Green Space Quality and Health: A Systematic Review

(1) Background: As cities densify, researcher and policy focus is intensifying on which green space types and qualities are important for health. We conducted a systematic review to examine whether particular green space types and qualities have been shown to provide health benefits and if so, which specific types and qualities, and which health outcomes. (2) Methods: We searched five databases from inception up to June 30, 2021. We included all studies examining a wide range of green space characteristics on various health outcomes. (3) Results: 68 articles from 59 studies were found, with a high degree of heterogeneity in study designs, definitions of quality and outcomes. Most studies were cross-sectional, ecological or cohort studies. Environment types, vegetation types, and the size and connectivity of green spaces were associated with improved health outcomes, though with contingencies by age and gender. Health benefits were more consistently observed in areas with greater tree canopy, but not grassland. The main outcomes with evidence of health benefits included allergic respiratory conditions, cardiovascular conditions and psychological wellbeing. Both objectively and subjectively measured qualities demonstrated associations with health outcomes. (4) Conclusion: Experimental studies and longitudinal cohort studies will strengthen current evidence. Evidence was lacking for needs-specific or culturally-appropriate amenities and soundscape characteristics. Qualities that need more in-depth investigation include indices that account for forms, patterns, and networks of objectively and subjectively measured green space qualities.


Introduction
Green spaces are a crucial aspect of urban cities. They protect against many of the harmful impacts of rapid urbanisation on health. They also permit social and economic benefits by providing preferential settings for relaxation, building social connections, engaging in physical activity and feeling closer to nature, including resident wildlife [1]. Therefore, urban greening is an important strategy for addressing complex global issues such as climate change, sustainable urbanisation and health inequality. This is recognised via the United Nations Sustainable Development Goal (SDG) 11 target 7, which states "by 2030, providing universal access to safe, inclusive and accessible, green and public spaces, in particular for women and children, older persons and persons with disabilities" [2].
Substantial research is dedicated to revealing the health benefits of green spaces [3]. While more green space tends to be good for health, such conclusions are not universally reported. Most research in this field tends to use measures of 'greenness' such as the normalised difference vegetation index (NDVI) to quantify green space exposure [4], ignoring substantial heterogeneity in the constituent qualities of green spaces that make them attractive for visiting and, in turn, support health and wellbeing. For example, green spaces may vary in terms of objectively measurable good qualities (e.g., presence of certain attractive elements, such as tree canopy, footpaths and seating) and others that are more subjective in nature (e.g., an emotional or spiritual connection to a particular green space). Bad qualities (e.g., proximity to a busy road and lack of accessibility) may discourage visitation and negate health benefits. Ignoring the constituent qualities that attract or discourage people to spend time in green spaces holds back the field from having more substantive impacts as a catalyst for improving community health and reducing inequities. Examining these qualities, both good and bad, may solve a missing link in our understanding of the relationship between green spaces and health [5].
Moreover, studying green space qualities has practical implications for urban planning. Driven by rapid densification, the compact, high-density city has become the dominant urban design worldwide. Not only does a compact city warrant multifunctional green spaces that can serve its diverse citizen population. It also presents a complex set of trade-offs between green space creation, regeneration and expansion on one hand, and the development of new, often competing land-use on the other (e.g., housing, infrastructure and commercial) [6]. Within space constrained contexts, modifying qualities of existing green spaces may offers an important way to maintain and improve quality of life in urban communities.
Research on the health benefits of green space qualities is still emerging and there are no consensus definition what green space quality is. We do not know which qualities can be modified, and which health benefits these modifications will bring (if any). To build capacities for research that attends to these issues, we conducted a systematic review to take stock of what research has been performed on green space qualities and health, with the broader aim of charting possible paths forward to strengthen the policy relevance of this research.
This systematic review aims to: (a) Evaluate whether improving certain qualities of green space provides health benefits to the population; (b) Identify and categorise all qualities of green space that have been investigated in previous primary studies; and (c) Explore the extent of variations in design characteristics of these studies.

Materials and Methods
The reporting of this review was guided by the updated Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guideline [7]. This review was not registered a priori, nor was a protocol published separately.

Search Strategy
We searched the following databases for articles from inception up to 8 December 2020: MEDLINE via Ovid, Embase via Ovid, PsycINFO via Ovid, CINALH via EBSCO and Scopus. No language or publication date restriction was applied. An updated search was performed on 30 June 2021. The search was supplemented by a manual search of the reference lists from relevant systematic reviews.
The search strategy was a combination of three components: (health outcomes AND green space quality AND green space types). For health outcomes, we used both generic and specific search terms to capture all dimensions of physical and mental health, drawing from previous systematic literature reviews on green space and health [8,9], obesity and physical activity [10,11], birth outcomes [12], mental health [13][14][15], puberty timing [16] and menopause [17]. For green space quality, we combined the word "quality" and other determinant terms adapted from audit tools used for assessing the physical environment of parks [18]. For green space types, we used both generic and specific search terms to capture all types of green space in both urban and rural settings. The full search strategy is available in Supplementary File S1.

Study Selection
We included all human studies meeting the following criteria: (a) Population: green space users of all ages and genders; (b) Exposure: In the context of our review, green space quality refers to any attribute that can affect willingness to use and interaction of users with that space, including but not limited to intrinsic characteristics (size or patterns), features (vegetation, facilities or amenities), conditions (maintenance or safety) or user perception of its usefulness or quality. All types of natural and man-made green environments, including parks, streetscape greenery, urban open spaces, playgrounds, coastal parks with vegetation, etc., were included as long as they were defined by authors as green space. Studies where participants viewed digitalised renderings or photographs of green spaces without actual exposure were excluded. Studies that did not investigate any aspect of green space quality were excluded. The percentage of overall vegetation coverage and "greenness" (e.g., the normalised difference vegetation index) were not eligible as they are considered measures of green space quantity, unless specific vegetation types were analysed (e.g., tree canopy); (c) Outcomes: Studies that investigated health outcomes, including but not limited to cardiometabolic, respiratory, reproductive, neurological and psychological health, and child development, were included. Studies that only measured behaviours (park usage, park-based activity, etc.) without assessing health outcomes were excluded; (d) Study design: All observational and intervention studies, including randomised, quasi-randomised and non-randomised trials. We excluded non-English language studies, study protocols, conference abstracts, dissertations, reviews, qualitative studies, editorials, case studies and opinion pieces.
All retrieved data were imported into Covidence (Veritas Health Innovation, Australia) to remove duplicates. Two reviewers (PYN and HR-A) independently screened all titles and abstracts in duplicate and excluded studies that did not meet the inclusion criteria. Studies that were included from title/abstract screening had their full text reviewed in duplicate by the same two reviewers and reasons for exclusion were noted. Disagreement was resolved by discussion with senior reviewers (XF and TA-B). All stages of study screening were conducted in Covidence.

Data Extraction and Quality Appraisal
One reviewer (P-YN) extracted the data using a standard data extraction form and a second reviewer (HR-A) validated 10% of the studies for accuracy. The data extracted included: study characteristics (location, time, settings), population's demographic and clinical characteristics, green space types, green space quality domains, health outcomes and corresponding measures of association. We also recorded the tools used to assess green space quality and health outcomes, effect measures reported, types of statistical analyses conducted and any adjustment for confounding factors. Based on the effect measures and 95% confidence intervals, we recorded the direction of effect for each study, i.e., whether the study presented some evidence of protective associations, some evidence of risk associations, or no significant associations at all.
One reviewer (P-YN) appraised the methodological quality of all included studies using the quality assessment tools for the appropriate study types [19] and the second reviewer (HR-A) validated 10% for accuracy. Because these tools do not provide for ecological studies, the existing tool for observational cohort and cross-sectional studies were adapted by adding 3 criteria addressing ecological fallacy, spatial autocorrelation and uncertainty in fitting spatial data [20,21]. Based on the list of applicable criteria, each study was given an adjusted quality score of 0-10 (Supplementary File S2). Disagreement was resolved with consensus via discussion with senior reviewers (XF and TA-B), if required.

Data Analysis
We used inductive categorisation to develop a set of domains of green space quality based on definitions reported in the included studies and stratified the findings of the studies based on these quality domains. Due to the heterogeneity of exposure, intervention and outcomes, meta-analysis was not conducted.

Results
In the initial search, we identified 30,220 records, and 7 additional records were added through manual searching. After removing duplicates, 23,745 studies were included for title/abstract reviews, from which 118 full texts were selected for further screening. Fifty full texts were excluded (Supplementary File S3). The final sample comprised 68 articles from 59 studies (Figure 1).
Cities and inner-city neighbourhoods were the predominant settings. Seven studies took place in multi-ethnic and/or socioeconomically deprived areas [29][30][31]37,[42][43][44]. One study specifically examined the differential impact of green space on children of South Asian descent versus Caucasian children [37]. The characteristics of included studies are summarised in Table 1.

Study Designs
Most included studies were cross-sectional (n = 32), followed by ecological studies (n = 16) and cohort studies (n = 15). Before-after (n = 1), quasi-experimental (n = 3) and case-crossover designs (n = 1) were rare ( Table 1). The latter were relatively newer approaches published from 2015 onwards ( Figure 2). All cohort studies were nested in existing longitudinal studies, usually with an additional cross-sectional survey for green space use and perceptions conducted after the initial survey waves. The followup time for longitudinal studies ranges from 2 to 18 years [56]. The quasi-experimental studies [55,57,87] had intervention and control groups selected in a non-random manner from two neighbourhoods with pre-determined green space qualities. The before-after study [22] was conducted among participants who participated in outdoor nature walks. The cross-over study [79] bi-directionally matched case days with the highest symptom severity scores to control days with the lowest scores, hence participants served as their own control. Among cross-sectional surveys, eight studies used convenience sampling by recruiting from park visitors [23,44,51,59,61,73,89,90]. The mean adjusted quality score among 68 articles was 0.49 ± 0.12 (scale 0-1).

Figure 2.
Published studies over the years, by study design.

Definition of Green Space
Most studies (n = 42) used a loose definition of green space to include any natural or open space, encompassing urban green space, private and community gardens, public open spaces, bushland and forest reserves, etc. Eleven studies included playgrounds and sports fields [25,[35][36][37]52,53,57,60,67,71,84]. Seven studies included streetscape greenery, which referred to any vegetation cover that gave the street a green appearance [52][53][54]68,69,74,81]. Forty-seven studies used data from a geographic information system (GIS) to identify green spaces or evaluate green space characteristics. One study examined neighbourhood vegetation as viewed from within the house [22]. The most common buffer size for GIS analysis was 0.5 mile (approximately 800 m), generally aligning with a 10-min walk [82]. Detailed definition of green space in each study is outlined in Table 1.

Green Space Qualities
Green space qualities were classified into 10 domains. Detailed definitions of green space qualities in each study is outlined in Table 3.

Environment/Land Cover Types
There was one before-after study, seven cohort studies, one case-cross over study, seven cross-sectional studies and six ecological studies under this domain. All studies used different land cover or environment classification, commonly via adopting definitions of the data sources, some adapting [39,53,83,85] or developing their own typologies [22,86]. Detailed definitions of environment types were outlined in Table 3.
Overall, a higher land-cover diversity in the neighbourhood was protective for chronic morbidities [53] and childhood asthma [56]. Some environment types were more likely to provide health benefits than others. Vegetation patches such as grassland and tree canopy was not associated with reduced sudden unexpected deaths, but formal green spaces such as greenways and forests were [85]. People who spent time outdoor recalled greater mental restoration following visits to coastal locations and rural green space than urban green space [86]. Some environment types ("broadleaf woodland", "arable and horticulture", "improved grassland", "saltwater" and "coastal" environment) were positively associated with prevalence of good health among UK citizens [83]. The observed relationship between land cover types and BMI varied across age and gender. A positive relationship with lower BMI was found with high coverage of impervious surfaces among middle-aged adults and high forest coverage among young adult males. In other age and gender groups, the relationships were non-significant [76]. More rigorous studies, however, did not report significant findings. In a before-after study, the environment type of an outdoor walk did not have significant influence on emotional states of participants [22]. A sibling matched case-control analysis of Scottish mothers and their children (1991)(1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010) found that infant birth weight was associated with the quantity of natural space around the mother's home, but was unrelated to specific types of natural space (parks, woods or open waters) [41].           Similarly, the type of vegetation within green space potentially modulated its health benefits. There was consistent evidence of forests being a protective factor for obstructive airway diseases [28,34], cardiovascular diseases [49], allostatic overload [59], psychological distress and general health [45,65,74] while grassland and herbaceous vegetation were not. On the other hand, some studies showed superior benefits of shrubs and grass compared to trees in improving mental health [64] or severe allergy [79]. In low-diversity areas, certain vegetation types presented higher risks for asthma or other allergic conditions, typically non-native shrubs [56] or coniferous trees [35]. No difference in benefits between vegetation types was observed in studies of memory and dementia [46,48], depression and anxiety [26,45]. In one study, all vegetation types were shown to be protective against autism, which was potentially driven by their shared function of buffering against traffic noise and air pollution buffering [33].

Natural Features
There was one before-after study, one cohort study, ten cross-sectional studies and three ecological studies under this domain. Natural features refer to characteristics of vegetation, animals, water bodies, and the overall naturalness of green space. Trees, flowers and fresh air [73] conferred restorative benefits to park visitors, with differential effects between genders. The higher density of trees among park vegetation was associated with lower rates of cardiovascular conditions [47,70] and a higher quality of life [24,51], but not overall general health [23]. The presence of dense shrubs, which implied lower security and safety, reduced the restorative benefits of parks [89]. Green spaces perceived as being more "natural", such as protected areas or bushlands, provided greater benefits on mental restoration [86] and physical health [52,83]. A combination of habitat, plant, bird and insect biodiversity exhibited restorative effects, but each biodiversity component alone did not [22,42,44]. Interestingly, neither quantity or diversity of neighbourhood vegetation alone was significant predictor of stress levels, but vegetation diversity could modify the relationship between vegetation quantity and stress levels [62].
Certain green space characteristics were potentially associated with health risks. Streetscape with tree species of high allergenicity was associated with an increase in local asthma hospitalisation rates in vulnerable populations [69]. Freshwater quality was identified as an indicator of poor health status [83].

Infrastructure and Amenities
There were nine cross-sectional studies, two quasi-experimental studies, one prospective cohort study and two ecological studies under this domain. Infrastructure and amenities refer to the availability of facilities for various purposes (recreation, resting, socialisation, etc.), the quality of paths within and leading to green space, and general maintenance. Park facilities did not reduce rates of depression [40], BMI [31,32,50,75] nor general health status of park users [23,24]. High maintenance was not associated with lower psychological distress [43] or BMI [50]. However, parks that function as recreational or sports venues may provide some cardiovascular and mental health benefits [71,84]. Mixed results were reported on the relationship between walking paths' conditions and quality of life [25,51]. A natural experiment was conducted in disadvantaged suburbs of Melbourne, Australia, tracking psychological wellbeing of park visitors for 3 years after adding refurbishments (playground equipment, walking paths and shade) to selected parks. When compared to control parks, park refurbishments did not improve emotional states of park visitors [55]. Similarly, in the Netherlands, neighbourhoods that implemented interventions to increase accessibility and useability of green space did not see an improved general health compared to control neighbourhoods [57].

Size
There was one prospective cohort study, six cross-sectional studies and four ecological studies and under this domain. Ten studies used spatial analysis to measure green patch size. Most studies found evidence for health benefits of larger green space for a wide range of outcomes: BMI [29,75], cardiovascular mortality [82], chronic morbidities [53], depression [42], general health status [23] and quality of life [30]. In a prospective cohort study in Perth (Australia), where residents were followed up after settling into a new neighbourhood, the increases in numbers of small parks, district parks and regional parks were each positively associated with mental wellbeing, but not the mid-sized local and neighbourhood open spaces [84]. However, some studies reported inconclusive evidence for these health benefits [24,32,78] 3.5.5. Shape, Pattern and Connectivity There were six ecological studies and two cross-sectional studies under this domain. While all studies used spatial analysis to quantify green space patterns, six studies combined health data at the spatial block level [63,67,76,77,80,82] while others conducted regression analyses using individualised data [29,30]. All studies reported positive correlation between indices measuring the shapes and distribution patterns of green patches and a wide range of outcomes, including BMI [29,76], paediatric quality of life [30], respiratory health [63,67,77] and all-cause mortality [82]. The indices include the fragmentation index (higher values indicate more fragmented green space areas), mean area of greens space (higher values indicate averagely larger green space areas), connectedness index (higher values indicate more connection between individual green spaces), aggregation/isolation index (higher values indicate more clustering of individual green spaces), shape irregularity index (higher values means more irregular shape of each green space, as opposed to round/oval shape). When stratified by gender, age and retirement status, differential benefits were observed for female and younger users [76].

Safety
There were six cross-sectional studies under this domain. The safety of green space was associated with better quality of life [23,25,51], reduced psychological distress [43] but did not have significant effects on BMI [50] of residents. In a mediation analysis, park crimes reduced the benefits of parks on mental health [72].

Cleanliness and Absence of Incivilities
There were three cross-sectional studies and one ecological study under this domain. Park cleanliness, either ranked by park visitors or assessed by trained auditors, was associated with lower rate of depression [42]. Evidence was inconclusive for BMI [50,78] or quality of life [24].

Peacefulness
There were three cross-sectional studies under this domain. A lower level of "nuisance" (defined as presence of dogs, dog fouling, or young people) was not correlated with better life satisfaction nor physical health among the elderly [25]. Park users did not consider a private environment in the park important in improving their mood states [73]. On the other hand, soundscapes in parks triggered positive feelings and reduced stress [61].

Perceived Quality/Satisfaction with Quality
There were four nested cohort studies, two cross-sectional studies, and one ecological study under this domain. In these studies, participants were asked to rank their perceived quality or aesthetics of green spaces, without a priori definition of factors to be considered. All studies examining "perceived quality" demonstrate positive association of green space's perceived quality with health. Women living near good-quality local parks had lower rates of postpartum psychological distress or serious mental illnesses [39]. The effect on postpartum weight gain was less clear, with significant benefits only observed in areas with high vegetation coverage (≥40%) [38]. Parents' satisfaction with green space was also linked to improved prosocial behaviour of their children [36,37]. Analysis of the Netherlands' population data found a modest increase in life expectancy among residents living near high-quality green spaces [66]. However, perceived aesthetics of parks was neither a predictor of mood states [73] nor BMI [50].

Combination of Features
There were one quasi-experimental study, two cohort studies, eight cross-sectional studies and two ecological studies under this domain. These studies use a mix of features from the previous domains to evaluate park quality. Detailed definitions of these composite scores were outlined in Table 3.
Three studies investigated both objectively measured and the perceived quality of green spaces, and compared their effects on health. When comparing two neighbourhoods with different socioeconomic status, the residents' perceived quality of a green space statistically mediated the relationship between its objective quality and neighbourhood satisfaction, but did not have any direct effect of wellbeing [87]. Only objective quality reduced psychosocial distress (K6-PDS questionnaire) in one study [60] while only perceived quality improved mental wellbeing (MHI-5 questionnaire) in another study [68].

Discussion
Overall, our review demonstrates evidence of health benefits associated with a wide range of green space qualities. Increasing research interest in green space qualities was demonstrated ( Figure 2) and this aligns with rising interest in urban greening to counter the health and climate impacts of urbanisation [6]. The COVID-19 pandemic may also have amplified attention on this topic from academics and policymakers, as communities in many countries have flocked to green spaces as a means of coping with lockdowns and socioeconomic disruption [92,93]. After excluding results with a study quality assessment score under 50 (N = 32), evidence showed consistent positive associations with health with the green space qualities we classified as "environment types", "natural features", "shape and connectivity", and "objective quality scores". Limited evidence was found on the health benefits of improving infrastructure or amenities in green spaces. Research gaps were identified for the following green space qualities: peacefulness, safety and absence of incivilities; needs-specific or culture-appropriate amenities, and soundscape characteristics.

Green Space Qualities
The most commonly assessed qualities of green spaces were the environment types of the natural space, as well as vegetation types and other natural characteristics. Our review shows that some environment types were linked to positive health outcomes more than others [41,83,86]. Health benefits were observed when the environment type facilitated age-and gender-appropriate physical activities. For example, middle-aged adults group preferred built facilities with paved paths for exercising whereas young adults prefer forested areas with unobstructed grounds for athletic, adventurous activities such as hiking, trail-running or mountain biking [76,82]. Therefore, preserving diversity in land cover types (e.g., structured versus natural) may be a potential option to enhance health benefits of green spaces, especially in dense urban areas with limited options for expansion. Moreover, green space designs might be optimized for health through tailoring to local community profiles, to bring people together and to enable them to do what they find nourishing. This requires consultation and it is likely that certain qualities may be a source of conflicting views. For example, accommodating for birds in green spaces may be viewed positively for their provision of restorative soundscapes and an enhanced feeling of connectedness with nature, but also negatively due to the timing of their sounds, impacts on property (e.g., droppings) and occasional swooping that may create a lack of felt safety [94].
Evidence indicated that some vegetation types may be more beneficial towards particular health outcomes than others. Tree canopy and forests were more consistently associated with better cardiovascular and respiratory health than grassland [47,59,67,79]. A reason may be that trees permit and promote restoration while also providing shade that helps to activate walking and active transportation (particularly in hot climates), whereas grass and shrubs might not convey the same range and levels of benefit [76]. Moreover, because of their foliage, evidence indicates that forests have the capacity to intercept airborne pollutants and buffer against traffic noise, alleviating oxidative stress and reducing risks of atherosclerotic diseases [95]. On the other hand, shrubs may impede visibility and reduce levels of felt safety, while large areas of open grass may reduce walkability (especially if it is walled or fenced-off, as can be the case for private green spaces like golf courses) [74,76,89]. Importantly, this may reflect an interaction between vegetation type and other contextual factors, such as levels of crime, nearby land-use and transport infrastructure. Further research that examines potential contingencies of association between vegetation types and health outcomes within the context of other land-uses is warranted.
Interestingly, many studies in the facilities/amenities domain show no statistically significant associations with physical or mental health, despite evidence that some of these qualities are associated with physical activity [90]. This might be because different types of facilities may result in different forms of behaviour, some of which may instead promote sedentary forms of leisure (e.g., seating) or detract from the perceived 'naturalness' appealed by certain park users (e.g., some sports facilities that use synthetic materials) [96]. Moreover, some studies may log the availability facilities but not their condition and usability. For instance, access to areas of parks and particular buildings may be difficult for people with functional limitations, while there may also be cultural or social factors that influence whether a particular facility is considered accessible [97].
Some qualities have a small evidence base, such as safety, tranquility or absence of incivilities. Most of this evidence focused on psychological wellbeing or quality of life using Likert-type rankings or the number of unwell days. Future studies on these aspects will benefit from using robust, validated questionnaires featured in other quality domains, such as MHI-5, PRS or PANAS [22,55,68,89]. Moreover, perceived safety of public spaces can be influenced by neighbourhood characteristics and social vulnerabilities, which need to be accounted for in future studies.
Some quality domains were not featured among the included studies. Availability of needs-specific amenities, such as for people living with particular disabilities, may encourage more inclusive park usage and increase the potential to reduce health inequity [3,98]. Tailoring park amenities and features to the local communities, such as instructions in multiple languages, accommodation (and celebration) of cultural traditions and rituals, etc., may be particularly important in multi-cultural neighborhoods [99,100]. One study examined the feelings evoked by soundscape [61], but the constituents of soundscape that provide therapeutic effects, such as sounds of nature, human activities or traffic noise, were not elucidated. Types of bird songs were previously studied, but as sound clips rather than actual exposure inside parks [89].

Health Outcomes
Physical health is the most commonly assessed set of health outcomes. Most studies showed evidence of potential benefits for anthropometric measures (BMI and obesity) and respiratory health (allergic diseases). Understandably, there are established frameworks explaining how green spaces reduce obesity via promoting physical activities [96,101] and protects against respiratory diseases via regulating temperature and air pollution [77]. Only 7 out of 34 studies on physical health examined associations with cardiovascular diseases.
Based on existing evidence, higher quality green space may reduce cardiovascular mortality and incidence of cardiometabolic diseases [47,70,71]. However, evidence for associations with specific cardiovascular diseases was small. Consistent evidence from this review indicated a range of probable mental health benefits linked with various green space qualities. This aligns with existing conceptual frameworks, which suggest green spaces can confer mental health benefits via reducing stressor exposures and replenishing mental resources for coping [3].
Although the evidence base is substantial for physical and psychological health outcomes, there is granularity in the quality of outcome measurement tools. For physical health, a number of studies relied on general health questionnaires such as SF-12 and SF-36. These have a low administrative burden and good internal validity [91], but responses may differ among age, education or ethnicity subgroups [102], which may explain conflicting findings among these studies. For mental health, some studies used self-ranked Likert-type questions, which lacked reliability and consistency compared to validated questionnaires like the MHI-5, K6-PDS or CED-S. A potential approach for future studies is to use quantifiable biological measure to validate subjective questionnaires, such as hair cortisol levels as a proxy for stress [62].
Few studies investigated child development. This could be the focus of future studies, as evidence suggests possible health benefits linked to reduced maternal stress during pregnancy [33] and opportunities for play and socialisation during time spent in green spaces [36,37].
Certain outcomes were not featured in the included studies. Vegetation types and structure influence their ability to regulate pollution and local climate, and thus will have differential effects on heat-related health risks [103]. Postpartum distress was examined [39], but the effects on antenatal depression or neonatal outcomes were not investigated. This is an important topic, as the greenness of the environment was associated with reduced risks of low birth weight and preterm delivery [104].

Quality of Study Designs
Overall, the level of evidence certainty for health benefits of green space quality remains low. This is due to two important reasons. Firstly, there was a high degree of heterogeneity in study designs, green space and green space quality definitions, and outcome measurements. Some studies use factor analysis to derive the qualities, which make it difficult to find out the definitions behind the derived terms, especially when the survey questionnaires were not included [40,52]. Many studies ask participants to rank certain qualities or report health outcomes on a Likert scale-type questions, without defining the quality being surveyed for the participants. These potentially introduce bias in response and are a major limitation among studies in this topic. Even with GIS methods, which are deemed more reliable and reproducible in quantifying green space exposure, variations in proximity radius and buffer zones make it difficult to compare results across studies.
Secondly, none of the included studies were randomised trials, which resulted in a lower overall quality of evidence.
Only 6/10 domains featured evidence from longitudinal cohort studies or interventional studies (before-after and quasi-experimental studies), namely the domains of environment types, natural features, infrastructures and amenities, size, perceived quality, and combination of features. Within each domain, cross-sectional and ecological studies often accounted for more than half of the evidence base. The prevalence of observational studies is characteristic of environmental health research, which faces intrinsic logistical and ethical challenges in designing rigorously controlled trials [105]. Nonetheless, observational studies have their limitations [106]. Cross-sectional surveys do not permit inference of causation. In our review, many cross-sectional surveys used convenience sampling, which could introduce selection bias due to seasonal weather, site of surveys or time of day. Longitudinal studies can factor in temporal relationship between green space exposure and health outcomes. They also enabled adjustment for factors that can influence health outcomes, such as demographic characteristics, measures of poverty and deprivation, and socioeconomic status (income, education and employment) (Table 1). However, many cohort studies in our review were nested in longitudinal health surveys that did not routinely collect data on green space quality, and only achieved so via a cross-sectional survey or geospatial analysis [36,38,39], again making it impossible to establish temporal causation. Although ecological studies echo the principles of environmental health policies, their generalizability is limited. By assuming that green space exposure applies uniformly to all individuals within a census tract or administrative area, these studies do not control for individual health and preference, and thus may lead to incorrect inferences ("ecological fallacy"). The use of multiple databases in GIS analysis, featured in many of our studies, also raises the possibility of spatial autocorrelation and mismatched data sources, etc. [34,83]. Before-after studies and quasi-experiments are pragmatic designs that support causal inference by establishing a clear temporal relationship between exposure and outcomes and controlling for confounding factors. They provide real world effectiveness of complex interventions, and are thus compatible with population policies [107].
It is important to note that, although longitudinal cohort studies and interventional studies were less prevalent, they have methodological strengths that cross-sectional and ecological studies do not. In our review, limiting analysis to these studies did not change the overall conclusion across all quality domains.

Future Directions
Innovative trial designs have been featured in this review, namely quasi-experimental studies using controlled parks or neighbourhoods [55,57]. In addition, controlled intervention design had been used in forest therapy trials, which allowed for robust pre-post measurements of cardiovascular outcomes such as blood pressure, heart rate and oxygen saturation [108]. However, high logistical demands often limited the duration of these trials and precluded studies of long-term (child development) or high-risk outcomes (childbirth, cardiovascular events). Studies nested in cohort follow-up studies [28,36,37,41,49,60,84] are a promising approach by leveraging on well-designed longitudinal studies with annual follow ups, comprehensive baseline data collection, and large sample sizes for robust statistical power. Where randomisation is not possible, study data could be analysed using interrupted times-series analysis, which adjusts for some effects of context and individual health variations over time [69].
Satellite imagery and GIS should still be part of the essential toolbox for green space quality studies, as long as GIS data is linked to patient-level data instead of being aggregated at ecological unit levels. GIS has proven useful in combining cartographical datasets, identifying and classifying land cover types. Recent advances in geospatial big data also introduced new approaches to assessing green space exposure, such as eye-level exposure (street view imagery) as opposed to overhead exposure (satellite imagery) [109]. In addition, GIS technology has enabled new indices for quantifying green space size, shape and connectivity [30,82]. By virtue of defined formulae, these indices were reproducible and reliable, and could be used in various statistical analyses.
Our findings showed that perceived green space quality, even without any judging criteria, can predict health benefits [36,37,39,66]. This is an important consideration, given that spatial environmental indicators (size, greenness, aesthetics) do not always corresponded with user perceptions [110]. Therefore, it is advisable for future studies to measure both perceived and objective quality when assessing health benefits. This approach has the dual benefits of ensuring internal validity of the subjective quality measurement, while accounting for any mediating effect of user perceptions on the objective quality [60,68].
Several studies used a composite quality score that aggregated across several domains (e.g., Public Open Space Tool). Although a composite score approach can reflect the complexity of green space quality, coverage can be restricted to attributes related to facilities, safety and cleanliness, which are shown in our review to have little association with health so far. RECITAL, the latest quality assessment index developed to address this gap, incorporates other quality domains such as suitability for activities, land cover types and biodiversity [111], which generally aligns with our classification. This index can be stratified into single-item or sub-section scores, allowing researchers to investigate specific aspects of quality, which is a shortcoming commonly associated with aggregated scales.
Comprehensive indices such as this should be explored in future studies. Last but not least, there is a need for a new index that aggregates qualities across networks of multiple green spaces of various shapes and attributes. This may be particularly salient within higher density contexts, where multiple smaller green spaces exist with each containing a small number of qualities, but larger ones that may incorporate many more qualities do not.

Strengths and Limitations
The strength of our review is its breadth of coverage, as we formulated our search strategy intentionally to capture across a range of health outcomes, potential qualities and green space types. Our review is the first to capture the diverse evidence conducted in this area and map them into domains of quality. Nonetheless, our review was not without limitations. As the concept of green space quality was not well-defined, we took a holistic approach but our review could still potentially miss out relevant studies that did not use conventional descriptors of quality. Our review only included studies written in English, and in view of more emerging research on park designs from China in recent years [112], publication bias due to exclusion of non-English articles was possible. Although our review was structured based on established protocols, the screening process was subjected to some degree of subjectivity due to a lack of standardized definitions in this topic.

Conclusions
Research on green space quality and health has increased in volume, especially since 2016. A high degree of heterogeneity was observed in study design, and the definitions of quality and outcomes measured. Environment types, vegetation types, and the size and connectivity of green spaces, were associated with physical and mental health outcomes, with differences by age and gender. The associations indicative of health benefits were more consistent in populations with more tree canopy, but not more grassland. Qualities such as safety, cleanliness and aesthetics tended to be investigated with weaker study designs. Both objective and subjective quality demonstrated positive effects on health outcomes. There is a need for more experimental studies or well-designed prospective studies that incorporate longitudinal measures of green space qualities and outcome-appropriate confounders. Green space indices should account for form, pattern, networks, and both objective and perceived qualities.   (Feng). Astell-Burt and Feng were also jointly supported by grant 1101065 from the National Health and Medical Research Council project and grant GC15005 from the Green Cities Fund-Hort Innovation Limited, with co-investment from the University of Wollongong Faculty of Social Sciences, the University of Wollongong Global Challenges initiative, and the Australian Government. The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and the decision to submit the manuscript for publication.
Institutional Review Board Statement: Ethical review and approval were waived for this study, as the review does not use personal data.
Informed Consent Statement: Patient consent was waived for this study, as the review does not use personal data.

Data Availability Statement:
The data presented in this study are available in the main article and Supplementary Files.

Conflicts of Interest:
The authors declare no conflict of interest.