The Association Between the Built Environment and Insufficient Physical Activity Risk Among Older Adults in China: Urban–Rural Differences and Non-Linear Effects

Abstract

The built environment has been widely recognized as a critical determinant of physical activity among older adults. However, urban–rural disparities and the non-linear effects of environmental features remain underexplored. Using interpretable machine learning (random forest model) on nationwide representative data from 2526 older adults in the China Health and Retirement Longitudinal Study (CHARLS) database, this study identified both common and distinct risk factors for insufficient moderate-to-vigorous physical activity (MVPA) across diverse urban and rural contexts. The results revealed a location-based gradient in physical activity insufficiency: rural areas < suburban areas < central urban areas. Rural older adults faced greater constraints from safety concerns and transportation accessibility limitations. In comparison, urban older adults would benefit from targeted improvements in built environment quality, particularly elevator accessibility and diverse public activity spaces. Furthermore, non-linear relationships were observed between built environment features and physical activity, elucidating the “density paradox”: while moderate urban compactness promoted active behaviors, excessive density (>24,000 persons/km²), perceived overcrowding, and over-proximity to specific facilities (<1 km) were linked to reduced MVPA. These findings underscore the necessity for differentiated policy interventions in urban and rural settings to address the distinct environmental needs of older adults. Meanwhile, in urban planning, it is crucial that we balance spatial compactness and functional diversity within optimal thresholds for creating sustainable and inclusive built environments. Although a compact design may enhance mobility, equal attention must be paid to preventing spatial disorder from over-densification.

1. Introduction

The concurrent progression of global aging and urbanization has precipitated a widespread epidemic of physical inactivity, now recognized as a major public health challenge. Approximately 5.3 million deaths annually are attributed to insufficient physical activity [1]. The World Health Organization (WHO) recommends that adults should engage in at least 150–300 min of moderate-intensity aerobic physical activity, 75–150 min of vigorous-intensity aerobic physical activity, or an equivalent combination of moderate- and vigorous-intensity physical activities (MVPA) per week to achieve significant health benefits [2]. Levels below these recommendations are generally classified as physical inactivity or insufficient physical activity. In 2022, nearly one-third (31.3%) of the global adult population, approximately 1.8 billion individuals, failed to meet these recommendations [3]. Notably, older adults exhibit the lowest physical activity levels across all age demographics. Data from the U.S. National Center for Health Statistics (NCHS) indicate that only 15.3% of older men and 10.8% of older women meet the recommended physical activity guidelines [4]. In developing countries, physical inactivity has emerged as a growing public health concern, driven by rapid urbanization, industrialization, and the increasing dependence on motorized transportation [5,6,7]. China exemplifies this trend, with adult physical activity levels declining by over 30% between 1991 and 2006 [8]. This decline is also characterized by significant urban–rural disparities. The International Collaborative Study of Cardiovascular Disease in Asia reported that 78.1% of rural residents in China engage in physical exercise, compared to only 21.8% of urban residents [9]. A nationwide survey conducted by the Chinese Center for Disease Control and Prevention revealed that 71% of older adults reported no leisure-time MVPA. Additionally, urban older adults, particularly those with higher educational attainment or household income, tend to have longer sedentary times (4.5 h/day) compared to their rural counterparts (4.1 h/day) [10].

China is currently undergoing a profound socioeconomic transformation, characterized by rapid urbanization and accelerating population aging. This dual transition coincides with a paradigm shift in urban development strategies, where the traditional growth-centric model emphasizing spatial expansion is being superseded by a quality-oriented approach focused on the retrofitting and enhancement of existing built environments. Within this transitional context, this study examines the urban–rural differences and non-linear associations between built environment features and physical activity among older adults, aiming to provide insights and practical recommendations for enhancing environmental quality and creating age-friendly communities in both urban and rural settings. While extant research has thoroughly examined physical activity determinants from both public health and urban planning perspectives, this study advocates for a more comprehensive and dynamic framework, emphasizing the complex interactions among personal attributes, social relationships, environmental contexts, and policy factors [11]. Furthermore, given the substantial heterogeneity in environmental contexts between urban and rural areas, capturing geographic variations is essential for the precise assessment of environmental exposures and their health impacts [12].

The existing research suggests that environmental features differentially influence physical activity among older adults in urban and rural settings. Older adults in rural areas frequently encounter structural and environmental barriers, including inadequate infrastructure, limited access to recreational facilities, perceived safety concerns, and seasonal challenges such as winter slipping hazards [13,14]. For instance, rural road development does not invariably enhance the mobility of older adults; rather, busy highways primarily used for industrial traffic may hinder their leisure physical activity. Pelletier et al. (2020) noted that one respondent preferred redesigning local alleyways into pedestrian-friendly pathways suitable for walking, jogging, and cycling [15]. Moreover, while rural residents are more likely to engage in physical activity through natural environments, urban dwellers show a greater dependence on recreational infrastructure for exercise opportunities. However, the role of natural settings in rural physical activity is also complex. While natural settings can serve as facilitators by offering proximity and opportunities, they may also pose barriers due to risks such as wildlife encounters and harsh weather conditions [15]. Urban and rural development in China has long been shaped by a dual economic and social structure, resulting in substantial disparities in resource allocation, economic development, and infrastructure quality between urban and rural areas [16,17]. These disparities underscore the necessity for further research examining the heterogeneous needs of aging populations across different geographical contexts. Such investigations are crucial for guiding the development of more equitable and age-friendly built environments that can accommodate diverse living circumstances.

In terms of urban areas, the compact city paradigm has gained substantial recognition in urban planning for its efficacy in promoting active lifestyles. Grounded in the theoretical framework of New Urbanism, this planning approach advocates for land use efficiency through key urban design elements, including increased residential density, well-connected street networks, mixed land use, and enhanced accessibility to public transportation and destinations, all of which help encourage residents to engage in more walking and cycling activities [18,19,20]. Designing pedestrian- and cyclist-friendly infrastructure, optimizing residential density, reducing the distance to public transportation, and enhancing environmental safety, convenience, and aesthetic appeal are essential strategies for creating healthier living environments [21]. A global study focusing on moderate-to-vigorous physical activity further underscored the importance of factors such as net residential density, intersection density, public transport accessibility, and the availability of parks [22]. Another study conducted in Hong Kong, a prototypical high-density metropolis in East Asia, demonstrated that older adults consistently maintain physical activity levels exceeding recommended guidelines despite the constraints posed by a high density and heavy traffic conditions. This phenomenon is primarily attributed to the high accessibility and well-designed pedestrian-friendly infrastructure [23].

However, emerging evidence suggests that high-density urban environments and mixed-use developments do not always yield the anticipated positive effects on physical activity. For instance, residents in high-density areas may exhibit an increased reliance on public transit for daily commuting, which could inadvertently reduce opportunities for moderate-to-vigorous physical activity (MVPA). Empirical studies indicated that older adults living in urban areas were less likely to engage in regular cycling compared to their counterparts in semi-urban or lower-density communities [24]. Overdensity may also exacerbate safety concerns for older adults. Deficiencies in urban infrastructure, such as poorly maintained pavements, heavy traffic, insufficient rest areas, and dangerous intersections, have been shown to increase older adults’ fear of falling or accidents [25]. A Japanese study revealed an unexpected negative correlation between residential density and land use diversity and leisure walking among women, primarily due to excess car and pedestrian traffic [26].

Therefore, although extensive research suggests that a compact urban design positively influences physical activity, it often overlooks the potential environmental risk exposure associated with excessive densification. Moreover, the observed inconsistencies in findings about the environment–behavior relationship may be attributable to methodological limitations, including small-scale sampling biases and contextual disparities between Asian (high-density, mixed-use) and Western (low-density, car-centric) urban layouts. Particularly, urban environments characterized by homogeneity and a low density often exhibit limited variability in built environment features, which may result in an underestimation of the strength of their associations. The threshold effect refers to a specific type of non-linear relationship where the impact significantly increases or decreases once a critical point is exceeded. Certain features of the built environment may not have consistent effects across the entire range, making it essential that we identify appropriate thresholds to better understand the complex interaction between environmental features and individual behaviors. In addition, given that older adults are particularly more sensitive and vulnerable to adverse features under high-density environments, targeted empirical studies are imperative to explore the subtle associations among compact urban design parameters, distinct age cohorts, and specific physical activity behaviors. Such a focused inquiry would enable a more precise assessment of a compact urban design’s applicability and identify the critical thresholds where its benefits may diminish or be offset by potential risks.

This study addressed existing research gaps by employing a systematic framework and nationally representative data to investigate the association between built environment attributes and insufficient moderate-to-vigorous physical activity (MVPA) among older adults in urban and rural settings. Recognizing the limitations of traditional linear regression models in capturing non-linear relationships, we employed advanced machine-learning algorithms to elucidate the complex interplay between environmental factors and physical activity behaviors. Using data from the China Health and Retirement Longitudinal Study (CHARLS), we examined a sample of 2526 adults aged 60 and older to develop a random forest model for predicting the risk of insufficient physical activity. Moreover, our analysis distinguished between urban and rural contexts, identifying key modifiable environmental factors that differentially influence physical activity levels across settings. Finally, the findings revealed non-linear associations between built environment features and older adults’ MVPA, providing both empirical evidence and practical implications for creating age-friendly and sustainable built environments.

2. Materials and Methods

2.1. Data Resources

The data were obtained from the China Health and Retirement Longitudinal Study (CHARLS) [27], a nationally representative longitudinal survey widely recognized for its rigorous methodology in monitoring the health status of older adults in China. The CHARLS was launched in 2011 and employed a rigorous stratified multistage random sampling strategy (county/district–village/community–households). Given that community environmental information is only available from the 2011 baseline community questionnaire, this study utilized a baseline survey covering 10,257 households and 17,708 individuals in 450 villages/communities across 150 counties in 28 provinces. Participants were middle-aged and older adults (≥45 years), distributed as 52.67% rural and 47.33% urban residents, with 40% aged ≥60 years and 52.1% female.

2.2. Sample Selection

Participants were excluded from the baseline survey data if they met any of the following criteria: (1) missing more than 20% of individual-level characteristic variables, (2) aged below 60, or (3) lacking key information on physical activity. The final analytical sample comprised 2526 individuals, with an average age of 68 years old. Among the study participants, 1017 (40.26%) did not meet the WHO-recommended threshold of 600 MET-minutes/week for moderate-to-vigorous physical activity (MVPA).

2.3. Variables

The outcome variable in this study was insufficient moderate-to-vigorous physical activity (MVPA), defined as a binary variable (meeting/not meeting WHO guidelines) based on International Physical Activity Questionnaire (IPAQ) data from the CHARLS 2011 baseline survey. The study included six categories of independent variables, comprising a total of 70 candidate variables: (1) basic demographic factors: gender and age; (2) socioeconomic factors: household income and years of education; (3) individual physical health factors: activities of daily living (ADL), functional mobility, number of chronic diseases, and pain intensity; (4) lifestyle factors: social interaction and fall experience; (5) housing environment factors: construction year, housing area, elevators, shower facilities, and flushable toilets; and (6) community environment factors: location, community economic status, proportion of older adults in the community, population density, distance to transit, destination accessibility, diversity of daily facilities, public activity spaces and facilities, and aesthetic and design. Descriptive statistics for all variables are presented in Appendix A 

Table A1. Descriptive statistics.

	Categories	Variables	Mean	Std	Min	Max
1	Outcome	Physical activity	0.40	0.49	0	1
		(MET)	7750.14	6321.56	60	20,160
2	Demographic Factors	Urban/Rural (01, 1 = urban)	0.39	0.49	0	1
3		Age	67.65	6.37	60	92
4		Gender (1 = female, 0 = male)	0.51	0.50	0	1
5	Socioeconomic Factors	Years of education	3.60	4.45	0	19
6		Individual income	5034.68	10,156.46	0	50,000
7		Household income	16,188.90	23,201.29	0	100,000
8	Health Status	Disability (1 = yes) (deleted)	0.04	0.19	0	1
9		Poor eyesight (1 = yes)	0.40	0.49	0	1
10		Poor hearing (1 = yes)	0.19	0.39	0	1
11		Poor sleeping (1 = yes)	0.54	0.50	0	1
12		Chronic diseases (1 = yes)	0.75	0.43	0	1
13		Number of chronic diseases	1.65	1.48	0	8
14		Pain intensity	1.74	1.10	1	4
15		BMI	0.37	0.48	0	1
16		ADL	3.86	0.30	1.91	4
17		Functional mobility	3.53	0.48	1.33	4
18	Lifestyle	Social interaction (type × frequency)	1.51	1.94	0	12
19		Fall experience	0.18	0.39	0	1
20	Housing Built Environmental Factors	House area	112.36	69.99	8	300
21		Years since construction	22.49	14.44	0	50
22		Old residential (1 = old)	0.55	0.50	0	1
23		Elevator (01, 1 = have)	0.80	0.40	0	1
24		Accessible facilities (1 = have)	0.24	0.43	0	1
25		Steps	2.17	1.12	1	5
26		Distance to toilet	7.31	10.61	0	30
27		Type of toilet (1 = toilet without a seat)	0.82	0.38	0	1
28		Flushable toilet (1 = no flush)	0.61	0.49	0	1
29		Bath facilities (1 = no bath)	0.68	0.47	0	1
30		Lack of electrification	0.18	0.38	0	1
31		No heating	0.91	0.29	0	1
32		No phone (deleted)	0.48	0.50	0	1
33		No wifi	0.89	0.32	0	1
34		Tidiness (1 = tidy)	3.07	1.05	1	5
35		Temperature (1 = hot)	2.90	0.49	1	5
Community Natural, Social, and Built Environmental Factors
36	Location	Location (1 = urban, 0 = suburban)	0.50	0.86	0	2
37	Economic and	Economic perception (1 = poor, 7 = rich)	3.76	1.36	1	7
38		Community economic conditions (per-capita net income)	6200.09	7610.91	0	30,000
39		Pension to persons older than 65 (1 = yes, 0 = no)	0.40	0.00	1
40		Tenant proportion	0.20	0.38	0	1
41		The proportion of the migrant population	0.07	0.15	0	0.92
42		The proportion of the population aged 65 and above	0.16	0.12	0.01	0.93
43	Weather	The proportion of rainy and snowy days	0.17	0.12	0.00	0.91
44	Density	Community population density	3381.00	10,719.00	0.03	272,727
45		Subjective (1 = crowded, 7 = sparse)	4.66	1.60	1	7
46	Road Conditions	Have roads passing through (1 = yes, 0 = no) (deleted)	0.93	0.26	0	1
47		Types of roads	1.42	0.68	1	5
48		Number of days with road impassable	0.09	0.24	0	1
49	Distance to Transit	Number of bus lines	2.51	4.40	0	20
50		Distance of bus stops	1.88	3.12	0	10
51	Accessibility of Medical Resources	Distance to the general hospital	31.22	20.85	0	50
52		Distance to the pharmacy store (deleted)	0.72	1.71	0	10
53		Distance to the healthcare facilities	2.50	3.45	0	10
54	Diversity of Daily Facilities	Number of kindergartens	0.85	1.34	0	10
55		Number of primary schools (deleted)	0.73	0.82	0	6
56		Number of post offices	0.18	0.44	0	3
57		Number of police stations	0.51	0.91	0	15
58		Number of banks	0.61	1.29	0	9
59		Number of convenience stores	10.49	13.52	0	50
60		Number of farmer’s markets or supermarkets	1.36	3.29	0	50
61	Destination Accessibility	Distance to the kindergarten (km)	1.89	3.01	0	10
62		Distance to the primary school (km) (deleted)	1.24	2.26	0	10
63		Distance to the post office (km)	3.78	3.48	0	10
64		Distance to the police station (km)	2.80	3.43	0	10
65		Distance to the bank (km)	3.23	3.40	0	10
66		Distance to the convenience store (km)	0.38	1.70	0	10
67		Distance to the supermarket (km)	3.44	3.64	0	10
68	Public Activity Spaces	Public activity spaces and facilities	3.42	3.34	0	13
69		Community barrier-free facilities (1–7, 7 = convenient)	2.00	1.47	1	7
70	Design and Aesthetics	The tidiness of the roads (1 = dirty, 7 = tidy)	3.89	1.49	1	7
71		Construction structure (1 = disorganized, 7 = organized)	3.41	1.61	1	7

. Certain variables, such as distance-related measures, were processed using capping to handle outliers, and missing values were imputed using a backward fill method. Finally, all variables were standardized before analysis to ensure comparability.

2.4. Methodology

The study employed six machine-learning models to predict insufficient MVPA among older adults in China. Model performance was comprehensively evaluated using multiple metrics, including accuracy, AUC, recall, specificity, precision, negative predictive value (NPV), and Brier score (Appendix A 

Table A2. Performance evaluation metrics of six machine-learning models for predicting insufficient physical activity among older adults in China (n = 2526).

Model	Accuracy	AUC	Recall	Specificity	Precision	NPV	Brier Score
RF	0.7322	0.7754	0.5544	0.8448	0.6936	0.7495	0.1852
CatBoost	0.7296	0.7636	0.5510	0.8427	0.6894	0.7476	0.1868
XGBoost	0.7084	0.7419	0.5510	0.8082	0.6454	0.7396	0.2216
LR	0.7018	0.7182	0.4864	0.8384	0.6560	0.7204	0.2048
SVM	0.6860	0.7228	0.4354	0.8448	0.6400	0.7025	0.2021
Decision Tree	0.6319	0.6084	0.5034	0.7134	0.5267	0.6939	0.3681

). The random forest model was selected for further analysis due to its superior performance in terms of accuracy and robustness. The random forest model utilizes bootstrap sampling from the original dataset to build multiple decision trees, aggregating predictions through majority voting (for classification) or averaging (for regression). By combining the bagging ensemble technique with decision tree algorithms, random forest effectively captures high-dimensional data complexity and exhibits strong robustness against multicollinearity [28]. The specific model construction process included the following steps:

(1)

Data preprocessing and feature selection. The dataset was randomly split into training (70%) and testing (30%) sets. From an initial pool of 70 candidate variables, 6 variables were excluded through a two-stage process: (a) literature review to identify clinically irrelevant variables; and (b) univariate logistic regression to eliminate statistically non-significant predictors (labeled as “deleted” in Appendix A Table A1). Variance inflation factor (VIF) diagnostics confirmed no multicollinearity among the final 64 features (all VIF < 5).

(2)

Model optimization and validation. The optimal hyperparameters (n_estimators = 500, min_samples_split = 5, max_features = ‘sqrt’) were determined through grid search, and 10-fold cross-validation was employed to examine the model’s robustness and generalization capability. For example, the national model demonstrated a mean AUC of 0.7155 (±0.0431) on the training set and 0.7322 on the independent test set, confirming its robust generalizability.

(3)

Model robustness verification. To further validate the model’s stability, we constructed a reduced model using only the top 30 most influential features as ranked by SHAP values. The comparative analysis demonstrated complete consistency in the ranking of the top 20 core variables between the reduced and full models, with only minor positional variations observed for variables ranked 11th–20th, thereby validating the robustness of the original model.

To interpret the contribution of each feature variable, this study employed SHapley Additive exPlanations (SHAP) values to achieve both global and local interpretability. SHAP values quantify the marginal contribution of each input variable to the prediction outcomes, explicitly indicating both the direction (promoting/inhibiting) and magnitude of their influence [29]. Based on the SHAP feature importance rankings of urban and rural models, this study further uncovered the distinct environmental needs of older adults in these two settings. Furthermore, non-linear association analyses of key environmental features were conducted to offer deeper insights into their complex relationships with insufficient MVPA.

3. Results

3.1. National Risk Prediction Model

Figure 1 presents the key risk factors associated with insufficient physical activity among Chinese older adults. Females and people with an advanced age are significantly associated with physical inactivity. Among socioeconomic factors, those with extremely low household incomes are at a higher risk of insufficient MVPA, potentially due to limited healthcare access and poorer health resources. Impaired functional mobility and a poor activities of daily living (ADL) capacity are negatively correlated with MVPA engagement. Social interaction exhibits a non-linear (U-shaped) relationship with MVPA. Low social engagement, often indicative of loneliness and social isolation, is linked to reduced physical activity, whereas excessive social participation may also involve more sedentary behaviors such as prolonged card-playing [30]. Built environmental factors significantly influence physical activity patterns. Housing environmental risk factors, including a restricted living space and the lack of elevator access, are key predictors of insufficient activity. Moreover, urban older adults face significantly higher risks compared to those living in rural areas. An excessive community population density is positively associated with physical inactivity. Additionally, the accessibility of daily amenities, such as general hospitals, supermarkets, post offices, and banks, demonstrates an inverse relationship with MVPA. However, proximity to primary healthcare facilities is positively associated with increased physical activity, likely through enhanced health awareness and self-management.

3.2. Urban–Rural Grouping Study

Given the substantial urban–rural development disparities in China, this study conducted an urban–rural grouping analysis to reveal differential associations between the built environment and physical activity among older adults across diverse environmental contexts. In accordance with the urban–rural classification standards established by China’s National Bureau of Statistics (NBS) [31], participants were categorized as urban or rural residents based on their primary place of residence. Among the 997 urban participants, the mean MVPA was 3007 MET-minutes/week (equivalent to approximately 6–12 h/week), with 531 individuals (53.3%) failing to meet the WHO recommendations. In contrast, rural participants (n = 1529) exhibited significantly higher mean MVPA levels (6286 MET-minutes/week, or 13–26 h/week), with only 486 (31.8%) falling below the WHO threshold. Independent-sample t-tests confirmed a statistically significant urban–rural disparity in MVPA insufficiency rates: urban older adults demonstrated a markedly higher prevalence (mean rate = 0.5326, 95% CI [0.5016, 0.5636]) compared to their rural counterparts (mean rate = 0.3179, 95% CI [0.2945, 0.3412]; p < 0.001). These findings underscore a significantly elevated risk of physical inactivity among urban-dwelling older adults in China. To address these disparities, this study further developed separate predictive models for urban and rural older adults, enabling the tailored identification of built environment needs for promoting MVPA in each setting.

In the urban model, the environmental variable “location” was operationalized according to the administrative classification of residential communities. Communities governed exclusively by urban neighborhood committees were categorized as core urban areas (location = 1), whereas those administered by village committees or exhibiting mixed governance structures (village and urban neighborhood committees) were classified as peripheral urban areas (location = 0). The analytical sample comprised 997 urban participants, with 618 residing in core urban areas and 379 in peripheral urban areas. Older adults in peripheral areas demonstrated significantly higher physical activity levels, with a mean MVPA of 5338 MET-minutes/week (95% CI [5012, 5664]; equivalent to 11–22 h/week). Of these, 137 participants (36.15%) failed to meet the WHO physical activity guidelines. In contrast, core urban residents exhibited markedly lower activity levels, averaging 1578 MET-minutes/week (95% CI [1452, 1704]; approximately 3–6 h/week), with 394 individuals (63.75%) classified as insufficiently active. Independent-sample t-tests also revealed statistically significant differences in MVPA insufficiency between groups, with a mean MVPA insufficiency rate of 0.3615 (95% CI [0.3129, 0.4101]) for peripheral urban older adults and 0.6375 (95% CI [0.5995, 0.6756]) for core urban residents (p < 0.001). These findings highlight substantial intra-urban disparities in MVPA between older adults living in core and peripheral urban areas, indicating a higher risk of insufficient physical activity among those residing in core urban areas in China.

Figure 2 and Figure 3 presented the prediction results for insufficient physical activity among urban and rural older adults, respectively. The results indicated that demographic characteristics, health status, and socioeconomic factors, as well as residential and community built environment factors, served as key predictors of physical activity among older adults in urban and rural areas. However, there were both common and distinct features between these two settings (Figure 4). Specifically, age emerged as the strongest predictor in both models. However, gender exhibited a more pronounced effect in the rural model, with older females demonstrating a higher propensity for physical inactivity compared to their male counterparts. This discrepancy may be attributed to the influence of traditional gender roles, where older men in rural areas are more likely to participate in agricultural or labor-intensive tasks, while women typically focus on lighter household chores, resulting in lower physical activity [32]. Functional mobility, activities of daily living (ADL), and the number of chronic diseases were common predictors for both urban and rural older adults. Household income and social interaction were also shared factors across both settings, while educational attainment was only relevant for rural older adults. In addition, economic conditions and the proportion of older adults in the community were identified as key predictors of insufficient MVPA in both urban and rural contexts, with economic conditions showing particular significance in urban areas.

In terms of housing environment factors, building age was a significant predictor across both urban and rural settings, while housing size and elevator availability were uniquely influential only among urban older adults. Regarding community built environment factors, population density and accessibility to essential services (e.g., supermarkets and banks) were common determinants in both urban and rural areas. Nevertheless, urban older adults exhibited a greater sensitivity to the distance to general hospitals and post offices, the diversity of public activity spaces and recreational facilities, and the number of convenience stores. In contrast, the diversity of activity facilities was less significant in rural areas, possibly due to the closer proximity to natural environments. Rural older adults, however, were more adversely affected by poor accessibility to kindergartens, police stations, and bus stops, and inadequate street sanitation conditions, both of which were associated with insufficient MVPA.

3.3. Non-Linear Association Analysis

To capture the complex interactions between environmental features and individual behaviors more accurately, moving beyond traditional linear assumptions, this study further investigated the non-linear relationships between key community environmental features and insufficient moderate-to-vigorous physical activity (MVPA) among older adults. The analytical framework employed distinct models for urban and rural contexts, with the left panel illustrating the urban model and the right panel presenting the rural model. As demonstrated in Figure 5, the analysis initially examined the non-linear effects of environmental features which were significant in both urban and rural contexts, including community socioeconomic status, the proportion of older adults, objective population density, subjective density perception, and the accessibility to supermarkets and banks. Subsequently, the study explored context-specific key environmental determinants. Urban-specific factors included the proximity to general hospitals, primary healthcare facilities, and public activity facilities. The rural-specific analyses revealed distinct non-linear patterns related to the accessibility to public transportation nodes (bus stops), and the proximity to community safety services (police stations).

Figure 5 illustrated the non-linear associations between key environmental factors and insufficient MVPA among older adults in urban and rural areas. Our analysis revealed a positive association between an elevated community socioeconomic status (measured by per capita disposable income) and the increased risk of physical inactivity. This association may be explained by behavioral adaptations accompanying economic development, such as an increased mechanization and vehicle dependence, which may reduce active transportation opportunities [6]. However, this study found that extremely poor urban communities (threshold [0, 0.1], corresponding to a per capita disposable income below 3000 CNY [≈400 USD] annually) exhibited high risks of insufficient physical activity. Economic disadvantage frequently coincides with substandard housing and constrained mobility options, potentially exacerbating physical inactivity among older adults [33].

The proportion of older adults in the community displayed divergent urban–rural associations. In rural areas, a higher proportion of older residents was correlated with a reduced MVPA, potentially reflecting an outmigration-induced community decline characterized by a deteriorating infrastructure, limited transportation options, reduced economic activity, and weakened social support systems, all of which contributed to physical inactivity among older adults. Conversely, in urban areas, a higher proportion of older adults was associated with increased activity levels. In China, a considerable number of urban communities with a higher proportion of older residents were composed of former “work-unit” compounds. These compounds represented a special neighborhood layout pattern that emerged during China’s unique urbanization process, where workplaces such as factories and government offices provided collective housing for their employees. These communities not only offered living spaces but also exhibited a “latent work-unit presence effect”, where close social ties were formed through past workplace relationships and contributed to a closely knit social structure. Such networks are beneficial for enhancing residents’ sense of belonging and encouraging self-organized physical activity among older adults [34].

The accessibility of banks and supermarkets showed a U-shaped relationship with the risk of insufficient MVPA in both urban and rural areas. Facilities located either very close (range [0, 0.1], less than 1 km) or very far (range [0.8, 1], more than 8 km) were linked to lower physical activity. An extremely close distance might reduce travel opportunities or time due to overcrowding, while an excessive distance could lead to a reliance on motorized transportation. Shopping is the most common reason for older adults to go out, accounting for 33.2% of trip purposes [12,35]. The results underscored that optimal accessibility to shops and commercial destinations was more effective in encouraging physical activity among older adults.

In rural settings, the community population density showed a negative association with an insufficient MVPA risk. A low population density and dispersed living patterns in rural areas often increase the reliance on motorized transport. Clustered and compact rural communities may provide better infrastructure and facilities that support physical and social activities. This finding suggested that developing compact rural communities might enhance opportunities for active ageing in rural communities.

However, in urban areas, the population density exhibited an S-shaped relationship with insufficient MVPA. In low-density neighborhoods (range [0, 0.1], below 3000 persons/km²), older adults demonstrated higher activity levels, likely due to the necessity-driven movement in areas with limited amenities. As population density increased and accessibility improved, the risk of insufficient physical activity tended to rise. With further increases in density (range [0.3, 0.8], approximately 9000 to 24,000 persons/km²), SHAP values showed a slight decline, suggesting that compact communities might help alleviate the insufficient MVPA among older adults. However, as the population density continued to increase (range [0.8, 1], exceeding 24,000 persons/km²), the SHAP values rose again, indicating a growing risk of insufficient MVPA, likely due to overcrowded and complex traffic environments. In the context of high-density East Asian cities, keeping the population density within a reasonable range to avoid overcrowding appears more conducive to promoting MVPA among older adults. In addition, we also found that subjective perceptions of environmental density showed a negative correlation with physical activity in both urban and rural contexts, with overcrowding perceptions being linked to reduced physical activity. Overall, both objective and subjective measures of density revealed an adverse impact of overcrowding on MVPA levels among older adults.

Figure 6 illustrates the non-linear effect of the urban key environmental features. The accessibility to general hospitals ranked high only in urban settings (fifth). A short distance to the general hospital ([0, 0.05], less than 2.5 km) was associated with higher inactivity risk, possibly due to overcrowding and complex traffic conditions. On the contrary, better access to primary healthcare facilities correlated with increased outdoor activities, likely due to improved health literacy and regular health check-ups. This finding aligns with prior research using the number of healthcare facilities as a measure [36]. Additionally, public activity spaces and facilities also played important roles for urban older adults, exhibiting a U-shaped relationship. Diverse recreational facilities showed significant protective effects against inactivity, but only when the availability reached the upper range ([0.6–1.0], minimum of eight facility types). Notably, our analysis demonstrated that economically disadvantaged communities exhibited a markedly lower diversity in public activity space facilities (as visualized through clustered blue data point distributions). This pattern may reflect systemic spatial inequities in urban resource allocation, where marginalized communities face compounded disadvantages.

Figure 7 depicts the non-linear associations between key environmental features and insufficient physical activity among older adults in rural settings. Long distances to bus stops and police stations were both significantly associated with physical inactivity, underscoring the critical role of safety perception and transportation infrastructure in rural communities. In rural areas, the distribution of distances to bus stops exhibited relative uniformity across the [0, 1] range, with this variable emerging as the ninth most influential predictor in the rural model. Better public transportation accessibility can effectively mitigate the risk of physical inactivity, particularly for older adults who may no longer drive or lack confidence in driving [37,38]. However, these two features showed less significance in the urban model. Recent research has indicated that urban residents exhibited sensitivity to bus stop proximity. Wu et al. (2021) observed that physical activity increased with bus stop density up to a threshold of 6 stops/km², beyond which the effect plateaued [39]. An excessive proximity to bus stops may discourage physical activity, as shorter distances reduce the need for walking [40]. Furthermore, perceived safety from crime was also important, particularly in low-density environments where violent crime can be a great barrier to engaging in physical activity [21,41]. Existing research has examined the positive correlation between perceived safety and physical activity, including walking and overall physical activity [12,42]. In conclusion, this finding underscores the importance of accessibility to public transportation and police stations in rural areas for maintaining mobility and reducing social isolation among older adults.

4. Discussion

4.1. Location-Based Gradient in the Insufficient MVPA Risk

This study utilized random forest models with nationwide representative data from the CHARLS baseline survey to develop a national risk prediction model for insufficient physical activity among older adults in China. Distinct urban and rural models were subsequently constructed to identify both shared and unique environmental determinants across different settlement types. Furthermore, the study examined the non-linear associations between key built environment features and physical inactivity. By providing empirical evidence from a high-density, transitioning East Asian context, this research offers valuable insights for creating age-friendly and inclusive built environments. The key findings and policy recommendations are summarized as follows.

Firstly, both the national and urban models revealed a location-based gradient in the risk of insufficient physical activity among older adults in China: rural < peripheral urban areas < core urban areas. Older adults in urban areas face a higher risk of insufficient physical activity than rural residents, with the risk being greater in central than in peripheral urban areas. Urbanization may bring unique challenges such as overcrowding, disorganized traffic conditions, and inadequate green spaces, making the built environment less supportive of physical activity [6]. In comparison, rural residents typically benefit from a high availability of open spaces, lower traffic density, and limited access to daily amenities, which often require more outdoor activities [43,44]. In addition, urbanization may also reshape social interaction patterns, potentially weakening social connections and perceived safety, particularly in socioeconomically segregated neighborhoods, further exacerbating inactivity risks [45]. In this study, the urban model found that older adults living in communities with extremely poor economic conditions, a low proportion of older residents, and limited social interaction are often linked to segregation and loneliness and faced a higher risk of inactivity.

Moreover, the analysis identified a distinct intra-urban gradient in MVPA insufficiency. Older adults residing in core urban areas exhibited a higher risk of insufficient MVPA compared to those in peripheral areas, potentially due to the excessive population density and environmental disorder. Previous studies have demonstrated that suburban communities are typically characterized by segregated land uses and enclosed street patterns, which may constrain opportunities for physical activity. In contrast, core urban areas generally exhibit greater compactness, featuring traditional grid-patterned street layouts, well-developed road infrastructure, higher population density, and better public service accessibility—characteristics generally considered conducive to active lifestyles [22,46,47]. However, this study, which focused on older adults in high-density East Asian contexts, uncovered a counterintuitive pattern: older adults residing in core urban areas exhibited the highest risk of physical inactivity. This phenomenon may be attributable to the adverse effects of excessive densification on MVPA among older adults. This negative association was validated in Section 3.3 and will be further discussed in Section 4.3.

4.2. Different Environmental Needs Between Urban and Rural Areas

Building upon the observed urban–rural disparities in physical activity patterns, this study examines differential environmental needs among older adults in urban and rural contexts. The analytical model reveals that rural older adults derive greater benefits from an enhanced accessibility to external resources, whereas urban older adults exhibit a stronger responsiveness to improvements in the quality of built environments and recreational facilities. In rural communities, infrastructural deficiencies and geographically dispersed amenities contribute to mobility constraints and environmental stressors. As illustrated in Figure 3 and Figure 4, a lower population density and poor accessibility to kindergartens, bus stations, and police stations were significantly correlated with insufficient physical activity among older adults. These features reflect great challenges in rural built environments, including excessive spatial fragmentation, infrastructure deficiencies, diminished safety perceptions, and limited service accessibility. Substandard road infrastructure and public transportation networks further exacerbate geographical barriers, potentially intensifying social isolation and mobility restrictions. Integrated public transit networks and pedestrian-oriented infrastructure that bridge resource disparities between well-served and underserved areas can effectively alleviate mobility disadvantages in low-walkability communities [23]. Targeted densification strategies and a compact community design remain imperative in rural settings, with core infrastructure modernization being prioritized under China’s rural revitalization initiative.

However, in urban areas, older adults place more emphasis on high-quality residential and community built environments. Key residential attributes, such as housing size and elevator accessibility, along with community features, including the diversity of public activity spaces and facilities, reflect urban older adults’ heightened demand for enhancing environmental quality. Elevators, as essential vertical transport systems in buildings, play a critical role in facilitating older adults to leave their houses and engage in outdoor activities more frequently [48]. Nevertheless, in high-density, high-rise urban environments, many residential buildings in socioeconomically disadvantaged communities either lack elevators entirely or suffer from frequent malfunctions, posing significant accessibility challenges [23]. During China’s rapid urbanization, most of the residential buildings constructed before 2000 were six-story slab buildings without elevators. As the infrastructure and the populations age, the lack of elevators has become a significant obstacle to daily mobility and physical activity among older adults. In recent years, China has actively promoted the retrofitting of elevators to address these challenges. Based on nationwide representative data, this finding uncovered the strong association between elevators and MVPA among older adults, providing empirical evidence to support the ongoing expansion of elevator retrofit policies within urban renewal initiatives.

Regarding urban community built environments, a high accessibility to primary health care facilities and diverse public activity facilities were found to promote physical activity among older adults in the specific threshold, further highlighting the significance of high-quality living environments in urban settings. Improved access to primary healthcare facilities enables older adults to better maintain daily health monitoring, undergo routine medical check-ups, and acquire health-promoting lifestyle information. In addition, the significant impact of public open spaces and recreational facilities on outdoor activities has also been further emphasized [12,49]. The proximity to parks showed a positive correlation with walking, but only within an effective distance of 1 km [30]. In this study, we found that one single type of activity facility is insufficient to encourage MVPA, especially in low socioeconomic communities where the quality and maintenance of spaces and facilities may be inadequate. Effective promotion was observed only when the diversity of public space and facilities was significantly high (range [0.6, 1], exceeding eight types of facilities), suggesting that a greater diversity may better meet the physical activity needs of older adults with varying health conditions. Overall, the findings emphasized the critical role of high-quality built environments in supporting physical activity for urban older adults.

4.3. Compact Design vs. Over-Densification

Last but not least, this study further revealed the non-linear associations between built environment features and MVPA insufficiency among older adults. Specifically, the results indicated that the positive effects of a compact urban design on mitigating insufficient physical activity were observed only within a specific density threshold (approximately 9000–24,000 people/km²). Over-density (>24,000 people/km²) and perceived overcrowding appeared to hinder physical activity levels. Similarly, accessibility to facilities, such as banks, supermarkets, and general hospitals, exhibited U-shaped relationships with MVPA, suggesting both an excessively low (range [0, 0.1], distance less than 1 km) and excessive proximity to these amenities may adversely impact health behaviors. By revealing the threshold effects and non-linear relationships, this study contributed empirical clarification for the “density paradox”: moderately compact urban development promotes physical activity within optimal thresholds, while over-densification exhibits negative associations with active behaviors. Residents in hyper-dense environments with an excessive facility proximity often encounter significant traffic congestion, poor sanitation, and spatial disorder. For older adults, in particular, who frequently face mobility limitations and health vulnerabilities, such conditions may pose substantial safety risks, including a fear of falls and traffic accident exposure [25,50].

These non-linear relationships further elucidated the intra-urban gradient patterns in MVPA insufficiency. Older adults dwelling in urban core areas exhibited significantly higher risks of physical inactivity compared to those in peripheral zones, a spatial disparity potentially attributable to over-densification and environmental disorder. Recent studies corroborate these findings, revealing similar non-linear effects between urban density and health behavior, especially for older adults. Cheng et al. (2020) found that population density and mixed land use enhanced older adults’ walking only within a specific range (6000–20,000 people/km²), beyond which walking decreased [30]. Likewise, Wu et al. (2021) [39] observed that higher population densities generally support walkability and service diversity. However, once population density exceeded 7000 people/km², the marginal effect on walking frequency became negligible [39]. Taken together, these findings shed light on the “density paradox” and underscored the importance of a balanced urban density and accessibility to promote active aging.

4.4. Planning and Policy Implications

In summary, based on a systematic framework and nationwide survey data from China, this study identified key risk factors for the insufficient physical activity among older adults and examined the diverse environmental needs across different urban and rural contexts. The results revealed a location-based gradient in the risk of insufficient MVPA: rural areas < suburban areas < central urban areas. Furthermore, this study uncovered significant differences in environmental needs between urban and rural settings. For rural older adults, environmental factors related to safety perception and external connectivity, such as police stations, public transportation, access to daily amenities, and road infrastructure, were particularly important. In contrast, urban older adults were more influenced by the quality of residential and community features, including housing size, elevator availability, the distance to the most commonly used primary healthcare facility, and the diversity of activity spaces and facilities. Moreover, the research also depicted non-linear associations between built environment features and insufficient MVPA, indicating that environmental variables impact health behavior differently beyond certain thresholds or tipping points. In China’s high-density urban landscapes, moderately compact development (approximately 9000–24,000 people/km²) effectively promotes physical activity, whereas an excessive density (>24,000 people/km²), perceived overcrowding, and hyper-proximity to supermarkets, banks, and general hospitals (range [0, 0.1], distance < 1 km) correlate with diminished activity levels.

In light of these findings, we propose the following policy implications: (1) In rural areas, promoting compact development and improving transportation infrastructure are essential strategies. Structured residential clustering through well-designed relocation programs can effectively address spatial fragmentation while optimizing land use efficiency and service accessibility in dispersed settlements. Concurrently, developing integrated urban and rural public transport systems and enhancing road infrastructure are also particularly crucial for rural older adults. (2) Urban areas require well-designed diverse public activity spaces and barrier-free facilities to comprehensively address the varied needs of older adults with differing health conditions. This study underscored that expanding the variety of recreational facilities, such as walking paths, exercise areas, and multipurpose sports venues, proved more impactful than simply increasing their number. Furthermore, promoting elevator retrofitting in aging neighborhoods can help eliminate mobility barriers and create more inclusive living environments. (3) Finally, based on identified non-linear threshold effects, moderate urban compactness (9000–24,000 people/km²) helps mitigate physical activity insufficiency. However, an excessive density may lead to spatial strain, necessitating the implementation of organic decentralization strategies in hyper-dense urban cores to mitigate the issues of overcrowding, resource hyper-concentration, and environmental congestion. In addition, for daily amenities requiring wide-area coverage (e.g., supermarkets, banks, and general hospitals), a tiered spatial distribution with moderate spacing intervals is recommended in order to strike a balance between service deserts and facility overutilization. Overall, maintaining a strategic equilibrium between neighborhood connectivity and appropriate development intensity is essential for creating sustainable communities and healthy aging.

5. Limitations

Despite these valuable insights, the study has several limitations. First, while the analysis incorporated a comprehensive set of community environment indicators, data constraints prevented access to precise GPS locations, limiting the ability to account for more granular environmental variables. Second, although the urban–rural comparison provided important insights, our analysis could not capture the broader spatial heterogeneity, such as disparities between coastal and inland regions or variations across economic zones. Third, the lack of longitudinal data on community environmental characteristics constrained our analysis in examining cross-sectional relationships, which prevented us from establishing causal relationships.

Additionally, this study utilized the 2011 baseline data since community environment measures were only collected during this wave. Nevertheless, the key findings retain important practical implications. First, China’s structural urban–rural disparity persists: while the rural revitalization strategy has improved infrastructure, rural areas continue to exhibit pronounced deficits in facility accessibility. Second, rapid urbanization has intensified spatial concentration effects in metropolitan areas, resulting in exacerbated overcrowding and diminished public activity spaces. Meanwhile, a critical concern is the persistent shortage of age-friendly housing, particularly the lack of elevator accessibility, which severely constrains mobility for the growing ageing population. Encouragingly, China’s ongoing urban renewal initiatives have demonstrated measurable progress in upgrading public space quality and facilitating elevator retrofits, though broader implementation is still needed.

Future research could address these limitations by integrating complementary datasets and maintaining a consistent tracking of neighborhood environmental metrics in national surveys. To advance this research field, we recommend combining machine-learning algorithms with geographically weighted regression or regional stratification analyses to better capture multi-scale spatial heterogeneity. Such improvements would enable a more in-depth exploration of dynamic trends and causal mechanisms underlying the relationship between MVPA and environmental factors.

6. Conclusions

This study analyzed a nationwide representative sample of 2526 older adults aged 60 and above in China from the CHARLS database, employing interpretable random forest models to predict insufficient moderate-to-vigorous physical activity (MVPA) and identifying both common and distinct risk factors for urban and rural older adults. This study (1) revealed a location-based gradient in the risk of insufficient MVPA: rural areas < suburban areas < central urban areas; (2) uncovered different environmental needs between urban and rural areas—rural older adults were more influenced by safety concerns and transportation accessibility, while urban older adults exhibited a stronger demand for upgrading the quality of the residential and community built environments, such as elevators and public activity facilities; and (3) analyzed non-linear threshold effects and elucidated the “density paradox”, demonstrating that a compact urban design yields health benefits only within specific density parameters (9000–24,000 persons/km²). Over-density (>24,000 people/km²), perceived overcrowding, and an excessive proximity to specific service facilities, such as supermarkets, banks, and general hospitals (range [0, 0.1], distance less than 1 km) were associated with insufficient MVPA. These findings provide empirical evidence for the development of sustainable and inclusive communities in rapidly urbanizing, high-density contexts characterized by pronounced urban–rural disparities. Policymakers and urban planners should adopt context-specific interventions to address the varying needs of older adults across urban and rural settings. While rural regions necessitate compact urban forms and transportation infrastructure to address accessibility challenges, urban environments would achieve greater age-friendliness through the provision of multifunctional public spaces and universal accessibility features. Crucially, optimizing built environments within evidence-based thresholds proves to be a more effective strategy. Future community planning needs to carefully balance spatial density and functional diversity to maximize health benefits for older adults.