Longitudinal associations between neighborhood park and open space access and children’s accelerometer-assessed physical activity: the evidence from the MATCH study

Abstract

Background:

Cross-sectional studies have shown positive associations between neighborhood park access and children’s physical activity (PA); however, research that examines the relationship longitudinally is lacking. This study investigates how neighborhood park access affects the longitudinal trajectory of PA in 192 children across three years.

Methods:

Accelerometer-assessed PA data of children (N=202) were collected across six semi-annual waves (seven days each) between 2014-18. Geographical Information Systems was used to measure neighborhood park access (i.e., coverage, density, and proximity) at baseline. Mixed-effects models examined associations of park access with children’s baseline and trajectory of moderate-to-vigorous PA (MVPA) minutes across three years, and whether associations differed by sex or weekends vs. weekdays.

Results:

Higher neighborhood park density and coverage were positively associated with children’s baseline MVPA min/day. Longitudinally, higher park density was associated with smaller decreases in children’s MVPA min/day, but not coverage. Park proximity was not associated with baseline or change in MVPA min/day. The above associations did not differ by sex or weekdays vs. weekends.

Conclusions:

Having access to more neighborhood parks protected against age-related declines in children’s PA. These findings suggest that neighborhood park density should be considered by urban planners when evaluating health impacts of their policies.

1. Introduction

Physical inactivity contributes to risks for obesity, cardiovascular disease, and all-cause mortality outcomes.¹ The United States National Physical Activity (PA) Guidelines indicate that substantial health benefits can be achieved in children (ages 6 through 17 years) by engaging in moderate-to-vigorous physical activity (MVPA) for at least 60 or more minutes each day.¹ However, objectively measured PA data show that only 22% of children currently meet this guideline.²

The built environment is a relevant domain that influences PA behavior, aside from individual and interpersonal factors.³ Of all built environment features, neighborhood parks and open space have been frequently examined due to their roles in providing exercise facilities and venues for leisure-time activities (e.g., team sports, unstructured play).⁴ Compared to adults, children and adolescents typically have access to fewer PA resources (e.g., rely on parental transport to sports facilities, gym age limits) and rely more on public open spaces (e.g. parks) to be active.⁵ Thus, their PA levels over time may be more influenced by the ability to access parks in their surrounding neighborhood compared to adults.⁶

A host of studies have shown positive associations between neighborhood park access measures, including proximity (i.e., shorter distance to the nearest park), density (i.e., higher total numbers of parks), availability (i.e., the presence of a park), coverage (i.e., higher amount of park areas), and size (i.e., larger sizes of parks) and higher PA in children.^7–11 However, due to their cross-sectional nature, these studies are limited in providing evidence that living near neighborhood parks can buffer against declines in children’s PA that typically come with age.¹² Additionally, results might be biased by the possibility of residential self-selection. That is, families who are more physically active may choose to live in the neighborhood with better access to neighborhood parks.¹³

Limited research has examined neighborhood park access as a predictor of children’s longitudinal trajectory of PA. A six-year longitudinal study of 214 adolescents in Denmark reported an increase in moveability index score (i.e., a composite measure of opportunities for PA of children including parks) between baseline and follow-up was associated with a reduced decrease in mean daily PA minutes.¹⁴ Another four-year European-based longitudinal study of 2488 children and adolescents aged 3 to 15 found that public open space availability predicted a slower decline in MVPA minutes per day.¹⁵ However, the former examined effects of residential environment changes on adolescent’s change in PA, while only 5% of subjects in the latter study provided repeated accelerometer measures across three data collection periods to allow accurate longitudinal trajectory modeling. Until today, no study has examined associations between long-term effects of multiple neighborhood park access measures on children’s PA changes over time across middle childhood, during which age-related decline occurs.

Several recent studies have examined pre-post changes in children’s PA after gaining access to neighborhood parks through natural experiment design.^16–18 For example, a study in Chattanooga, Tennessee, reported that a policy to increase the numbers of parks/recreational sites in 2010 led to more than doubled the odds of engaging in MVPA for their children/youth participants in 2014 compared with their 2010 counterparts.¹⁷ Although these studies suggest that improved neighborhood park access may lead to increased PA, they cannot be used to understand children’s long-term PA vulnerabilities in neighborhoods where dramatic built environment or policy changes are not possible. In this situation, if lack of park access is shown to be a risk factor for faster rates of decline in children’s PA, then other targeted educational or programmatic interventions may be necessary.

To address these gaps, the current study examined the association of neighborhood park access with changes in PA in children over three years. We aimed to examine whether neighborhood park access (i.e., density, coverage, and proximity) were associated with baseline and longitudinal trajectories of children’s PA and whether such associations differed by sex and on weekdays versus weekends. Since previous studies have reported age-related declining trends in PA across childhood,¹⁹ we hypothesized that children who lived in neighborhoods with higher park density (i.e., higher number of parks), higher park coverage (i.e., higher proportions of parklands), and lower park proximity (i.e., shorter distances to closest parks) would show slower declining trends in PA as they got older. Also, studies have shown that girls’ longitudinal PA levels may benefit more from park access,^14,15 and that the availability of recreational facilities is more closely tied to children’s PA on weekend days.²⁰ Therefore, we hypothesized that these associations would be stronger in girls and during weekends.

2. Methods

2.1. Study Participants

Participants were part of the Mothers’ and Their Children’s Health (MATCH) study, which recruited a longitudinal cohort of mother-child dyads and focused on how maternal behavior and stress affect children’s obesity risk. An ethnically- and racially-diverse group (N=202) (56.8% Hispanic) of mothers and children (age 8-12 years at baseline) were recruited and enrolled from the Los Angeles area. Data were collected at six semi-annual waves spaced approximately six months apart between 2014-18. The MATCH study received ethical approval from the Institutional Review Board of the University of Southern California. Additional details on study design, including recruitment procedures, inclusion/exclusion criteria, and data collection procedures, can be found in Appendix 1 and the study protocol paper.²¹

2.2. Measures

2.2.1. Physical Activity

A GT3X Actigraph accelerometer was used to measure physical activity (PA). Children were instructed to wear the accelerometer on their waist for seven consecutive days at each wave. The accelerometer was worn all the time except sleeping, bathing, or swimming on the right-hip with an adjustable belt attached.²¹ Accelerometer data were set up to collect body movement data in activity count units for each 30-second epoch. Non-wear times (i.e., >60 continuous minutes of zero activity counts) and non-valid days (i.e., < 10 hours of wearing time) were removed from analyses.²¹ The study used age-specific thresholds for children to identify moderate-to-vigorous physical activity (MVPA; Freedson prediction equation above 4 METS) and sedentary activity (<100 counts per minute).^22,23 Daily MVPA minutes were calculated by summing total minutes of all MVPA bouts in a day that were greater than 10 minutes in duration, allowing for up to 2 minutes of non-MVPA activity. This approach removes short bursts of MVPA (e.g., from running to a bus) from the daily total, given evidence that neighborhood park access is more likely to be associated with a higher number or longer duration of MVPA bouts.^24–26

2.2.2. Neighborhood Park Access

We generated spatial boundaries of a child’s residential neighborhood via “Sausage Network Buffers (SNB)” approach (see Figure 1), which represents roaming ranges one could realistically travel to around their residence.²⁷ To generate the buffer, we applied Geographic Information Systems (GIS) software ArcGIS (Esri, Redlands, CA, USA) to trace street lines from the home residence along with the network up to 500 meters and buffer those streets by 100 meters along both sides of streets. The 500 m distance, which corresponded to approximately a 5- to 10-minute walk, was selected based on the lower end of the range of 400 to 1,600 m recommended by previous studies and in consideration of relative limits of the mobility of youth compared to adults.²⁸ The 100 m buffer width was determined based on visual inspection of the average depth of parcels along with major street networks in the study, which might contain destinations (e.g., transit stops, open spaces, parks) for PA. Within each SNB, three commonly-used neighborhood park access measures were calculated: including the park density (i.e., the total number of parks that are completely or partially within the buffer area), park coverage (i.e., the total area of all parks within the buffer divided by the buffer area) and park proximity (i.e., street network distance to the nearest park in km).¹² Park data used came from the Countywide Parks and Open Space GIS data published by Los Angeles County Parks and Recreation Department dated January 2019.

Figure 1.

An illustration of 500 m residential neighborhood using “Sausage Network Buffer” (SNB) approach.

Note. The SNB is generated via the “Service Area Analysis” tool of ArcGIS 10.1 software (Esri, Redlands, CA, USA). The tool traced street lines along with the network up to 500 meter and buffer those streets by 100 m along both sides.

2.2.3. Covariates

We included sex and ethnicity at Wave 1, mean child body mass index (BMI) z-score across waves, and annual person household income per wave as covariates. Besides, we generated two age covariates: baseline age and person-mean centered age, to account for the nested person-wave structures. Furthermore, we included daily accelerometer wear time, total valid days for wearing accelerometers per wave, and whether a day was weekday versus weekend day to adjust for individual device wearing behaviors. Finally, we included two neighborhood characteristics reported by past studies to contribute to overall PA in youth as covariates - neighborhood walkability index score and poverty rate.^29,30

2.3. Statistical Analyses

Demographic and descriptive statistics were calculated. Due to the hierarchical structure of the data (i.e., day measurements nested within people), we applied linear mixed-effects models to estimate effects of baseline (i.e., main effects) neighborhood park access measures, their interaction effects with child age (i.e., two-way interactions), and whether interaction effects differed by sex and weekends (i.e., three-way interactions) in predicting MVPA minutes per day. Total MVPA min/day showed a right-skewed distribution; therefore, we log-transformed to ensure normality. Since neighborhood park density, coverage, and proximity were moderately to strongly correlated (r = .63-.68), we tested each variable in a separate model to avoid multicollinearity.³¹ In a stepwise fashion, we adjusted all models for children’s baseline age, sex, ethnicity, the person-mean centered age, annual household income per wave, mean person BMI z-score across waves, weekdays vs. weekends, daily wear time, valid wear day, neighborhood walkability index z-score, and neighborhood poverty rate. We also included two more dummy variables as covariates– whether the participant had at least three valid waves and whether the participant had all six waves – to control for potential influences of the number of valid data collection waves (i.e., those who had higher compliance) on longitudinal trends in PA. We included a random intercept and a fixed slope for each participant in all models to allow baseline PA to vary among individuals. We determined significance at p<.05. We performed analyses using R studio software (RStudio, Boston, MA, USA) with the lme4 package.³²

3. Results

Descriptive characteristics of the sample are presented in Table 1. Overall, the mean age was 10.12 years (SD=0.90). More than half of the children were Hispanic (56.80%), and less than half were boys (46.90%). There was a total of 7,288 accelerometer-measurement days of PA from 202 children. Among them, we removed 2,867 non-valid days with less than 10 hours of wear time, including six children without any valid days. We excluded 18 data collection days from six children who had reached 14 years old by Wave 6 to prevent biased estimation at older ages. Moreover, since we were interested in long-term associations of neighborhood park access on PA, we excluded 259 data collection days (N=4) from children who moved. These efforts resulted in a final data analysis sample of 4,144 days (N=192).

Table 1.

Descriptive statistics of covariates, physical activity outcomes, and neighborhood park access predictors of study participants (N = 192).

Variable	n(%) or mean(SD)
Outcome
Daily MVPA min	38.08 (24.24)
Predictors
Park density1	0.62 (0.83)
Park coverage2	1.76% (4.18%)
Park proximity3	0.80 (0.47)
Covariates
Child’s age (baseline)	10.12 (0.90)
Child’s sex (baseline)
Male	90 (46.90%)
Female	102 (53.10%)
Child’s ethnicity (baseline)
Hispanic	109 (56.80%)
Non-Hispanic	83 (43.20%)
Annual household income (baseline)
≤$34,999	51 (26.56%)
$35,000 - $74,999	40 (20.83%)
$75,000 - $104,999	53 (27.60%)
≥$105,000	48 (25.00%)
Mean Body Mass Index z-score4	0.54 (1.06)
Weekend days	1063 (26.65%)
Total person valid days	22.76 (11.17)
Daily accelerometer wearing min	817.77 (55.46)
At least 3 valid waves5	155 (80.73%)
At least 6 valid waves	87 (45.31%)
Neighborhood walkability index z-score6	−0.12 (6.84)
Neighborhood poverty rate7	14.69% (9.56%)

Note. MVPA = moderate-to-vigorous physical activity; Daily MVPA min = daily total minutes of bouts with greater than 10 minutes duration allowing for up to 2 minutes of non-MVPA activity.

¹Total number of parks in their neighborhoods.

²Percentages of lands that are park uses in their neighborhoods.

³Distance in km to the nearest neighborhood park based on street network distances.

⁴For each person taking an average of age and gender-specific BMI z-scores that using EpiInfo 2005, Version 3.2 (CDC, Atlanta, USA).

⁵A valid data collection wave requires at least one valid day (≥10 hours) of accelerometer data.

⁶Neighborhood walkability index z-score = 2*land-use mix z-score + residential density z-score + intersection density z-score.

⁷Data comes from American Community Survey (ACS) 2012-2016 census-tract level data (US Census Bureau, 2016).

The average daily moderate-to-vigorous physical activity (MVPA) across all six waves was 38.08 min (SD=24.24), which was higher in boys (49.21 min/day) than in girls (28.43 min/day). About one-fourth of data collection days were on weekends (25.70%), which had similar daily MVPA mins (37.98) compared to weekdays (38.08). For neighborhood park access, children on average had less than one park in their neighborhoods (Mean=0.62, SD=0.83), and parks covered 1.76% (SD=4.18%) areas of their neighborhoods. The mean distance to the nearest park was 0.80 km (SD=0.465).

The intraclass correlation coefficient (ICC) indicated that approximately 25.01% of the variance in daily MVPA minutes could be explained by differences between participants, while the remaining 74.99% of the variance could be due to within-person fluctuations or measurement error, respectively. Model results are presented in Table 2.

Table 2.

Results of three mixed-effect models in testing the hypotheses of main, two-way interactions (neighborhood park access measures and age) effects, and three-way interaction (neighborhood park access measures, age, and sex/weekends) in predicting daily MVPA minutes (log-transformed).

	Daily MVPA minutes (log-transformed)
	N
Level-1 (Day)	4144
Level-2 (People)	192
	Model A		Model B		Model C
	Estimate (SE)	95% CI	Estimate (SE)	95% CI	Estimate (SE)	95% CI
Main effects
Age (baseline)	4.42 (0.79) ***	2.88 – 5.96	−0.25 (0.07) ***	−0.38 – −0.12	−0.25 (0.07) ***	−0.39 – −0.12
Age (wave-based)	−0.30 (0.03) ***	−0.36 – −0.25	−0.31 (0.03) ***	−0.36 – −0.25	−0.3 (0.03) ***	−0.36 – −0.25
Sex (Male)	0.80 (0.12) ***	0.57 – 1.03	0.80 (0.12) ***	0.57 – 1.03	0.8 (0.12) ***	0.57 – 1.03
Mean BMI z-score	−0.12 (0.06) *	−0.23 – −0.00	−0.12 (0.06) ***	−0.23 – −0.01	−0.12 (0.06) *	−0.23 – −0.01
Hispanic/Latino	−0.19 (0.12)	−0.43 – 0.05	−0.19 (0.12)	−0.43 – 0.05	−0.21 (0.12)	−0.45 – 0.03
Annual household income per wave	−0.03 (0.04)	−0.11 – 0.06	−0.02 (0.04)	−0.11 – 0.06	−0.03 (0.04)	−0.11 – 0.06
Daily accelerometer wear time (minute)	0.00 (0.00) ***	0.00 – 0.00	0.00 (0.00) ***	0.00 – 0.00	0.00 (0.00) ***	0.00 – 0.00
Weekend day	−0.72 (0.05) ***	−0.83 – −0.62	−0.72 (0.05) ***	−0.83 – −0.62	−0.72 (0.05) ***	−0.83 – −0.62
Total valid days for wearing accelerometer per wave	−0.01 (0.01)	−0.03 – 0.01	−0.01 (0.01)	−0.03 – 0.01	−0.02 (0.01)	−0.04 – 0.00
Neighborhood walkability z-score	0.00 (0.01)	−0.01 – 0.02	0.00 (0.01)	−0.02 – 0.02	0.01 (0.01)	−0.01 – 0.02
Neighborhood poverty rate (%)	−0.01 (0.01)	−0.03 – 0.00	−0.01 (0.01)	−0.03 – 0.00	−0.01 (0.01) *	−0.03 – −0.00
Park density	0.14 (0.07) *	0.01 – 0.28
Park coverage			3.13 (1.45) *	0.30 – 5.97
Park proximity					−0.25 (0.13)	−0.50 – 0.00
Two-way interaction effects
Park density x age	0.08 (0.04) *	0.01 – 0.15
Park coverage x age			1.09(0.74)	−0.35 – 2.54
Park proximity x age					−0.11(0.07)	−0.23 – 0.02
Three-way interaction effects
Park density x age x weekends	0.08 (0.08)	−0.08 – 0.24
Park density x age x sex	−0.12 (0.07)	−0.26 – 0.02
Park coverage x age x weekends			3.48(1.78)	−0.01 – 6.97
Park coverage x age x sex			−1.46(1.49)	−4.39 – 1.47
Park proximity x age x weekends					−0.08(0.15)	−0.37 – 0.21
Park proximity x age x sex					0.08(0.13)	−0.18 – 0.34

^*p < 0.05,

^**p < 0.01,

^***p < 0.001

Note. BMI = body mass index; park density = the total number of parks within 500 m neighborhood; Park coverage = total amount of park areas within 500 m neighborhood / total areas of 500 m neighborhood; Park proximity = shortest distance to the nearest park.

Boys recorded higher mean daily MVPA minutes, while higher age at baseline, person-mean centered age, mean body mass index (BMI) z-score across waves and weekends were associated with fewer mean daily MVPA minutes. Moreover, after controlling for all covariates, neighborhood park density (i.e., number of parks) was associated with higher mean MVPA min/day (b=0.14; p<0.05; 95% Confidence Interval [CI]: 0.01^~0.28). Neighborhood park coverage (i.e., proportions of parklands) was associated with higher mean MVPA min/day (b=3.13; p<0.05; 95% Confidence Interval [CI]: 0.30^~5.97). However, there was not a significant association between neighborhood park proximity and PA.

After adding two-way interactions to each model (see Table 2), rates of longitudinal change in daily MVPA minutes differed by neighborhood park density (b=0.08; p<0.05; 95% CI: 0.01^~0.15). Predicted trajectories for MVPA minutes related to age from 8 to 13 years were estimated (see Figure 2), with stratifications by park numbers available in the neighborhood. According to Figure 2, the decline in MVPA min/day was steeper in children that had a lower neighborhood park density (i.e., fewer numbers of parks in their neighborhoods). Whereas the higher park density (i.e., most neighborhood parks in their neighborhoods) group had almost no decline in MVPA min/day. The interaction effects for neighborhood park coverage and proximity were not significant.

Figure 2.

The significant age and park density interaction in predicting daily MVPA minutes.

Note. Person-mean centered age = subtracting participants’ age at each wave by their person mean of age; MVPA = moderate-to-vigorous physical activity.

The further addition of three-way interactions (see Tables 2) showed that interactions between the three neighborhood park access measures and age did not differ by weekends or child sex at baseline. However, the interactions between neighborhood park coverage, age, and weekends approached significance (b=3.48; p=0.051; 95% CI: −0.01^~6.97), which is worth mentioning. In exploring this interaction, Figure 3 showed that as children got older, those who had lower park coverage (i.e., lower proportions of parklands) in their neighborhoods had a more significant decrease in MVPA min/day, and this occurred primarily during weekends.

Figure 3.

The longitudinal trends of daily MVPA minutes during weekdays and weekend days as a function of park coverage and age.

Note. Person-mean centered age = subtracting participants’ age at each wave by their person mean of age; MVPA = moderate-to-vigorous physical activity; Low/Mid/High park coverage was plotted using cut-offs of 33% and 66% of park coverage distributions among participants.

4. Discussion

Parks are important features of the neighborhood built environment and provide children opportunities to engage in PA. The current study examined whether multiple neighborhood park access measures were associated with children’s longitudinal PA trajectory and explored how these associations varied by sex and weekdays versus weekends. At baseline, we found significant positive associations for relationships of park density (i.e., number of parks) and park coverage (i.e., proportions of parklands) with children’s daily MVPA minutes. However, park proximity (i.e., distance from home to the closest park) was not related to children’s daily MVPA minutes. Longitudinally, we found those children who had access to more parks within 500 m of home had a slower decline in daily MVPA minutes. However, such longitudinal associations were not detected for park coverage or proximity. Lastly, our analyses indicated interactions between all three neighborhood access measures and children’s change in daily MVPA minutes did not vary by sex, but there was a trend for a difference for weekdays versus weekends for park coverage. We discussed these findings below.

Our finding of the cross-sectional association of neighborhood park density at baseline was consistent with previous research on park availability in facilitating children’s PA.^33–35 Longitudinally, our finding was consistent with two previous studies that reported negative associations between neighborhood recreational facility densities and age-related decline in PA.^14,15 Kaczynski et al. hypothesized that neighborhoods with higher park density are likely to offer more diverse opportunities for PA through providing different types of parks, which may explain results observed in our study.³⁶ Future studies should explore the mechanisms through which park density within a neighborhood may influence children’s longitudinal changes in PA in samples of different sociodemographic compositions.

Our finding of the cross-sectional association between park coverage at baseline and children’s daily MVPA outcomes was also consistent with previous research.^10,11 Higher proportions of parklands in the neighborhood might offer more diverse opportunities for children’s park-based PA by providing more park features (e.g., sports field, basketball court); thus, more daily MVPA minutes would be accumulated.³⁶ Longitudinally; however, our study discovered park coverage was not associated with children’s PA outcomes. This finding suggests proportions of parklands in a neighborhood alone may not be sufficient in attenuating children’s longitudinal PA trajectories. Veitch et al. indicated children’s ability to access neighborhood parks might be influenced by parental time and inclination to take children to parks.⁵ In a study of 801 9–15 years old youth across 10 US cities, Bradley et al. reported less parental monitoring was associated with a faster decline in their MVPA minutes.³⁷ Therefore, changes in parenting practice as children age might counteract the positive influences of having access to larger parks on their age-related PA outcomes. Despite the above explanations, no known past studies have examined the longitudinal associations between park coverage and children’s PA outcomes, and more evidence remained to be provided by future studies.

Our null finding between park proximity and children’s daily MVPA outcomes added to previous studies that reported mixed results.^8,11,38 Stewart et al. indicated the inconsistency might be attributed to the total PA being examined rather than park-related PA, which was the case of our study.³⁹ Lackey and Kaczynski further pointed out such consistencies may also be due to the mismatch between objective (e.g., GIS-based) versus perceived (e.g., survey) approaches in measuring proximity.⁴⁰ The park proximity measure in our study was calculated objectively as the shortest route from home to the nearest park, which may not match the perceived actual route to the park (e.g., other possible longer routes that may be safer, more walkable, or have higher aesthetics) that could be associated with their MVPA outcomes. Furthermore, children, compared to adults, may have limited mobility walking to the park.⁵ As a result, a distance threshold for a typical child to be willing to walk to a neighborhood park may exist in order for park proximity to affect their PA. This insignificant finding, in combination with positive associations between park density and children’s longitudinal PA outcomes above, suggests that neighborhood parks may need to exist within at a closer distance (e.g., 500 m) in order to better promote longitudinal PA in children in comparison to the status quo in our sample.

Longitudinal associations between three neighborhood park access measures and children’s MVPA minutes did not differ by sex in our study, which is inconsistent with a previous longitudinal study.¹⁵ However, this earlier study measured public open spaces rather than parks, which could also include other spaces such as public squares, plazas, schoolyards, and playgrounds. Klinker et al. reported boys and girls might have different preferences in places for leisure-domain PA;⁴¹ thus, it might be other outdoor features, rather than parks, that led to sex-differences in children’s longitudinal PA changes. Also, in our study, we found longitudinal associations between the three neighborhood park access measures and children’s MVPA outcomes did not differ on weekdays versus weekends. The three-way interaction among park coverage, age-related PA, and weekends vs. weekdays did approach significance, in that those with the highest park coverage had the slowest decline in daily MVPA mins during weekends. It is likely that, on weekends, factors such as daily school schedules and parental transportation to activities, may no longer influence children’s park access. Thus, children who have access to a higher amount of park areas can utilize parks to perform certain types of PA activities (e.g., sports), which typically requires larger park sizes, and usually accumulate more MVPA minutes over time. We suggest future studies to explore further using weekend days as intervention opportunities to slow down age-related PA decline in children.

Our study has a few limitations. First, our study did not collect the data on qualitative or perceived neighborhood park characteristics (e.g., safety, amenities, maintenance) that previous studies reported as predictors of PA and might act as mediators of pathways from the neighborhood park exposure to PA behaviors.^18,34,42 However, we did control for proxy neighborhood measures. Second, the study assumed neighborhood park access remained unchanged over the data collection periods (approximately three years). New park construction or expansion projects might happen during the period and affect our neighborhood park access measures. Third, our neighborhood park access measures may also be subject to the “uncertain geographic context problem” (UGCoP). Specifically, associations between neighborhood park access and PA might be affected by the spatial methods that were used to delineate the extents of the contextual neighborhood (e.g., residential neighborhood).⁴³ Our study defined “neighborhood” as 500 m sausage network buffers (i.e., 5-10 minutes of walking distance) around children’s residences per recommendations of previous studies and characteristics of the study population,²⁸ however, other definitions of neighborhoods might influence study results. Fourth, the accelerometry data were unable to differentiate park-related PA (e.g., leisure walks in a park, park commutes) from PA that occurred in other spatial contexts (e.g., commuting walk to schools, PE classes). Without fully separating these two types of PA, and the mismatch between the neighborhood park access metrics measured and PA outcome utilized may occur. Fifth, despite the longitudinal analyses, due to the study design, we could not infer causality. Lastly, all children in this study lived in the urban Los Angeles area, and about half were of Hispanic origin. The results may not be generalizable to children living in other geographic areas and/or of different racial and ethnic compositions.

To the best of our knowledge, our study is the first in the US to that examines longitudinal associations between multiple neighborhood park access measures and PA in children. A major strength of our study is the use of accelerometers to collect repeated monitor-based PA measurements across six data collection waves. As a result, we are able to overcome recall biases inherent to self-report PA measurements and provide greater insight into PA trends across a longer duration. In addition, we provide insights into the relative importance of three different neighborhood park access measures in driving longitudinal PA changes in children. Furthermore, our study investigated sex and weekdays vs. weekends differences in interaction effects. Our results suggest neighborhood parks have an overall protective effect against age-related PA decline regardless of sex and day of the week. Future studies should continue exploring this topic to allow comparisons of results across children of different sociodemographic compositions.

5. Conclusions

Neighborhood parks are important locations for children to engage in PA. Specifically, neighborhood park density might be critical in promoting PA in children and slowing age-related PA decline. As a result, future policies and interventions for planning healthy neighborhoods and promoting PA in children should consider adding a greater number of parks to the proximity of their residential neighborhoods. Particularly, it is critical to eliminate “park deserts” (i.e., no park presence) in order to achieve the most effective intervention outcomes.

APPENDICES

Appendix 1.

Details of the MATCH study design characteristics.

Study Design Characteristics	Details
Data Collection	The present study examined data from all six assessment waves. Each wave consisted of in-person visits, during which mothers and children completed separate questionnaires. Assessments occurred from 2014 to 2018 during the fall (mid-August through mid-December) and spring (January through May) to avoid data collection during the summer months and winter holidays when unusual patterns of physical activity may be expected.
Inclusion Criteria	(a) child currently in 3rd – 6th grade (b) child resides with mother at least 50% of time (c) ability of both mother and child to speak and read in English or Spanish
Exclusion Criteria	(a) use of medication for thyroid or psychological condition (b) health condition limiting ability to be physically active (c) child enrolled in a special education program due to concerns about reduced understanding of assent (d) currently using oral or inhalant corticosteroids for asthma (e) pregnancy (f) child underweight (BMI < 5th% for age and sex) due to concerns about these children having a different set of health concerns (g) mother works more than two evenings (between 5-9 pm) during the week, or more than one 8–12-hour weekend shift

Note. MATCH = Mothers and Their Children’s Health study.