Urbanization, loneliness and mental health model - A cross-sectional network analysis with a representative sample

The study investigates how urbanization relates to mental health by examining the roles of physical and social environmental factors and the mediating influence of loneliness. Urbanization is rising globally and has been associated with elevated rates of anxiety and mood disorders, though findings are mixed and mechanisms unclear. Mental health risk factors in urban settings span social domains (e.g., low social capital, segregation, weak neighborhood cohesion) and physical domains (e.g., noise, pollution, oppressive architecture, safety). Green spaces may protect mental health through proximity, quality, and access, while poor neighborhood architectural conditions are linked to worse physical health. Loneliness—an aversive emotional state with evolutionary roots—has been posited to increase in crowded, urban conditions and is associated with social withdrawal and mental health problems; it may mediate links between social/physical environments and mental health. Building on prior frameworks of urban mental health and complexity science, the authors propose a theoretical model wherein neighborhood cohesion and spatial/physical environmental factors influence stress, anxiety, and depression and physical health, with loneliness as a mediator, and individual sociodemographics as additional determinants.

The paper reviews evidence that urban environments can be associated with higher prevalence of anxiety and mood disorders, although not consistently across studies. Social urban risk factors include low socioeconomic status, reduced social capital, segregation, and weak neighborhood cohesion; greater social cohesion generally relates to better mental and physical health and may protect against anxiety and depression. Neighborhood cohesion, a form of social capital comprising neighborhood belonging and social cohesion, has been linked to reduced mental health problems. Physical urban risk factors include air and noise pollution, oppressive building forms (e.g., high-rises), and higher perceived physical threats, whereas use and access to public and green spaces are associated with better mental health; the diversity, quality, proximity, and number of green spaces matter. Poor neighborhood architectural conditions correlate with poorer physical health. Spatial cohesion is framed as distances to and frequency of use of green and urban public spaces. Loneliness emerges as a contemporary public health challenge, associated with overcrowding, higher population density, reduced social engagement, and built environment deficits; it can mediate relationships between social support and depression, and between green space use and mental health. Some studies suggest green space reduces loneliness (via social connection and solace), and housing disrepair is associated with loneliness. The review underscores gaps and mixed findings regarding direct effects of urbanicity on mental health and loneliness, motivating the current model-driven, network-analytic approach.

Design and setting: Online cross-sectional survey conducted June 2–20, 2023 in the Metropolis GZM (Upper Silesia and Dąbrowa Basin), Poland. Sampling and participants: Representative sample of 3,296 residents aged 18–90 years recruited via Poland’s ARIADNA research panel; participation was voluntary with prize-based incentives. The sample was representative by age, sex, employment and student status across 26 urban, 2 urban-rural, and 13 rural municipalities. Response rate was 100%. Mean survey completion time was ~12 min 87 s. Initial N=3301; five participants aged 95–99 were excluded per age criteria. Sample size calculation used a 99% confidence level, SD=0.5, margin of error=2.81 for a population of 2,229,806, yielding a required N=2106; the final N≈3300 increases stability for network models and meets SEM-based CMB minimums (>300). Measures: - Mental health: Perceived Stress Scale (PSS-10; 10 items, 0–4; sum 0–40; α=0.86; high stress ≥20). Generalized Anxiety Disorder (GAD-7; 7 items, 0–3; 0–21; cutoff 10; α=0.92). Patient Health Questionnaire (PHQ-9; 9 items, 0–3; 0–27; cutoff 10; α=0.89). - Loneliness: Revised UCLA Loneliness Scale (R-UCLA 3; 20 items, 1–4; sum 20–80; α=0.94). - Physical health: Self-rated general health (GSRH; single item, 1=excellent to 5=poor; higher=worse). Physical activity: weekly aerobic activity time aligned with WHO guideline threshold (≥150 min vs <150 min). - Social (neighborhood) factors: Neighborhood Cohesion Scale with two subscales—social cohesion (8 items; 5-point Likert; sum 8–40; α=0.80) and neighborhood belonging (7 items; 5-point Likert; sum 7–35; α=0.91). - Spatial cohesion: Self-reported frequency of use of public green spaces and urban areas during the last week (1=not at all to 5=every day). Distance to nearest public green/urban areas (1=<300 m, 2=300–500 m, 3=500–1000 m, 4=1–2 km, 5=>2 km). - Physical environment: Urbanization (1=village, 2=small town <20K, 3=town 20K–99K, 4=city 100K–300K; dichotomized for network: 0=village/small town/town; 1=city). Personal space (m² per person). Building type (0=house; 1=block of flats/tenement). Poor neighborhood condition assessed via REAT 2.0-derived items (1=very good to 4=poor/needs immediate repair), summarized as building poor quality. - Sociodemographics: Sex, age, monthly net income, education, marital and parental status, employment. Common method bias (CMB): Addressed by randomized question order and varied scales. Tested via CFA with and without a Common Latent Factor (CLF). The baseline CFA (65 indicators; 7 latent variables) fit acceptably: χ²(1940)=13,379.63, p<0.001; CFI=0.913; RMSEA=0.042. Adding CLF significantly improved fit, indicating some response bias. Items with standardized coefficient differences >0.2 between CLF and non-CLF models were flagged (primarily neighborhood cohesion items); other items showed no bias. Caution is advised interpreting neighborhood cohesion results. Statistical analysis: Descriptive statistics summarized sociodemographics and key variables. Group differences across urbanization levels were assessed via one-way ANOVA with Tukey HSD post hoc tests and effect sizes (Cohen’s d). Pearson correlations described bivariate associations. Network analysis estimated a mixed graphical model (MGM) using nodewise regression with Lq-penalization; EBIC model selection (γ=0.25), pairwise model (k=2). Layout used Fruchterman–Reingold algorithm. Centrality metrics emphasized strength and expected influence (EI), with closeness and betweenness also reported. Stability assessed via nonparametric and case-drop bootstrapping (1,000 samples); correlation stability (CS) coefficients computed. Predictive accuracy indices (RMSE, R² for continuous; concordance correlation coefficient for categorical) reported in Supplementary Materials.

Sample characteristics: N=3296; age 18–90 (M=44.70, SD=16.47); 52% men; urbanicity distribution included 46.8% city residents (100K–300K), 21.6% village, 6.6% small town, 25.0% town. Most (72%) were physically inactive per WHO threshold. Prevalence and descriptive indices: - High perceived stress: 52% (low stress: 19%). - Anxiety risk (GAD-7 ≥10): 36%. - Depression risk (PHQ-9 ≥10): 39%. - Means (SD): Stress 18.90 (6.84) [0–40]; Anxiety 7.95 (5.64) [0–21]; Depression 8.88 (6.64) [0–27]; Loneliness 41.70 (9.46) [20–75]; Self-rated health 3.04 (0.77) [1=excellent to 5=very poor]. ANOVA by urbanization (very small effects): - Stress: F(3,3292)=8.83, p<0.001, η²=0.008. Rural > towns (d=0.14, p=0.038) and cities (d=0.20, p<0.001); small towns > cities (d=0.24, p=0.005). - Anxiety: F(3,3292)=7.79, p<0.001, η²=0.007. Rural > towns (d=0.13, p=0.043) and cities (d=0.20, p<0.001); small towns > cities (d=0.20, p=0.024). - Depression: F(3,3292)=9.00, p<0.001, η²=0.008. Rural > cities (d=0.22, p<0.001); towns > cities (d=0.20, p=0.026). Overall, city residents exhibited better mental health indices than rural and small-town residents. Correlations and selected network edge weights: - Strong positive interrelations among mental health nodes: Anxiety–Depression (weight≈0.70); Stress–Anxiety (≈0.41). Loneliness correlated strongly and positively with mental health indices and with worse self-rated health. - Neighborhood belonging–Social cohesion positive (≈0.53); Social cohesion–Loneliness negative (≈−0.17). - Poor neighborhood condition–Neighborhood belonging negative (≈−0.29). - Spatial/usage interrelations: Green-space use–City-space use frequency (≈0.86); Place of residence (city)–Blocks of flats (≈0.65); Proximity green–Proximity urban public spaces (≈0.65). - Age showed small negative associations with stress (≈−0.08) and was central in paths linking urbanicity with other dimensions. Network centrality and structure: - Betweenness: City (urbanicity) had the highest betweenness, acting as a bridge between domains; Blocks (housing type) and Age also relatively high; Sex and Income near zero. - Closeness: Age highest; Sex and Income lowest. - Expected Influence (EI): Anxiety highest positive EI, followed by Depression and Stress—indicating these nodes exert the strongest influence within the network. - Stability: Case-drop bootstraps showed high stability (CS≈0.75 for betweenness, closeness, EI, and strength), with 95% of bootstrap samples maintaining ≥0.7 correlation with original centrality estimates even after dropping up to 75% of data. Model-confirmed pathways: - Physical environment → Neighborhood cohesion → Loneliness → Mental health: relationships between physical environment and mental health were consecutively mediated by neighborhood cohesion and loneliness. - Urbanicity showed no direct link to mental health or loneliness in the network but functioned as a central connector across domains (sociodemographics, physical environment, spatial cohesion), explaining mixed findings in prior literature. - Spatial cohesion related strongly to physical health and physical environment factors; frequency of green-space use related directly to better self-rated health and, via physical health, to lower stress. - No direct link between spatial cohesion and loneliness; instead, links were explained via physical health, age, and neighborhood cohesion. - Stress was linked to younger age; neighborhood cohesion did not directly relate to urbanicity but increased with age and was inversely associated with loneliness.

The study validates a theoretical, interdisciplinary model in which urban physical and social environmental factors affect mental health through neighborhood cohesion and loneliness. Despite ANOVA indicating differences across residential settings, the network analysis shows urbanicity is not directly associated with mental health or loneliness; instead, it serves as a bridging node connecting sociodemographics, the physical environment, spatial cohesion, and health domains. This helps reconcile prior inconsistent findings on urbanicity’s mental health impact by highlighting indirect, system-level pathways and the role of age. Poor neighborhood architectural conditions undermined neighborhood belonging, which in turn related to social cohesion and lower loneliness, ultimately linking to better mental health outcomes. Neighborhood belonging emerged as more connected than interaction-based social cohesion, suggesting that fostering pride and attachment to place may be particularly impactful. Contrary to some prior work, there was no direct link between spatial cohesion and loneliness; the pathway appeared to run through physical health, age, and social cohesion. Frequency of green-space use related to better self-rated health and, indirectly, to lower stress, underscoring health benefits of green-space engagement. Mental health indicators were strongly interrelated, with anxiety exerting the largest expected influence, indicating that addressing anxiety may yield broad benefits across the network. Age played a pivotal role (highest closeness), shaping neighborhood belonging and health connections, whereas sex and income were less central. Overall, the findings emphasize that interventions should target social cohesion (especially belonging) and the built environment, alongside access and proximity to green and urban public spaces, to indirectly improve mental health.

The analysis confirms that physical environmental factors influence mental health primarily via neighborhood cohesion and loneliness. Urbanicity functions as a bridge across key domains but is not directly associated with mental health or loneliness, indicating complex, indirect relationships. Spatial cohesion contributes to better physical health, which in turn benefits mental health; frequent use and closer proximity to public green and urban areas support physical activity and self-rated health and are associated with reduced stress via physical health. Practical implications include strengthening neighborhood belonging and social cohesion; repairing and maintaining neglected neighborhood buildings (e.g., removing vandalism/graffiti); and urban planning that ensures shorter distances to green and public spaces to promote use. Among mental health indicators, anxiety is the most influential node and should be prioritized in interventions. The study advances an integrative, network-based understanding of urbanization, loneliness, and mental and physical health in a large, representative metropolitan sample.

The cross-sectional design precludes causal inference; longitudinal studies are needed. All measures were self-administered, raising potential for response biases. Common method bias testing indicated potential bias in neighborhood cohesion items, warranting cautious interpretation. The study was conducted in a highly urbanized Polish region (Metropolis GZM), which may limit generalizability and lacks broader cultural context. The urban–rural categorization in the network may be too coarse to capture nuanced density effects on mental health.