The roles of residential greenness in the association between air pollution and health: a systematic review

While a growing body of literature suggests beneficial impacts of greenness on several health outcomes, relatively few studies have examined greenness as an effect modifier to impacts of air pollution on health outcomes, and results from the existing studies are inconclusive. We performed a comprehensive, systematic review of previous literature on greenness as a potential effect modifier for associations between particulate matter air pollution and health. After initial screening of 7814 studies, we identified 20 eligible studies. We summarized findings on study characteristics based on several criteria: health outcome, air pollution exposure, source of air pollution data, study location, study period, and median year of the study period. We evaluated characteristics of effect modification by greenness on air pollution and health associations based on the number of greenness metrics applied, type of greenness metric (e.g. normalized difference vegetation index, land use), data source for greenness, and spatial resolution and buffer size. We also summarized evidence for effect modification by greenness based on strength and direction of evidence for each study and overall evidence of effect modification by greenness by several study characteristics. Our systematic review showed that only a limited number of studies have been conducted on greenness as an effect modifier for air pollution-health associations. We found differences in several study characteristics such as greenness assessment (e.g. greenness metrics applied, spatial resolution, and data sources) across studies. Collectively, the studies provide suggestive evidence for the hypothesis that areas with high greenness have lower impacts of air pollution on health, although some studies reported inconsistent findings. The findings from our review provide valuable knowledge on how greenness affects associations between air pollution and health and could help identify critical areas for future study.


Introduction
More than half of the world's population currently lives in urban areas and the world's urban population is expected to increase to 68% by 2050 (United Nations 2018). Urbanization can bring a number of benefits such as economic growth and development, but also adverse impacts such as urban sprawl, pollution, and environmental degradation. Greenness is known to have environmental benefits such as reduced air pollution, noise, temperature and urban heat island effect by filtering of air pollutants and increased deposition and dispersion; absorption, refraction, deflection and masking of sounds; and decreased local temperature due to shade and vegetation (Janhäll 2015, Van Renterghem et al 2015, Aram et al 2019. Many previous studies investigated the direct effect of greenness on health and suggested that exposure to green spaces is associated with better health outcomes such as reduced mortality and improved mental health (Gascon et al 2016(Gascon et al , 2018. However, in addition to the direct impacts on health, green space appears to modify the association between other environmental conditions and health outcomes. The mechanisms of the modifying effects of greenness on the associations between air pollution and health include general health enhancement through increased physical activities, stress reduction, changes in the chemical composition and mixture of pollutants, more social interaction, and other factors impacting health in ways that affect susceptibility (Saebø et al 2012, Markevych et al 2017, Franchini and Mannucci 2018. While a growing body of literature has suggested beneficial impacts of greenness on several health outcomes such as mortality, cardiovascular outcomes, mental health, physical activity, asthma and allergyrelated symptoms, and birth outcomes , Fong et al 2018, Zhan et al 2020, Hartley et al 2021, fewer studies focused on greenness as a potential effect modifier to the impacts of air pollution on health. Recent studies suggested lower impacts of air pollution on health for those living in areas with high greenness and access to green spaces, including studies on mortality and preterm births (Asta et al 2019, Kim et al 2019, however, the results are inconclusive. For some health outcomes such as allergic diseases, some studies reported negative impacts meaning higher associations between air pollution and health were observed in areas with higher levels of greenness (Cariñanos and Casares-Porcel 2011). Inconsistent findings may result from several factors such as different definitions of greenness and different metrics used to assess greenness across studies. Therefore, a comprehensive review of existing research on the impacts of greenness on the association between air pollution and health is needed to better understand the current state of the scientific literature on the complex relationship of greenness with air pollution and health outcomes.
In this study, we systematically reviewed previous literature on greenness as a potential effect modifier for associations between particulate matter air pollution and health outcomes. While previous literature reviews have focused on the direct impact of greenness on health outcomes such as mental health, mortality, and birth outcomes (Fong et al 2018, Zhan et al 2020, to the best of our knowledge, no previous systematic review synthesized the scientific evidence on the role of greenness as an effect modifier in the air pollution-health relationship. Such scientific evidence is important for policy makers to inform evidence-based interventions and policy to address the public health burden from air pollution and for management of greenspaces. Also, our study informs our understanding of gaps in knowledge and can identify research needs and directions for future study.
The search was conducted on 22 December 2020. The systematic search was conducted with consideration of the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-analyses) guidelines (Moher et al 2015).

Selection criteria
We selected studies meeting the following inclusion criteria. Studies had to: (a) be population-based; (b) consider exposure to particulate air pollution; (c) explore mortality or morbidity; (d) examine effect modification by greenness on the association between particulate matter air pollution and health; (e) be peer-reviewed; (f) be written in English; and (g) be published through November 2020. Both single-city and multicity studies were included.
Screening of the articles was performed by two independent reviewers (CCL, HMC, JYS, KCF, and SH). Inconsistency regarding decisions of whether to include or exclude studies were resolved by a third reviewer (MLB). After we excluded studies by screening of titles and abstracts based on the inclusion criteria, we reviewed the full text of remaining articles. After the first initial screening (i.e. title and abstract), two reviewers independently performed full-text screening to determine whether the article was eligible for inclusion and, if so, performed data extraction. Figure 1 provides a flow diagram for the identification and selection of studies. After duplicate studies were removed, we initially identified and screened 7814 studies. Of these, 117 studies were identified for full-text screening, and of these 20 studies were eligible for inclusion in this review. We extracted information of each article's study location, time frame, study design, health outcomes, air pollutant, air pollution exposure metric (e.g. daily average, annual concentration), increment of exposure used in presentation of effect estimates, lag structure (i.e. delayed effects of air pollution on health outcomes) (e.g. previous day, average of previous days), covariates adjusted in the model, effect modification Records excluded: (n = 3, non-research article including review, workshop report, commentary, etc.) (n = 2, no greenness) (n = 1, no health outcomes of interest) (n = 11, no air pollutant exposure of interest) (n = 80, no effect modification by greenness, direct effect of greenness) Eligible studies after full screening (n = 20) Figure 1. Flow diagram of literature selection process for systematic review. Note: Screening prior to 'full-text' screening was based on the title and abstract. factors studied, greenness metric (e.g. normalized difference vegetation index (NDVI), land use) and results for main findings regarding greenness as an effect modifier of the air pollution and health relationship. We used results from the key findings presented by study authors, as originally reported. Some studies could contribute more than one result, for example if they investigated multiple health outcomes.
We summarized findings on study characteristics based on several criteria: (a) health outcome: mortality, hospital admissions/emergency room visits, and other health outcomes; (b) air pollution exposure: PM 2.5 (particulate matter with aerodynamic diameter ⩽2.5 µm), PM 10 (particulate matter with aerodynamic diameter ⩽10 µm), and other; (c) source of exposure data: modeled, monitoring, and other; (d) study location; and (e) study time period. Study period was used to calculate the median year of the study period for each study. We evaluated characteristics of effect modification by greenness on the air pollution-health association based on (a) number of greenness metrics applied (i.e. single, multiple); (b) type of greenness metric (e.g. NDVI, land use); (c) source of data for greenness (e.g. moderate resolution imaging spectroradiometer (MODIS), Landsat); and (d) spatial resolution and buffer size (e.g. 250 m, county-level) for greenness estimate. For each study, we summarized evidence of effect modification of the particulate matter and health association by greenness based on the direction and strength of evidence based on the statistical significance and several study characteristics. Statistical significance was based on the main findings presented by study authors, as originally reported. If the authors did not state the statistical significance, we used numerical results (i.e. 95% confidence intervals, p-value of 0.05). We also summarized overall evidence of effect modification by greenness using the following categories: no evidence, weak evidence, limited/suggestive evidence, and strong evidence based on the quantity of studies providing consistent evidence compared with conflicting findings.

Results
A total of 7814 unique articles were identified from the systematic search (figure 1). After first screening these studies by title and abstract, 7482 papers were excluded. Non-English papers were also excluded (215 papers). After the initial screening (i.e. title and abstract), 117 papers remained for full-text review. The primary reasons for excluding studies at this stage were: no effect modification by greenness (e.g. study examined the direct effect of greenness), no air pollution exposure of interest, non-research article (e.g. review, commentary), no investigation of greenness, and no health outcomes of interest. The full-text screening resulted in 20 eligible studies for inclusion in our review. Table 1 provides the characteristics of the studies included in this review: study period and location, study design, health outcome, air pollutant, air pollution exposure metric, lag structure, exposure increment used to present results, covariates adjusted in the model, potential effect modifiers considered, greenness metric, and main findings. Table 2 shows summary characteristics across the 20 eligible studies for health outcomes, air pollutant, source of air pollution exposure data, study location, study period, and median year of the study period. The most represented country was the United States with six of the 20 studies. The remaining studies were conducted in China (four studies), the Netherlands (two studies), South Korea (two studies), Spain (two studies), and other countries (Canada, Greece, Iran, Italy). No study was conducted in South America or Africa. Most study time periods were less than five years (45%). Some studies examined multiple health outcomes and/or air pollution exposures. Among the 20 studies, 10 studies investigated mortality. Of the ten studies that examined mortality, seven considered non-accidental mortality. Two studies investigated hospital admissions/emergency room visits and eight studies assessed associations of air pollution with other health outcomes (e.g. preterm birth). For air pollution exposure, of the 20 studies, 16, 10, and 1 studies evaluated PM 2.5 , PM 10 , and other pollutants (e.g. coarse particles), respectively. Regarding the source of data for air pollution, 11 studies used modeled estimates and 10 studies used values from monitoring stations, with 1 study using estimates from both types of data sources. Table 3 summarizes characteristics of the greenness metrics used in the studies based on whether a single or multiple greenness metric was applied, type of greenness metric, source of data for greenness, and spatial resolution and buffer size. Most studies (75%) applied a single greenness metric. NDVI was the most commonly used metric to quantify greenness (16 studies), followed by amount of green space/vegetation coverage based on land cover/land use data (4 studies). Other greenness metrics applied in the studies were distance between home and the boundary of the nearest green space, neighborhood walkability index, streetscape greenery, and landscape metrics. Nine studies used NDVI from MODIS, with spatial resolution of 250 m × 250 m, followed by NDVI from Landsat images with 30 m × 30 m spatial resolution (seven studies) and land use/land cover database/map (four studies). Some studies used distance from the nearest green space identified from a shapefile provided from the Municipality (one study), US EPA Smart Location database (one study), City Statistical Yearbook (one study), and street view image database (one study). Studies examined greenness at different spatial buffers. Eight studies applied multiple buffers, with others using various single buffer sizes (e.g. 250 m, 500 m, 1000 m) and administrative geographic areas (e.g. county level, census block group level, district). Table 4 provides a summary of evidence for effect modification by greenness on the association between particulate matter and health. Of the 20 studies, 8 studies (40%) found evidence of lower risk of air pollution on health outcomes in areas with high greenness. Three studies (15%) found higher risk of air pollution with high greenness and four studies (20%) did not find evidence of effect modification by greenness. Five studies (25%) showed different findings of the effect modification of air pollution and health associations by greenness depending on several factors such as air pollutants, urban/rural region, cause of death, and socioeconomic status (SES). For example, one study found higher risk of health impacts from PM 10 and PM 2.5 in areas with less greenness in rural regions, whereas estimated particulate matter health effects were higher in areas with more greenness in urban regions. Some studies found beneficial impact of greenness only in low SES areas.

Summary of evidence for effect modification by greenness on the association between air pollution and health
We provided a summary of evidence for effect modification by greenness based on several study characteristics: health outcome, air pollutant, source of data for air pollution, study location, greenness metric (number and type), data source for greenness, and spatial resolution (table 5). Regarding health outcomes, we found suggestive evidence of lower risk with high greenness for the association between air pollution and mortality. Of the ten studies that investigated the risk of air pollution on mortality, five studies showed significant evidence of lower risk with high greenness, while one study reported significant evidence of higher risk with high greenness. For hospital admissions/emergency room visits or other health outcomes, there exists weak or no evidence of effect

Discussion
Our systematic review found that only a limited number of studies have investigated greenness as an effect modifier of the association between air pollution and health and that those studies included a range of study designs and characteristics such as the way in which greenness was assessed (e.g. greenness metrics applied, spatial resolution, and data sources). This review also observed suggestive evidence of the beneficial impacts of living in areas with high greenness with lower air pollution-health associations, although findings were inconsistent across studies.

Knowledge gaps in the evidence of effect modification by greenness and critical areas for future study
Although evidence on the direct effect of greenness and health outcomes is well documented , Fong et al 2018, Zhan et al 2020, Hartley et al 2021, very few studies have been conducted to examine the effect modification by residential greenness on the association between air pollution and health outcomes. Findings of the studies we reviewed suggest that higher greenness was associated with lower risk of air pollution exposure for several health outcomes such as mortality, although results were not fully consistent. A recent study suggested an increasing trend in PM 10 effects on birth outcomes with an increasing distance from green areas (Asta et al 2019).
Another study in Canada evaluated the role of residential greenness in modifying associations between long-term exposure to PM 2.5 and non-accidental and cause-specific mortality. They found that the strength of observed associations between PM 2.5 and mortality decreased as greenness increased (Crouse et al 2019).
Other studies also observed that risk of air pollution exposure on various health outcomes such as hospital admissions (Heo and Bell 2019) and infantile atopic dermatitis (Lee et al 2018) was lower for people living in areas with more green space.
Our findings of overall evidence of effect modification by greenness based on several study characteristics can inform understanding of knowledge gaps and help identify research needs and direction for future study. For example, we observed that the most examined health outcome was mortality, for which we found suggestive evidence of lower risk with higher greenness, and there were few studies investigating the modifying effect by greenness for other health outcomes such as hospital admissions/emergency room visits, for which we found weak evidence of effect modification. This indicates that further research is needed on morbidity outcomes. Our findings also showed that most studies were based on areas in Asia, North America, and Europe. Current epidemiologic evidence is very limited in many other regions, especially in South America or Africa, and studies are needed in these areas. The influence of greenness on the association between air pollution and health could differ by location due to several factors such as differences in pollution mixtures (e.g. PM 2.5 chemical composition, co-pollutants), population characteristics, indoor/outdoor activity patterns, housing, and type of greenness. Variability in findings across studies may further relate to differences in population characteristics, air pollution composition, and vegetation, thus studies are needed in many regions and types of settings. Scientific evidence from this review has important implications for policy makers developing urban planning and health policy strategies, as we identified potential benefits from greenspace.

Assessment of greenness
In our assessment of greenness as an effect modifier on air pollution-health associations, we found differences in several characteristics such as greenness metrics, data source, and spatial resolution and buffer across studies. To assess greenness, most studies applied NDVI from MODIS (Kioumourtzoglou et  or assessed greenness at the administrative boundary (e.g. county-level, census block group level) but the associations did not vary by spatial resolution or buffer size based on sensitivity analyses within these studies and comparison across studies. Greenness was commonly measured for the area around the subject's residential address (James et Table 5. Summary of evidence for effect modification by greenness on the association between particulate matter and health, by study characteristics.
Note: If the study reported mixed findings, we included information on each case separately but for some common study characteristics, we used the color of stronger evidence; We summarized overall summary of evidence using the following categories: no evidence, weak evidence, limited/suggestive evidence, and strong evidence based on the quantity of studies providing consistent evidence compared with conflicting findings.

Heterogeneity in modifying effect of greenness by several characteristics
Findings were not consistent across studies. Some studies did not find evidence to support effect modification by greenness for the estimated effects of air pollution on health outcomes, with null or opposite findings of higher effect with higher greenness ( Our findings suggested that greenness may modify the associations between air pollution and mortality differentially depending on the cause of death. Kim et al (2019) found that the effect of PM 10 on cardiovascular mortality was stronger in districts with a lower level of greenness, whereas the effect of PM 10 on non-accidental mortality was higher in districts with higher greenness. The authors suggested that possible explanations for these seemingly contradictory findings are: high greenness areas produce allergenic pollens, which subsequently interact with air pollutants; increased exposure to pesticides; transmission of infections by arthropod vectors; and excessive exposure to UV radiation, which could reduce the immune defenses (Lohmus andBalbus 2015, WHO 2016). Further studies are needed to understand the underlying mechanisms and factors that contribute to these variations, with attention to the multiple pathways through which greenness could modify the air pollution and health relationship.
In our review, we found differences in the estimated modifying effect by greenness between urban and rural areas. For example, de Keijzer et al (2017) found that effects of PM 10 and PM 2.5 were higher in areas with less greenness in rural regions, whereas those effects were higher in areas with more greenness in urban regions. The different patterns between the urban and rural areas may result from differences in population characteristics (e.g. indoor/outdoor activity patterns, health behaviors), PM 2.5 chemical composition relating to different sources, type and diversity of greenness, and other environmental factors that differ between urban and rural areas.
Previous literature suggested that health benefits related to greenness may differ by socioeconomic status. Yitshak-Sade et al (2019) observed that PM 2.5 -related CVD mortality risk was lower in highly populated greener neighborhoods with sociodemographic features that are highly correlated with lower socioeconomic status. Another study reported that in low SES areas, green space seems to alleviate effects of air pollution on the prevalence of high blood pressure and diabetes (Groenewegen et al 2018). Intersectionality between SES and other population characteristics relating to vulnerability (e.g. race/ethnicity, rurality) may contribute to the health disparities through multiple pathways. Thus, further research is needed considering relevant factors and complex interaction in relation with underlying mechanisms.

Possible mechanisms of modifying effect by greenness
There are several possible mechanisms through which greenness could modify the air pollution and health relationship, such as changes to the air pollution mixture (e.g. filtering of pollutants), different exposure patterns, and increased overall health (e.g. physical activity, mental health and wellbeing) in ways that affect susceptibility. Greenness may reduce the adverse effect of air pollution by filtering of air pollutants and increased dispersal of pollutants through improving ventilation, improving the immune system through psychological restoration for stress and anxiety, and increasing physical activity and social cohesion (Nowak et al 2006, Nieuwenhuijsen et al 2014, Markevych et al 2017. Another possible mechanism is the environmental biodiversity hypothesis. This suggests that contact with nature is beneficial for human microbiota, which influence immune tolerance and immunomodulatory capacity (Rook 2013, Kuo 2015. A study suggested that increasing greenness and lowering air pollution might have a synergistic effect on mortality and interact nonlinearly with mortality (Ji et al 2020). Other studies reported mixed findings on whether greenness mediates air pollution-related exposure pathways, which indicate the complex relationships between greenness, air pollution, and health outcomes (de Keijzer et al 2017, Kim et al 2019). Further studies are needed to understand the underlying physiological mechanisms of how greenness could modify the association between air pollution and health.

Study limitations
Our review includes several limitations. We were unable to quantitatively combine estimates through meta-analysis due to the small number of studies with the same category of health outcome, particulate matter exposure, and assessment of greenness. We elected not to assess quality of the identified studies. Instead, we provided detailed information for each individual study such as health outcome, air pollution exposure, exposure data source, greenness metric and data source, and spatial resolution and buffer size for greenness measurement. This information from our review provides valuable knowledge on how greenness affects associations between air pollution and health outcomes and helps identify critical areas for future study. Publication bias may exist. Studies that did not find statistically significant results or found unanticipated findings may be selectively less likely to be submitted or published. To the best of our knowledge, this is the first systematic review of effect modification by greenness on the association between air pollution and health outcomes. The findings from our review extend our understanding of greenness as an effect modifier, provide vital information for policy makers, and inform future studies.

Conclusion
Overall, our review suggests that greenness may have positive effects on the association between air pollution and health outcomes, although some findings vary, such as by type of health outcome, and results were inconsistent. Future research considering multiple relevant factors and pathways is needed to better understand the characteristics, mechanisms, and potential mediators in relation to greenness on the association of air pollution and health outcomes.

Data availability statement
No new data were created or analyzed in this study.