Optimizing Urban Environments for Sustainable Development: Strategies and Practices to Enhance Mobility Among Older Adults

Abstract

Population aging is an irreversible global trend, with China experiencing an aging trajectory far exceeding that of most low- and middle-income and high-income countries. Emerging evidence highlights the urban built environment (BE) as a critical determinant of older adults’ travel behavior (TB), exerting a greater effect than commuter populations. However, findings on BE-TB associations remain inconsistent. This study systematically reviewed 46 studies from 1666 articles retrieved from Web of Science and China National Knowledge Infrastructure databases, applying CiteSpace 6.4.R1 for keyword co-occurrence and temporal clustering analysis. This review synthesizes research trends, theoretical frameworks, key determinants, and methodological approaches by integrating knowledge from multiple fields such as urban planning, transportation engineering, public health, and social policy. It provides a comprehensive perspective on how the BE influences the TBs of the aging population. This article can contribute to improving the quality of life for older adults, promoting intergenerational harmony, reducing healthcare costs, fostering economic development, and encouraging green transportation. By identifying critical gaps and future research directions, our findings offer insights to inform strategies for promoting healthy aging and sustainable urban development.

1. Introduction

With the intensification of global population aging, the scale, severity, and speed of aging are particularly pronounced in China [1]. As this shift intensifies, the burden of aging-related diseases escalates. At the same time, the demand for sustainable urban development is becoming increasingly urgent. In response, China has prioritized “active aging” as a national strategy [2]; it aims to contribute to the sustainable development of cities by promoting the health and well-being of older adults. Mobility—fundamental to independence and well-being in older adults—is shaped by both physiological constraints (such as reduced physical function) and socio-economic factors (e.g., increased leisure time). Addressing the mobility needs of aging populations through built environment (BE) interventions has emerged as an imperative for fostering healthy aging and sustainable urban development.

The urban BE, encompassing man-made surroundings such as population density, land use diversity, transportation infrastructure, and green spaces, plays a crucial role in shaping travel behavior (TB) [3]. Studies suggest that these factors exert a stronger influence on older adults’ TB than on working-age populations [3,4]; they also directly impact the efficiency of resource utilization and environmental sustainability in cities [5]. For instance, Ewing and Cervero [6] identified density, design diversity, accessibility, and safety as key determinants of urban TB, while Enayat et al. [7] established direct links between BE characteristics and travel choices. Cerin et al. (2017) revealed strong correlations between BE features and walking behavior [8], and Li et al. highlighted how BE shapes individual mobility patterns, affecting spatial activity distribution and temporal constraints [9].

Empirically, the BE is often evaluated using frameworks such as Cervero and Kockelman’s “3D” model [10]—density; diversity; and design—later expanded by Ewing and Cervero (2001) to include “distance to transit” and “destination accessibility” [11]. Chinese researchers have further identified land use, spatial form, transportation systems, green spaces, and open areas as critical, modifiable elements [12]. However, uneven spatial distribution and socio-economic disparities limit older adults’ access to these resources, exacerbating mobility inequities, and they also hinder the progress of sustainable urban development [13]. Despite growing evidence, inconsistencies remain regarding the BEs influence on older adults’ TB. Additionally, there is a scarcity of literature that analyzes research trends and hotspots in the characteristics of elderly travel behavior based on bibliometrics. Using CiteSpace visualization tools, we systematically organized and analyzed empirical studies from 2000 to 2025, mapping the 25-year research evolution, hotspots, and development trends in this field.

Therefore, this study systematically reviews 46 studies from both international and domestic sources, analyzing research trends, theoretical frameworks, methodologies, and key determinants. This comprehensive analysis not only clarifies the developmental trajectory of research on the characteristics of elderly travel behavior but also deepens the understanding of the role of the built environment in shaping elderly travel behavior. It further highlights future research directions to support healthy aging and sustainable urban development. It provides strong support for constructing a greener, more equitable, and more sustainable urban environment.

2. Methods

Search Strategy. The study retrieved literature from the Web of Science (WOS) and China National Knowledge Infrastructure (CNKI) databases, covering the period from 1 January 2001 to 20 March 2025. A keyword-based search approach was used to identify relevant articles. For WOS, search items included “built environment”, “green spaces”, “the elderly”, “older people”, “seniors”, “travel behavior”, “travel pattern”, and “travel choice”. For CNKI, combinations of keywords such as “built environment”, “elderly”, and “travel” were employed alongside terms like “density”, “design”, “public transport accessibility”, “destination accessibility”, “diversity”, and “green spaces”. This strategy yielded 1693 articles, which were managed using EndNote X9 software. After removing duplicate entries, 538 articles remained for further screening.

Inclusion Criteria. Articles were included if they met the following criteria:

• The article type was primary research article;
• The language of publication was English or Chinese;
• the study population focused on elder adults;
• the BE exposure variables included “5D” elements or green spaces;
• TB characteristics examined walking, cycling, or public transportation use.

Exclusion Criteria. Articles were excluded if they met any of the following conditions:

• The article type was a review article;
• The language of publication was neither English nor Chinese;
• The study population did not align with the definition of older adults used in this study;
• The BE exposure was assessed only through subjective measures.

Literature Screening. The screening process consisted of two stages. First, two researchers independently reviewed titles and abstracts, with disagreements resolved by consultation with senior researchers specializing in this field. Second, full texts of the selected articles were reviewed to ensure they met the inclusion and exclusion criteria. The process resulted in the final literature database used for analysis; a total of 46 articles ultimately met the inclusion criteria. The detailed workflow and screening results are presented in Figure 1.

Analysis Strategy. This study employed a two-step approach to examine the literature and uncover insights into the relationship between BE and TB among older adults. Step 1: Bibliometric analysis: A bibliometric analysis was conducted to identify research trends, hot spots, and frontier developments. The final dataset of 54 articles was analyzed using CiteSpace, a bibliometric tool designed for citation visualization. Keyword co-occurrence and keyword centrality maps were generated to systematically examine the evolution of research topics and emerging trends. CiteSpace integrates scientometrics and data visualization to reveal structural patterns in scientific knowledge, enabling the identification of research hotspots, evolutionary pathways, and interdisciplinary intersections. These capabilities were leveraged to uncover knowledge gaps and potential research directions. Step 2: Thematic content analysis: A thematic analysis was conducted through a systematic review of the full text of selected articles. Key information—including research design, target population, sample size, BE factors, TB characteristics, and primary findings—was extracted (see

Table A1. General characteristics of the studies included in this review.

No.	Reference	City	Sample Size	Data Source	Study Population (Years)	Research Methods	Research Design Type
1	Chen et al., 2023 [69]	Chong qing	391	Data on pedestrian-vehicle collision accidents involving elderly pedestrians in Chongqing’s Yuzhong District from 2010 to 2021	Seniors (>60)	Extreme Gradient Boosting Model (XGBoost)	Cross sectional
2	Xu et al., 2023 [62]	Chong qing	2698	Large-scale resident travel survey conducted in Chongqing in 2016	Seniors (>65)	Multiscale Geographically Weighted Regression Model	Cross sectional
3	Yang et al., 2023 [65]	Wuhan	1,098,390	Data from subway card swiping in Wuhan City for a continuous week in March 2019	Seniors (>65)	Gradient Boosting Decision Tree Model	Cross sectional
4	Li et al., 2023 [58]	Dalian	597	Questionnaire survey	Seniors (>60)	Multilevel Linear Model	Cross sectional
5	Shi et al., 2022 [66]	Guiyang	6185	Survey data on the travel patterns of Guiyang residents in 2021	Seniors (>60)	Gradient Boosting Decision Tree Model	Cross sectional
6	Yang et al., 2023 [53]	Hong Kong	12,620	Data from the household interview survey conducted in the large-scale travel habit survey by the Transport Department of Hong Kong from 2011 to 2012	Seniors (>65)	Multilevel Binary Logistic Regression Model, Geographically Weighted Binary Logistic Regression Model	Cross sectional
7	Wu et al., 2022 [73]	Nanjing	590	Survey data on resident travel patterns in Nanjing in 2013	Seniors (>60)	Random Forest Model	Cross sectional
8	Yang et al., 2022 [42]	Xiamen	93,861	Survey on Residents’ TBs in Xiamen in 2015	Seniors (>60)	Multilevel Binary Logit Model, Multilevel Negative Binomial Regression Model	Cross sectional
9	Liu et al., 2021 [70]	Xiamen	93,812	Survey on Residents’ TBs in Xiamen in 2015	Seniors (>60)	Extreme Gradient Boosting Decision Tree Model	Cross sectional
10	Wang et al., 2020 [46]	Dalian	204	Questionnaire survey	Seniors (>60)	Multiple Regression Model	Cross sectional
11	Liu et al., 2018 [43]	Shang hai	606	Questionnaire survey	Seniors (>60)	Linear Regression Analysis Model	Cross sectional
12	Feng et al., 2016 [3]	Nanjing	969	Survey Data on Residents’ TBs in Nanjing in 2012	Seniors (>60)	Ordered Logit Model (OLM), Ordi-nary Least Square (OLS)	Cross sectional
13	Wang et al., 2022 [55]	Chong qing	325	Questionnaire survey	Seniors (>60)	Multinomial Logistic Regression Model	Cross sectional
14	Li et al., 2021 [59]	Dalian	597	Questionnaire Survey on Walking Efficiency of Elderly People in Dalian in 2019	seniors	Multilevel Linear Model	Cross sectional
15	Feng et al., 2016 [44]	Nanjing	969	Survey on Travel Patterns of Nanjing Residents	Seniors (>50) and Retirees	Ordered Logit Model, Ordered Probit Model, Linear Regression Model	Cross sectional
16	Zang et al., 2019 [50]	Hong Kong	3961	Survey on Travel Characteristics of Hong Kong in 2011	Seniors (>65)	logistic regression model	Cross sectional
17	Cheng et al., 2019 [14]	Nanjing	4474	Survey data on residents’ travel patterns from household interviews in Nanjing in 2013	Youth (>6) and seniors (>60)	Zero-Inflated Ordered Probit Model, Cox Proportional Hazards Model	Cross sectional
18	Wang et al., 2019 [74]	Beijing	229	Activity-travel diary data of 229 households in Beijing before and after relocation	all age groups	Dual-Wave Structural Equation Model	Longitudinal
19	Yang et al., 2022 [61]	Wuhan	-	March 2018 Wuhan Subway Smart Card Data	Seniors (>60)	geographically weighted regression (GWR)	Cross sectional
20	Xiong et al., 2024 [40]	Guiyang	436	Questionnaire survey	Seniors (>65)	Mixed Geographically Weighted Regression Model	Cross sectional
21	Guo et al., 2023 [60]	Xiamen	93,861	Survey on Residents’ Travel Patterns in Xiamen in 2015	Seniors (>60)	BO-LightGBM	Cross sectional
22	Du et al., 2021 [37]	Beijing	915	Conducted face-to-face interviews at nine top-tier hospitals in Beijing in 2019	Seniors	Multinomial Logit Model	Cross sectional
23	Yang et al., 2022 [33]	Hong Kong	10,700	Extracted data on elderly (outdoor) walking behavior from the Survey on Travel Characteristics of Hong Kong in 2011	Seniors (>65)	Logistic Regression Model, Geographically Weighted Logistic Regression Model	Cross sectional
24	Mercado et al., 2009 [47]	Hamilton	16,190	Data from the Hamilton CMA in Canada	All age groups	Multilevel models	Cross sectional
25	Cheng et al., 2020 [15]	Nanjing	702	2013 Nanjing Household Travel Survey	Seniors (>60)	Random Forest Model	Cross sectional
26	Shi et al., 2022 [68]	Nanjing	20 w	Smart Card Data of Nanjing in 2019	Seniors (>60)	Gradient Boosting Regression Tree Model (GBRT)	Cross sectional
27	Liu et al., 2020 [75]	Denver	9248	Survey records of RTD subway users in Denver, Colorado	Youth (19–64) and Seniors (>65)	Multinomial probit model	Cross sectional
28	Cerin et al., 2020 [76]	Hong Kong	909	Project on Active Lifestyles and Environments for Chinese Elders	Seniors (>65)	Generalized Additive Mixed Model (GAMM)	Cross sectional
29	Hou 2019 [38]	Singapore	25,922	2012 Household Travel Survey through Interviews in Singapore	Youth (20–54) and Seniors (>55)	Individual TB Model, Zero-Inflated Ordered Choice Model	Cross sectional
30	Ning et al., 2021 [41]	Qingdao	331.4 w	Qingdao Subway Smart Card Data	All age groups	Negative Binomial Regression Model	Cross sectional
31	Hatamzadeh et al., 2020 [51]	Rasht	600	Rasht Household Travel Survey	Seniors (>60)	Binary Logit Model	Cross sectional
32	Yang et al., 2022 [34]	Xiamen	11,732	Survey data on residents’ travel patterns in Xiamen in 2015	Seniors (>65)	Simultaneous equations model	Cross sectional
33	Yang et al., 2022 [52]	Chiba City	2003	The 6th Tokyo Metropolitan Area Person Trip Survey	Seniors (>65)	Ordered logistic model, Duration models	Cross sectional
34	Zhao et al., 2023 [56]	China	12,439	National survey data from 119 townships in China	All age groups	Multinomial Logit Model (MNL)	Cross sectional
35	Yang et al., 2021 [16]	Hong Kong	101,385	Survey on Travel Characteristics of Hong Kong in 2011	Seniors (>65)	Random Forest Model	Cross sectional
36	Büttner et al., 2024 [48]	Munich	11430	Data from the Munich National Mobility Survey	Youth (18–64) and Seniors(>65)	Regression Model	Cross sectional
37	Yang et al., 2020 [54]	Hong Kong	19,703	Survey data on travel characteristics of Hong Kong in 2011	Seniors (>60)	Logistic regression model, Multilevel logistic regression model, Geographically weighted logistic regression model	Cross sectional
38	Perchoux et al., 2019 [49]	Luxembourg	471	LuxCohort questionnaire and VERITAS questionnaire	Seniors (>65)	multilevel logistic regressions	Cross sectional
39	Lu et al., 2018 [79]	Hong Kong	720	Questionnaire survey	Seniors (>65)	Multilevel mixed models	Cross sectional
40	Wang et al., 2020 [45]	Netherlands	66,880	Dutch National Travel Survey (2015–2017)	All age groups	Tobit Regression Model	Cross sectional
41	Ma et al., 2022 [78]	Dalian	533	Questionnaire survey	Seniors (>60)	Zero-Inflated Poisson Regression Model (ZIP)	Cross sectional
42	Yang et al., 2018 [17]	America	104,613	2009 National (US) Household Travel Survey	Adults (45–64)	linear regression models and logistic regression models	Cross sectional
43	Hong et al., 2024 [57]	Shang hai	4429	2021 Shanghai Jiading District Resident Travel Survey	Seniors (>60)	Mixed Logit Model	Cross sectional
44	Zhu et al., 2025 [71]	Shantou	1109	Data from the 2021 Shantou Resident Travel Survey	Seniors (>60)	eXtreme Gradient Boosting (XGBoost)	Cross sectional
45	He et al., 2025 [72]	Guiyang	463	Questionnaire survey	Seniors (>60)	eXtreme Gradient Boosting (XGBoost)	Cross sectional
46	Liu et al., 2024 [63]	Guang zhou	21,897	Survey data from the Guangzhou Transportation Bureau in 2017	Seniors (>60)	Generalized Additive Mixed Model (GAMM)	Cross sectional

). This synthesis provided a comprehensive assessment of the theoretical frameworks, influencing factors, mechanisms, and research methodologies employed in studies examining the BE-TB relationship in older adults.

3. Results

3.1. Visual Analysis of Literature

Figure 2 displays a keyword co-occurrence network map generated by CiteSpace based on the WOS literature database. The keywords in each article summarize the core themes of the literature, and this co-occurrence network vividly illustrates the rich connotation and extensive extension of research on the built environment and related fields. Each point in the figure represents a keyword. The size of a node indicates the frequency of the keyword’s appearance in the literature—the higher the frequency; the larger the node. The connecting lines between nodes represent co-occurrence relationships, meaning these keywords appear together in one or more articles. The thickness of the lines indicates the frequency of co-occurrence—the higher the frequency; the thicker the line. The parameters of the map were set as follows: the time span was from 2000 to 2025, segmented into 1-year slices for analysis; the g-index (k = 25) was used to ensure the retention of high-impact literature in each time period, combined with LRF = 2.5 (look-back years) and L/N = 10 (maximum links per node) to control network density; LBY = 5 was set to capture burst growth of keywords; a network comprising 140 nodes and 535 edges was generated, with a density of 0.055, and all nodes belonged to the same connected component, reflecting the high relevance of research topics.

From the analysis, it can be observed that the most frequently occurring keywords include “built environment”, “health”, “active travel”, “behavior”, “associations”, “active aging”, “elderly people”, and “impacts”. Among these, “built environment” serves as the core keyword, occupying the central position in the network. The nodes for “built environment”, “active travel,” and “health” are the largest, indicating that these three keywords are the research hotspots in this field. Furthermore, the thickest connecting line is between “built environment” and “active travel”, further highlighting the widespread attention given to the role of the built environment in shaping active travel. As an important outcome variable in studies on residents’ travel behavior, active travel has sparked extensive discussion. Existing research shows that sociodemographic characteristics and built environment factors are commonly used to explain active travel among older adults [14,15,16,17]. As global populations age, the travel needs of older adults—shaped by social participation; physical health; quality of life; and well-being—have drawn considerable scholarly attention [18,19,20]. Active travel modes, particularly walking and cycling, are widely recognized for their role in enhancing physical activity levels among older adults, providing substantial benefits in mitigating age-related health challenges [18,21]. Research [22] indicates that regular physical activity among older adults plays a significant role in preventing physical and mental illnesses, as well as delaying cognitive and functional decline. Studies from the United States and Australia conclude that the accessibility and quantity of green spaces promote behavioral activity and are positively correlated with healthy travel among residents [23]. Street network density and connectivity have been found to encourage healthy behaviors. On one hand, they enhance the safety and accessibility of travel for older adults, thereby increasing residents’ willingness to travel. On the other hand, higher street network density and connectivity contribute to providing safer walking environments and fostering social interactions among older adults [24].

In summary, this keyword co-occurrence network not only reveals the hotspots and frontiers in built environment and related research fields but also demonstrates the complexity and diversity of such studies. Through interdisciplinary research, an in-depth exploration of the intricate relationships between the built environment and human behavior/health can provide scientific foundations and practical guidance for creating more livable, healthy, and sustainable urban environments. Future research could further expand and deepen these areas by focusing on the needs and characteristics of specific populations, as well as the interaction mechanisms between the built environment and other social, economic, and cultural factors.

Figure 3 presents a time-clustered keyword co-occurrence map generated by CiteSpace based on the WOS literature database, revealing the top 30 keywords with the strongest citation bursts during the period 2000–2025. The analysis highlights the evolutionary trends of research hotspots. In the early stage (around 2019), research hotspots primarily focused on transportation-related topics, as evidenced by the bursts of keywords such as “distance traveled”, “commuting distance”, and “work”. This reflects the growing emphasis on travel distance, commuting patterns, and work-related transportation issues during this period, driven by accelerated urbanization and increasing transportation demands. These findings underscore the broader societal concern for transportation efficiency and travel experience at the time. Over time, the research focus gradually expanded. In 2017, keywords such as “quantitative analysis”, “qualitative analysis”, “form”, “old people”, and “trip generation” experienced significant bursts, indicating a growing emphasis on the application of analytical methods. From a methodological perspective, recent years have seen increasing attention to the integration of machine learning, which addresses the limitations of traditional models, including underfitting, complexity, and inefficiency in processing large-scale datasets. Machine learning techniques offer advantages such as flexibility, automated feature extraction, ensemble learning, and scalability, while also enhancing interpretability and adaptability. Nevertheless, traditional models remain relevant due to their solid theoretical foundations, ease of application, and well-established interpretability. Looking ahead, hybrid approaches that combine traditional statistical methods with machine learning are expected to drive methodological advancements in the field. The emergence of keywords such as “land use”, “density”, “accessibility”, “associations”, and “neighborhood design” during 2018–2019 marked a shift in built environment research from macro-level analysis to in-depth exploration of specific environmental elements. On one hand, studies began focusing on quantifiable physical characteristics like land use patterns, development intensity, and spatial accessibility. On the other hand, scholars paid greater attention to the systemic relationships between these elements and their comprehensive effects at the neighborhood scale. This transition reflects both the refinement of research methods and the academic community’s deeper understanding of the multidimensional and multi-level impact mechanisms of the built environment. Furthermore, research has increasingly focused on neighborhood design, recognizing its pivotal role in supporting older adults’ daily activities and social interactions—particularly when age-related physical limitations emerge. Consequently, optimizing neighborhood environments has become a key strategy for fostering active social engagement [25]. By 2022, the emergence of burst keywords like “China”, “spatial heterogeneity”, “seniors”, “household”, “neighborhood”, and “people” reflected an expansion of research into China-specific contexts, spatial variations, and interconnected domains involving aging populations, family structures, and community dynamics. This evolution demonstrates a continued diversification and deepening of research themes within the field.

Figure 4 presents the keyword analysis results derived from the CNKI database. The parameters for the mapping analysis were configured as follows: The time span was set from 2000–2025 with annual segmentation (1-year slices); the g-index (k = 25) was employed to filter high-impact literature for each time period, combined with LRF = 2.5 (retrospective linkage of 2.5 years) and L/N = 10 (maximum links per node) to optimize the network structure; LBY = 5 was applied to identify burst trends in keywords. The resulting network comprised 39 nodes and 96 edges, with a density of 0.1296 (indicating relative compactness), and all nodes belonged to a single connected component, demonstrating high thematic focus in the research.

In the figure, “built environment” and “older adults” emerge as central nodes, interconnected with numerous other keywords, highlighting the research emphasis on how the built environment influences older adults’ travel behavior. The color-coding of the nodes allows for the identification of research hotspots during different periods. Early studies primarily examined the statistical correlations between the built environment and older adults’ travel behavior. Subsequently, the emergence of keywords such as “influencing factors” and “impact mechanisms” indicates that research has evolved beyond merely establishing correlations to delve deeper into specific factors and underlying mechanisms. Meanwhile, travel behavior has been expanded to include physical activity, active travel, and walking efficacy, indicating a refinement of the dependent variables under study. Additionally, keywords such as “urban planning”, “urban transportation”, and “rail transit” in the figure reflect the consideration of factors like urban planning and transportation infrastructure. Keywords like “quality of life”, “health”, “nonlinearity”, and “mediation effects” suggest that the built environment may influence older adults’ quality of life and health status by affecting their travel behavior. The inclusion of keywords such as “walking time” and “mobility” further demonstrates that the research has incorporated temporal and spatial dimensions into its analysis.

In summary, this research field demonstrates a clear interdisciplinary trend, integrating urban planning, transportation engineering, public health, and other disciplines. With its focused examination of older adults as a vulnerable population group, the research pays particular attention to their travel needs and behavioral characteristics, fully reflecting social and humanistic concerns. Collectively, these studies contribute significantly to establishing more supportive urban living environments for aging populations.

3.2. Theoretical Frameworks for Examining BE-TB Associations

3.2.1. Theoretical Basis

Theoretical perspectives on the BE-TB associations among older adults span multiple disciplines. These frameworks provided insights into the complex interactions shaping mobility patterns.

Environmental Behavior Theory underpins studies of environmental influences on human actions, encompassing three perspectives: environmental determinism, interactionism, and mutual permeability. Environmental determinism [26] posits that human behavior is predominantly shaped by external environmental factors, assigning a passive role to individuals. While this theory highlights the influence of the external environment, it is criticized for its one-dimensional focus, overlooking the role of human needs and agency. Interactionism [27] and mutual permeability theory [28] challenge the limitations of environmental determinism. Interactionism views humans and their environment. As distinct yet interdependent elements, suggesting that behavior emerges from their interaction. In contrast, mutual permeability theory posits that humans and the environment form an integrated system of mutual influence, emphasizing their inseparability. Together, these perspectives affirm that the environment significantly shapes human behavior, albeit through complex, multifaceted interactions.

Social-Ecological Theory provides a multidimensional framework for understanding the interplay between individuals, social structures, and the socio-physical environment. Widely applied in health promotion, this theory posits that health-related behaviors are influenced by interdependent factors at multiple levels [29]. This model emphasizes the dynamic interactions among the individual attributes (e.g., education, gender, and age), social factors (e.g., social networks), physical environment (natural and built), and institutional frameworks. These interacting influences operate synergistically, amplifying their impacts on health-related behaviors [30].

Health Behavior Theory explores how individuals make decisions to maintain or improve their well-being, drawing from multiple behavioral models, including the Health Belief Model, Cognitive Information Processing Theory, the Theory of Reasoned Action, Social Learning Theory, Social Support Theory, the Theory of Behavioral Change, and Self-Efficacy Theory [31]. These models collectively emphasize that health behaviors are shaped by perceptions, attitudes, and subjective feelings. They provide a theoretical foundation for understanding how older adults navigate health-related decisions.

The theory of Planned Behavior [32] posits that behavioral intention is the primary driver of individual actions. It is shaped by three core determinants: behavioral attitudes, subjective norms, and perceived behavioral control. Behavioral intention serves as the strongest predictor of actual behavior, while perceived behavioral control can independently affect actions by determining the extent to which an individual feels capable of executing specific actions.

3.2.2. Approaches for Quantifying the TB and BE

Quantifying TB and BE metrics forms the foundation for investigating the BE-TB relationships. As reviewed, questionnaire surveys remain a widely used method for gathering data on residents’ travel habits. These surveys are typically administered on-site and may incorporate complementary approaches, including direct observations, interviews, and recordings. For instance, Cheng et al. [14] explored the impact of the BE on active travel frequency and duration using travel survey data from Nanjing. Yang et al. [33] applied logistic regression models to explore associations between BE and elderly walking behavior in Hong Kong. Yang et al. investigated the BE factors that were significantly associated with walking frequency and duration among elderly individuals in Xiamen, using data from the 2015 Xiamen Residents’ Travel Survey data and associated geographical dataset [34].

Questionnaire surveys are also valuable for capturing subjective evaluations of the BE, including perceptions and satisfaction levels. For instance, Li et al. [35] investigated the nonlinear effects of walking accessibility on the happiness of rural elderly individuals in Jintang County, Sichuan Province, collecting data on sociodemographic characteristics, subjective happiness, BE perceptions, and physical activities. Similarly, Wang [36] combined subjective evaluations of the BE with objective data to analyze the spatiotemporal differentiation of BE characteristics on elderly walking behavior in Nanjing. Du et al. [37] conducted face-to-face interviews to compare the determinants of travel mode choices between elderly and non-elderly patients seeking healthcare in Beijing. Hou [38] utilized data from the 2012 Household Interview Travel Survey (HITS) in Singapore to study how the planned polycentric urban form in Singapore affects the non-work travel frequency and mode choices among elderly (55 and above) and non-elderly (aged 20–54) residents. Zhang et al. [39] collected elderly residents’ opinions on accessibility, safety, and road quality using a face-to-face questionnaire survey, offering valuable insights into elderly mobility and infrastructure needs.

Spatial computation and analysis, primarily leveraging Geographic Information System (GIS) technology, play a critical role in processing, analyzing, and visualizing objective BE characteristics. These methods facilitate the spatial distribution, transportation network structures, land use patterns, and their influence on elderly mobility. In a study examining the spatial heterogeneity of BE impacts on elderly walking distance in mountainous cities, Xiong et al. [40] used satellite maps, GIS panoramic maps, and Baidu Maps to classify the terrain conditions in Guiyang’s urban area, categorizing streets into five levels based on slope gradients. Cheng et al. [14] employed the Baidu Map API to geocode respondents’ home addresses and used ArcGIS buffer analysis to exact BE variables surrounding each household. To further investigate BE influences on public transportation use, Ning et al. integrated Baidu Map-derived Points of Interest (POI) data to assess BE factors influencing passenger volume and the modal share of elderly and student subway users [41]. Meanwhile, Yang et al. collected vector data—including POIs; Areas of Interest (AOIs); and administrative boundaries—from OpenStreetMap and employed ArcGIS spatial analysis to quantify BE characteristics at the community level; identifying features conducive to elderly-friendly environments [42].

3.2.3. Statistical Strategies for Examining BE-TB Associations

In the field of research on the built environment and the travel behavior of the elderly, research methods have evolved from traditional statistical models to modern machine learning algorithms. Each of these methods has its own characteristics, and when conducting research, an appropriate model should be selected based on the research objectives and data characteristics.

Liu et al., 2018 [43], Feng et al., 2016 [44], and Yang et al., 2018 [17] employed the linear regression model in their research. The linear regression analysis model is a classic statistical method, with principles that are intuitive and easy to understand. The linear regression model is generally used as a baseline model for more complex models, and the improvement effect of new models is evaluated by comparing the performance of different models. In addition, Wang et al., (2021) [45], employed the Tobit regression model, as it is specifically designed to handle situations where the dependent variable is censored or constrained.

Wang et al., 2020 [46], Mercado et al., 2009 [47], Baquero et al., 2024 [48], and Perchoux et al., 2019 [49] employed the multiple regression model (MRM) in their research. The MRM is slightly more complex than the linear regression model because it requires considering the influence of multiple independent variables simultaneously. At this point, it may be necessary to take into account the interactions between independent variables and the issue of multicollinearity.

Yang et al., 2018 [17], Zang et al.,2019 [50] employed the logistic regression model in their research. The logistic regression model demonstrates strong robustness when dealing with outliers. It maps the output of linear regression to probability values through the logistic function, thereby enabling the prediction of binary classification problems. This transformation allows the model to retain the intuitiveness of linear regression while also possessing the capability to handle classification tasks. In addition, there are many extension models of logistic regression. Hatamzadeh et al., 2020 [51] used the Binary Logit Model because it is well-suited for handling binary dependent variables. Yang et al., 2022 [52] used the Ordered Logistic Model because it is suitable for situations where the dependent variable is an ordered multinomial variable. Yang et al., 2023 [53], Yang et al., 2022 [42], Perchoux et al., 2019 [49], Yang et al., 2020 [54], used the Multilevel Logistic Regression Model because it is suitable for considering individuals nested within higher levels and can analyze situations with binary dependent variables. Wang et al., 2022 [55], Du et al., 2021 [37], Zhao et al., 2023 [56], used the Multinomial Logit Model because it is suitable for situations where the dependent variable has three or more categories and there is no inherent order between the categories. Hong et al., 2024 [57], used the Mixed Logit Model because it can model the random effects of individuals when choosing different options and can address the Independence of Irrelevant Alternatives (IIA) problem present in traditional Logit models. Hong et al., 2024 [57], used the Mixed Logit Model because it can model the random effects of individuals when choosing different options and can address the Independence of Irrelevant Alternatives (IIA) problem present in traditional Logit models.

Li et al., 2023 [58] and Li et al., 2021 [59] utilized the multilevel linear model (MLM) in their research. The core advantage of the multilevel linear model lies in its ability to handle nested data structures, effectively avoiding statistical biases.

Yang et al., 2023 [60], Yang et al., 2022 [61], Yang et al., 2022 [33], and Yang et al., 2020 [54] employed the Geographically Weighted Regression model (GWR) in their research. By establishing local regression equations at each point within the spatial domain, GWR assigns a unique regression equation to each regression point (sample point). Consequently, it can reveal the spatially varying relationships between the dependent and independent variables, allowing for the exploration of spatial variability and its associated driving factors at specific scales.

Xu et al., 2023 [62] and Xiong et al., 2024 [40] utilized the Multiscale Geographically Weighted Regression model (MGWR) in their research. On the one hand, MGWR relaxes the assumption that spatial variation processes change uniformly across the same spatial scale, effectively addressing the issue of modifiable areal unit problems that may arise in traditional geographically weighted regression. On the other hand, the multi-bandwidth approach yields spatial process models that are closer to reality and more useful [40].

Liu et al., 2024 [63], employed the Generalized Additive Mixed Model (GAMM). GAMM can accommodate various types of distributional assumptions and provide data-driven, customized nonlinear methods to identify threshold effects of environmental attributes [64], while also taking into account the nested structure of the data.

Yang et al., 2023 [65], and Shi et al., 2022 [66], utilized the Gradient Boosting Decision Tree model (GBDT) in their research. GBDT does not assume a linear relationship between predictor variables, allowing it to effectively predict any form of nonlinear relationship between independent and dependent variables. The threshold effects exhibited by this nonlinear association can identify the range of influence of independent variables, thereby more accurately assisting in planning practices. Additionally, GBDT can capture the relative importance of independent variables, which helps planners reasonably determine intervention measures under limited conditions. Furthermore, GBDT adjusts the weights of predictor variables through stage-wise learning from the data, resulting in high predictive capability [67].

Shi et al., 2022 [68], employed the Gradient Boosting Regression Tree model (GBRT) in their research. The enhancement of the GBRT belongs to the ensemble learning approach, which is a combined algorithm of regression trees and gradient boosting [67]. Due to its strong capability in exploring nonlinear relationships and variable interactions, GBRT has been increasingly applied in various research fields, including transportation [65]. The primary difference between GBDT and GBRT lies in their respective focuses on solving classification problems and regression problems.

Chen et al., 2023 [69], Liu et al., 2021 [70], Zhu et al., 2025 [71], and He et al., 2025 [72] employed the Extreme Gradient Boosting model (XGBoost) in their research. Compared to traditional linear regression models, XGBoost does not require adherence to data assumptions and exhibits stronger tolerance for missing data. The algorithm takes into account the interactions between predictor variables, addresses the issue of multicollinearity, and provides more accurate predictions. It can handle highly discrete variables more appropriately and offer insights into the relative importance of variables while fitting complex nonlinear relationships between them.

Wu et al., 2022 [73], Cheng et al., 2020 [15], and Yang et al., 2021 [16] utilized the Random Forest model in their research. Compared to general mathematical models, Random Forest offers the following advantages, high learning accuracy of the algorithm; proficiency in handling high-order relationships between variables, effectively uncovering nonlinear relationships among features; no need for strict statistical assumptions, allowing flexible application to data with or without specific distributions; and proven suitability for small datasets [15].

Wang et al., 2019 [74] used cross-lagged panel models because they can establish relationships between variables and their historical values, and are a type of structural equation model specifically designed for longitudinal studies. Liu et al., 2020 [75] used the Multinomial Probit Model because it can handle multi-category discrete choice problems. Cerin et al.,2020 [76] used GAMMs because they can model data with various distributional assumptions, account for dependency in error terms due to TPU-level clustering (participants sampled from selected TPUs), and estimate complex dose–response relationships of unknown form [77]. Ma et al., 2022 used [78] the Zero-Inflated Poisson Regression Model because it can handle situations with an excess of zero values.

3.3. Empirical Evidence on Environmental and Individual Determinants for TB

This study includes a total of 46 pieces of literature for review and summarizes the general characteristics of these documents and the independent variables studied (see Table A1 and

Table A2. Characteristics of built environment elements included in the studies of this review.

No.	Built Environment Elements						Research Variable	Dependent Variable
	Density	Diversity	Design	Distance	Destination	Green Area
1	√	√	√	√	√		Number of Public Transport Stops, Pedestrian Overpasses, Zebra Crossings and Underpasses, Population Density, Recreational and Entertainment Venue Density, Daily Life Service Facilities Density, Medical Facilities Density, Land Use Entropy Index, Distance to CBD, Road Network Length	Accident frequency of elderly pedestrians
2	√	√	√	√	√	√	Population Density, Vegetable Market Density, Supermarket Density, Park and Plaza Density, Land Use Mix, Road Network Density and Intersection Density, Street Greenery Rate, Bus Stop Density, Distance to City Center	Walking time of the elderly
3	√	√	√	√	√		Number of Parks and Plazas, General Hospitals, Shopping Centers, Catering Facilities, Primary Schools and Kindergartens, Farmers’ Markets and Senior Activity Centers, Permanent Resident Elderly Population, Building Floor Area Ratio, Land Use Mix, Road Network Density, Intersection Density, Distance to City Center, Distance to Suburban Center, Bus Stop Density	Travel distance, frequency, time, duration, and direction of the elderly
4	√		√	√		√	NDVI, Intersection Density, Road Network Density, Residential Area Openness, Time to Bus Stops, Public Service Facilities, Landscape Facilities, Bus Stop Density	Convenience of public transportation for the elderly
5	√		√	√	√		Number of Intersections, Vegetable Markets, Supermarkets, Pharmacies, Parks, Chess and Card Rooms, Distance to Bus Stop, Distance to City Center	Decision-making on travel modes for the elderly
6	√	√	√	√		√	Population Density, Land Use Mix, Intersection Density, Metro Accessibility, Bus Accessibility, Recreational and Sports Facility Accessibility, Park Accessibility, Green View Ratio	Travel tendency (whether they traveled in the past 24 h)
7	√	√	√	√		√	Proximity to Metro Stations, Proximity to Public Service Facilities, Number of Intersections, Number of Public Bicycle Stations, Greenery Proportion, Residential Density, Non-Motorized Traffic Network Density, Land Use Mix	Slow travel time of the elderly group
8	√	√	√	√	√		Population Density, Land Use Mix, Intersection Density, Bus Route Density, Distance to Commercial Facilities	Travel tendency and frequency of the elderly
9	√	√	√	√	√		Population Density, Floor Area Ratio, Land Use Mix, Road Intersection Density, School Density, Park Density, Shop Density, Restaurant Density, Distance to City Center, Bus Stop Density	Whether the elderly engaged in vital travel within 24 h
10	√	√	√		√	√	Number of Comprehensive Parks, Community Parks, Leisure Facilities, Service Facilities, Streets Adjacent to Green Spaces and Road Intersections, Quantity of Scenic Recreation Green Spaces, Adjacent Street Nature to Green Spaces, Type of Green Space, Minimum Proximity Distance, Spatial Distribution Density, Quality of Green Spaces, Scale of Green Spaces	Green space utilization rate of the elderly
11	√	√	√	√			Number of Public Service Facilities such as Commercial, Medical, Cultural, Recreational, and Public Restrooms, Land Use Mix, Road Network Density, Intersection Density, Bus Route and Stop Density	Travel purpose, frequency, time consumption, and mode
12	√	√		√	√		Number of Bus Stops, Sports Stadiums, Movie Theaters, Museums, Cultural Centers, Chess Card Rooms, Parks and Plazas, Population Density, Diversity of Land Use, Distance to the Nearest Subway Station, Distance to the Nearest Large Shopping Mall	Travel time, frequency, and distance
13	√	√	√	√		√	Road Network Density, Block Edge Length, Intersection Density, Road Segment Node Ratio, Land Use Mix, Commercial Facility Density, Cultural and Entertainment Facility Density, Educational Facility Density, Medical Facility Density, Public Transport Terminal Density, Park and Green Space Area Ratio, Green Open Space Density	Frequency of walking trips
14	√		√	√		√	Nighttime Light Radiation Brightness Value, Building Floor Area Ratio, Public Service Facility Density, Intersection Density, Road Network Density, NDVI	Walking efficiency
15	√	√	√	√	√		Population Density, Land Use Mix, Car Traffic Accessibility, Public Transit Service Accessibility, Distance to the Nearest Subway Station, Number of Road Intersections, Average Road Width, Distance to the Nearest Supermarket/Convenience Store or Farmers’ Market	Activity participation and daily travel distance of the elderly
16	√	√	√	√		√	Urban green exposure level, Population density, street connectivity, Land use mix, Distance to the nearest subway station, Number of retail stores	Travel willingness, travel frequency, total travel distance, total travel time, total walking time, and number of motorized trips
17	√	√		√	√		Number of bus stops, bike-sharing stations, parking lots, Population density, Land use mixture, Distance to the nearest shopping mall, Distance to the nearest convenience store, Distance to the nearest market, Distance to the nearest park/square, Distance to the nearest chess/card room, Distance to the nearest gym/sports center, Arterial density, Distance to the nearest metro station	Frequency and duration of active travel
18	√			√	√		Distance to the city center, Population density, Commuting distance, Metro accessibility	Travel preferences, total travel time, number of private car trips, number of public transport trips, and number of non-motorized trips
19	√	√	√	√	√		Number of intersections, general hospitals, parks, commercial facilities, and bus stops, Elderly population density, Building Floor area ratio, Degree of land use mixture, Distance from the city center, Distance from the sub city center, Station area road length	Subway passenger flow and travel distance of the elderly
20	√	√	√	√	√		Population density, residential density, Road network density, Slope, land use mix, Destination accessibility, Distance to the nearest park and bus stop, Density of leisure and entertainment venues, Density of shopping malls, Density of bus stops	Walking distance of the elderly
21	√	√	√	√	√		Population density, Land use mix, Intersection density, Public transit line density, Distance to the nearest commercial, school, and recreational facilities, Density of commercial facilities, Density of recreational facilities, Density of school facilities	Tendency for active travel (whether to go out by biking or walking)
22	√	√	√				Main road length, Land use entropy, Development density, Parking easiness	Choice of travel mode
23	√	√		√		√	Population density, Land use mix, Intersection density, Number of bus stops, Green view ratio	Tendency for walking among the elderly
24	√		√				Low commercial and low residential, Low commercial and high residential, High commercial and low residential, High commercial and high residential	Travel distance
25	√	√	√	√	√		Population density, Land use mix, street connectivity, Number of bus stops and shared bike stations, Distance to the nearest square and park, Distance to the nearest chess and card room	Daily walking time
26	√			√	√		Employment density, Distance to the nearest bike station, Distance to the nearest bus stop, Distance to the nearest chess/card room, Distance to the nearest convenience store, Distance to the nearest park/open square, Distance to the nearest restaurant, Number of Sport/gym, Number of tourist attractions	Travel time of the elderly
27	√	√		√			Bus connectivity, interchange density, Land use mix degree, Parking capacity, Population density, Urban station	Entry and exit methods, travel purposes, station choices, and travel times of the elderly using rail transit
28	√		√				Residential density, Civic and institutional destination density, Recreation density, Public transport density	Walking frequency and duration
29	√	√			√		Regional accessibility to commercial/activity centers, Neighborhood land use characteristics (LUDIV, etc.), Number of community service facilities, cultural facilities, sport, recreational facilities, food courts/restaurants, religious establishments, Proportion of the elderly population	Frequency of non-work-related travel using different modes of transportation
30	√		√	√			Number of bus stops, road intersections, metro station entrances or exits, elementary and middle schools, tertiary hospitals, supermarkets, squares, parks, and scenic spots in 800m buffe, Whether the station is in the main urban area	Subway passenger flow for different groups of people
31	√	√			√		Land use mix, connectivity, Population density, Whether the trip destination is to the CBD, Whether the trip destination is to a main area of business	Whether to choose walking
32	√	√		√	√		Population density, Land-use mix, Intersection density, Distance to commercial center, Bus route density	Walking frequency and walking duration
33	√	√	√	√	√		Population density, Road density, Number of bus stops, Distance to the nearest rail transport station, Number of medical centers, Number of parks, Distances to the nearest gym/sports center, Land use mix	Frequency and duration of walking and cycling
34	√	√	√	√	√		Distance to urban center, Distance to highways, Access bus stop, Bus frequency, Settlement size, Design compact, Land use mix entropy, Population density, Provision of local, Services Intersection density, Availability of parking	Choice of travel mode
35	√	√	√	√	√	√	Land use mix, Intersection density, Bus accessibility, Recreational facility accessibility, Street greening	Walking preference of the elderly towards streetscape greening
36	√		√	√		√	Population density, Points of Interest density, Cycling infrastructure, Street intersection density, Number of green area infrastructure, benches and toilets, The population served at 300 m walking distance from public transport stops	Whether bicycles are used and the frequency of use
37	√	√	√	√		√	Population density, Land use mix, Intersection density, Number of bus stops, Number of recreational and sports facilities in the neighborhood, Street greenery, Number of parks, NDVI	travel propensity
38	√		√	√		√	Number of amenities, Number of public transport stops, Street connectivity, Greenness index	Travel purpose
39	√	√	√	√			Population density, Land-use mix, Street intersection density, Presence of MTR station, Number of bus stops, retail shops, and recreational facilities	Physical Activity
40	√	√	√			√	NDVI, Crossing density, Land use mix, Residential building density	Walking type
41	√	√	√	√	√		Number of bus stops, Distance to the nearest bus stop Distance to the nearest recreational facility, Population density, Land use mix entropy, Road density	Walking frequency
42	√		√		√		Population density, Intersection density, Distance to the nearest park, and walkscore	the total number of daily trips, travel purpose diversity, total travel distance, maximum distance traveled, whether a person has at least one active travel (by walking or by bicycle), and total distance traveled by active modes
43	√	√	√	√	√		Land use mix degree, road network density, population density, density of public service facilities POI, road network distance to the city center, distance to the nearest subway station, and bus stop density	Mode of transportation
44	√	√		√	√	√	Land use mix, population density, residential density, NDVI, number of bus stops, number of intersections, and number of parks	Travel time
45	√	√	√	√	√		Slope, population density, bus stop density, Shopping store density, road network density, residential density, distance to the nearest park, distance to the nearest bus stop, land use mix degree, leisure and entertainment density	Walking distance
46	√		√		√	√	Road surface ratio, street obstacles, Street safety, Street greenery, Street design, Street vitality, Density of community centers, Density of various POI (Points of Interest), Intersection density, building density, Population density	Active travel time

Note: √ indicates that the study included an analysis of this factor.

). This section focuses on exploring the built environment factors and personal economic attributes that influence the travel behavior of the elderly, as well as the research differences across different cities. The results of the literature analysis are as follows:

3.3.1. BE Determinants for Elderly Travel

As reviewed, BE elements, including population density, land use mix, public service facilities, public transportation facilities, and intersection density, have been associated with elderly travel. Other factors such as green spaces, slope, and road network density also play important roles in shaping elderly mobility (Figure 5).

Population density is among the most studied BE factors, showing positive correlations with elderly travel propensity [53] and trip frequency [33]. This is likely because high-density areas typically feature more comprehensive transportation infrastructure (such as public transit, pedestrian walkways, and bicycle lanes), which further enhances travel convenience. Moreover, in high-density areas, residents generally enjoy easier access to various destinations (such as shops, schools, recreational facilities, etc.), reducing both travel time and costs—factors that collectively promote travel behavior. Additionally, walking propensity [33], walking duration [16], and walking distance [40]. It can be inferred that high-density areas typically place greater emphasis on pedestrian-friendly infrastructure (such as sidewalks and greenways) and generally feature more developed and frequently utilized public transit systems, which collectively encourage non-motorized travel modes. However, it negatively correlates with travel distance [40] and cycling frequency and duration [16]. This can be attributed to residents’ closer proximity to various destinations (e.g., shops, schools, workplaces), which reduces travel time and costs, thereby shortening trip distances. Additionally, improved public transport convenience and accessibility have led to decreased cycling mode share. Cheng et al. [15] highlighted significant nonlinear and threshold effects, noting that increased population density enhances walking only up to a certain point.

Land use mix is another critical factor with positive effects on travel propensity [53], walking propensity [33,45], walking duration, and walk frequency [34], and vital TBs [40] and a preference for metro travel. This finding aligns with previous research results. One compelling explanation is that high land-use mix typically features a diversity of destinations or activity types, indicating the presence of adequate service facilities nearby to meet various needs of older adults (e.g., dining, leisure, entertainment, and shopping). However, it negatively correlates with elderly walking as a travel mode [62] and cycling frequency and duration [79]. A plausible explanation is that areas with high land-use mix tend to exhibit elevated noise levels and crowding, which may negatively impact older adults’ psychological well-being and consequently reduce their willingness to walk or cycle. Notably, some studies report insignificant impacts of land use mix on elderly TBs, such as rail transit passenger flow [65]. The BE impacts on travel propensity often exhibit spatial heterogeneity. For instance, Yang and Zhu [53] observed bidirectional effects of land use mix on travel propensity, with positive impacts in a city’s western areas and negative impacts in the east. Cheng et al. [15] further identified threshold effects of land use mix on walking duration.

Public service facilities encompass a wide range of amenities that are significantly associated with TB in older adults. For instance, the density of markets and parks or squares promotes vital travel and subway usage among older adults; for instance, the density of food markets, parks, and public squares has been shown to promote active travel (e.g., walking and cycling) and metro usage among older adults [79]. Leisure and entertainment venues positively impact walking distance [16]. It can be inferred that this is because food markets, parks, public squares, and similar amenities are typically located near metro stations or in areas with good transportation accessibility, which increases older adults’ willingness to choose metro travel. Furthermore, higher densities of food markets, parks, public squares, and recreational facilities enable older adults to more conveniently access daily necessities, thereby increasing their travel demands and frequency. While the number of schools, hospitals, supermarkets, squares, parks, and tourist attractions near subway stations significantly increases the proportion of older adults using the subway for travel [41]. Proximity to markets, parks or squares, and chess and card rooms contributes to active travel patterns [14], whereas greater distance from commercial centers negatively affects walking frequency and duration, longer distances imply that older adults must expend more time and energy to reach their destinations. Additionally, the decline in physical functioning associated with aging further discourages travel behavior among this population [34]. However, some studies suggest that certain public service facilities have no significant impact on mobility, such as the number of parks, which was found to have no measurable effect on elderly TBs [57].

Public transportation facilities also play a vital role in shaping elderly TB, often demonstrating nonlinear effects. For example, a higher level of bus service promotes vital travel among older adults [79], while bus route density positively influences walking frequency and duration [34]. The distance to transportation infrastructure has a positive effect on trip frequency [5], and accessibility to bus stops enhances walking propensity [34]. Moreover, proximity to bus stops and the number of bus stops or subway stations encourage walking as a preferred mode of transportation [54,62]. This may be attributed to the high convenience and accessibility of bus stops, which enable older adults to more easily reach them on foot, thereby encouraging walking as their preferred travel mode. Conversely, Perchoux et al. [49] reported that a high number of bus stops inhibited walking as a travel mode. A plausible explanation is that an excessive concentration of bus stops may complicate pedestrian environments (e.g., increased traffic flow, crowded sidewalks), which reduces walking comfort and safety, consequently discouraging walking as a travel mode. Additionally, some research indicates that bus stop density has no significant impact on rail transit usage by older adults [80].

Intersection density has been shown to positively affect older adults’ TB in many studies. It promotes travel propensity [53], walking time [34], walking frequency [34,40], and the likelihood of choosing walking as a travel mode [45], given that older adults may experience osteoporosis and physical decline, a higher density of road intersections—indicating stronger street connectivity—can facilitate more frequent and prolonged walking activity among this population. However, intersection density can also have negative effects. For example, it may exacerbate the constraints of age and years of residence on the convenience of bus travel [59]. Some studies have found no significant impact of intersection density on walking propensity [34].

Urban green spaces, as a vital component of human settlements, provide environmental support for healthy activities among older adults and demonstrate significant health promotion benefits [8,9]. The density and quality of green open spaces are important for promoting walking behaviors among older adults. High-quality green environments, such as streetscape greening, have been positively associated with walking propensity and vital travel, often exhibiting threshold effects [46,79]. The proportion of park and green space area positively influences walking frequency [40], while the green view rate of streets enhances travel propensity [53]. This can be attributed to the fact that higher green view indices more effectively facilitate social interactions among older adults, with community parks and gardens serving as vital venues for such activities. However, some studies indicate that normalized difference vegetation index (NDVI) may inhibit walking as a travel mode [45]. A plausible reason could be that an exceptionally high Green View Index frequently indicates a lack or inadequacy of alternative attractions for senior citizens. And the number of parks and NDVI have been found to have insignificant impacts on travel propensity in some contexts [57]. Proximity to residential streets and diverse amenities can regulate the frequency of green space use by older adults [46].

Street connectivity is another vital factor. Perchoux et al. [49] identified a positive correlation between street connectivity and walking frequency, with a threshold effect indicating diminishing returns beyond a certain point. Road network density has been shown to positively moderate bus travel among older adults, higher street network density enables broader coverage of bus routes, reducing service gaps and consequently increasing the likelihood of older adults choosing public transit [59]. Additionally, Wang et al. [45] found that high intersection density, road network density, and block length within a unit area collectively promote walking frequency among older adults. This may be attributed to the increased road connectivity and shorter block lengths, which allow older adults to more conveniently select pedestrian routes while minimizing detour distances, ultimately enhancing walking frequency. Although limited evidence, slope showed a significant negative correlation with walking distance among older adults, as demonstrated in a study of Guiyang City [40], Steeper terrain significantly increases the physical exertion required for walking, particularly for older adults experiencing declining physical function who may struggle with sustained uphill walking. This reduces their willingness to walk, consequently decreasing walking distances. Residential building density positively influences walking inclination, with individuals in denser areas more likely to walk [45], This aligns with the impact mechanism of land-use mix.

The inner circle represents the dependent variable, while the outer circle illustrates the independent variables. Pink segments in the outer circle denote independent variables positively associated with elderly TB, whereas blue segments indicate variables with negative effects.

3.3.2. Main Individual Determinants for Elderly Travel

Sociodemographic factors play a crucial role in understanding elderly TB, encompassing both individual and family attributes. Variations in cultural background, gender, age, and other factors contribute to the heterogeneity observed in elderly travel patterns [54,60]. Among these, gender and age are the most frequently analyzed factors. For example, Xiao et al. [81] used a discrete-continuous model to analyze commuting preferences, finding that age and gender significantly influence commuting mode and time choices. Relatively speaking, males face greater work pressure and are more concerned with commute duration, preferring faster and more convenient modes of transportation such as cars for commuting. Females, with a lower perceived value of time, tend to opt for lower-cost commuting methods like buses. Furthermore, compared to young people, older adults are more inclined to choose slower modes of transportation such as bicycles, which may be related to their slower pace of life and the need for physical exercise. Similarly, Ingvardson and Nielsen [82] conducted a large-scale travel survey, revealing that public transportation usage declines with age. Martín and Páez [83] observed that women are more likely to walk or cycle than men, with younger individuals favoring cycling and older adults preferring walking. Age further shapes TB. In Hong Kong, individuals aged 70–79 undertake fewer mandatory trips during evening peak hours, while those aged 80 and above engage in more discretionary trips before the morning peak [84]. In Nanjing, age negatively correlates with travel distance, whereas education positively influences travel frequency and distance. Gender differences are also evident: elderly women travel shorter distances than men, make more maintenance trips, and engage in fewer discretionary trips [44] Moreover, suburban elderly residents are more likely to use private cars for accessing medical resources [60], though private care usage declines with age [85].

Economic characteristics also significantly impact travel mode choice. Fu and Juan [86] classified travel groups based on socioeconomic attributes, showing that activity-travel relationships vary across groups. Ao et al. [87] found that rural household registration and higher income levels were positively associated with car usage among rural Sichuan residents. Educational attainment, household income, and possession of driver’s licenses are positively associated with elderly mobility [25].

3.3.3. Regional Differences

Among the 46 pieces of literature, three studies focused on entire countries, namely the Netherlands, the United States, and China. The remaining studies were conducted at the city level, with three from Chongqing, two from Wuhan, four from Dalian, three from Guiyang, seven from Hong Kong, six from Nanjing, four from Xiamen, two from Shanghai, and two from Beijing. Additionally, there was one study each from Shantou, Guangzhou, Qingdao, Hamilton, Denver, Singapore, Rasht, Chiba City, Munich, and Luxembourg. It can be seen that most current studies are concentrated in coastal cities or high-density cities, with very little empirical research conducted in the western regions. Due to differences in economic and geographical conditions, various cities exert different influences on the travel behavior of the elderly. For instance, research conducted in Guiyang demonstrates that high population density has a negative impact on the walking distance of the elderly [72]. In contrast, a study in Nanjing found that population density can to some extent increase the walking activity of the elderly [15]. Additionally, research in Xiamen revealed that the density of bus routes has a positive impact on the walking frequency and walking time of older adults [34]. However, the study conducted in Guiyang indicated that higher convenience of public transportation services within a region may to some extent reduce the walking activity of the elderly [72].

4. Discussion

Through in-depth and systematic literature analysis: this paper comprehensively reviews 46 pieces of literature in related fields, aiming to accurately present the cutting-edge developments and latest advancements in this research area. This detailed review not only unveils the intricate internal connections between the BE and TB of the elderly but also provides us with a multi-dimensional and in-depth perspective for understanding this relationship.

The results of the visual analysis clearly demonstrate that in recent years, the research scale in this field has been gradually refined, shifting from a macro urban level to a micro community level. The research mechanisms have also become increasingly in-depth, evolving from studying correlations to exploring nonlinear relationships and mediating effects. This transformation profoundly reflects the increasing severity of the aging trend and highlights the crucial role of communities as core areas of daily life and key venues for social interaction among the elderly. Meanwhile, the research dimensions have shown a trend of continuous expansion and deepening. The scope of research has extended from the traditional domain of physical health to the realm of mental health, reflecting a profound concern for the overall well-being of the elderly. The evaluation indicators for the built environment have also evolved, gradually progressing from the initial “3DS” system to the “5DS” and even more comprehensive indicator systems, which signifies a deepening understanding of the complexity of the built environment. Additionally, the research areas have gradually expanded from high-density coastal cities to low-density inland cities, ensuring the universality and applicability of the research findings.

Currently, researchers are actively engaged in constructing an interdisciplinary comprehensive theoretical framework with the aim of comprehensively and accurately assessing the diverse impacts of the built environment on individual travel behavior. This framework has successfully encompassed multiple key dimensions of the built environment, such as street greenery, land use mix, and accessibility of transportation facilities, and has conducted more refined and in-depth classification and exploration of these elements. At the methodological level, research is gradually breaking free from the constraints of traditional basic correlation and regression analysis methods and instead adopting more complex machine learning algorithms and spatial analysis models, such as the Random Forest model, to explore the nonlinear relationships between the built environment and the travel behavior of the elderly. Furthermore, innovations in research methods are also reflected in the diversification of data types and analysis approaches. By integrating survey data, open-source data, and multi-source data and comprehensively utilizing descriptive statistical methods, mathematical models, and interdisciplinary comprehensive analysis, research is provided with a richer and more multi-dimensional perspective.

Research findings consistently indicate that numerous built environment factors, such as land use mix, public service facilities, public transportation facilities, intersection density, population density, street greenery rate, and street connectivity, have a significant impact on the travel behavior of the elderly. Additionally, this paper delves into the mechanisms through which personal attributes influence the travel behavior of the elderly. Specifically, gender and age, as fundamental demographic characteristics, exert a significant influence on the travel patterns of the elderly. Economic characteristics, such as income level, also significantly affect the choice of travel modes among the elderly. It is noteworthy that factors such as education level, household income, and possession of a driver’s license are positively correlated with the travel ability of the elderly, which further enriches our understanding of the factors influencing their travel behavior.

5. Conclusions

This paper systematically reviews and synthesizes existing literature to summarize research trends, theoretical backgrounds, influencing factors, and methodologies in studying the associations between the BE and elderly TBs. This study employs the CiteSpace bibliometric tool to conduct a detailed analysis of selected articles, thereby elucidating the mechanisms by which the BE influences elderly TB. These findings provide valuable insights for understanding current research trends and identifying future research directions in this field while also contributing to healthy aging and sustainable urban development.

5.1. Strategies to Promote Mobility for the Elderly

Based on the review in this paper, the following four strategies are proposed to ensure or promote the travel of the elderly:

(1)

Aging-friendly Community

Aging-friendly community development serves as the foundation for ensuring the mobility of the elderly. By optimizing the pedestrian environment within communities, such as constructing barrier-free sidewalks, adding rest benches, and providing sunshade facilities, safe and comfortable walking conditions can be provided for the elderly. Additionally, more public activity spaces, such as parks, squares, and green areas [88], should be planned within communities to offer leisure and socializing venues for the elderly, encouraging them to engage in outdoor activities. The rational layout of living facilities is also crucial, with markets, pharmacies, medical centers, etc., being located as close to residential areas as possible to shorten the daily travel distances for the elderly. Meanwhile, promoting aging-friendly residential design, such as installing elevators, adding handrails, and reducing step heights, can further facilitate the daily mobility of the elderly and enhance their quality of life.

(2)

Planning and Improvement of Public Transportation

The planning and improvement of public transportation are at the core of ensuring mobility for the elderly. Firstly, the coverage of bus stops should be increased to ensure that the walking distance for the elderly to reach a bus stop is within a reasonable range (e.g., within 800 m). Secondly, the design of bus routes should be optimized to increase coverage of important destinations such as hospitals, parks [89], and community activity centers, based on the travel needs of the elderly. Enhancing the quality of bus services is also crucial, such as increasing the frequency of bus departures, extending operating hours, and providing barrier-free facilities (e.g., low-floor buses, wheelchair ramps [90]), to ensure the convenience and safety of elderly travelers. Additionally, promoting community micro-circulation bus routes can effectively address the “last-mile” travel challenges faced by the elderly, further enhancing their travel experience.

(3)

Preferential Policies for the Elderly

The Third Plenary Session of the 20th Central Committee of the Communist Party of China proposed to actively address population aging and improve the policy mechanisms for the development of elderly care services and industries. The Central Economic Work Conference clearly stated the need to expand inclusive elderly care services. Formulating and implementing preferential policies for the elderly is an important means to promote their mobility. Firstly, free or discounted public transportation fares can be offered to the elderly to reduce their travel costs and encourage them to use public transportation more frequently. Secondly, for elderly individuals with mobility issues, exclusive travel services can be provided, such as on-demand community buses and barrier-free taxis, to meet their special needs. Additionally, providing travel subsidies for low-income elderly can alleviate their financial burden and enhance their willingness to travel. Finally, promoting elderly-friendly intelligent travel tools (such as one-click taxi hailing and real-time bus inquiry [19]) can help them plan their trips more conveniently, improving the efficiency and experience of their travel.

(4)

Optimizing Built Environment Using Big Data and Machine Learning Models

Numerous studies have examined the differences in travel behavior between older adults and younger populations [13]. Findings indicate that the built environment has a greater impact on older adults than on younger individuals. Therefore, policy formulation must account for age-related disparities and optimize the built environment to meet the mobility needs of older adults. First, collect and analyze travel data of older adults (e.g., public transit card swipes, GPS trajectory data) to identify their travel patterns (e.g., trip timing, distance, frequency) and understand their mobility demands. Mobile signaling data and social media data can be leveraged to detect activity hotspots (e.g., parks, hospitals, community centers) and determine high-demand areas. Next, researchers can employ machine learning algorithms (e.g., random forest, support vector machines, neural networks) to develop predictive models that assess the influence of built environment factors on older adults’ travel behavior. Classification algorithms (e.g., decision trees) can categorize built environment types (e.g., high-density residential areas, low-density retirement communities) to identify those best suited for older adults. Finally, based on analytical results, targeted optimization strategies can be proposed to enhance age-friendly urban planning.

5.2. Future Perspective

(1)

Expanding research to diverse urban contexts. Current research predominantly focuses on major cities such as Beijing, Shanghai, and Nanjing, with limited attention to border cities and other regions. Future studies should conduct comparative analyses across a wider range of urban settings, incorporating diverse regional development levels and lifestyles. Such efforts will enhance the understanding of overall travel patterns and health outcomes among older adults in different contexts.

(2)

Promoting interdisciplinary integration. The study of elderly TB has gathered interest from disciplines including urban planning, sociology, psychology, geography, and transportation. Future research should further integrate these disciplines, combining insights with public policy frameworks to provide a more holistic understanding of elderly TB.

(3)

Enhancing BE adaptations. Municipal departments should prioritize BE factors that significantly influence elderly TB. Strengthening barrier-free facilities, optimizing public transportation, implementing aging-friendly modifications, and addressing specific needs can enhance the safety, convenience, and comfort of travel, thereby improving the quality of life and social participation of older adults.

(4)

Understanding health-related attitudes and perceptions. Health-related attitudes, such as health awareness and preferences for active travel, likely influence TB. Future research should investigate how perceptions and experiences with transportation options affect travel choices among the elderly. Key questions include: What role do transportation-related attitudes play in shaping TB? How does health awareness influence mode choice? How do perceptions and attitudes towards travel evolve over time, and to what extent do they affect TB?

(5)

Currently, research predominantly focuses on eastern coastal cities or high-density urban areas, while studies on western cities or low-density regions remain relatively scarce. However, as China’s urbanization process accelerates and the population ages, the travel issues faced by the elderly in western cities and low-density areas have become increasingly apparent. Therefore, in future research endeavors in this field, there should be an appropriate increase in the proportion of studies dedicated to western cities or low-density regions. This not only contributes to promoting sustainable and balanced development across the country but also facilitates a more comprehensive understanding of the characteristics and patterns of elderly travel behavior in diverse geographical locations and population densities. Furthermore, it provides a scientific basis for formulating more targeted and effective policies to address elderly travel needs, thereby better satisfying their travel demands and enhancing their quality of life.