Risks and Benefits of Green Spaces for Children: A Cross-Sectional Study of Associations with Sedentary Behavior, Obesity, Asthma, and Allergy

1Exposure to greenness has been associated with both health benefits and risks. An emerging body of evidence has linked exposure to greenness with improving both perceived and objective physical and mental health and well-being, probably through increasing physical activity, reducing psychophysiological stress, enhancing social contacts, and reducing noise, heat, and air pollution levels. In children, for example, exposure to greenness has been associated with reduced sedentary behavior and obesity, which is of public health importance, considering the emerging pandemic of childhood obesity worldwide. However, evidence of a beneficial influence of exposure to greenness on physical activity and obesity has not been consistently reported by all studies. Similarly, the available literature on the association of exposure to greenness with allergy and asthma in children is inconsistent. Although some studies have reported positive associations between greenness and allergic conditions or exacerbation of asthma in children in relation to greeness, others have reported no associations or even evidence of protective effects.

2These inconsistencies might reflect variation in health influences related to differences in the timing of exposure (e.g., early-life exposure vs. exposure later in life) or the type of greenness (e.g., parks, forests, and residential surrounding greenness) evaluated by different studies. They also might reflect potential conflicting effects of greenness on these health outcomes. For example, greenness can increase the risk of allergic conditions and asthma through production of pollen and fungal spores or through exposure to pesticides/fertilizers. On the other hand, green spaces may help reduce the prevalence of these conditions by reducing sedentary behaviors and obesity, and improving air quality. Green spaces can also enhance the biodiversity of the living environment at both macrobiota and microbiota scales, which in turn has been inversely associated with dysfunctions of the immune system such as atopy. Few available studies have simultaneously investigated these conflicting influences, particularly in association with different types of greenness. Such studies are needed to provide a more holistic view of the potential harms and benefits of exposure to greenness and help health professionals and policy makers better incorporate the research evidence into recommendations, targeted interventions, policies, and urban planning.

3The overarching aim of the present study was to simultaneously investigate the potential health benefits and risks associated with exposure to different types of greenness in children. Toward this aim, we evaluated associations of residential surrounding greenness and proximity to green spaces (separately for parks and forests) and indicators of sedentary behavior, obesity, asthma, and allergy in schoolchildren.

4This study was carried out in the context of Positive Health Effects of the Natural Outdoor Environment in Typical Populations of Different Regions in Europe (PHENOTYPE). Our analysis was based on a population-based sample of schoolchildren in the 4th to 6th year of primary schools (9–12 years old) across Sabadell, Spain, in 2006. Sabadell is a city with about 200,000 inhabitants located 20 km northwest of Barcelona in the Catalonia autonomous region, Spain. It has a Mediterranean climate characterized by hot and dry summers, mild winters, and maximum precipitation and vegetation during autumn and spring.

5To recruit study participants, we contacted 58 primary schools, of which 53 agreed to participate in the study. Data collection was conducted during June 2006. Questionnaires and informed consent forms were provided to the parents of the children. The overall response rate was 58% (3,322 questionnaires).

6The study was approved by the ethics committee of the research center, and written informed consent was provided by all participants.

7Parents completed written questionnaires concerning sociodemographic characteristics, other potential covariates, and the outcomes evaluated for this study:

8Respiratory outcome. Respiratory health status of children was assessed using the validated International Study of Asthma and Allergies in Childhood (ISAAC) questionnaire translated into Spanish and Catalan. We asked about “asthma ever” (“Has your child ever had asthma?” followed by the question “Has it been diagnosed by a physician?”), asthma medication in the preceding 12 months and wheezing in the preceding 12 months. “Current asthma” was defined as asthma ever plus having had wheezing or having used asthma medication in the preceding 12 months. We used current asthma as our respiratory outcome.

9Allergic outcome. We applied the ISAAC questionnaire (“In the past 12 months, has your child had a problem with sneezing, or a runny, or blocked nose accompanied by itchy-watery eyes when she/he did not have a cold or the flu?”) to characterize current allergic rhinoconjunctivitis used as our allergic outcome.

10Sedentary behavior. Noneducational screen time has been used as an indicator of sedentary behavior in schoolchildren. In our study, parents were asked to report the average noneducational screen time children spent in front of television, computer, and/or video game console separately during working days and weekends. According to the American Academy of Pediatrics recommendation of limiting noneducational screen time to 1–2 hr/day, we constructed a binary (yes/no) variable (hereafter referred to as “excessive screen time”) indicating whether the child spent > 1 hr during each working day and > 2 hr during each weekend day on watching television, playing video games, and/or working with computer. We used excessive screen time as a surrogate of sedentary behavior.

11Body anthropometry. We calculated body mass index (BMI) as weight/height squared. We used World Health Organization reference values to transform BMI into z-scores according to age and sex. To determine the overweight/obese status, we applied age- and sex-specific BMI cut points recommended by the Childhood Obesity Working Group of the International Obesity Taskforce to define overweight and obesity. Because of the small number of obese children (n = 83), we pooled overweight and obese participants and constructed a binary variable (hereafter referred to as overweight/obesity) to indicate whether the participant was overweight or obese (yes) or not (no). We used BMI z-scores (continuous variable) and obesity/overweight (binary variable) as indicators of body anthropometry.

12We used two measures of exposure to greenness to address different aspects of such an exposure. We used residential surrounding greenness as a surrogate for general outdoor greenness of the living environment of study participants, and used residential proximity to green spaces as a surrogate for the access to green spaces. We used ESRI ArcGIS Desktop (version 10; ESRI, Redlands, CA, USA) to process the relevant maps.

13Residential surrounding greenness. To measure surrounding greenness, we used the Normalized Difference Vegetation Index (NDVI) derived from the Landsat 4–5 Thematic Mapper (TM) images at 30 m × 30 m resolution. NDVI is an indicator of greenness based on land surface reflectance of visible (red) and near-infrared parts of spectrum. It ranges between –1 and 1, with higher numbers indicating more greenness. To achieve maximum exposure contrast, we looked for available cloud-free Landsat TM images during springs/autumns (i.e., the maximum vegetation period of the year for our study region) of 2005–2007 (the relevant years to our study period) from the NASA’s Earth Observing System Data and Information System (EOSDIS) website (https://earthdata.nasa.gov/). On the basis of this search we generated our NDVI map using the image obtained on 18 May 2007 (see Supplemental Material, Figure S1).

14For each participant, surrounding greenness was abstracted as the average of NDVI in buffers of 100 m, 250 m, 500 m, and 1,000 m around her/his geocoded address of residence. The residential addresses at the time of interview were geocoded to the exact address according to the postal code, street name, and house number.

15Residential proximity to green spaces. To determine green spaces, we used the Urban Atlas map (2007) detailing land use and land cover for large urban areas (> 100,000 inhabitants) across Europe. The Urban Atlas, developed by the European Environment Agency, provides data separately on green urban areas (hereafter referred to as parks) and forests (see Supplemental Material, Figure S2). Based on this distinction, we constructed two binary variables (yes/no) indicating whether the child’s residential address was located within 300 m separately from a park or forest. The 300 m distance was selected according to the European Commission general recommendation for access to green space based on the concept of having a green space within a 15-min walk from home.

16Main analyses. We used separate logistic mixed-effects models with a random school effect to estimate associations for each exposure and dichotomous outcome, and used linear mixed-effects models to estimate associations between each exposure and BMI z-scores. Random intercepts were used to adjust for potential confounding by unmeasured school characteristics. All analyses were adjusted for indicators of individual level socioeconomic status (SES), including higher educational achievement by either parent (none or primary/secondary/university), and the type of school (public/private). In addition, models were adjusted for area-level SES using quintiles of the Urban Vulnerability index, a measure of neighborhood SES at the census tract level (median area of 0.08 km2 for the study area) based on 21 indicators of urban vulnerability grouped into four themes: sociodemographic vulnerability (5 indicators), socioeconomic vulnerability (6 indicators), housing vulnerability (5 indicators), and subjective perception of vulnerability (5 indicators). For this study we used the Urban Vulnerability index based on 2001 Spanish Census. The analyses for the respiratory and allergic outcomes were further adjusted for child’s sex and age, exposure to environmental tobacco smoke at home (yes/no), having older siblings (yes/no), and parental history of asthma (none/either/both). Having an older sibling has been associated with reduced risk of asthma and rhinitis. We adjusted the analyses of overweight/obesity for sport activity at school or sport facilities (> 6 hr/week, yes/no), and having siblings (yes/no). The analyses for sedentary behavior were controlled for child’s sex and age and having siblings. To facilitate the comparison of estimated associations for the surrounding greenness across different buffer sizes, we reported the results for 1 interquartile range (IQR) increase in average NDVI in each buffer size based on all study population. Accordingly, for each subject we divided the average NDVI value in each buffer by the IQR of average NDVI values for all population in that buffer (0.076 for 100-m buffer, 0.105 for 250-m buffer, 0.120 for 500-m buffer, and 0.097 for 1,000-m buffer). Stata statistical software (version 12; StataCorp, College Station, TX, USA) was used to carry out the analyses.

17Sensitivity analyses. We conducted a variety of sensitivity analyses for each outcome. For the respiratory outcome, we ran additional models with further adjustment for a history of eczema (yes/no) and for breastfeeding history (any/none), and also estimated associations with cases limited to those confirmed by a physician. For the allergic outcome, we also evaluated further adjustment for a history of eczema and for breastfeeding history. For sedentary behavior we evaluated other definitions of excessive screen time (e.g., noneducational screen time > 1 hr/day for both working days and weekends); did analyses excluding participants not able to do physical exercise due to allergy or respiratory health problems (n = 3), other health problems (n = 2), and other reasons (n = 30); and evaluated effect modification by sex. For overweight/obesity, we ran additional models with further adjustment for attending sport classes at school, and models that excluded participants unable to do physical exercise (as defined above).

18Further analyses. Network buffers. For the main analyses we abstracted residential surrounding greenness using NDVI averages in a number of circular buffers around the home address of study participants. Circular buffers could address a number of our hypothesized mechanisms such as stress restoration through visual access to greenness, reduction in noise and air pollution, and dispersion of pollens; however, for physical activity, network buffers could be more relevant. To explore this, we developed network buffers of 250 m and 500 m around the home address of each study participant and averaged NDVI over these buffers to abstract a new set of residential surrounding greenness. We repeated the main analyses using this alternative set of exposures.

19Effect modification by SES. We tested the statistical significance of multiplicative interactions between measures of green exposure (residential surrounding greenness in a 250-m buffer, residential proximity to a park, and residential proximity to a forest) and measures of SES (parental education, type of school, and quartiles of the Urban Vulnerability index) in relation with outcomes using likelihood ratio tests comparing models with and without interaction terms.

20For all aforementioned analyses, the statistical significance was considered at p < 0.05.

21We were able to geocode home addresses for 3,178 (95.7%) of 3,322 participants who returned questionnaires. Those participants whose home addresses were not geocoded (n = 144) were not statistically significantly different from the rest of participants with regard to the outcomes and covariates (data not shown). Descriptive statistics for the characteristics of study participants and the prevalence of our investigated outcomes are presented in Table 1. Maps of population density and Urban Vulnerability index values at census tracts across Sabadell are presented in Supplemental Material, Figures S3 and S4, respectively.

22Of the 225 children who reported ever having asthma, 209 (92.9%) were diagnosed by physician. Excessive screen time was positively associated with BMI z-score [unadjusted regression coefficient of 0.14; 95% confidence interval (CI): 0.05, 0.22], and the relative prevalence of overweight/obesity [unadjusted odds ratio (OR) = 1.32; 95% CI: 1.09, 1.59] and current asthma (unadjusted OR = 1.39; 95% CI: 0.98, 1.98). Furthermore, each 1-unit increase in BMI z-score was positively associated with the relative prevalence of current asthma (unadjusted OR = 1.40; 95% CI: 1.15, 1.70) and current allergic rhinoconjunctivitis (unadjusted OR = 1.17; 95% CI: 1.06, 1.28).

23The median of average NDVI in larger buffers around participants’ home addresses was larger than the median for smaller buffers. The medians (IQRs) of average NDVI values across buffers of 100 m, 250 m, 500 m, and 1,000 m were 0.038 (0.076), 0.061 (0.105), 0.095 (0.120), and 0.125 (0.097), respectively.

24With regard to residential proximity to green spaces, 1,399 (44.0%) and 342 (10.8%) participants lived within 300 m of a park or a forest, respectively. The median of average NDVI values in all buffer sizes was higher for participants living within 300 m of a park or forest compared to those living further away (see Supplemental Material, Table S1).

25Main analyses. Higher residential surrounding greenness in all buffer sizes was statistically significantly associated with lower relative prevalence of overweight/obesity and excessive screen time (Table 2). Consistently in the adjusted models, higher residential surrounding greenness was associated with lower BMI z-scores, which was statistically significant for surrounding greenness in 100-m buffer and nearly statistically significant for 250-m buffer (Table 2). On the other hand, the associations between residential surrounding greenness across 100-m and 250-m buffers and current asthma and current allergic rhinoconjunctivitis were nearly null (Table 2). For surrounding greenness in larger buffers of 500 m and 1,000 m, we observed slightly higher relative prevalences of current asthma and current allergic rhinoconjunctivitis, but associations were not statistically significant (Table 2).

26For residential proximity to a forest, we observed a similar pattern of associations to residential surrounding greenness in that it was associated with a lower relative prevalence of excessive screen time, nearly statistically significant lower relative prevalence of overweight/obesity, and non-statistically significant lower BMI z-scores, but it was not associated with current asthma (Table 3). There was a positive association between residential proximity to a forest and current allergic rhinoconjunctivitis which was not statistically significant (Table 3).

27Estimates from adjusted models were negative but not statistically significant for living near a park and associations with excessive screen time, overweight/obesity, and BMI z-scores (Table 3). On the other hand, living near a park was associated with an OR of 1.60 (95% CI: 1.09, 2.36) for current asthma and an OR of 1.10 (95% CI: 0.90, 1.35) for current allergic rhinoconjunctivitis (Table 3).

28Sensitivity analyses. The findings for current asthma stayed consistent after further adjustment of analyses for the history of eczema or breastfeeding or after limiting the cases to those confirmed by a physician (n = 131) (data not shown). Similarly, the findings for current allergic rhinoconjunctivitis were robust to additional adjustment of analyses for the history of eczema or breastfeeding (data not shown). Our observed associations for excessive screen time did not change notably after using other indicators of excessive screen time or excluding those participants not able to do physical exercise (data not shown). We did not observe any statistically significant interaction of sex and measures of exposure to greenness in association with excessive screen time (data not shown). Further adjustment of overweight/obesity analyses for attending sport classes at school or excluding the participants unable to do physical exercise did not result in a notable change in findings (data not shown).

29Further analyses. Network buffers. The Spearman’s correlation coefficient between NDVI averages in circular and network buffers was 0.87 for 250-m buffers and 0.89 for 500-m buffers. There was no notable difference between the findings for the residential surrounding greenness based on network buffers (see Supplemental Material, Table S2), and those of main analyses based on circular buffers in terms of direction and strength of the associations.

30Effect modification by SES. For the school type (i.e., private vs. public), none of the interaction terms were statistically significant (data not shown). For parental education, only the interaction with residential proximity to forests in relation with sedentary behavior was statistically significant (p = 0.02). After stratifying this analysis based on parental education we observed stronger associations for children whose parents had more education. The OR of excessive screen time associated with residential proximity to a forest was 0.91 (0.53, 1.58), 0.63 (0.41, 0.95), and 0.26 (0.11, 0.61), respectively, for children whose parents had primary education, secondary education, and university education, respectively. For Urban Vulnerability index, only the interaction with residential proximity to parks in association with overweight/obesity was statistically significant (p = 0.02). After stratifying this analysis based on quintiles of Urban Vulnerability index, we did not observe a clear trend in associations between residential proximity to parks and overweight/obesity across the strata.

31This study is one of the first to estimate simultaneously the potential benefits (i.e., reducing sedentary behavior and overweight/obesity) and harms (i.e., increasing risk of asthma and allergy) associated with exposure to greenness in children. To our knowledge, this study is also the first to separate and compare associations of residential proximity to natural and urban green spaces and asthma, allergy, and sedentary behavior. We observed that higher residential surrounding greenness was associated with lower BMI z-score and lower relative prevalence of overweight/obesity and excessive screen time, but was not associated with current asthma and current allergic rhinoconjunctivitis. Similarly, residential proximity to a forest was inversely associated with relative prevalence of excessive screen time and overweight/obesity (nearly statistically significant) but not with current asthma. Residential proximity to a forest was associated with an increased risk of current allergic rhinoconjunctivitis which was not statistically significant. On the other hand, living near a park was associated with higher relative prevalence of current asthma, while its associations with BMI z-score and relative prevalence of current allergic rhinoconjunctivitis, obesity/overweight, and excessive screen time were generally weak and did not attain statistical significance.

32We found negative associations of residential surrounding greenness and proximity to a forest with excessive screen time. These findings are in line with those of previous studies reporting higher physical activity associated with higher neighborhood greenness or better access to green spaces. Lower screen time has been associated with lower BMI and lower relative prevalence of overweight/obesity, which was supported by our data showing positive associations between excessive screen time and BMI z-score and overweight/obesity. Consistently, we found that lower BMI z-scores and lower relative prevalence of overweight/obesity were associated with residential surrounding greenness and residential proximity to forests—the same measures of exposure to greenness for which we also observed a negative association with excessive screen time. These findings are consistent with those of other studies reporting reduction in BMI z-score and overweight/obesity in children in association with higher neighborhood greenness (defined as NDVI average over a buffer of 1,000 m around residential addresses) or proximity to green spaces.

33The available evidence on the association of exposure to greenness and allergy and asthma in children is inconsistent. Similarly, we observed different patterns of association for current asthma and current allergic rhinoconjunctivitis in relation with our different measures of exposure to greenness. Although the associations between current asthma and residential surrounding greenness and proximity to a forest were almost null, there was a higher relative prevalence of current asthma associated with living close to a park. Similarly, current allergic rhinoconjunctivitis was not associated with residential surrounding greenness, but it was positively associated with residential proximity to a park or to a forest, although none of the associations attained statistical significance. Sedentary behavior including noneducational screen time and obesity have been associated with higher relative prevalence of asthma and allergic conditions in children. The negative association of residential surrounding greenness and proximity to forests with excessive screen time and overweight/obesity can therefore partly explain why these exposures were not associated with current asthma and current allergic rhinoconjunctivitis. Differences in estimated associations of asthma with residential proximity to forests (null) and parks (positive) could also be partly explained by a difference between forests and parks in terms of their flora and the allergenicity of their pollens. For example, compared to forests with their predominantly native flora, parks in Catalonia have mainly “exotic” species such as Platanus sp. that are anemophilous (wind-pollinated) trees and have been shown to cause pollen allergies among local people. Another example is Parietaria, a weed that thrives abundantly in urban areas of the Mediterranean costal regions. Parietaria pollen is known to have strong allergenicity particularly for people 10–30 years of age. Furthermore, in our setting, forests were mostly located on the edges of the city while parks were located mostly in the middle of the city (see Supplemental Material, Figure S2). This could be suggestive for some potential differences in the living environment (e.g., in terms of degree of urbanity or air pollution levels) experienced by those living next to a forest and those living next to a park that in turn could have contributed to differences in our evaluated outcomes. However, we did not observe any association between ambient levels of air pollutants (nitrogen dioxide and particulate matter ≤ 10 μm in diameter) at home address of participants and current asthma and allergic rhinoconjunctivitis, as described in Supplemental Material (see Supplemental Material, “Air pollution and asthma and allergy”).

34Our study faced some limitations. Our cross-sectional study had a limited capability for a causal inference regarding our evaluated associations. For our main analyses, we could not rule out the likelihood of self-selection bias: Families with higher SES may have been more likely to select living in greener neighborhoods, or families of children suffering from asthma/allergy may have been more likely to choose living away from green spaces. Likewise, generalizability of our findings could have been affected by selection bias in that those families who agreed to participate in the study might be different from those who did not participate with respect to a number of characteristics. We did not have data to evaluate such potential differences. However, there was no statistically significant (p = 0.87) difference between public and private schools in their response rates, which might suggest that SES was less likely to play a major role in participating in the study.

35We used remote sensing–derived NDVI to assess surrounding greenness. Application of this objective measure of greenness enabled us to take account of small-scale green spaces (e.g., home gardens, street trees, and green verges) in a standardized way; however, NDVI does not distinguish between different types of vegetation, which could be relevant to our investigated associations. By using an NDVI map obtained at a single point of time, we effectively assumed that the spatial distribution of NDVI across our study region remained constant over the study period (2006). The findings of our previous study across the same region supports the stability of the NDVI spatial contrast over seasons and years.

36Furthermore, we could not address the impact of the quality of green spaces in our analyses of residential proximity to green spaces. Quality characteristics of green spaces such as aesthetics, biodiversity, walkability, sport/play facilities, safety, and organized social events have been suggested to affect the use of green spaces for physical activity. Moreover, we did not have data on use of smartphones by our study participants, which could be relevant to our characterization of sedentary behavior. However, because the data were collected in 2006, we do not expect that a notable portion of our study sample used smartphones. Also, although the validated ISAAC questionnaire was used to identify children with allergic rhinoconjunctivitis, such identification by parents was prone to subjectivity (e.g., attributing common cold to allergic rhinoconjunctivitis or vice versa), and we did not have data on objective tests (e.g., pollen sensitization) to validate parental reports for this outcome. Additionally, we did not have data on severity, frequency, and duration of episodes of allergic rhinoconjunctivitis, which could be of importance in our analysis. Similarly, we did not have data on ethnicity, residential history, and built environment characteristics, such as access to fast food outlets or street connectivity, that could be relevant to our analyses. Moreover, although we adjusted our analyses for indicators of SES at both individual and area levels, we did not have data on other relevant SES indicators such as family income, and residual SES confounding could not be ruled out.

37We observed two separable patterns of associations between our evaluated health outcomes and exposure to different measures of greenness. Residential surrounding greenness was negatively associated with BMI z-scores, overweight/obesity, and sedentary behavior, but was not associated with current asthma or allergic rhinoconjunctivitis. Similarly, residential proximity to forests was negatively associated with sedentary behavior, overweight/obesity, and BMI z-scores, and was positively associated with current allergic rhinoconjunctivitis but not current asthma. On the other hand, residential proximity to parks was positively associated with current asthma and weakly associated with allergic rhinoconjunctivitis, but was not associated with significantly reduced sedentary behavior or lower overweight/obesity. These observations could indicate critical roles of the use of green spaces and their flora in determining the direction of the equilibrium between health benefits and risks of exposure to greenness; however, because we did not have data on these factors, we could not directly examine their roles. Furthermore, our observed lower relative prevalence of overweight/obesity and sedentary behavior associated with exposure to greenness, if confirmed by future studies, can be of public health importance, considering the current pandemic of childhood obesity and sedentary behavior worldwide and their notable burden due to the wide range of adverse health outcomes associated with these conditions. Further longitudinal studies are warranted to replicate our findings in other settings. We advise future studies to take account of use, quality, and type of plants in green spaces as well as seasonality in vegetation when evaluating health risks and benefits of contact with greenness.