Figure 4. SHAP summary bar plot.
Figure 4. SHAP summary bar plot.
Land 14 02064 g004
Land 14 02064 g005
Figure 5. SHAP beeswarm plot.
Figure 5. SHAP beeswarm plot.
Land 14 02064 g005
Table 1. Personal Characteristics.
Table 1. Personal Characteristics.
Mean (Tiesdell & Allmendinger)/%
Personal characteristics
Overweight
Yes42%
No58%
Sex
Male37%
Female63%
Married
Married80%
Unmarried20%
Education level
Primary or below3%
Secondary (Junior high/vocational high/general high/technical school)62%
Undergraduate33%
Postgraduate or above3%
Age (years)57.03 (13.89)
Log personal income10.20 (1.70)
Table 2. Behavioral Habits and Built Environment Variables.
Table 2. Behavioral Habits and Built Environment Variables.
VariablesMean (Tiesdell & Allmendinger)/%
Lifestyle attributes
Exercise Freq
No exercise18%
≤1/month8%
Several times/month8%
Several times/week18%
Daily48%
Drinking Freq
Former drinker8%
<1/week or none81%
1–2/week5%
3–4/week1%
Nearly daily5%
Smoking Freq
Non-smoker70%
Non-smoker, passive smoke8%
Former smoker6%
Few cigarettes/week2%
Daily (few)10%
Daily (≥20/day)4%
Sleep Quality
Much worse than usual2%
Worse than usual12%
About the same37%
As good as usual49%
Life Satisfaction
Very dissatisfied2%
Dissatisfied4%
Neutral14%
Satisfied54%
Very satisfied27%
Built environment
NDVI (1000 m buffer)0.16 (0.03)
Distance to nearest park (m)651.01 (333.55)
Green View Index (1000 m buffer)0.22 (0.06)
Park area within 1000 m (m2)78,102.14 (55,370.08)
Population density (persons/km2)37,744.85 (19,631.22)
Green space ratio (%)26.89 (15.71)
Public transport accessibility (min)4.18 (2.66)
Table 3. RI of predictors for overweight.
Table 3. RI of predictors for overweight.
VariablesAll Sample-RI, %All Sample-RankingMale-Married-RI, %Male-Married-RankingMale-Unmarried-RI, %Male-Unmarried-RankingFemale-Married-RI, %Female-Married-RankingFemale-Unmarried-RI, %Female-Unmarried-Ranking
Built Environment Variables
Total37.79 47.41 49.05 47 48.24
Park area within 1000 m (m2)5.37106.7286.8667.0746.874
Population density5.5467.1826.33116.8468.482
NDVI (1000 m buffer)5.4687.0636.8376.55106.4810
Green space ratio5.25136.29137.3736.6496.579
Distance to nearest park (m)5.6556.8766.9755.89156.786
Green View Index (1000 m buffer)5.4197.0146.6297.5716.3711
Public transport accessibility (min)5.11166.3118.0826.45116.697
Lifestyle attributes variables
Total26.64 33.71 32.84 31.49 29.13
Exercise frequency 5.21146.29125.48136.6486.3512
Drinking frequency5.7548.4619.7516.1125.5713
Smoking frequency5.11156.6295.62126.07136.815
Sleep quality5.3115.95146.7486.775.0215
Life satisfaction score5.27126.39105.24145.98145.3814
Sociodemographics variables
Total35.57 18.88 18.11 21.51 22.63
Gender (1 = male)12.231016016016016
Education level7.325.15154.42157.528.721
Marital status (1 = married)4.7517017017017017
Log personal income5.5176.8576.36107.1636.688
Age (squared)5.7836.8857.3346.8457.223
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Zhong, X.; Xiao, Y.; Huang, Y. Nonlinear Associations Between Built Environment and Overweight: Gender and Marital Status Differences in Urban China. Land 2025, 14, 2064. https://doi.org/10.3390/land14102064

AMA Style

Zhong X, Xiao Y, Huang Y. Nonlinear Associations Between Built Environment and Overweight: Gender and Marital Status Differences in Urban China. Land. 2025; 14(10):2064. https://doi.org/10.3390/land14102064

Chicago/Turabian Style

Zhong, Xiaohua, Yang Xiao, and Yihui Huang. 2025. "Nonlinear Associations Between Built Environment and Overweight: Gender and Marital Status Differences in Urban China" Land 14, no. 10: 2064. https://doi.org/10.3390/land14102064

APA Style

Zhong, X., Xiao, Y., & Huang, Y. (2025). Nonlinear Associations Between Built Environment and Overweight: Gender and Marital Status Differences in Urban China. Land, 14(10), 2064. https://doi.org/10.3390/land14102064

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop