For the needs of this study, five cities from different Balkan countries were selected for the questionnaire survey. These are Niš (Serbia), Plovdiv (Bulgaria), Athens (Greece), Bucharest (Romania) and Štip (North Macedonia). The selected cities vary in size, built environment characteristics and dedicated pedestrian and cyclist infrastructure.

Niš is the largest city and industrial center in the south-east of Serbia, with a population of around 250,000 citizens which also makes it the third largest city in the country. The private car still comprises the dominant transport mode. The public transport network primarily relies on buses. Dedicated infrastructure for walking and cycling has not been a priority for the city so far.

Plovdiv is the second largest city in Bulgaria with a population of 348,000 inhabitants, 52% of which are women. Plovdiv boasts the most extensive cycling network, which is continuously expanded to further promote the use of active transport modes. The total length of the cycle lanes is 60 km, covering 58.8% of the city’s area. Additional factors including the wrongly positioned bus stops and the ongoing infrastructure reconstruction activities, also hinder cycling continuity.

Athens, the capital of Greece, has an urban population of 3,059,764 and a metropolitan population of 3,638,281 inhabitants, making it the largest city of the country. Athens is a major transportation hub and has an extended public transport network that serves the city and its metropolitan area. The Athenian metro is the most reliable public transport mode. In addition, trips are served by a tram and an extensive bus network, but their performance is affected by traffic congestion. Although sustainable transport options are offered, Athenians comprise a car-dependent population, with the proportion of more than 50% of the total trips made by private car. In general, the quality of pedestrian infrastructure is poor with narrow or even non-existent sidewalks, obstacles on them (ranging from kiosks and restaurant seats to parked vehicles) and bad surface conditions. Considering the cycling network, its quality is even worse.

Bucharest, the capital of Romania, is a vibrant metropolis with a population of approximately 2 million residents, making it the largest city in the country. Public transportation is widely used, with an extensive network of buses, trams, and the metro system facilitating daily commutes for over 1.2 million people. Active mobility is also on the rise, with increased investments in cycling infrastructure and pedestrian zones, including over 25 km of dedicated cycle lanes. Despite these improvements, walking and cycling rates remain lower than the EU average, with only about 5% of trips made by bicycle and 10% by walking, compared to the EU averages of 8 and 12%, respectively.

Štip is the smallest participating city with 44,866 inhabitants. The city is located in the eastern part of the Republic of North Macedonia and is the 7th largest city in the country. The city is connected to its suburbs and neighboring municipalities through a bus network, and its pedestrian and cycling infrastructure spans approximately 7 km. The terrain in Štip is diverse, with flat areas along the Bregalnica and Otinja rivers, and hilly regions in other parts of the city. On the other hand, hilly terrains in some city areas reduce the uptake of active transport modes. For example, cycling, as a recreational activity, is popular around the fortress Isar or the Štip Kale. However, reaching the city’s most famous landmark requires climbing more than 400 stairs.

A questionnaire survey was designed to identify knowledge gaps regarding physical activity, mental well-being, and the built environment in the Balkans region. The tool used was a revealed preference questionnaire, which was designed following a mixed-methods approach, incorporating both quantitative and qualitative questions. The questionnaire was divided into five sections: “Neighborhood and Mobility Attributes,” “Physical Activity,” “Nutrition,” “Well-being,” and “Socioeconomic Factors.”

Data on mobility attributes was collected through questions revealing participants’ modal choice preferences. Data on neighborhood attributes is classified into two categories: perceived attributes and objective characteristics of the built environment. The perceived attributes were collected utilizing the ALPHA environmental questionnaire (Spittaels et al., 2009). This involves a standardized set of questions addressing various elements of the neighborhood environment such as land uses, walking and cycling infrastructure, prevailing conditions including safety and security and so on. To explore the objective built-environment elements, participants were asked to indicate the nearest intersection or landmark to their home [their exact home address inquired, in accordance with the General Data Protection Regulation (GDPR)]. Objective built environment variables, such as link density, street-length density, intersection density, and link-node ratio, were calculated within a 600-meter catchment area utilizing GIS software. The physical activity level of the participants was estimated using the short form of the International Physical Activity Questionnaires (IPAQ) (Lee et al., 2011). Through IPAQ self–reported physical levels and attributes of engagement in moderate and vigorous PA, including the number of days and time (minutes) spent for walking, sitting, and screen time are recorded. Furthermore, three questions on eating habits provided data on participants’ nutrition. Participant well-being was measured using the Warwick-Edinburg Mental Well-Being Scale (WEMWBS). This scale is aimed at identifying the perceived mental well-being, through a set of 14 different statements referring to personal feelings and different aspects of well-being. Last, data regarding socioeconomic variables was collected in the last section of the questionnaire and included information on age, marital status, household size, number of children, work status, gross household income, financial status, and private transport mode ownership and accessibility.

The questionnaire was initially designed in English, while prior to being distributed in the five cities it was translated to the respective languages. A set of guidelines was also created for clarification reasons. In total the questionnaire consisted of 49 questions, and participants required approximately 10–15 min to complete it.

Data was collected over a three-month period, through on-line distribution. Online distribution included posts at various social media channels and groups, as well as to participants known to the interviewers or interviewees. Efforts were made to ensure a heterogeneous sample, including participants with diverse sociodemographic characteristics, to capture a wide range of perspectives. Participants who had difficulty understanding the questions were invited to an online meeting to receive additional information and guidance on how to answer them. As such, the sampling method comprised a combination of convenience, snowball and quota sampling methods. For the needs of the study, participant gender and age were predefined, i.e., participants had to be young women, aged between 17 and 40. The final sample consisted of 1,222 women, with their distribution across the different cities being 49% (597) from Bucharest, 17% (205) from Niš, 15% (135) from Athens, 11% (191) from Štip, and 8% (94) from Plovdiv.

The sample composition in total and per city is illustrated in Table 1.

As anticipated, the sociodemographic characteristics of the survey participants differ across the five cities/countries. In particular, the predominant age group varies across the pilot cities, with the sample concentrated in the 17–30 age range. Considering employment, the prevalent groups are students and women who work. Marital status is fairly evenly distributed across the different groups, while only a small proportion of women have children. Most women in the sample own a private car, and the majority report a middle-income financial status category.

Following data collection, the responses provided in text format were first translated from the local languages into English. Two rounds of data cleaning and validation were then conducted to create a valid dataset. The next step involved variable coding, which was primarily required for categorical variables. The dependent variables were MPA, VPA and mental well-being, while the independent were the built environment and neighborhood subjective variables, the built environment objective variables and the different socioeconomic characteristics.

The index utilized to represent moderate and vigorous PA was calculated according to the guidelines of the short version of IPAQ, with the coefficients for moderate and vigorous PA being 4 and 8, respectively. The total Metabolic Equivalent of Task (MET) was calculated as the sum of MET-minutes per week from walking, moderate, and vigorous activities. The MET coefficients used were 3.3 for walking, 4.0 for moderate physical activity (MPA), and 8.0 for vigorous physical activity (VPA) based on IPAQ protocol. To estimate mental well-being index, the methodology reported in Stewart-Brown et al. (2011) was adopted. Participants rated different elements of their well-being utilizing a 5-point Likert scale for each question. The mental well-being index was then computed by summing the scores of all 14 questions for each participant. Figure 1 shows the different levels of MPA, VPA, and well-being among various countries included in this study. To compute the objective built environment parameters, GIS software (ArcMap 10.4) was used. Shapefiles from the pilot cities, i.e., Athens, Štip, Plovdiv, Bucharest, and Niš were downloaded from OpenStreetMap, and participants’ “home” locations (indicated by the nearest intersection or landmark) were transferred from Google Maps to ArcMap. The disaggregated built environment variables, which were link, intersection, and street-length density, within a pre-defined catchment area were calculated in GIS using network and spatial analysis tools. The catchment area was defined as a 600-meter radius, measured along the street network, in accordance with urban transportation literature (Masoumi et al., 2019; Li et al., 2019).

Kolmogorov–Smirnov and Sharipo-Wilk tests were applied for the assessment of the variable distributions, which would determine the statistical analysis that would be adopted. A p-value smaller than 0.05 suggests that the null hypothesis, which is that the variable follows a normal distribution, should be rejected. The computed p-values (Table 2) were below 0.05, hence the three dependent variables – namely, moderate and vigorous PA, and mental well-being among young women in the Balkans are not normally distributed.

Multiple linear regression (MLR) modelling was performed to identify the subjective and objective built environment and neighborhood factors that affect the engagement in moderate and vigorous PA, and the mental well-being of the sample, and provide answers to the study’s research questions. The subjective built environment and neighborhood variables were derived from the questionnaire based on the perceptions and attitudes of participants reported in the relevant questionnaire section. These variables include the availability of different routes for cycling, the pleasantness of these routes, land use diversity, the pleasantness of the walking environment, the availability of green spaces and trees near pedestrian areas, shops in the neighborhood, and segregated cycling routes in the neighborhood. The objective built environment variables are the link-node ratio, link density, and street-length density, as specified in the previous section. However, only those exhibiting significant associations in the MLRs were kept in the final models.

To obtain better results, socioeconomic variables were treated as control variables in the MLR models. The control variables for MPA, as indicated from the statistical analysis, were age, household size, number of children, income, and employment status. Additionally, age, household size, number of children, and income were the control variables in the VPA model. Finally, the only controlled variable in the MLR for mental well-being was income. Different socioeconomic variables were fixed in the multiple linear regression (MLR) models to achieve the best results, characterized by the highest R-squared value, more significant independent variables, and statistically significant control variables.

The five pilot cities exhibited variations in the objective built environment characteristics, which also represent street connectivity. Higher values of link density, link-node ratio, and intersection density are associated with greater street connectivity in the neighborhood. The average link-node ratios were found to be 2.859, 1.476, 1.741, 1.743, and 2.553 for Athens, Bucharest, Niš, Plodiv and Štip, respectively.

MPA and VPA present distinct patterns and mean values across the different countries. In particular, the distribution between MPA and VPA is rather similar in the explored cities in Greece, Bulgaria and Romania. In contrast, participants from Serbia engage in substantially more VPA than MPA, while women from North Macedonia engage in more MPA than VPA. Despite variations in MPA and VPA levels across countries, the average mental well-being score remains relatively consistent, ranging from approximately 49–56 (Table 3).

To explore the differences of MPA, VPA and mental well-being between countries, the Kruskal-Wallis test was performed. Tables 2, 4, 5, illustrate the results.

Results indicate that almost all pairwise comparisons show statistically significant differences considering MPA and VPA. These are reflected in Tables 2, 4, where p-values exhibit values equal or lower than 0.05. Moreover, although the differences of the mental well-being scores across the five cities were relatively small, the statistical analysis suggests that these differences are statistically significant for most pair-wise comparisons (Table 5).

Two MLR models were developed separately for MPA and VPA. Table 6 summarizes the MPA model and the related ANOVA F-test. Most of the socioeconomic variables (age, household size, number of children, gross household monthly income, and employment status) affect MPA, however, as the aim of the study is to isolate the effect of the built environment on MPA, we treat them as control variables, and their effect is not elaborated in the present study. The only objective built environment variable in the model is street-length density, which is marginally significant (p = 0.097). It is positively correlated with MPA, suggesting that street connectivity increases the MPA levels of young women. Two subjective built environment variables were found to significantly affect MPA. The availability of shops near the living place presents an 8% reduction in MPA probably because shopping, i.e., walking around the shops is associated more with light PA rather than MPA (p = 0.016). Moreover, the presence of biking routes in the living neighborhood is, as expected, positively correlated with MPA, and is associated with 8% more MPA among the respondents of this study (p = 0.019). It should be noted that there is no multicollinearity among the model’s independent variables, as the VIF values ranged between 1 and 1.3. The ANOVA F-test results indicate that the model is valid (p < 0.001).

Table 7 presents the MLR model for VPA. In the case of vigorous PA, there is a smaller number of socioeconomic factors that affect the variable levels. These are age, household size, number of children, and gross household monthly income, and were treated again as control variables. Two objective built environment variables, those of link density and link-node ratio were found to be significant in the model (p = 0.001 and p = 0.042, respectively). Both represent street network connectivity and both show that there is a positive relationship between connectivity and VPA. When link density, which is represented by ratio of intersections to the area, increases VPA also increases among the respondents. As the link-node ratio increases, represented by the ratio of the number of streets to the number of intersections, both connectivity and vigorous physical activity (VPA) decrease. Moreover, specific subjective built environment variables appear to affect VPA. These are the availability of shops in the neighborhood and the availability of biking routes. Availability of shops in the neighborhood is associated with 6% lower VPA among the respondents (p = 0.089). On the other hand, availability of cycling routes in the neighborhood is significantly correlated with 11% more VPA (p = 0.002). The model is valid as the p-value of the ANOVA F test is <0.001. The model has an R² of 6% and there is no collinearity between the independent variables, as the VIF values are between 1 and 1.3.

MLR modelling was employed to analyse the associations between built environment variables and mental well-being. The respondents’ well-being scores were treated as the dependent variable. The model achieved the best fit by treating gross household and monthly income as control variables. Five subjective built environment factors were identified as having an impact on mental well-being levels, while objective built environment variables related to street network were not found to be influential (Table 8).

The availability of cycling routes in the neighborhood is strongly correlated with perceived well-being (p < 0.001). The correlation is positive, and the model results show that the availability of biking routes in the neighborhood is associated with a 12% higher well-being score. The diversity of land use, reflecting the presence of different urban functions in the living neighborhood, is significantly and positively associated with perceived well-being (p = 0.006). This indicates that a diverse land-use environment results in an 8.5% higher well-being score, particularly among women. The attractiveness of cycling routes and their surrounding environment is marginally positively correlated with well-being (p = 0.072), with a 7% increase in mental well-being associated with more attractive cycling routes. Similarly, the attractiveness of the neighborhood for walking comprises yet another significant independent variable (p = 0.009). Its positive coefficient indicates that the attractiveness of the area for walking results to a 10% increase in mental well-being among the respondents. Finally, the presence of green areas and trees in the neighborhood shows a marginally significant association with mental well-being (p = 0.077), corresponding to a 6% increase in the reported well-being. The p-value of the ANOVA F-test is below 0.001, indicating model validity. R² is 7.8%, meaning that the independent variables in the model explain 7.8% of the variation in the mental well-being of young women. As seen in Table 8, there is no sign of multicollinearity among the independent variables, as the VIF values range from 1 to 1.5.

One of the objective built environment variables affecting MPA levels of women in the Balkan countries was found to be street-length density, which is marginally significant and positively correlated with MPA, indicating that areas with densely connected networks promote more MPA. This finding aligns with previous research, which has shown that community design with higher street connectivity promotes walking and cycling and encourages active travel by offering more direct and safer routes (Saelens and Handy, 2008; Frank et al., 2005). Similarly, Sallis et al. (2012) demonstrated that urban areas with high intersection density and accessible infrastructure encourage greater PA participation. However, the relationship between street connectivity and MPA in our study appears inconsistent, as some countries with higher street connectivity (e.g., Greece) report lower MPA and VPA levels and vice-versa (e.g., Serbia). This suggests that not only street connectivity, but additional factors also affect the PA levels (e.g., weather conditions, pedestrian areas, sidewalks, lighting, overall safety etc.). For example, in our study, in Athens, pedestrian infrastructure is identified as low quality with narrow or non-existent sidewalks and bad surface conditions, making the streets unattractive for walking, running, and cycling. Sallis et al. (2012) reported that the availability of high-quality infrastructure for walking and cycling (e.g., sidewalks, bike paths, crosswalks) promotes active travel PA among the population, hence the lack of such infrastructure discourages PA. Additionally, high temperatures during summer months can be an additional burden and further discourage PA. A similar situation is observed in Bucharest, where, despite the relatively good infrastructure and street connectivity, and the increased investments in cycling infrastructure and pedestrian zones (including over 25 kilometers of dedicated cycling lanes), participants from Romania exhibited the lowest reported minutes for MPA and the second lowest for VPA. On the contrary, Niš (Serbia) is reported to have the second lowest values for street connectivity, but still exhibits the highest reported minutes for VPA, and a higher level of MPA compared to Greece and Romania. Although the dedicated infrastructure for walking and cycling is not recognized as a priority for the city of Niš, it is described as a relatively pedestrian-friendly city, that offers a network of sidewalks and pedestrianized areas, especially in its central part. Cycling also gained popularity among residents, as a result of the design of dedicated cycling lanes and the adoption of a bike-sharing scheme. These examples suggest that additional objective factors (infrastructure, climate, air pollution etc.), as well as habits, cultural, and social factors also play a role when considering PA behaviors (Sallis et al., 2012). Societal barriers are especially present in Balkan countries. Pojani et al. (2017), based on results from a study conducted in the Netherlands, Albania and Kosovo, reported that Balkan cities face societal barriers such as the high status associated with car ownership and the perceived incompatibility between cycling and the female gender. This discourages people from cycling, particularly limiting PA engagement resulting from utilitarian walking or cycling trips. The location of places for walking and cycling is also important. Zhang et al. (2019) identified significant positive associations between park-based leisure PA, park lighting, and walking and cycling trails. Among the contributing subjective built environment variables, the availability of shops in the neighborhood was found to be significantly associated with an 8% decrease in MPA. Although proximity to shops encourages walking, which is a light-intensity physical activity, participants may not include shopping time when reporting MPA. Previous studies have noted similar trends, where accessibility to retail destinations is linked to increased walking, but not necessarily with higher-intensity physical activities (Giles-Corti et al., 2005).

On the other hand, the presence of cycling routes near the living area is positively correlated with MPA, supporting the notion that dedicated cycling infrastructure promotes active travel and physical activity (Winters et al., 2010; Pojani et al., 2017). Living close to a natural outdoor environment suitable for different forms of PA was found positively correlated with the PA levels of adults in a study performed in different European cities (Triguero-Mas et al., 2017). The observed 8% increase in MPA due to the presence of cycling routes is consistent with prior findings, highlighting the importance of cycling infrastructure in active commuting (Buehler and Pucher, 2012; Pojani et al., 2017).

Two objective built environment variables, link density and link-node ratio, were found to significantly affect vigorous physical activity. The positive relationship between link density and VPA suggests that areas with a higher number of intersections per unit area facilitate more vigorous activity, likely by promoting walkability and access to various destinations (Ewing and Cervero, 2010). However, the positive association between the link-node ratio and VPA suggests that as the number of streets divided by intersections increases, connectivity weakens, potentially reducing accessibility and engagement in vigorous activity. This finding is somewhat counterintuitive and warrants further investigation.

Similarly to the MPA model, the subjective built environment variables function in the same manner. The availability of shops is negatively correlated with VPA, reinforcing the idea that shopping-related movement is generally of lower intensity. The availability of cycling routes shows a significant positive relationship with VPA, with an 11% increase in vigorous PA, further confirming the role of cycling infrastructure in promoting active travel (Buehler and Pucher, 2012).

Overall, these findings highlight the importance of both objective and subjective built environment factors in shaping physical activity behaviors. Higher street connectivity and cycling infrastructure appear to facilitate moderate and vigorous physical activity, whereas proximity to shops may promote lighter activity such as walking rather than MPA or VPA. These results suggest that urban planning and transportation policies should prioritize the development of connected street networks, with available sidewalks, safe pedestrian areas, and dedicated cycling infrastructure to encourage higher levels of physical activity among young women. In this regard, it should not be forgotten to consider issues that arise from unplanned urban expansions leading to environmental and health hazards, including air pollution, diminished safe spaces for walking and cycling, and so on (Agarwal et al., 2007).

Positive practices, such as bike-sharing schemes, community support in creating green spaces and urban zones that encourage PA, subsidies for bicycles and other micro-mobility transport modes purchase, and other activities like forming walking and cycling groups can serve as effective strategies. In this regard, in PA Factsheets from 2021, WHO included urban planning in one of the policy responses aimed at improving PA levels among EU member countries. Within these recommendations, it is suggested that “all population groups should have access to infrastructure conducive to active leisure time.” Guidelines have been prepared within the European Commission-funded project Improving Infrastructures for Leisure-time Physical Activity in the Local Arena (IMPALA), including sports facilities, infrastructure, and urban “green” and “blue” spaces” (WHO, 2021 PA factsheet, pp. 11). This has initiated the design and implementation of specific actions and policies in the EU countries. For example, in Greece a national bicycle strategy and a national accessibility plan have been designed. Within this framework, a “bicycle observatory” was developed to monitor the implementation of the strategy, assess cycling infrastructure and record the number of daily bicycle trips (WHO, 2021, pp. 108).

The relationship between built environment variables and the mental well-being of young women in the participating countries comprised the second research question of the present study. The respective MLR model for mental well-being highlights significant associations between subjective built environment variables and perceived well-being. Notably, no objective built environment variable related to the street network was found to be significant. Instead, five subjective built environment variables contribute significantly to perceived well-being, explaining 7.8% of the variance. They are the availability of cycling routes, the extent of land use diversity, the attractiveness of cycling routes, the overall attractiveness of the neighborhood and the presence of green areas and trees.

The availability and attractiveness of cycling routes in the neighborhood were identified as relevant and important for mental well-being. These factors were also found to be associated with PA, highlighting the link between well-being and PA. This finding is in line with previous studies noting that access to active transport infrastructure enhances physical activity levels (Mueller et al., 2015) and mental health by reducing stress and promoting outdoor engagement (Ma and Dill, 2015). To comprehend outdoor engagement and its impact on well-being, it is worth emphasizing the links between outdoor activities and spending time outdoors with well-being. Engagement in outdoor activities encourages positive emotions and enthusiasm providing pleasure, enjoyment, and personal satisfaction, comfort and intense emotions, satisfaction from personal achievements, vital strength and experience from nature (Clough et al., 2016; Tsaur et al., 2015; Crust et al., 2013). Spending time outdoors enhances awareness of personal health and its importance for overall well-being. It fosters a commitment to develop healthy habits and cultivates a healthy and active lifestyle as a personal goal, offering numerous benefits for both mental and physical well-being (Deenihan and Caulfield, 2014). Evidence on the positive correlations of outdoor activities on different dimensions of health and well-being can be found in several studies. Examples include revitalization, and reduction in tension, confusion, anger, and depression (Eigenschenk et al., 2019), greater enjoyment (Crust et al., 2013), improvement of individual well-being and increased perception of personal health and stronger nature-relatedness (Puhakka et al., 2018). This is also confirmed by the results of the present study, which showed that the presence of green areas and trees is associated with a 6% increase in mental well-being. This finding supports the growing body of evidence indicating that exposure to green spaces and tree-lined streets enhances psychological well-being by reducing stress and promoting relaxation (Gascon et al., 2015), and by increasing enjoyment (Crust et al., 2013).

The attractiveness of cycling routes and the overall attractiveness of the neighborhood for walking and cycling were also found to be positively associated with well-being, as they increased the well-being scores by 7 and 10%, respectively. These results reinforce the notion that aesthetically pleasing and accessible pedestrian and cycling environments contribute to mental health benefits (Mitchell et al., 2015), emphasizing the links between aesthetics and well-being.

The presence of diverse land uses in the neighborhood comprises yet another significant predictor associated with an 8.5% increase in well-being. Land use diversity facilitates access to commercial, recreational, and institutional establishments that are located within close proximity, thereby increasing the likelihood of utilitarian active travel, and hence PA levels and well-being. As indicated in our study, the different cities offer a variety of available facilities and spaces that meet PA needs and support mental well-being. This aligns with research emphasizing the psychological benefits of mixed land use, which provides greater opportunities for social interaction, active travel, and access to essential services (Sugiyama et al., 2008). Areas with a diverse mix of land uses encourage walking and cycling by providing essential services close to residential areas (Frank et al., 2005), affecting positively PA levels as people are more likely to walk or bike to nearby destinations (Saelens and Handy, 2008). It also supports community engagement and improves mental well-being by offering various public spaces and social opportunities (Sugiyama et al., 2008).

Overall, these findings highlight the importance of land use planning and urban design in shaping physical activity behaviors and mental well-being. Well-connected street networks and dedicated cycling infrastructure appear to facilitate both moderate and vigorous physical activity. Additionally, aesthetically pleasing and diverse land uses, including green spaces and pedestrian-friendly environments, contribute to improved mental well-being. These results underscore the need for policymakers and urban planners to prioritize active transport networks and mixed land-use designs to support both physical and mental health outcomes among young women.

By assessing the differences in the effects and coefficients of subjective and objective built environment features on levels of physical activity and mental well-being, the results showed that perceived environment variables had stronger associations than objective ones. This finding suggests that mobility behavior and well-being among young women in Balkan countries may be more closely linked to their perceptions of urban design, such as perceived neighborhood safety, mixed land use, attractiveness, and aesthetic quality than to the actual physical characteristics of the environment.

It is also important to note that the objective built environment features used in this study primarily reflected street-network characteristics, rather than detailed measures of land use or urban design elements. Therefore, to better understand the differences between subjective and objective impacts of the built environment, it is necessary to conduct direct, one-to-one comparisons between corresponding variables in future research.

This study presents interesting findings on the effects of built environment on the PA levels and mental well-being of young women from Balkan countries. However, it has specific limitations that may reduce its representativeness and generability potential.

The structure of the utilized sample is identified as one of the study limitations. This study was conducted in five different cities, with a sample that is diverse in terms of socio-economic structure, comprising mainly students in Romania and North Macedonia, employed women in Serbia, and mixed groups in Greece and Bulgaria. Therefore, the generalization of the findings should be made with care.

Using self-reported tools for physical activity and mental well-being, as well as subjective reports of some variables of the built environment, can be another limitation. Supplementing self-reported tools with objective measures (e.g., accelerometers, GPS tracking, or psychological assessments) could enhance data accuracy.

The cross-sectional design is another identified limitation. The study is based on a cross-sectional dataset, meaning it captures a single point in time. As a result, causality cannot be established between land use diversity, physical activity and mental well-being. Longitudinal or experimental studies would provide stronger causal inferences. Longitudinal data can provide a more accurate and comprehensive understanding of how built environment features influence PA and well-being in Balkan countries. So, the current study suggests that future studies should examine the correlates and determinants of physical activity and mental well-being using longitudinal designs. Such an approach would allow researchers to better capture causal relationships over time and account for seasonal, social, or environmental changes that may influence behavior and health outcomes.

While control variables were included, other factors; such as personal motivation, weather conditions, safety concerns, air pollution, and cultural norms may also influence PA levels and well-being, particularly among women. These factors were not directly accounted for in the model, presenting an additional limitation. Although the aim of this paper was to examine the relationship between subjective and objective built environment features and physical activity and well-being, it is important to acknowledge that contextual factors and cultural variation can significantly shape perceptions of the built environment. This paper suggests that incorporating some of these variables in future studies would provide a deeper and more comprehensive understanding of the research problem.