Association between residential greenspace structures and frailty in a cohort of older Chinese adults

Frailty is a geriatric syndrome characterized by increased vulnerability due to cumulative declines across physical, psychological, and social domains, leading to higher risks of hospitalization, falls, depression, and mortality. With rapid urbanization and population aging in China, understanding environmental determinants of frailty is increasingly important. Prior studies largely used NDVI to quantify greenness and found associations with frailty-related outcomes. However, NDVI does not capture structural characteristics of greenspaces such as area, shape complexity, and connectivity that may influence health via physical activity, social interaction, psychological restoration, and reductions in environmental stressors. This study investigates whether specific greenspace structural characteristics—area-edge (LPI), shape complexity (shape index), and connectivity/proximity (patch cohesion index)—are associated with frailty in older Chinese adults using the CLHLS cohort. The hypothesis is that higher values of these structural indices are protective against frailty, and effects may vary across demographic and socioeconomic subgroups.

Epidemiologic evidence links higher residential greenness (often measured by NDVI) with better health outcomes and lower frailty risk, including studies from Hong Kong and China showing lower likelihood of frailty and potential mediation via physical activity, disease burden, and cognition. Mechanistic pathways include enhanced social cohesion, psychological restoration, increased physical activity, and reduced exposures to air pollution, noise, and heat. Emerging literature suggests greenspace structure matters: larger greenspaces and more complex shapes influence thermal regulation and microclimate and may impact mental health and cause-specific mortality. Larger area/edge has been associated with less depression and stress; complex shapes with higher preference and satisfaction; and connected greenspaces with reduced air pollution and improved respiratory and cardiovascular outcomes. Despite these insights, associations between specific greenspace structures and frailty remain underexplored, motivating the current study.

Study design and population: Data were drawn from the China Longitudinal Healthy Longevity Survey (CLHLS), a nationally representative cohort of older adults across 22 provinces (approx. 85% of China’s population). The analytic period was 2008–2014, including baseline interviews (2008/09) and follow-ups (2012, 2014). From 16,072 individuals, exclusions were applied for missing demographic characteristics (N=3,073), missing frailty index (N=4,215), and missing NDVI (N=8) at baseline, yielding 8,776 participants. Exposure assessment—Greenness structures: Following landscape ecology conventions, three non-redundant structural characteristics were selected to capture independent aspects of greenspace: area-edge, shape, and proximity/connectedness. For each characteristic, one representative index was computed at county level using FRAGSTATS 4.2 from Advanced Land Observing Satellite (ALOS) data with 100 m × 100 m grid resolution: (1) Largest Patch Index (LPI; area-edge; % of landscape in the largest patch); (2) Mean Shape Index (shape; shape complexity relative to circle/square of same area); (3) Patch Cohesion Index (cohesion; physical connectedness/aggregation of the patch type). Spearman correlations verified low collinearity (criterion |rho|<0.72). In-study ranges and means: LPI 0.01–53.7 (mean 8.6), shape 1.0–41.2 (mean 8.5), cohesion 84.9–100.0 (mean 97.8). Higher values indicate larger area, more complex shapes, and greater connectedness, respectively. Exposure assessment—Greenness quantity (NDVI): Contemporaneous NDVI around each residential address was derived from MODIS, removing negative values (water). NDVI was calculated at death date for decedents or last interview date for others. NDVI showed no strong correlation with structure indices and served as a complementary measure of overall vegetative density. Outcome—Frailty: Frailty Index (FI) was computed as the proportion of 39 self-reported health deficits present, covering functional status, cognitive function, self-reported and interviewer-rated health, mental health, sensory abilities, heart rhythm, and chronic diseases. Items scored 0/1 except the number of serious illnesses (0/1/2). FI ranges 0–1; higher indicates greater frailty. FI was analyzed as continuous and dichotomized at 0.21 (non-frail ≤0.21; frail >0.21). Change in FI from baseline to last follow-up was categorized as no change/decrease vs increase. Covariates: Baseline covariates included study entrant year; age group (65–79, 80–89, 90–99, 100+); sex; literacy; annual household income; financial independence (independent vs dependent); BMI; marital status (married and living with spouse vs other); smoking (current, former, non-smoker); alcohol consumption (current, former, non-drinker); physical activity (current, former, never); residential location (city, town, rural); geographic region (7 macroregions); and 3-year average PM2.5 (2006–2008) at 0.1° resolution from a satellite-based model. Statistical analysis: Primary hypothesis tested whether higher LPI, shape, and cohesion are associated with lower frailty, with heterogeneity across subgroups (sex, age, literacy, residence, marital status, FI change). Cross-sectional analyses at baseline used multivariable linear regression for continuous FI and logistic regression for frailty status, adjusting for covariates. To address nonlinearity, NDVI and greenness structure indices were categorized into quartiles (Q1 reference). Longitudinal analyses among followed participants used GEE models (both continuous and binary FI) with greenness structure indices. NDVI (quartiles) was additionally analyzed to reconfirm associations with overall greenness. Sensitivity analyses and subgroup analyses: Alternative indices were tested to assess robustness: Edge Density (ED, area-edge), Area-weighted Mean Fractal Dimension (FRAC, shape), and Percentage of Like Adjacencies (PLADJ, proximity). Subgroup analyses stratified by sex, age, literacy, urban/rural residence, marital status, PM2.5, and FI change. Software and data: Analyses conducted in R 3.6.3. FRAGSTATS 4.2 used for landscape metrics. CLHLS data from Peking University Open Research Data; greenspace data from ALOS; GEE package from CRAN; code available on GitHub.

- Sample: 8,776 baseline participants; mean baseline indices: LPI 7.93, shape 8.11, cohesion 97.6; mean FI 0.17. Nearly half were men (47.1%), 60.9% lived in rural regions, and 37.6% were married and living with a spouse. - Cross-sectional associations: Higher NDVI associated with lower frailty (e.g., linear coefficient for FI = -0.014, 95% CI: -0.022 to -0.007, P<0.001; logistic OR 0.794, 95% CI: 0.674–0.937 in Q3 vs Q1). Greenness structure indices showed significant inverse dose–response relationships with FI. In adjusted linear models, within the fourth quartile, each 0.1-unit increase in LPI, shape, and cohesion was associated with 0.026 (95% CI: -0.033 to -0.019), 0.028 (95% CI: -0.035 to -0.021), and 0.025 (95% CI: -0.032 to -0.018) lower FI scores, respectively. In adjusted logistic models, participants in the highest quartile (Q4 vs Q1) had lower odds of frailty: LPI OR 0.676 (95% CI: 0.579–0.789, P<0.001), shape OR 0.650 (95% CI: 0.556–0.760, P<0.001), cohesion OR 0.635 (95% CI: 0.541–0.744, P<0.001), corresponding to 32%, 35%, and 37% lower odds, respectively. - Longitudinal associations: No significant associations were found between greenness structure indices and frailty over follow-up in GEE models. - Subgroups: Stronger cross-sectional protective associations were observed in females, centenarians (100+), illiterate individuals, city residents, unmarried individuals, and participants with increased frailty over time. Those with increased frailty appeared to benefit more from higher cohesion. - Sensitivity analyses: Results were consistent when using alternative structural indices (ED, FRAC, PLADJ), supporting robustness of findings.

The study addressed whether greenspace structural characteristics are associated with frailty in older adults. Cross-sectional analyses demonstrated inverse, dose–response relationships between larger greenspace areas (LPI), more complex shapes, and greater connectivity (cohesion) with frailty, suggesting that structural qualities of greenspaces may confer health benefits beyond overall greenness measured by NDVI. Potential mechanisms include increased opportunities for physical activity afforded by larger and better-connected green areas, enhanced social cohesion and psychological restoration associated with accessible and diverse green environments, and reductions in environmental stressors such as air pollution and heat, to which greenspace structure may contribute by minimizing fragmentation and improving ecological connectivity. Subgroup findings imply that vulnerable populations—women, the oldest-old, those with lower literacy, unmarried individuals, and urban residents—may be particularly sensitive to greenspace structural benefits, potentially due to baseline differences in frailty, environmental exposures, or access to supportive resources. However, longitudinal analyses did not confirm these associations, potentially due to substantial attrition (deaths and loss to follow-up), selection of healthier survivors, time-varying confounding, or suboptimal modeling of temporal dynamics. Overall, the results support the relevance of structural greenness metrics in understanding environmental influences on frailty and inform urban planning for healthy aging, while emphasizing the need for longitudinal designs that can better ascertain causality.

Greenspace structural characteristics—greater area-edge (LPI), more complex shapes, and higher connectivity—were cross-sectionally associated with lower frailty among older Chinese adults, complementing evidence based on NDVI. These findings suggest that urban planning emphasizing larger, well-connected, and structurally diverse greenspaces may help mitigate frailty, particularly among vulnerable subgroups. However, longitudinal analyses did not show significant associations, so causal inference remains uncertain. Future research should incorporate time-varying exposure measures across multiple periods, address attrition and selection bias, include detailed data on vegetation types and individual greenspace use, model potential mediators and confounders (e.g., neighborhood safety, social networks), and examine combined effects of NDVI and structural metrics to clarify mechanisms and inform targeted interventions.

- High attrition over follow-up: 5,921 individuals lacked 2014 FI due to death or loss to follow-up, potentially biasing longitudinal analyses toward healthier survivors and reducing power. - Single time-point exposure for structural indices: Greenspace structure was assessed only in 2008, limiting the ability to capture changes over time. - Limited exposure detail: No information on specific vegetation types, individual time spent in greenspaces, or activity patterns could be obtained from satellite data. - Potential unmeasured confounding and mediators: Factors such as neighborhood safety, social networks, and time-varying socioeconomic conditions were not fully accounted for. - Cross-sectional findings susceptible to bias, including residential self-selection (less frail individuals choosing greener areas) and model specification limitations in longitudinal analysis.