Nature and Children’s Health: A Systematic Review

This review followed Preferred Reporting Items for Systematic Reviews procedures and guidelines¹⁹  and was registered in the International Prospective Register of Systematic Reviews systemic review database (Registration number: CRD42018087552).²⁰ 

We developed a comprehensive search strategy of relevant databases in consultation with a University of Washington librarian. The electronic database search was initiated in PubMed by using the first iteration of the search terms; subsequent searches were conducted in Cumulative Index to Nursing and Allied Health Literature, PsycInfo, ERIC, Scopus, and Web of Science databases. In all searches (Supplemental Information 1), the first element included nature terms (eg, green space) and the second element included child health outcome terms (eg, physical activity). Broad search terms were selected, and truncated terms and several variations of the same term were included to capture all relevant articles. Because different databases use different search methods, a revision of the initial search string was often required to fit each database’s search method to maximize results. We excluded articles that (1) included only participants 19 years of age or older, (2) involved nonhuman subjects, (3) were written in languages other than English, (4) were systematic literature reviews, (5) lacked text describing natural environments, (6) had exposures that combined nature contact and a nonnature environmental factors, (7) examined outcomes not related to child health, or (8) the exposure was measured prenatally (Supplemental Information 2). The database search was completed on February 23, 2021.

Initial study screening and selection was performed by using Covidence, a Web-based platform for managing systematic literature reviews that facilitates removing duplicates, screening citations for relevance, abstract and full text review, study extraction, and risk of bias assessment.²¹  We established criteria for study inclusion and exclusion by consensus. Criteria are specified in Supplemental Information 2. All titles and abstracts of selected studies were screened by 2 authors to determine whether to include or exclude for full text review. If a study received conflicting votes or if the reason for exclusion differed between the 2 reviewers, consensus was reached through discussion with a third author.

All studies retained after the title and abstract screening phase then underwent full text review by 2 authors. The full text review was the final step for determining if a study would be included or excluded from the final literature review. If the reviewers disagreed or if the reason for exclusion differed between the 2 reviewers, consensus was reached through discussion with a third author.

All studies selected for inclusion in the full text review were independently evaluated for quality by 2 authors using the Mixed Methods Appraisal Tool (MMAT), a reliable and valid quality assessment tool for multiple study designs including quantitative randomized controlled trials (RCTs) and quantitative nonrandomized trials (eg, experimental, observational, etc).²²  The MMAT includes 5 items, with different questions for RCTs and quantitative nonrandomized study designs. We did not generate an overall MMAT score to define high- or low-quality studies because we deemed an overall score alone is insufficient to identify methodologic strengths and weaknesses. Instead, in accordance with recommendations from the MMAT developers, we examined which MMAT questions met criteria to receive a “yes,” “can’t tell,” or “no.” MMAT questions were stratified by study design and “yes” answers indicated a more rigorous methodologic approach. Consensus for quality assessment was reached between 2 reviewers.

In reviewing the Cochrane tools for bias assessment, Risk Of Bias In Non-randomized Studies of Interventions (ROBINS-I) for nonrandomized designs and Risk of Bias 2 (RoB 2) for RCTs,^23,24  we determined these tools were excellent but not feasible given their extensive criteria for the large number of articles included in our review (n = 296). Thus, we examined other tools that rigorously identified common and relevant sources of bias and had established reliability and validity. We had originally intended the MMAT to be used only to assess study quality.²²  On review, we recognized that many of the MMAT questions assessed relevant areas of bias. For example, appraising whether participants were representative of the target population assesses selection bias, and evaluating whether outcome data are complete assesses attrition bias (see Supplemental Table 6 for detailed description). This approach has been used in other recent systematic reviews.^25,26  All studies were reviewed for bias by 2 reviewers using the MMAT.²²  Risk of bias was scored as low (“yes” on MMAT, indicating higher quality), unclear (“can’t tell” on MMAT, indicated uncertainty in quality), or high (“no” on MMAT, indicated lower quality). Consensus was reached by at least 2 reviewers for all articles.

We used a previously published framework to categorize nature exposures²⁷ : gardening, green space activity (eg green space not near residence or school), residential green space, school green space, greening intervention, wilderness experience, nature walk, and “other.” Studies in which researchers evaluated mixed spaces were categorized as other. Measures of nature exposure included self or proxy-report measures, intervention-associated exposures, and objective measures such as land-use data and Normalized Difference Vegetation Index, an approach used to estimate the density of green on an area of land.²⁸  Studies with multiple nature exposures were included; each exposure was included independently.

We grouped health outcomes and behaviors into 7 categories: (1) physical activity; (2) cognitive, behavioral, and mental health; (3) BMI; (4) cardiovascular and metabolic measures; (5) asthma and allergy; (6) academic and learning; and (7) other. For mental health outcomes, we included a range of outcomes from psychological well-being to mental health diagnoses.²⁹  We limited our academic and learning outcomes category to those most related to classroom performance. Studies with multiple outcomes were included; each outcome was included as an independent outcome.

We used standard definitions of study designs, to maintain consistency across reviewers.³⁰  Because there is no accepted standardized metric that formally summarizes strength of evidence for systematic literature reviews, we relied on the clinical and methodologic expertise of our authorship team to determine the variables critical to assessing methodologic strength of evidence. The a priori variables included (1) number of “yes” responses on the MMAT, (2) number of studies within each nature exposure or health outcome category, (3) sample size (n) within each study, (4) strength of association, (5) consistency in results, (6) variability in outcomes, and (7) study design. Strength of evidence was determined by consensus by the full authorship group and ranked on an eight-point scale: low (−), low, low (+), moderate (−), moderate, moderate (+), high (−), or high. Final strength of evidence classification indicates the strength of a positive association between nature and health in childhood (ie, health promoting).

Our search yielded 10 940 studies for title and abstract screening; 296 were selected for inclusion (Fig 1). Among included studies, the majority were observational (n = 225, 76%), and of these, 204 (69%) were cross-sectional. Only 13 (4%) were RCTs. Half of the studies examined the presence of residential green space (n = 147, 50%) as the nature exposure. The most common health outcomes and behaviors were physical activity (n =108, 32%); cognitive, behavioral, or mental health (n = 85, 25%); and BMI (n = 45, 13%) (Fig 2). No strength of evidence summary is reported for studies in the “other” nature exposures^31–51  or health outcomes and behaviors^{33,40,52–82}  categories because of high variability in exposures and outcomes evaluated.

Overall, observational studies (n = 225) had the lowest risk for bias (Fig 3). Although 44% were at high risk of selection bias because of participants not being representative of the target population, there was low risk of bias in all other bias categories evaluated. For experimental studies (n = 58), 95% of studies were at high risk of selection bias because of participants not being representative of the target population and 55% of studies were at high risk of confounding bias. For RCTs (n = 13), 46% had a high risk of selection bias because of groups not being comparable at baseline; 62% had a high risk of detection bias because of outcome assessors not being blinded to intervention assignment.

To assess methodologic limitations and opportunities to strengthen evidence in future studies, we summarized the results of studies meeting criteria for a “yes” (more rigorous) categorization for each MMAT question and stratified results by study design (Tables 1 and 2). For the RCTs (n = 13), no studies had outcome assessors that were blinded to group allocation, and 31% (n = 4) of studies had groups that were comparable at baseline. For experimental studies (n = 58), 2% (n = 1) included participants representative of the target population and 33% (n = 19) adequately accounted for confounding. Observational studies (n = 225) had similar challenges. Only 48% (n = 108) enrolled participants representative of the target population and 79% (n = 178) appropriately measured the of outcome of interest. In most studies, regardless of design, the intervention (or exposure in observational studies) was administered as intended.

Studies of residential green space were largely observational (97%, n = 143), and approximately half used geographic information system mapping as an objective assessment of nature exposure (Table 3).^* Overall, n = 86 (59%) reported positive associations between nature exposure and improved health, n = 50 (34%) reported mixed or null results, and n = 11 (7%) found negative associations. Most of the negative associations were reported for asthma and allergy-related conditions (n = 8, 5%), which had considerable diagnostic heterogeneity (eg, conjunctivitis, asthma, allergy, etc) and type of allergen (eg, grass, trees, etc). Physical health outcomes and behaviors were most common (n = 110, 75%), followed by mental health (n = 30, 20%) and academic or learning outcomes (n = 7, 5%). Researchers in the single RCT found that increased residential park density was associated with a decrease in BMI z scores in overweight rural youth (8–14 years).⁹³ 

Studies evaluating school green space were 90% observational and 10% experimental (no RCTs) (Table 3).^† Researchers of 1 large longitudinal cohort study (n =94 997) reported that children attending schools in neighborhoods with greater green space density (eg, parks) had improved fitness during the school year.²²⁴  Overall, associations between school green space and child health were mostly positive (n = 33, 83%); n = 4 (10%) had null or mixed findings and n = 3 (7%) reported negative associations. The most common outcomes were physical health (n =22, 55%), academic or learning outcomes (n = 12, 30%), and last, mental health (n = 4, 10%).

For general green space activity, 80% of the studies were observational (Table 3).^73,250–283  Most associations were positive (n = 19, 54%), n = 16 (46%) found null or mixed results, and no negative associations were reported. The most common outcomes were physical health (n =27, 77%), then mental health (n = 6, 17%), and last, academic or learning outcomes (n = 2, 6%). The single RCT reported that preschool-aged children attending a 10-week nature program reported lower stress and changes in fecal serotonin and gut microbiome diversity.⁷³ 

Wilderness experience (n =  11)^{58,59,284–291}  and gardening (n = 12)^{52–56,72,77,79,292–295}  were both categorized as moderate (−) (Table 3). Although there was strong and consistent evidence supporting a positive impact of nature exposure on health outcomes and behaviors in these categories, both contained a limited number of studies with additional noteworthy methodologic weaknesses. All studies evaluating wilderness experience were experimental but generally had small sample sizes (median n = 58), and only 27% accounted for confounders. Gardening studies had wide variation in the types of outcomes, making it difficult to assess consistency in results; 84% were experimental or RCTs and 42% reported null or mixed results. Two RCTs randomized schools to gardening interventions; positive associations were found for school gardens increasing moderate-to-vigorous physical activity (accelerometry)²⁹⁵  and science knowledge.²⁹³ 

Classroom interventions (n = 17)^{233,296–311}  were categorized as low (+) strength of evidence (Table 3). Greening interventions (n = 9)^{57,68,312–318}  and nature walks (n = 4)^319–322  were both categorized as low strength of evidence primarily because of the small number of studies within each category. Most (77%) of the greening intervention studies were experimental or RCTs; 56% reported null or mixed results. All 4 nature walk studies were experimental and reported positive associations, but median sample size was small (n = 42) and none adjusted for appropriate confounders. The single RCT in the greening intervention category reported decreases in depressed affect, antisocial behavior, and moderate-to-vigorous physical activity (accelerometry) after renovating outdoor preschool spaces to include more natural elements.³¹³  Although the authors noted that lower moderate-to-vigorous physical activity was unexpected, activity was measured only during the scheduled outdoor period (20-minute interval), and the reduction was small (1.32 minutes, SE = 0.37, P < .001).³¹³  In 1 RCT in the “other” category, researchers found that psychological well-being was improved for adolescents in an outdoor versus indoor space after completing stressor tasks³⁷ ; researchers in another reported that views of green landscapes versus rooms without windows improved attention performance and stress recovery in high school students.⁴⁰ 

When evaluating the strength of evidence stratified by health outcome (Table 4), we incorporated all outcomes including when multiple health outcomes and behaviors were reported within individual studies. Thus, our final sample size (n = 343) is larger than the total study number (n = 296).

Researchers in most studies evaluating physical activity were observational (n = 88, 81%), used objective measures such as accelerometry (n = 59, 55%), and reported positive associations (n = 71, 66%).^‡ Null or mixed results for physical activity were found in n = 33 studies (31%), and n = 3 studies (3%) reported negative associations. The most common nature exposures for physical activity were residential green space (n = 45, 42%) and general green space activity (n = 25, 23%). The one RCT in this category found that low-income elementary schools randomly assigned to gardening interventions reported children had increased moderate-to-vigorous physical activity (accelerometry) compared with control schools.²⁹⁵ 

Of the studies evaluating cognitive, behavioral, and mental health, 58% were observational (n = 49), but 37% (n = 31) were experimental and 6% (n = 5) were RCTs.^§ Positive associations were found in 71 studies (83%), and 12 studies (15%) reported mixed or null results; 2 studies (2%) reported negative associations. Within this broad category, attention and depression were the most common outcomes. Studies revealing positive associations between nature contact and mental health most commonly used the reliable and valid Strengths and Difficulties Questionnaire.³²⁵  The primary nature exposure used to assess cognitive, behavioral, and mental health outcomes was residential green space (n = 34, 42%). Most wilderness exposure studies also evaluated behavioral and mental health (n = 8, 10%). All 5 RCTs reported positive associations with well-being,⁵⁷  mental fatigue recovery,⁴⁰  cognitive tasks and emotional status,⁴⁹  performance tasks and perceived restorativeness,³²²  and perceived stress.⁷³ 

Studies evaluating childhood weight status were categorized as having only moderate (−) strength of evidence because most were observational (n = 41, 91%).^ǁ Roughly one-third (n = 16, 36%) reported positive associations, but more than half (n = 27, 60%) reported mixed or null results and n = 19 (42%) were null. Two negative associations were reported (4%). Residential green space was the most common nature exposure evaluated (n = 34, 76%). Of the 2 RCTs, researchers in 1 found that increased residential park density, as part of a larger behavioral weight management intervention that included dietary education, was associated with a decrease in BMI z scores in overweight rural youth (8–14 years).⁹³  In the other, researchers reported null results for the effect of a school-based gardening intervention on BMI.²⁹⁴ 

Academic and learning outcomes (n = 27),^¶ asthma and allergy outcomes (n = 26),^# and cardiovascular and metabolic outcomes (n = 16),^** were all categorized as low (+) principally because of the small number of studies within each category. Academic and learning outcomes varied widely with the most common outcome being grades and test scores (n = 14, 52%). Many of these studies reported positive associations (n = 22, 81%), n = 5 (19%) reported mixed or null results, and one study reported that green space was negatively correlated with academic achievement.²¹⁷  Researchers in both RCTs found no difference in achievement for children in outdoor compared with indoor settings.^293,301  Studies in the asthma and allergy category were all observational. Positive associations were reported for n = 13 studies (50%); negative associations were found in n = 8 (31%) studies, indicating that green space may exacerbate allergies or asthma in certain scenarios in childhood. For cardiovascular and metabolic outcomes, n = 7 (44%) of studies reported positive associations and n = 9 (56%) reported null or mixed results. Most were observational (n = 9, 56%) and focused on blood pressure (n = 8, 50%). The remaining studies reported heterogenous outcomes. All 3 RCTs reported null results for the associations between dish gardens in hospitalized patients and blood pressure,⁴⁹  school-based gardening intervention and blood pressure,²⁹⁴  and time in outdoor and indoor environments for heart rate and heart rate variability.³⁷ 

Authors in 15 studies reported negative associations between nature exposures and health outcomes or behaviors, primarily in residential green space (n = 11, 73%) (Table 3). Most negative associations were reported for asthma and allergy-related conditions (n = 8, 53%), which had considerable heterogeneity both for diagnosis (eg, conjunctivitis, asthma, allergy, etc) and for the type of allergen (eg, grass, trees, etc) (Table 4).^†† Three studies (3%) evaluating physical activity reported negative associations: living in an urban green neighborhood reduced walking or bicycling,¹⁴⁵  renovation of an outdoor space at a preschool decreased moderate or vigorous physical activity,³¹³  and living in greener neighborhoods was associated with higher inactivity.²²⁵  Researchers in 2 studies (2%) examining cognitive, behavioral, or mental health reported negative associations. In one, more green space play time was associated with increased behavioral difficulties,²¹⁴  and increased neighborhood parks or playgrounds was associated with more depressed feelings and mood.¹⁷⁴ 

Our systematic review included a substantial body of literature, covering a wide range of nature exposures and outcomes, and most often found positive associations between nature exposure and children’s health. The evidence for an association between nature and health is currently most compelling for outcomes related to physical activity and cognitive, behavioral, and mental health. For nature exposures, the evidence was strongest for presence of green space near the child’s home, school, and for activity done in green spaces, highlighting the importance of ready access to nature. We believe our review warrants pediatricians, public health practitioners, and policy makers claiming a solid and growing body of evidence for policies and practices that promote environmental justice and equitable nature contact for children where they live, play, and learn.

Although outdoor time is associated with health benefits and daily outdoor play is recommended for children,²  we excluded studies that only reported outdoor time and did not specifically describe nature contact. Although focusing only on outdoor time is simple, all outdoor spaces are not comparable: a parking lot is not a park, an urban playground without natural elements is not a garden. The purpose of this review was to assess the literature specifically focused on contact with the natural world for children’s health. We evaluated the full breadth of quantitative evidence with a systematic search and consensus-based review process.

Although many types of nature exposures were studied, most researchers examined nature or green space contact close to children’s homes or schools, with moderate (+) to high (−) strength of evidence for these nature exposures. Most studies focusing on residential green space used land-use data or Normalized Difference Vegetation Index and found positive associations with many physical and mental health outcomes for children. Although there were fewer studies focused on school green space, the strength of evidence was relatively strong because of consistently positive associations for outcomes such as physical activity, weight status, and test scores. These findings are supportive of greening schoolyards to increase access to daily opportunities for children to move and play in nature-rich settings located in their neighborhoods.³²⁸  The strength of evidence was low to moderate for other nature exposure categories primarily because of the low number of studies and small sample sizes.

A consistent limitation related to nature exposure was that researchers did not measure access to, quality of, or actual use of these parks or green spaces. Researchers in previous studies have found that park proximity alone is insufficient to predict use, and characteristics such as amenities and programming also need to be considered.^329,330  Future research on park proximity and green space access would be of greater value if investigators incorporated measures of actual use of the space. In addition, many studies did not describe the natural features of the exposure. If future studies differentiated between natural characteristics such as green space versus “blue space” (eg, water features) or exposure to tree canopy versus smaller plants or grass, it might help delineate which types of nature contact, and how much, are most impactful.³³¹ 

Of the specific health outcomes and behaviors, compelling evidence was found for most studies focused on the relationship between nature contact and physical activity, a health behavior with numerous known benefits for children including healthy weight and cardiorespiratory fitness.³³²  Although outdoor time has been found to be a consistent correlate of children’s physical activity, existing literature does not typically focus on the independent contribution of nature-rich outdoor spaces. Studies in adult populations support the concept of “green exercise;” physical activity in natural environments is associated with health benefits beyond physical activity, particularly mental health.^333–335  Although there were several studies of childhood overweight and obesity that reported positive associations with nature contact, most reported null findings. Given the global challenges related to physical inactivity,³³⁶  obesity,³³⁷  and poor mental health, our review lends support to increasing equitable access and use of parks and green spaces.

We also found a strong evidence base for the relationship between nature contact and children’s cognitive, behavioral, and mental health. Most studies found a positive association for a range of outcomes, including attention and mood, and a few researchers examined longitudinal outcomes for >6 months. Studies included participants from preschool-aged to adolescents, and nature exposures close to home and in the wilderness. This not only highlights the diversity of the current literature but also suggests a range of potential mental health benefits from nature contact. Our findings are consistent with those of another recent review that was focused only on mental health outcomes in adolescents and found beneficial associations between green space exposure and numerous mental health outcomes.³³⁸ 

There are a number of methodologic limitations inherent to our review. First, there was wide variability across studies in participant samples, nature exposures, and outcome measures, which made it difficult to aggregate and draw meaningful conclusions on the state of the evidence for certain categories. Second, risk of selection bias was moderate for observational studies and RCTs but high for experimental studies; experimental studies had moderately high risk for confounding bias. Third, most of the literature were observational. Although causality cannot be rigorously assessed, we found a low risk of confounding bias for observational studies. Fourth, because the nature exposures and health outcomes and behaviors studied were evaluated by using many different methodologies and instruments, we attempted to construct a measure of overall strength of evidence for each category of nature exposure and health outcome. Although we used a priori criteria and consensus from the entire authorship group, we acknowledge that this type of collective judgment is not wholly objective. Fifth, we used a reliable and validated tool, the MMAT, to assess study quality and bias. Although the MMAT is suited for quantitative study designs, we acknowledge that this approach also requires some subjective decision-making. Sixth, the literature search included articles published in English; thus, it is possible that relevant articles have been missed.

The coronavirus disease 2019 pandemic and attention to the health consequences of systemic racism have highlighted the importance of equitable access to safe, outdoor places.^339–341  Pandemic-related school closures and social distancing recommendations restricted opportunities and spaces for children to play. Although some communities had access to green spaces, others have not, largely because of historical environmental injustices that resulted in disparities in access to nature. Families of color and those living in poverty have less access to nearby parks and green space compared with those more privileged^342,343 ; a recent study found that park access increased the odds of outdoor activities in children.^344,345  A notable limitation of extant literature in our review is that few studies examined the impact of nature exposure on marginalized communities most at risk for inequitable access to green space and health disparities. This limitation is particularly important because of “equigenic effects”: the idea that an environmental factor (eg, nature contact) could not only disrupt the existing relationship between socioeconomic disadvantage and health but could have a greater effect on those most disadvantaged.^15,16 

The available evidence reveals a positive relationship between nature contact and children’s health. Although there is a need for higher quality evidence, the total weight of the current evidence supports advocacy for practices and policies that address issues of equitable access to nature. Pediatric health care providers, legislators, and the public need to recognize that nature contact, especially access close to children’s homes, is a potent environmental and social determinant of health with the potential to prevent disease and promote health equity in children. Along with access to safe and high-quality education, health care, housing, neighborhoods, water, and food, the opportunity for children to access nature-rich environments must now be designated a public health priority.

We thank Kelley Scholz for her contributions to this literature review.

MMAT

Mixed Methods Appraisal Tool

RCT

randomized controlled trial

RoB 2

Risk of Bias 2

ROBINS-I

Risk Of Bias In Non-randomized Studies of Interventions