Greenspace exposure and children behavior: A systematic review

Evidence on the association of greenspace exposure with child behavior is accumulating


Introduction
The ongoing urbanization has led to an increase in the number of children being born and raised in urban areas (Nations, 2014).Urban living is often associated with a stressful and sedentary lifestyle, increased exposure to urban-related environmental hazards such as air pollution, noise, and heat, and limited access or exposure to natural environments (Vlahov and Galea, 2002;Gregg et al., 2003).Natural environments, including greenspace (has been defined by the U.S. Environmental Protection Agency as "land that is partly or completely covered with grass, trees, shrubs, or other vegetation" (US, 2021)), have been suggested to improve mental and physical health and wellbeing (Barton and Rogerson, 2017;Roberts et al., 2019).Exposure to greenspace has been suggested to be associated with improved overall mental and physical health of children, including behavioral outcomes (McCormick, 2017;Nicole, 2018;Parmes et al., 2020).The ability of greenspace to exert such benefits has been suggested to be through socio-behavioral and environmental pathways (Markevych et al., 2017).Socio-behavioral pathways encompass an increase in physical activity, social cohesion, and attention restoration and a decrease in stress (Markevych et al., 2017).Additionally, greenspace could exert its benefits through environmental pathways, including mitigating the exposure to urban-related environmental hazards such as air pollution, noise, and heat and enriching microbiota diversity (Selway et al., 2020;Nieuwenhuijsen, 2021).As a result, an increasing number of studies have evaluated the association of greenspace exposure with neurodevelopment, including behavioral outcomes in children (McCormick, 2017).
Behavioral problems are common in children.For example, it has been estimated that one out of seven children aged two to eight years in the United States suffers from different types of behavioral, mental, or developmental problems (Data and Statistics on Children's Mental Health[cited, 2021).Evidence that children spend less time in nature than previous generations (Clements, 2004) has prompted researchers to explore the connection between exposure to greenspace and the global rise in the prevalence of mental and behavioral problems (Putra et al., 2020).Previous studies assessing this association were mainly of experimental design looking at the "therapeutic effects" of short-term contact with greenspace on neurodevelopmental impairments such as attention deficit-hyperactivity disorders (ADHD) in children who were affected by these conditions (Faber Taylor and Kuo, 2011;Taylor et al., 2001).More recently, a growing number of epidemiological studies have investigated the association of long-term exposure to greenspace with neurodevelopment, including behavioral development, in children from the general population (McCormick, 2017;Putra et al., 2020).Higher levels of surrounding greenspace have been reported to be associated with better behavioral development (Liao et al., 2020).Inverse associations have been reported between higher residential greenspace and ADHD symptoms and incidence (Amoly et al., 2014;Balseviciene et al., 2014).Increasing exposure to nature through adding greenspace to schoolyards promoted the social behaviors of children (Raney et al., 2019).A major part of the evidence on greenspace and children mental health suggests a beneficial role of greenspace exposure on children's behavioral difficulties (Vanaken and Danckaerts, 2018).However, some other studies did not find such association between neighborhood greenspace and outcomes such as selfregulation in children (Mueller and Flouri, 2020).
In general, the evidence on the association between exposure to greenspace and children's behavior is increasing rapidly and in some cases is equivocal.Inconsistencies might arise from how green space exposure is measured, outcome assessment, and other methodological issues such as study design and statistical methods (Putra et al., 2021).We aimed to systematically review the existing evidence on the association between long-term exposure to greenspace and behavioral problems in children.We also aimed to explore the methodological issues in the studies in terms of exposure and outcome assessment to better understand the sources of heterogeneities, to propose further research insights.

Methods
The review was conducted according to the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) statement (Page, 2021).We developed a systematic review protocol before the initiation of the review, and all the review team members adhered to the protocol that was set at the beginning of the review.

Inclusion and exclusion criteria
We included original articles written in English and reporting on observational studies with quantified estimates of the association between long-term exposure to greenspace and at least one of the behavioral outcomes in the general population until the age of 18 years (the age group in this study was all children as defined those up to 18 years old) (Table S1).Studies were excluded if the exposure variable was not quantified (e.g.binary variables such as presence or absence of a park without further information about the distance or satisfaction with neighborhood greenspace [yes/no] without a rating on a scale), or the study population only consisted of patients or non-healthy populations (e.g.children with autism at baseline).We also excluded review articles, semi-or natural experimental studies, and qualitative studies.

Information sources
We searched three databases to find relevant articles: PubMed (National Library of Medicine), Scopus, and Web of Science (including WOS, KJD, RSCI, SCIELO).The searches were conducted on November 29, 2021.The content of search strategy used for PubMed, Scopus, and Web of Science databases are provided in Supplementary tables S2-S4 respectively.In addition to the databases search, we also conducted a manual search of the references lists of the retrieved original studies or review articles (Putra et al., 2020;Vanaken and Danckaerts, 2018;Islam et al., 2020) to identify additional studies.The identified studies in reference checking step underwent abstract and full-text evaluation for the assessment of the eligibility of inclusion in the review.The overview chart of the search strategy is presented in Fig. 1.

Search strategy
To build a more comprehensive search queries, and to gain a broader view on the keywords in the context, and capture articles within the scope of the review, at first we searched relevant systematic reviews on the greenspace and natural environment (as the exposure of interest), and different childhood behavioral problems (as the outcome of interest) (Putra et al., 2020;Vanaken and Danckaerts, 2018;Islam et al., 2020).We then made the search queries based on the PECO (Population, Exposure, Comparison, and Outcome) framework (Morgan et al., 2018), by combining the keywords on the exposure (e.g.green space(s), greenness, normalized difference vegetation index (NDVI)); outcome (e.g.neurodevelopment(al), motor development, behavio(u)ral problems, hyperactivity, Attention Deficit Hyperactivity Disorder (ADHD), Strengths and Difficulties Questionnaire (SDQ)); and the study population (e.g.child, children, adolescent(s), adolescence, teenager(s), teen, childhood, early life, prenatal, and postnatal) (Table S1).

Study selection
The study selection process was conducted in the Rayyan online platform (Ouzzani et al., 2016).After duplicate removal, the title and abstract of the articles were screened to exclude the articles that did not meet the review criteria.Afterwards, the full texts of the remaining articles were examined by two teams of reviewers (CK, MB, PK, and LMB) independently to decide which articles met the inclusion criteria and were to be included in the review.In the case of disagreements, another reviewer (PD) checked the paper for inclusion or exclusion in the review.

Data extraction and data items
After completion of the selection process, relevant data on study characteristics (design, location, time of the study), study population (number, age, and sex of participants), exposures (type of exposure, source of exposure data, exposure assessment method, exposure allocation method), outcomes (the type of outcome, method of outcome measurements, outcome data sources), statistical methods (statistical models, level of adjustments, covariates used in the models), and results (main finding, effect sizes) were extracted.We checked the study names (if available) and population in the selected papers to avoid the overlap of the study population in the papers.

Study risk of bias assessment
The risk of bias (ROB) of the included studies was assessed by the Newcastle-Ottawa Scale (NOS), as recommended in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins et al., 2019).The NOS includes three domains: (i) selection (four items), (ii) comparability (one item), and (iii) ascertainment of exposure/outcome (three items) (Wells et al., 2000).For cohorts and case-control studies, all items except the comparability can earn one scoring star (the comparability item can earn a maximum of two stars).The sum of the earned stars (maximum of nine) is the ROB score of each study.As NOS has not originally been developed for environmental epidemiology studies, we slightly modified the NOS items (especially for the exposure ascertainment domain) to be applicable in the field of environmental epidemiology (Sakhvidi et al., 2020).For greenspace exposure assessment we used "the location of exposure allocation" as a selected aspect for quality assessment and considered exposure assessment via geocoding at the residential location as the highest exposure assessment quality.Table S5 presents the detailed description of each item in the risk ROB assessment of each study design.For cross-sectional studies, we used a modified version of the NOS for cohort studies (Sakhvidi et al., 2020).Different cut-offs have been reported to categorize the studies based on the NOS score.While the definition of cut-offs is not stated in the original NOS guideline, we defined five categories of risk of bias according to the NOS score which also is reported in other studies (including: "very good " for NOS score ≥8, "good "for NOS score six or seven; "fair "for NOS score four or five, and "poor "for NOS score <4) (Sakhvidi et al., 2020).The ROB evaluation was performed by two reviewers (M.Z and A.M), and in the case of disagreement, the third reviewer (P.D) resolved the disagreement.

Effect measures
We extracted the reported associations (e.g.relative risk (RR), hazard ratio (HR), odds ratio (OR), incidence rate ratio (IRR), standardized incidence ratio (SIR), or correlation coefficient) and their corresponding confidence intervals as the acceptable effect sizes in this study.In the case of reporting crude and adjusted associations for specific exposure-outcome pairs, we preferred the most complete adjusted one which did not include the mediators in the model.

Synthesis methods
Because of the small number of studies, heterogeneity in outcome assessment and reporting, and also using different exposure measures which were not feasible to convert to each other, we were unable to perform a meta-analysis.Instead, we reported the proportion of significant protective/detrimental, suggestive for protective/ detrimental association and null associations for each outcome for the description of the findings.

Table 3
Risk of bias assessment for the selected case-control studies in the review.For details on each item description please see Table S5.
grasslands (Maes et al., 2021).One study used an audit, reporting the results based on the percentage of green land cover within an approximate half-mile distance from the participants' homes or schools (Meidenbauer et al., 2019).
Three studies found a protective association between greenspace exposure and ADHD diagnosis (Amoly et al., 2014;Baumgardner et al., 2010;Markevych et al., 2018).In one study, all the associations were null (Reyes et al., 2013), and in the rest of the studies, the findings were different depending on the exposure definition, mostly inclined toward the protective role of this exposure.

Externalizing and internalizing disorders
Six studies (Bijnens et al., 2020;Feng and Astell-Burt, 2017a;Feng and Astell-Burt, 2017b;Lee et al., 2019;McEachan et al., 2018;Jimenez et al., 2021), including two longitudinal (Feng and Astell-Burt, 2017a;Jimenez et al., 2021), and four cross-sectional studies (Bijnens et al., 2020;Feng and Astell-Burt, 2017b;Lee et al., 2019;McEachan et al., 2018) reported the association between exposure to greenspace and internalizing and/or externalizing problems.Four of these studies used SDQ (Feng and Astell-Burt, 2017a;Feng and Astell-Burt, 2017b;McEachan et al., 2018;Jimenez et al., 2021) and two applied CBCL (Bijnens et al., 2020;Lee et al., 2019) to characterize the internalizing and/or externalizing problems.Three studies reported the results for percent of land cover (Bijnens et al., 2020;Feng and Astell-Burt, 2017a;Feng and Astell-Burt, 2017b;Lee et al., 2019;McEachan et al., 2018;Jimenez et al., 2021), and the remaining two reported for surrounding greenspace based on NDVI (McEachan et al., 2018;Jimenez et al., 2021) and MSAVI (Lee et al., 2019).Two of the studies had fair quality (Feng and Astell-Burt, 2017b;McEachan et al., 2018) and the four remaining had good quality (Bijnens et al., 2020;Feng and Astell-Burt, 2017a;Lee et al., 2019;Jimenez et al., 2021).All of the 12 extracted associations were suggestive of the beneficial role of greenspace on externalizing disorders (three of the associations were significant).Similarly, for internalizing disorders most of the reported associations (9 out of 12 reported associations) were suggestive of a protective role of greenspace exposure.In three studies also at least one significant protective association was found between greenspace exposure and internalizing disorders.Two associations were suggestive for the negative role of greenspace on externalizing disorders, none of them were significant.

Discussion
We conducted a systematic review of the available evidence on the association of greenspace exposure (only in the studies that greenspace was reported based on the quantified estimates) with behavioral development in children.Most of the available studies were conducted from 2015 onwards, were carried out in Europe, and had a cross-sectional design.In general, most of the available studies had fair and good quality.Nine different behavioral outcomes (mostly total difficulties measured by SDQ) were reported across the studies.Overall, the reported findings were suggestive of the potential association of the greenspace exposure with reduced risk of behavioral problems.However, diversity of exposure-outcome assessment and reporting prevented us from doing a meta-analysis.

Exposure assessment
Exposure to the greenspace in the reviewed studies was characterized based on different approaches and performance (in terms of temporal and spatial resolution), mostly at residential address locations.Few studies (Liao et al., 2020;Yang et al., 2019) reported associations for school or kindergarten greenspace or the mixing approach (constructing the mixed weighted index by combining home and school greenspace).Part of the heterogeneity in the findings of the studies could be due to differences in the exposure assessment and allocation approaches (for example, residential versus school).Additionally, most of the studies did not consider longterm exposure to greenspace.For example, exposure assessment focused on specific periods such as school-age could consequently overlook other potential windows of susceptibility such as prenatal and preschool periods, which could be relevant to behavioral development (Richardson et al., 2017;Dadvand et al., 2015).Exposure to greenspace during pregnancy and early years of age has been considered very scarcely.Only three studies in this review included the exposure during the pregnancy period or from birth in their exposure assessments (Bijnens et al., 2020;Donovan et al., 2019;Richardson et al., 2017).Accordingly, it is difficult to draw firm conclusions on the relevance of these potentially critical windows of exposure.As the plasticity of the child's brain is considerably higher in the first years of life, any environmental exposure, including greenspace, could induce changes in the developing brain.Richardson et al. (Richardson et al., 2017) findings suggest that any beneficial influences are more likely to occur at younger ages, whereas Donovan et al. (Donovan et al., 2019) findings imply a non-significant association between exposure during the early life (prenatal period to age two years) greenspace exposure and ADHD.As the time window of exposure to greenspace could be a determining factor in the observed associations, future longitudinal studies with repeated measurement of exposure from early life (pregnancy period and first years of life) by considering both long-term and short-term exposure to greenspace could shed light on possible different relevant windows of susceptibility to exposure and their role in the observed heterogeneities.
Most of the studies in this review used surrounding greenness and proximity to green spaces to measure exposure.Greenspace use and quality were merely considered in three reviewed studies (Amoly et al., 2014;Thygesen et al., 2020;Andrusaityte et al., 2020).The quality and type of greenspace can modify the observed health benefits of greenspace (McEachan et al., 2018;Richardson et al., 2017;Van Aart et al., 2018).Selection of appropriate dataset is necessary to achieve the best performance.The effectiveness of different greenspace measurement methods may vary according to the different types of greenspace.Additionally, the accuracy and effectiveness of greenspace dataset in estimating the exposure could depend on different factors including the data sources, spatial and temporal resolution, and classes of greenspace.Therefore, the most appropriate assessment method and dataset may vary in different datasets, different study areas and different outcomes (Liao et al., 2021).Associations between greenspace and health outcomes are also suggested to depend on the greenspace measurement method (Zhou et al., 2021).Surrounding greenspace has been suggested to be measured more accurately by NDVI, and land use categories may be less suitable for health outcomes (Zhou et al., 2021).
Reporting the results for different surrounding greenspace measures (e.g.NDVI, MSAVI, etc.), for different types of green space instead of merely relying on the abundance of vegetation regardless of their types, and taking account of quality characteristics of the green spaces in addition to their availability can increase the breadth of our understanding about the possible associations and underlying mechanisms in future studies (Knobel et al., 2021).

Outcome assessment
The majority of the included studies applied questionnaires/tools that were filled by parents and hence were prone to subjectivity.Applying objective measures for characterization of the outcomes including using computerized tests and neuroimaging techniques, implementing the assessment tools by professionals (e.g.pediatric psychologists or psychiatrists), or relying on more than one source (e.g.child, parents, and/or teachers) for questionnaires can decrease the likelihood of the outcome misclassification in future studies.Using different outcome assessment methods (self-administrated questionnaire, direct interview, or diagnosis/ evaluation by healthcare professionals) can have different validity and reliability in characterizing the behavioral problems and introduce heterogeneity in the findings.However, at the same time, these different tools can provide information on different aspects of the behavioral problems and, as such, can complement each other in providing a robust assessment of the outcomes.

Effect modification, confounding, and mediation
The effect of the natural environment on health could be moderated by the factors such as socioeconomic status (SES) (Balseviciene et al., 2014) (e.g.parental education (Markevych et al., 2014)), sex (Markevych et al., 2014), ethnicity (McEachan et al., 2018), and the degree of urbanity, among others.In addition to demographical characteristics such as ethnicity, sex, and age, future studies also need to control their analyses for family history of behavioral problems and wider SES variables.In addition, to be a potential effect modifier for the association between the greenspace exposure and behavior, SES can also be an influential confounder for this association.As such, studies evaluating this association need to properly control their analyses for SES indicators at both household (e.g., parental education, income, employment status) and neighborhood (e.g., census-based indices of deprivation) levels to minimize the likelihood of the residual SES confounding.Our included articles mainly controlled the analyses for indicators of household SES without taking neighborhood SES into account, which remains an area for improvement in future studies.Most of the studies have not explained the mechanisms through the mediator's analyses.We recommend future studies to formally analyze the mediation role of air pollution, physical activity, obesity, and environmental microbiota in the association between greenspace exposure and behavior problems.

Underlying mechanisms
The mechanisms underlying health benefits of greenspace are yet to be established.Additionally it is not clear to what extent the available proposed mechanisms for the health benefits of greenspace could be applicable for each of the behavioral outcomes.In general, reducing stress, restoring attention, enhancing social cohesion, increasing physical activity, enriching microbial input from the environment, and mitigating exposure to urban-related environmental hazards such as air pollution, noise, and heat have been suggested to be involved in beneficial role of greenspace on health (Markevych et al., 2017;Dadvand and Nieuwenhuijsen, 2019;Mueller et al., 2020).According to the stress reduction theory, natural environments can cause an overall sense of emotional wellbeing and calming effect by regulating human emotional responses to the environment and lowering neurophysiological stress.Moreover "attention restoration theory" describes the positive effect of exposure to the natural environment through redirecting attention away from attentional tasks that are necessary for daily urban life (Lei, 2018).Supportive environments can also alleviate the effects of fatigue and restore mental acuity.Furthermore, while exposure to air pollution and noise has been associated with an increased risk of behavioral problems (Forns et al., 2016), greenspace could mitigate such exposures (Dadvand et al., 2012;Dadvand et al., 2018).Further to the aforementioned, indirect mediated pathways, greenspace could also directly influence neurodevelopment through promoting discovery, risktaking, engagement, and control and mastery, inciting basic emotional states such as a sense of wonder and bolstering sense of self (Dadvand et al., 2015;Bowler et al., 2010;Kahn and Kellert, 2002;Kellert and Wilson, 1993).
The findings on the beneficial association of the greenspace exposure with ADHD outcomes could be explained by a number of pathways, including stress reduction, attention restoration theory, increase in physical activity and social cohesion, enrichment of the gut microbial diversity, and mitigation of hazardous environmental factors such as air pollution and noise (Nieuwenhuijsen, 2021).Stress can exacerbate ADHD symptoms (Combs et al., 2015).Therefore, environments that promote stress reduction could be beneficial for reducing ADHD symptoms (McCormick, 2017;Faber Taylor and Kuo, 2011).Inattention is one of the main criteria for ADHD diagnosis, and reports on the positive influence of the exposure to greenspace on improved attention could be discussed through attention-restoration theory, which, suggests that greenspace could induce restoration from mental fatigue caused by directed attention needed in everyday tasks (Stevenson et al., 2018).Studies have reported an association between the gut microbiota composition in infancy and subsequent behavioral outcomes (Loughman et al., 2020).Children with ADHD have been reported to have lower gut microbial diversity (Cenit et al., 2017).At the same time, surrounding greenspace has been reported to increase microbial diversity and alter human microbiota composition (Selway et al., 2020;Bowyer et al., 2021).Therefore, nature and greenspace exposure could be beneficial for behavioral development through the enrichment of the personal microbiota.
Exposure to environmental pollutants such as air and noise pollution has been reported to be associated with increased behavioral problems (Dadvand and Nieuwenhuijsen, 2019;Combs et al., 2015).Greenspace can mitigate exposure to these environmental pollutants (Stevenson et al., 2018;Loughman et al., 2020) and through this pathway it could improve children behavior.

Strength and limitations and future research directions
The reported associations for the exposure-outcome pairs were not sufficient to do a meta-analysis of the evidence.Even in the case of sufficient pairs of exposure-outcome, the use of different and non-convertible effect sizes prevent us to do a meta-analysis.Outcome assessment tools in the included studies were different, and the form of presentation of the results for the same instrument in some cases was different (e.g.use of cut-off for SDQ instead of reporting quantitative score).In addition, outcome definitions for the specific behavioral problem were different across the studies and different inclusion and exclusion criteria were applied for population and outcome definition across the studies, which reduced our ability to directly compare the results of the studies.The classification and diagnostic procedure of the outcomes are particularly important and diagnostic procedures should be given in more detail in future studies.Additionally, from the methodological point of view, we did not register the systematic review before conducting the review.However, we strictly adhered to the defined protocol for the review.
The available literature on the association of exposure to greenspace and children's behavior is nearly almost all coming from the North American and European courtiers.Low-and middle-income countries and more specifically African, South American, and Middle Eastern countries are underrepresented in the literature.None of our reviewed studies formally evaluated the potential mechanisms underlying their findings, which remain an open question to be evaluated by future studies.Shedding light on such mechanisms could support the causal nature of the associations observed by our reviewed studies.Considering different hypothetical pathways of greenspace on behavioral health, future studies suggested considering different indicators of exposure to greenspace which are relevant to different proposed mechanisms.For example, while residential proximity to or use of green spaces could be more relevant to physical activity, mitigating urban-related hazards surrounding greenspace could be more relevant.We found that most of the available studies used greenspace availability indicators (surrounding greenness, percentage of land cover) as a measure of exposure, however using other groups of greenspace exposure indicators such as greenspace accessibility, and especially the use of greenspace might be more appropriate to explore the possible association.
Given the small number of studies with comparable exposure and outcome metrics, we were not able to conduct meta-regression analyses to identify the outcome-specific sources of heterogeneity and compare these sources among different outcomes.This remains as an open question for the future reviews on this topic.

Conclusions
In this review, we found that existing literature points out a beneficial association between exposures to greenspace with several behavioral outcomes in children.However, the available evidence remains limited and non-conclusive for some behavioral outcomes.Of 29 studies included in this review, 17 reported cross-sectional associations, which had limited capacity in establishing causality.Additionally, we did not find any study from Africa or middle-east to which the findings in other regions might not be generalizable.Our findings highlight the importance of a change in school and living environments to enhance children's exposure to greenspace toward considering greenspace as one of the target elements in urban environmental planning.Further studies on the behavioral outcomes with currently scarce or non-conclusive findings such as bullying and aggressiveness are necessary.Further observational (longitudinal studies with multiple measures of exposure and repeated assessment of outcome) should be planned to investigate this association in different cities and urban contexts and disentangle the mechanisms that would help to understand this relationship.

Table 1
Characteristics of the included studies in the review.

Table 1
(continued) Normalized Difference Vegetation Index; ADHD: Attention Deficit Hyperactivity Disorder; DMS-IV: Diagnostic and Statistical Manual of Mental Disorders Fourth Edition; SDQ: Strengths and Difficulties Questionnaire; ICD: International Classification of Diseases; CBCL: Child Behavior Checklist; MSAVI: Modified Soil-Adjusted Vegetation Index; BASC-2: Behavior Assessment System for Children Second Edition; SAVI: Soil-Adjusted Vegetation Index.Risk of bias assessment for the selected longitudinal studies in the review.

Table 5
Detailed exposure assessment methods and exposure data sources in the included studies.

Table 4
Risk of bias assessment for the selected cross-sectional studies in the review.