Playgrounds are for children: Investigating developmentally-specific “Green Space” and child mental health

Abstract

“Green space” effects on health have been amply demonstrated, but whether specific features of green space exert differential effects remains unknown. Driven by developmental psychology theory highlighting the importance of play for young children, we investigated the association between parks with playgrounds, as a subtype of “green space”, and young child mental health. After adjusting for individual race, sex, income, and cumulative risk as well as neighborhood disadvantage, we find that children (N=95) living near parks with playgrounds have better mental health than those near parks without playgrounds ($\stackrel{\mathrm{ˆ}}{\mathrm{\beta }}=-1.22$ (95% CI: −3.82 – 1.38, p=0.36) Child Behavior Checklist scale). While not statistically significant, the observed difference, which is equivalent to 11% of the baseline mean score, is meaningful. Our results, while only suggestive, indicate that moving beyond “green space” to investigate developmentally-specific features may be worthwhile.

1. Introduction

1.1. Neighborhoods and health

That neighborhoods impact health is no longer debated [1, 2]. Features such as housing [3], social cohesion [4], air quality [5, 6], neighborhood disadvantage [7], and crime [8] are associated with health and health behaviors [1, 9, 10].

One neighborhood characteristic associated with health is green space. A large body of work documents the relationship between neighborhood “green space” and stress [11], mental health [12–16], and social cohesion [17]. Such studies have yet to illustrate whether specific features of green space exert differential effects [7, 8]. Yet, this information may inform the development of targeted public health policies [18]. In this study, we investigate the association between early childhood mental health and playgrounds in parks as a specific feature of green space.

1.2. Theory: Early childhood (ages 0–5), play, and green space

Developmental psychology theory highlights the unique needs and milestones of young children (ages 0–5). Children at these ages develop relational patterns and neurobiological functioning that influence long-term outcomes [19–21], and early childhood is a critical period for the development of social and emotional skills that provide the foundation to lifelong well-being [22]. From a life course perspective, supporting mental health in early childhood is a critical goal for public health [19, 20].

Play is one important means through which young children practice skills, build relationships, and learn autonomy [23, 24]. Biological studies illustrate play’s adaptive functions in training for unexpected events, practicing skills, and developing social flexibility [25]. Similarly, clinical research demonstrates its therapeutic effectiveness in the face of adversity and trauma [23, 26, 27] and in supporting physical and mental health [28]. Play reduces stress, including chronic stress [26]. It “prepares the child for the activities of adult life” [29] by providing a means to practice skills such as planning, self-regulation, and problem-solving in voluntary, pleasurable, and non-threatening environments [23, 25, 30].

Green space exposure is associated with a range of health and mental benefits, including reduced stress, fatigue, and anxiety [11, 12]. For children specifically, proximity to green space is associated with improved behavior [14, 31], reduced risk of mental illness [14, 15], and better attention and cognitive development [32]. Surrounding greenness is associated with lower prevalence of overweight/obesity, presumably due to more physical activity [32]. These effects have been observed for proximity to green space independent of usage of space [33].

By combining the benefits of green space with play [16, 34, 35], playgrounds within parks may offer developmentally meaningful opportunities for young children to benefit from a neighborhood environment. There thus exists the potential for public health to leverage playgrounds as a place-based mechanism for promoting mental health in early childhood. To our knowledge, no studies to date have examined the association of playground access with early-childhood mental health.

1.3. The present study

The present study tests the hypothesis that access to parks with playgrounds, as a type of “green space”, offers benefits to young children over and above parks alone. We leveraged a unique, multi-level dataset created by merging individual level measures of child mental health and risk with contextual data on neighborhood parks and playgrounds to test this hypothesis.

2. Methods

2.1. Study population

Participants were a community sample of children aged 36–39 months living within walking distance of a park in the city of Seattle (n=95). Participants were recruited as part of a larger, longitudinal investigation of child mental health [36]. Children and their mothers needed to be proficient in English and not be diagnosed with a developmental disability to participate.

2.2. Measures

Independent variables

A binary variable indicated the participant’s access to a playground (0 = participant lived within walking distance of a park without any playground; 1 = participant lived within walking distance of a park with a playground). Parks and playgrounds were identified using data from the City of Seattle’s Parks and Recreation Department [37, 38]. Distance was calculated in primary analysis as distance in meters “as the crow flies” between the participant’s home address and the center of the park or playground using the R package “sf”[39]. Walking distance was defined as within 400 meters, roughly a quarter of a mile [40]. In sensitivity analysis, we calculated distance using street connectivity using “ggmap” [41] and Google API as well as alternate specifications of walking distance (Supplement). All geospatial manipulation was performed in R 4.0.2 [42].

Dependent variable

Child mental health was assessed using mothers’ report of children’s internalizing and externalizing symptoms on the Child Behavior Checklist (CBCL, 4–18 years [43]) 9 months following study enrollment. Mothers rated the presence of each symptom on a 0–3 response scale. For our analysis, we used the raw overall adjustment score (alpha=0.78) calculated as an average of internalizing (alpha=0.83) and externalizing (alpha=0.82) scales which were correlated 0.36.

Control variables

Family cumulative risk was calculated as the sum of eight factors. Annual income was converted to a categorical variable with 14 levels. Child race and sex were reported by parents at baseline. Because the numbers of participants representing each racial and ethnic group were relatively small, we coded race/ethnicity as a dichotomous variable. The Neighborhood Deprivation Index (NDI) was calculated from 2010 census data using the packages “tidycensus”[44], “psych” [45], and command “ndi” [46]. The “ndi” command performs principal components analysis to summarize variables previously associated with health outcomes, following methodology by Messer et al [47]. See Appendix A for detailed information.

2.3. Analytical methods

Stepped linear regression models were used to investigate whether parks with playgrounds demonstrate positive associations with child mental health over and above parks alone. The first model included only the playground variable. The second model added individual control variables (cumulative risk, income, race, and sex), and the third added NDI. Differences in the playgrounds coefficient between the three models indicate that the raw association was confounded by covariates. Due to a lack of significant geographic clustering and no repeated measures, random effects models were rejected in favor of ordinary least squares linear models.

2.4. Secondary analysis

Since our sample includes only Seattle residents within walking distance of a park, we anticipated that our modest sample size (n=95) might yield inadequate power to identify a significant effect. To investigate sample size and power, we created replications of each participant 3, 5, and 10 times to produce simulated samples of 285, 475, and 950. As simple replications of the original dataset, the association between independent and dependent variables in the simulated dataset was identical to the original dataset. We then ran our model on these simulated samples to test whether statistical significance could be reached in larger samples.

3. Results

In the full sample, 55.1% of children were female and 62.5% were White (Table 1). On average, compared to participants near a park with no playground, participants near a park with a playground were slightly more likely to be male, non-White, lower income and greater cumulative risk. On average, census tracts with a playground had higher neighborhood disadvantage scores. The distribution of child adjustment scores differed slightly between the two groups, with children living near a playground having fewer adjustment (CBCL) problems on average (Figure 1).

Table 1:

Sample characteristics, stratified by exposure (park with or without a playground)

	Full sample (n=99)	By exposure (playground)
		No (n=40)	Yes (n=59)
% female	55.1%	60.0%	51.2%
% White	62.5%	72.5%	55.4%
Income category [mean(SD)]	8.2 (4.4)	8.9 (4.03)	7.72 (4.58)
Cumulative risk [mean(SD)]	0.83 (0.72)	0.77 (0.63)	0.87 (0.78)
NDI [mean(SD)]	0.03 (0.83)	−0.06 (0.86)	0.09 (0.81)
CBCL [mean(SD)]	11.2 (6.4)	11.5 (6.6)	11.1 (6.4)

^#unique census tracts: 56

Income is a categorical variable that ranges from 0–14, where higher numbers indicate higher income. Cumulative risk and NDI are both continuous variables where higher values indicate greater disadvantage.

SD – standard deviation

Figure 1.

Distribution of child adjustment problems score, stratified by exposure (park with or without a playground).

Regression results are summarized in Table 2. Regression coefficients suggest that, among children living near any park and adjusting for individual and contextual covariates, those with a playground have 1.22 (CI: −3.82, 1.38) fewer overall adjustment problems than those without, though results do not reach statistical significance. The observed difference of 1.22 points on the CBCL attributable to playground access is equivalent to 11% of the baseline mean score, or nearly 0.2 standard deviation units.

Table 2:

Regression results of separate OLS regressions models comparing participants near parks with or without a playground feature

	Overall child adjustment problems
Predictors	Estimates	Estimates	Estimates
Park has a playground feature	−0.39 (−3.07, 2.30)	−1.02 (−3.62, 1.59)	−1.22 (−3.82, 1.38)
Sex (male)		0.90 (−1.65, 3.44)	0.86 (−1.67, 3.38)
Income		−0.14 (−0.51, 0.24)	−0.16 (−0.53, 0.22)
Race (White)		0.22 (−2.64, 3.08)	0.27 (−2.57, 3.11)
Cumulative risk		1.89** (0.18, 3.60)	1.70* (−0.02, 3.42)
Neighborhood Disadvantage Index			1.21 (−0.34, 2.76)
Observations	95	93	93
R2	0.001	0.119	0.143

The referent group for sex is female and for race is non-White. Income is a categorical variable that ranges from 0–14, where higher numbers indicate higher income. Cumulative risk and NDI are both continuous variables where higher values indicate greater disadvantage.

^*p<0.1

^**p<0.05

^***p<0.01

Results of our secondary analyses indicate that point estimates would reach statistical significance at the 0.05 level with a sample of 475 participants. Sensitivity analysis using street connectivity to measure distance produced similar results ($\stackrel{\mathrm{ˆ}}{\mathrm{\beta }}:-1.36$, 95% CI: −4.05, 1.32).

4. Discussion

To our knowledge, our study is the first to test whether playground access appears associated with child mental health over and above “green space” parks. Our results provide preliminary support for this hypothesis, which is supported mechanistically by the developmental psychology literature. A positive association between playground access and child mental health would have clear policy implications. Healthy social and emotional development during early childhood is a critical public health goal [19, 20]. Supporting child mental health is consistent with the public health mandate of prevention. Public investment in outdoor play areas may have important public health impact for moderate cost.

Our unique, multi-level dataset included important neighborhood and individual factors, such as family cumulative risk and neighborhood deprivation, and is the first to our knowledge that simultaneously captures “green space” features and child mental health.

Our study is limited by its sparse information on certain key factors, cross-sectional nature, and modest sample size. At this time, alternative explanations due to confounding cannot be ruled out. The observed relationship may reflect other features correlated with playground access, such as daycare quality on the neighborhood level or social capital on the individual level. Similarly, our findings may be explained by reverse causation, or that parents who prioritize child mental health self-select into neighborhoods with playground access. Future research may incorporate these considerations.

Data on specific green features was sparse. Standard methods for measuring green space (ex. the normalized difference vegetation index, or NDVI [48]) do not capture features such as playgrounds or quality. Locally maintained data that collect this information, such as the City of Seattle’s Parks and Recreation dataset used for this study, have limited coverage. This limitation led us to focus our analysis only on participants within Seattle who had access to a park, which reduced our sample size with implications for study power. However, our secondary analysis indicated that, if the observed effect size holds, a future study of under 500 individuals may have sufficient power to detect a significant effect.

While our dataset provided information on children’s access to a park or playground, we lacked data on whether children used these amenities and their frequency or intensity of use. This information would enhance our investigation by permitting us to determine whether our observed effect can be explained by the mechanism of young children engaging in the developmentally crucial act of playing.

Future research will require novel data resources to address these limitations. These may be developed via partnerships with local governments to document features of public parks or by using machine learning to categorize online images of green space [49]. Merging multiple administrative datasets may provide the multidimensional covariates needed to rule out confounding. Incorporating questions on playground use into existing panels could be a cost-effective alternative to primary data collection, while a natural experiment study based on residential mobility could potentially identify causal links.

In summary, this study tests an innovative hypothesis about health and place by applying developmental psychology theory to a public health problem. Our unique, multi-level dataset incorporates detailed measures of child mental health, a specific “green space” feature, and key individual and contextual confounders. By focusing on playgrounds, we highlight a feature of the built environment that is modifiable and offers the potential for positive public health impact. Our results, while suggestive, indicate that moving beyond “green space” to investigate developmentally-specific features may be worthwhile.

Table 3:

p-value and confidence intervals for simulated datasets

Original analysis (“crow flies” distance ≤ 400m)			Secondary analysis (street connectivity ≤ 10 minute walk)
Sample size	p-value	95% CI	Sample size	p-value	95% CI
Original sample			Original sample
95	0.355	$\stackrel{\mathrm{ˆ}}{\mathrm{\beta }}=-1.22$ (−3.82, 1.38)	97	0.315	$\stackrel{\mathrm{ˆ}}{\mathrm{\beta }}=-1.36$ (−4.05, 1.32)
Simulated datasets			Simulated datasets
285	0.099	(−2.66, 0.23)	291	0.073	(−2.86, 0.13)
475	0.032	(−2.33, −0.10)	485	0.020	(−2.51, −0.21)
950	0.002	(−2.00, −0.43)	970	0.001	(−2.17, −0.56)