Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2021 May 1.
Published in final edited form as: Health Place. 2020 Apr 10;63:102330. doi: 10.1016/j.healthplace.2020.102330

Are low income children more physically active when they live in homes with bigger yards? A longitudinal analysis of the NET-Works Study.

Jonathan M Miller 1, Yingling Fan 2, Nancy E Sherwood 3, Theresa Osypuk 4, Simone French 5
PMCID: PMC7306417  NIHMSID: NIHMS1584192  PMID: 32543422

Abstract

This study prospectively examined the relationship of home yard-size to objectively measured physical activity over three years among a cohort of 531 low-income pre-school-aged children. An adjusted total-effect association of 12.72 additional minutes per week of moderate/vigorous physical activity (MVPA) was observed for each additional hectare of yard-size. The direct-effect association, adjusting for previous year MVPA, was not statistically significant. This study provides evidence that the private or semi-private space around the house may impact children’s’ physical activity. Public health and urban planning practitioners should consider these results to identify built environment solutions for improving MVPA among low-income minority children.

1. INTRODUCTION

Physical activity protects against heart disease, obesity and many cancers (2018 Physical Activity Guidelines Advisory Committee, 2018; Dixon, 2010; WHO, 2009), yet population levels of physical activity in the United States remain low (Hallal et al., 2012; Troiano et al., 2008). Physical activity level tracks across ages (Bauman et al., 2012), therefore establishing healthy physical activity habits at an early age may support activity later in life.

Recently public health researchers, urban planners and policy makers have focused on the role of access to public parks in promoting physical activity among children (Christian et al., 2015; Davison and Lawson, 2006; French et al., 2017). However, for preschool-aged children in particular, access to private outdoor spaces may be as important as public parks, with the most accessible private or semi-private space being the home yard around the child’s house or the common yard owned by the homeowners’ association if the child’s home lives in an apartment or condominium complex. Following the time-geography theory that temporal constraints translate into spatial limitations, Fan and Khattak showed that spatial accessibility of parks is more strongly related to joint recreation by two members of a household than to solo recreation (Fan and Khattak, 2009). We expect much physical activity among pre-schoolers will be supervised or joint activity with the parent. The findings of Fan and Khattak lead us to hypothesize that easy access to green space will predict pre-school physical activity and that in addition to access to parks, the size and configuration of the private or semi-private green space around the child’s house may be predictive of physical activity. A yard with adequate size may reduce the need to travel to parks with children and may allow for more unstructured physically active play than in public spaces.

Yet, recent density-conscious housing development trends have moved toward larger houses on smaller lots, and less available yard space (McGill, 2016; Terrazas, 2015). The two available studies have found a positive association between yard availability or yard-size and parent-reported child physical activity or outdoor play (Marino et al., 2012; Spurrier et al., 2008). However, these studies are cross sectional and did not use an objective measure of physical activity. We are aware of no study that has examined the prospective association between yard-size and child physical activity using objective measures of both physical activity and yard-size. Data on the impact of the home environment, including home yard space, on children’s physical activity is an important and substantial gap in the research literature (Christian et al., 2015). There are several factors that could impact any effect of home environment on childhood physical activity, including: the size and shape of the yard, whether the yard is fully private as with single family residences or semi-private as with many multi-family residences, and the socioeconomic characteristics of surrounding neighborhood environment. One important step in describing any possible relationship of home environment to childhood physical activity is to describe the relationship between the size of the yard and physical activity, and additionally whether any relationship between yard size and physical activity differs by the type of residence - single family or multi-family.

This study aimed to estimate whether home yard-size is associated with physical activity in low-income racially/ethnically diverse pre-school-aged children followed longitudinally for 36 months. The study also aimed to test whether the association differed by the type of residence, for example: apartments as compared to single family homes. It was hypothesized that larger home yard-size would be positively associated with child physical activity over three years.

Although this is an observational study, and therefore cannot account for all possible confounders, this study takes into account many possible confounders of the hypothesized relationship, and therefore represents a unique step towards describing the possible causal relationship between yard size and childhood physical activity.

2. METHODS

2.1. Sample

This study is a secondary analysis of data collected as part of the NET-Works study, a randomized controlled trial to prevent childhood obesity (French et al., 2018; Sherwood et al., 2013). Participants for the NET-Works study (n=534 parent-child dyads) were children aged 2–4 years at baseline in the Twin Cities of Minnesota, who were at risk for obesity (BMI z-score >=50th percentile) and whose yearly family income was below $65,000. Child parent dyads were recruited from primary care clinics and managed health care systems (Sherwood et al., 2013).

Baseline data collection occurred between 2012 and 2014 and at annual intervals (12, 24, and 36 months post- randomization). The NET-Works intervention was a multi-level intervention based on the social-ecological model and aimed to support a family and home environment that encouraged food availability, family meals and healthy play, and discouraged television viewing. Further details on the sample, intervention and measurements for the NET-Works study have been published previously (French et al., 2018; Sherwood et al., 2013). All study procedure were approved by the Institutional Review Board at the ____, and parents provided informed consent.

2.2. Measures

2.2.1. Outcome:

Child Moderate to Vigorous Physical Activity.

Child moderate to vigorous physical activity (MVPA) was measured at all four waves of data collection with Actigraph GT3X or GT3X+ accelerometers for 7 days. Data collection was set at 40 hertz for the GT3X+ and 1 second for the GT3X. Accelerometer measurements were included if children had at least 4 valid days (3 weekday and 1 weekend day) wearing the accelerometer for more than 6 hours. Non-wear time was ignored in converting accelerometer counts to MVPA. As wear time was not expected to be associated with yard size and therefore not to confound the main effect of interest, wear time was not adjusted in statistical analyses. Accelerometer counts on a 15 second epoch were converted to MVPA minutes per week from accepted age-appropriate cut-points based on Evenson et al. (2008), Trost et al. (2011) and Choi et al. (2011). More details on the accelerometer measurements have been provided previously (French et al., 2018, 2017; Sherwood et al., 2013).

2.2.2. Exposure:

Yard-size.

Participants were linked to parcels by geocoded home addresses reported at each of the four waves of follow-up. Yard-size in hectares of the child’s home at each wave of data collection was calculated in ArcGIS Pro (Version 2.0.0) by subtracting building footprints from parcel boundaries. This method has been used previously in real-estate market analyses (Kenyon Henderson and Song, 2008). Building footprints were estimated from Lidar remote sensing data (MnGeo, 2013). Parcel boundaries were retrieved from the 2013 MetroGIS Regional Parcel Dataset (MetroGIS, 2014). MetroGIS is a regional GIS data sharing initiative administered by the Metropolitan Council—the regional governmental agency for the 7 counties surrounding Minneapolis and St. Paul, Minnesota. The building footprint layer was subtracted from the parcel layer using the Erase command in ArcGIS and the area in hectares of the resulting polygon was calculated for each parcel.

2.2.3. Modifier:

Residence Type.

The type of residence the participants lived in was assessed by trained research staff during a neighborhood block audit conducted in person at each wave of data collection, described previously (French et al., 2018; Sherwood et al., 2013). Staff followed a standard protocol to record residence type as: ‘House’, ‘Apartment’, ‘Duplex’, or ‘Townhome/Condominium.’

2.2.4. Modifier:

Moving.

If participants reported a different address from the previous wave of data collection, they were categorized as movers.

2.2.5. Covariates.

Covariates were chosen a priori from variables shown previously to associate with MVPA (Bauman et al., 2012; De Craemer et al., 2012; Hinkley et al., 2008; Sallis et al., 2000; Trost et al., 2002; Xu et al., 2015).

Age.

Child age was parent-reported at baseline and follow up visits.

Parent physical activity.

Adult MVPA across 7 days for the primary caregiver was measured with accelerometers using the same protocol as child MVPA and modeled as minutes per week. Adult specific cut-points based on Trost et al. (2005), Matthews (2005) and Troiano et al. (2008) were used to convert accelerometer counts from a 1 minute epoch to MVPA.

Household Education and Income.

Household education was modeled as the highest reported parent education attainment in the household on a range from “less than high school” to “advanced degree”. Parents reported household income on a range from “$14,999 or less” to “$65,000”.

Child sex.

Child sex was parent-reported at baseline as male or female.

Intervention treatment group assignment.

Treatment group assignment (NET-Works treatment or control) was included as a covariate.

Race.

Child race was parent-reported at baseline as non-Hispanic White, non-Hispanic Black, Hispanic, Multi-Racial or other Ethnicity/Race. Although race has a complex relationship with education and socioeconomic status, it was included as a covariate and possible confounder in addition to education and income because of the history of discriminatory real-estate practices in the United States that may limit choices in housing, and because of evidence that race may be linked to levels of childhood physical activity (Bauman et al., 2012; Uijtdewilligen et al., 2011).

Neighborhood Poverty.

The proportion of each participant’s census tract in poverty was linked from the US Census Bureau’ American Community Survey 5-year estimates at each wave of follow-up.

2.3. Statistical Analysis

The average yard-size in hectares for each residence type was calculated and compared with ANOVA. The average child MVPA for each residence type was estimated using generalized estimating equations (GEE) to account for multiple (up to 4) observations of MVPA per child. An autoregressive working correlation matrix was used for the GEE and the model was adjusted for the covariates listed above.

The association between yard-size and MVPA was estimated using GEE specifying an autoregressive correlation matrix (Figure 1). The first GEE model (Model 1) represented an estimate of the direct-effect association of one-year lagged yard-size on MVPA, meaning that the association was adjusted for one-year lagged MVPA as well as child age, race and sex, household income and highest education attainment, adult MVPA, residence type, whether the participant moved between waves, and treatment group assignment. The second GEE model (Model 2) represented an estimate of the total-effect association of one-year lagged yard-size on MVPA, meaning that the association was adjusted for child age, race and sex, household income and highest education attainment, adult MVPA, residence type, whether the participant moved between waves, and treatment group assignment, but not for one-year lagged MVPA. Taken together these models could represent a product-based method to estimate mediation (Baron and Kenny, 1986). An important assumption of this method that is that the one-year lagged yard-size exposure has its effect prior to the measurement of one-year lagged MVPA. Any estimation of indirect effects using a product-based method would require the additional assumption of no interaction between the exposure (one-year lagged yard-size) and the potential mediator (one-year lagged MVPA) (Valeri and VanderWeele, 2013; Vanderweele, 2014). An interaction term between yard-size and one-year lagged MVPA was used to test the assumption of no exposure-mediator interaction. As estimations of the indirect effects were beyond the scope of this analysis, the product-based method was used to estimate the direct and total effects. These estimates represent causal effects only under the assumption that all possible confounding of the exposure (one-year lagged yard-size) on the outcome and of the mediator (one-year lagged MVPA) on the outcome was controlled (Vanderweele, 2014).

Figure 1:

Figure 1:

Open in a new tab

Directed Acyclic Graphs (DAGs) representing the associations estimated in A) Model 1 and B) Model2. MVPA stands for Moderate to Vigorous Physical Activity. ta+1 indicates that MVPA is measured one year after the yard-size measurement. β refers to the GEE coefficient for Yard-size on MVPA adjusting for the covariates show in the DAG.

To test for possible effect modification of yard-size on child’s MVPA by residence type, an interaction term of residence type by yard-size was added to both of the models of yard-size on MVPA. To test for possible effect modification of yard-size on child’s MVPA by moving residences, an interaction term of yard-size by whether the child moved between waves of data collection was added to both of the models of yard-size on MVPA. Tests for non-linearity of MVPA levels over yard-sizes were conducted by including quadratic and cubic terms for yard-size into statistical models. Sensitivity analyses were conducted for outliers by removing observations below the 5th percentile of yard-size (0.021 hectares) and again removing observations above the 95th percentile of yard-size (5.3 hectares). Interactions were considered statistically significant if P < 0.05.

Cohort retention was 92% at 36 month follow up (N=493). Accelerometry and yard-size data was complete for 434; 418; and 386 children at follow ups 12, 24 and 36 months, respectively. There were 326 participants that provided valid accelerometer and yard-size measurements at all 4 waves of data collection. An additional 158 participants provided valid accelerometer and yard-size measurements at 2 or 3 waves of data collection. Therefore, there were 484participants who could be analyzed in the GEE models. To account for possible selection bias due to loss to follow-up, response propensity weights were constructed using logistic regression. Baseline variables that correlated at greater than 0.1 or less than −0.1 with probability of responding at a wave of follow-up were added to a multiple logistic regression model to generate predicted odds of responding to a wave of follow-up for each participant. The inverse of these predicted odds was used as a response propensity weight for each of the models (Little, 1986). All analyses were conducted using SAS (version 9.4) in 2018.

3. RESULTS

3.1. Descriptive Characteristics

Given the inclusion criteria described above, the NET-Works study recruited a uniquely diverse and lower socioeconomic status sample. Male and female children were equally represented. Participants represented a diversity of ethnicities/races, but a majority of participants identified as Hispanic. The majority of households reported high school education or less, and 80.9% of households at baseline reported incomes less than $35,000 per year. At wave 4, 59.2% of households reported incomes less than $35,000. The average percent poverty of the participants’ census tract was 22.7% at baseline and 19.5% at wave 4. The proportion of families living in single family houses was 35.1% at wave 1 and 44.3% at wave 4. Many families (49.2%) moved at least once over the study. The mean size of the family’s yard was 0.7 hectares at wave 1 and 0.8 hectares at wave 4 (Table 1). Note that, for the participants who lived in an apartment, townhome or condominium complex (51.1% at baseline, 42.0% at wave 4), the yard space was not entirely private, but shared among homeowners in the complex; for example, 35.5% of the 394 unique apartment, condo or town home complexes in the sample had yard-sizes greater than 0.4 hectares – indicating that this may be semi-private space shared among the housing units.

Table 1.

NET-Works Demographic Variables by Visit for Preschool-Aged Children, 2012 to 2016

Visit 1 (Baseline) (n=531) Visit 4 (3 year) (n = 386)
Age: Years (mean, Range) 3.4 (2.1 to 4.3) 6.4 (4.9 to 7.6)
Sex: (n, %)
 Male 261 (49.2%) 191 (49.5%)
 Female 270 (50.9%) 195 (50.5%)
Ethnicity/Race: (n, %)
 Non-Hispanic White 67 (12.6%) 52 (13.5%)
 Non-Hispanic Black 95 (17.9%) 57 (14.8%)
 Hispanic 312 (58.8%) 233 (60.4%)
 Multi-Racial or Other Ethnicity/Race 57 (10.8%) 44 (11.4%)
Parent Education Level: (n, %)
 High School or Less 295 (55.5%) 215 (56.1%)
 Some College 135 (25.4%) 88 (23.0%)
 Bachelors Degree 62 (11.68%) 49 (12.79%)
 Advanced Degree 39 (7.34%) 31 (8.09%)
Household Income: (n, %)
 Less than $35,000 432 (80.9%) 255 (59.2%)
 $35,000 to $49,999 53 (9.9%) 77 (17.9%)
 $50,000 or More 49 (9.2%) 99 (23.0%)
Moderate to Vigorous Physical Activity: (mean, Range)
 Child MVPA (min/wk) 758 (212 to 1517) 720 (152 to 1632)
 Parent MVPA (min/wk) 147 (1 to 1048) 153 (4 to 801)
Home Residence Characteristics
Neighborhood Poverty: Percent (mean, Range) 22.7% (1.6% to 59.6%) 19.5% (1.0% to 60.3%)
Residence Type: (n, %)
 House 185 (35.1%) 170 (44.3%)
 Apartment 238 (45.2%) 132 (34.4%)
 Duplex 73 (13.9%) 53 (13.8%)
 Town Home/Condominium 31 (5.9%) 29 (7.6%)
Yard Size: Hectares (mean, Range) 0.7 (0.001 to 19.5) 0.8 (0.003 to 19.5)
Number of Moves: (n, %)
 None NA 158 (40.9%)
 1 NA 141 (36.5%)
 2 NA 58 (15.0%)
 3 or more NA 29 (7.5%)
Open in a new tab

Pre-schoolers’ mean MVPA was 758 minutes per week at wave 1 and 720 minutes per week at wave 4. Parent mean MVPA was 147 minutes per week at wave 1 and 154 minutes per week at wave 4 (Table 1).

3.2. Yard-size and MVPA by Residence Type

The average yard-size was 0.04 hectares for duplexes, 2.8 hectares for townhomes or condominiums, 0.5 hectares for single-family homes, and 1.0 hectares for apartments (P<0.0001). Children who lived in townhomes or condominiums recorded 76 more minutes per week of MVPA as compared to children living in apartments or duplexes and 67 more minutes per week as compared to children living in single family houses, though these adjusted differences did not achieve statistical significance (test of difference: P=0.1) (Table 2).

Table 2.

Yard size and average NET-Work pre-school participant MVPA by residence type, 2012 to 2016.

Yard Size: Hectares 95% CI MVPA: Minutes/Weeka 95% CI
Residence Type
 House 0.5 0.3, 0.6 752 724, 780
 Apartment 1.0 0.8, 1.1 743 715,771
 Duplex 0.04 0,0.3b 743 705, 780
 Town Home/Condo 2.8 2.4, 3.1 819 762, 877
Open in a new tab

MVPA-Moderate to vigorous physical activity.

a

Weekly minutes of MVPA were estimated with Generalized Estimating Equations and adjusted for child sex, race and age, household education and income, adult MVPA levels, treatment group assignment and neighborhood poverty.

b

Lower confidence level based on a normal distribution was −0.2 hectares, so confidence level was floored at 0, rather than the impossible negative number.

3.3. Estimates of the Effect of Yard-size on weekly MVPA

In models adjusted for covariates and previous year’s weekly MVPA and weighted for response propensity (Model 1, direct-effect model using a mediation framework), the estimated direct effect of previous year’s yard-size on weekly MVPA was positive but not statistically significant (2.92; 95% CI: −2.73, 8.58). When previous year’s yard-size was examined in relationship to weekly MVPA, adjusted only for covariates and weighted for response propensity (Model 2, total-effect model using a mediation framework), each additional hectare of yard-size was associated with 12.72 (95% CI: 4.21, 21.23) more minutes per week of MVPA among children. These covariate adjusted models were both closer to null than crude models. The crude association for the direct-effect model (model 1) was 8.03 minutes per week (95% CI: 4.13, 11.93) and for the total-effect model (model 2) was 18.72 minutes per week (95% CI: 11.32, 26.11). Under the assumption that previous year’s MVPA is mediating the relationship between previous year’s yard-size and current MVPA, the size of the yard was positively associated in a total-effect model with more weekly MVPA one year later. Using a product-based method for mediation, most of this association was an indirect effect of 9.80 minutes per week mediated through previous year’s MVPA. While more robust modeling of mediation (Valeri and VanderWeele, 2013; Vanderweele, 2014) was beyond the scope of this analysis, there was no statistical evidence of an interaction between previous year’s MVPA and previous year’s yard-size (p = 0.67) (Table 3).

Table 3.

Associations of NET-Works pre-school child participant MVPA with Yard Sizea, 2012 to 2016

Model 1. MVPA: Minutes/Week One Year Lag Adjusted for Prior MVPAb 95% CI Model 2. MVPA: Minutes/Week One Year Lag 95% CI
Yard Size
 Per Hectare 2.92 −2.73, 8.58 12.72 4.21, 21.23c
Yard Size (Remove Biggest 5%of Yards)
 Per Hectare 23.77 8.21, 39.33c 31.96 10.40, 53.53c
Yard Size (Remove Smallest 5%of Yards)
 Per Hectare 2.03 −4.15, 8.21 11.27 2.57, 19.96c
Open in a new tab

MVPA-Moderate to vigorous physical activity.

a

Models adjusted for child age, race and sex, household income and highest education attainment, adult MVPA, residence type, and treatment group assignment, whether the participant moved between waves and neighborhood poverty, and weighted using the response-propensity method.

b

Model additionally adjusted for MVPA at the previous year

c

Association is statistically significant at p < 0.05

Sensitivity analyses that removed participants with the biggest and smallest yards gave some evidence for non-linearity in the association of yard-size and MVPA. While removing participants below the 5th percentiles of yard-size did not substantially change the strength of the association of yard-size with MVPA, when removing participants above the 95th percentiles of yard-size, the association of yard-size with MVPA increased substantially (Table 3). The graphical form of a cubic model (Appendix Figure 1) further illustrates this non-linearity: stronger associations of yard-size with MVPA occur on the part of the curve where yards are smaller. However, non-linear associations should be interpreted cautiously as neither the quadratic nor the cubic terms achieved statistical significance.

Lastly, the impact of yard-size on MVPA did not differ significantly by type of residence or whether the participants moved for any of the effect modification models tested (Appendix Table A).

4. DISCUSSION

This study examined the prospective association between yard-size and weekly MVPA among a low-income cohort of preschool-aged children. The results showed that yard-size was significantly positively associated with higher child MVPA levels one year later. While previous work has focused on the impacts of having nearby parks on physical activity in children (Christian et al., 2015; Davison and Lawson, 2006; French et al., 2017), the present study provides evidence that physical activity can be supported by the presence of private or semi-private greenspace around a child’s home. This study is unique in its prospective design and use of objective measures of both yard-size and MVPA, adding to the small literature (Marino et al., 2012; Spurrier et al., 2008) that examines the relationship between yard space and physical activity.

Another unique characteristic of the study is its low-income, racially/ethnically diverse cohort of preschool-aged children. Declines in MVPA are characteristic of the transition of children into adolescence, and are of particular concern among lower income and ethnically/racially diverse children (Bauman et al., 2012; Troiano et al., 2008; Uijtdewilligen et al., 2011). Additionally, living in urban neighborhoods, which may have less available public green space, these children maybe at particularly high risk for declines in MVPA. This study showed that in an urban setting among low-income diverse children, yard-size was a minor, though potentially modifiable, variable in total-effect models that could support healthy levels of MVPA.

Our findings are particularly relevant since housing development is trending towards building larger houses on smaller lots to serve housing density objectives (Terrazas, 2015). The impact of smaller yard-sizes suggests that families may need to take advantage of other spaces for physical activity – like public parks. However, previous findings have suggested that low-income minority families face greater barriers to park use like not feeling welcome, cultural and language restrictions, and concerns about program schedules, pricing or facility mismatch (Das et al., 2017). Low-income minority children may be further disadvantaged in terms of physical activity resources if their physical activity locations are limited to public parks. Even though previous studies have shown that park accessibility promotes joint parent-child activity, private and semi-private home yard space provides an important and complementary setting for child physical activity that they can access without direct parent time and effort (Fan and Khattak, 2009; French et al., 2017).

A diverse range of urban planning strategies can be used to increase the affordability and availability of homes with private and semi-private yards among low-income minority families. Municipal governments can adopt policies to encourage home yard-size considerations in affordable housing development projects. Across the nation, housing voucher programs exist to assist low-income families to afford decent housing in the private market (Sard, 2001), and could incorporate home yard-size considerations. In addition, zoning codes can be used to define minimum building setbacks and maximum impervious surface ratio to secure minimum yard spaces.

We did not hypothesize a priori any non-linear shape to the association of yard size with MVPA in children. While there was no statistical evidence for a quadratic or cubic relationship, the sensitivity analysis removing the 5% of observations with the largest yards showed a much stronger relationship of yard size with MVPA in the smaller 95% of observations. This finding was not expected, but may indicate that there is a diminishing marginal impact for increases in yard size at larger size. However, future studies will be needed to further explore this possible non-linearity of the association, particularly accounting for possible differences between fully private space around single-family homes and semi-private space around multi-family homes.

This study had many notable strengths and some limitations. Strengths include longitudinal, objective measures of yard-size and physical activity, over 90% cohort retention in follow up, and the unique sample of low-income children and parents. One notable limitation of this study is that many multifamily housing units like apartments, town homes and condominiums were recorded as a single parcel in the parcel dataset. Because of this limitation, apartments, town homes, and condos were recorded as having much larger yard sizes than single family houses and duplexes in this study. Further, the average yard area in this study was 0.7 hectare, much larger than the recent estimate of 0.06 hectares of yard for new construction of homes,(McGill, 2016) which may limit generalizability to other geographic regions. In particular the relationship between lot size and socioeconomic status may be complex in the United States, where some wealthy areas like New York city may have smaller lots even though larger lots within the same community are generally associated with high home value and therefore higher socioeconomic status.

Condominiums and townhomes had the biggest yard footprint, and this limits our ability to fully describe the access to private space for physical activity for children living in these multifamily developments as the space around condominiums and townhomes may not be fully private. In some cases, children may only have access to the part of the development immediately near their condominium or townhome. In other cases, children may have access to the entire grounds of the development. The height of the multifamily development may also impact the accessibility of the private or semi-private greenspace as children on higher floors may be less likely to descend the stairs to get to green-space. Some condominium complexes may have playgrounds or pools, essentially giving children access to a private park. This mechanism could help explain a previous finding by Aarts et al. (2010) that 4 to 6 year old boys living in rental properties showed more physical activity than boys living in owned properties. While our study did not find a statistical difference in physical activity by residence type, the mean physical activity was highest for children living in condominiums or townhomes. Future work should aim to better characterize the private or semi-private space available for physical activity in multifamily developments, including the resources available within the entire condominium or townhome complex beyond resources available within children’s’ immediate dwelling unit to measure and test effects of private space resources on children’s’ physical activity. It could also be the case that the self-governance of condominium and townhome developments promotes interactions between neighbors, and the proximity to other families with young kids might have a role in supporting physical activity of young children.

The present study was conducted in the Minneapolis-St Paul metropolitan area, so results may not be generalizable to other regions. In Minneapolis, 97% of homes are located within a 10 minute walk from a park (The Trust for Public Land, 2018), so access to public spaces for physical activity may be different than in other cities. As an observational study, there is potential for unmeasured confounding. We attempted to control for possible confounders suggested by prior literature that were measured in this dataset. In particular, we believe that adjusting for the physical activity levels of the child’s parent represents one of the larger potential confounders and we are aware of few studies that have objectively measured physical activity from both parents and pre-school aged children. It is likely that the current study could not fully adjust for residential self -selection bias. We adjusted for several demographic, socioeconomic factors and neighborhood factors, which are commonly associated with choice of residence. However, it is still possible that yard-size may be correlated with other unmeasured variables (e.g. wealth) that cause both housing choice and children’s physical activity.

While a full mediation analysis was beyond the scope of this study, we estimated direct and total effects for yard-size on MVPA one year later. Although we found that yard-size was associated with higher physical activity one year later in the total-effect model, we also found that the strength of this association decreased substantially after adjusting for past-year physical activity in the direct-effect model. As previous year’s MVPA was measured at the same time as previous year’s yard-size, previous year’s MVPA could be either a mediator or a confounder of the estimated effect between yard-size and MVPA. The mediator explanation for this finding requires that one-year lagged yard-size causes higher one-year lagged MVPA in children; while the confounder explanation requires that the child’s MVPA causes larger yard-size, as would be the case if parents chose to move to a bigger house because their child was more active. While neither of these explanations can be ruled out, it is reasonable to assume that most families had been living in their current home for some span of time at the time MVPA was measured, therefore making it reasonable to also assume that the previous year’s yard-size temporally preceded the previous year’s MVPA.

5. CONCLUSION

Using a prospective design, this study showed that yard-size around a child’s residence is positively associated with their later level of physical activity. City planners should consider policies that maximize yard-sizes in urban settings where low-income diverse families live.

HIGHLIGHTS.

  • Larger yard size was related to more physical activity in young children

  • Children living in condominiums had the highest levels of physical activity

  • The relationship of yard-size with physical activity lessened at larger yard-sizes

ACKNOWLEDGEMENTS:

The NET-Works Study was supported by grant U01HD068890 from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD). Dr. Miller is supported by grant T32CA163184 from the National Cancer Institute (NCI). The content is solely the responsibility of the authors and does not necessarily represent the official views of the NICHD, NCI; or the National Institutes of Health.

Appendix Table A.

Sensitivity analysis association estimates of Child MVPA with Yard Sizea

MVPA: Minutes/Week One Year Lag Adjusted for Prior MVPAb 95% CI MVPA: Minutes/Week One Year Lag 95% CI
Yard Size - Fully Adjusted (Weighted)
 Per Hectare 2.92 −2.73, 8.58 12.72 4.21, 21.23c
Yard Size - Crude/Unadjusted (Weighted)
 Per Hectare 8.03 4.13, 11.93c 18.72 11.32, 26.11c
Yard Size - Fully Adjusted (Unweighted)
 Per Hectare 1.94 −3.10, 6.97 11.43 2.53, 20.33
Yard Size - Fully Adjusted (Remove Biggest 5% of Yards)
 Per Hectare 23.77 8.21, 39.33c 31.96 10.40, 53.53c
Yard Size - Fully Adjusted (Remove Smallest 5% of Yards)
 Per Hectare 2.03 −4.15, 8.21 11.27 2.57, 19.96c
Yard Size - Fully Adjusted (Remove Biggest and Smallest 5% of Yards)
 Per Hectare 22.00 6.16, 37.84c 27.19 4.85, 49.53c
Interaction by Residence Type
 P-Value 0.68 0.58
Interaction by whether Participant Moved
 P-Value 0.75 0.82
Interaction with One Year Lagged MVPA (Exposure-Mediator Interaction)
 P-Value 0.67 NA
Quadratic Term for Yard Size
 P-Value 0.14 0.15
Cubic Term for Yard Size
 P-Value 0.17 0.13
Open in a new tab

MVPA–Moderate to vigorous physical activity.

a

Models adjusted for child age and sex, household income and highest education attainment, adult MVPA, residence type, whether the participant moved between waves, proportion of the participant’s census tract in poverty and treatment group assignment, and weighted using the response-propensity method.

b

Model additionally adjusted for MVPA at the previous year

c

Association is statistically significant at p < 0.05

Appendix Figure 1:

Appendix Figure 1:

Open in a new tab

Cubic relationship between yard-size in hectares and weekly minute of moderate to vigorous physical activity (MVPA) among pre-school aged children in the NET-Works study from 2012 to 2016. The curve represents estimates from a regression model of MVPA on cubic yard-size adjusted for prior year MVPA and child age, race and sex, household income and highest education attainment, adult MVPA, residence type, whether the participant moved between waves, and treatment group assignment, and weighted using the response-propensity method.

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

DECLARATION OF INTEREST: The authors declare no conflicts of interest.

Contributor Information

Jonathan M. Miller, Division of Family Medicine and Community Health, University of Minnesota..

Yingling Fan, Humphrey School of Public Affairs, University of Minnesota..

Nancy E. Sherwood, Division of Epidemiology and Community Health, University of Minnesota.

Theresa Osypuk, Division of Epidemiology and Community Health, University of Minnesota..

Simone French, Division of Epidemiology and Community Health, University of Minnesota..

REFERENCES

  1. 2018 Physical Activity Guidelines Advisory Committee, 2018. 2018. Physical Activity Guidelines Advisory Committee Scientific Report. Washington D.C. [Google Scholar]
  2. Aarts MJ, Wendel-Vos W, Van Oers HAM, Van De Goor IAM, Schuit AJ, 2010. Environmental determinants of outdoor play in children: A large-scale cross-sectional study. Am. J. Prev. Med 39, 212–219. 10.1016/j.amepre.2010.05.008 [DOI] [PubMed] [Google Scholar]
  3. Baron RM, Kenny D. a., 1986. The Moderator-Mediator Variable Distinction in Social. J. Pers. Soc. Psychol 51, 1173–1182. 10.1037/0022-3514.51.6.1173 [DOI] [PubMed] [Google Scholar]
  4. Bauman AE, Reis RS, Sallis JF, Wells JC, Loos RJF, Martin BW, 2012. Correlates of physical activity: Why are some people physically active and others not? Lancet 380, 258–271. 10.1016/S0140-6736(12)60735-1 [DOI] [PubMed] [Google Scholar]
  5. Choi L, Liu Z, Matthews CE, Buchowski MS, 2011. Validation of accelerometer wear and nonwear time classification algorithm. Med. Sci. Sports Exerc 43, 357–364. 10.1249/MSS.0b013e3181ed61a3 [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Christian H, Zubrick SR, Foster S, Giles-Corti B, Bull F, Wood L, Knuiman M, Brinkman S, Houghton S, Boruff B, 2015. The influence of the neighborhood physical environment on early child health and development: A review and call for research. Heal. Place 33, 25–36. 10.1016/j.healthplace.2015.01.005 [DOI] [PubMed] [Google Scholar]
  7. Das KV, Fan Y, French SA, 2017. Park-Use Behavior and Perceptions by Race, Hispanic Origin, and Immigrant Status in Minneapolis, MN: Implications on Park Strategies for Addressing Health Disparities. J. Immigr. Minor. Heal 19, 318–327. 10.1007/s10903-015-0339-1 [DOI] [PubMed] [Google Scholar]
  8. Davison KK, Lawson CT, 2006. Do attributes in the physical environment influence children’s physical activity? A review of the literature. Int. J. Behav. Nutr. Phys. Act 3 10.1186/1479-5868-3-19 [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. De Craemer M, De Decker E, De Bourdeaudhuij I, Vereecken C, Deforche B, Manios Y, Cardon G, 2012. Correlates of energy balance-related behaviours in preschool children: A systematic review. Obes. Rev 13, 13–28. 10.1111/j.1467-789X.2011.00941.x [DOI] [PubMed] [Google Scholar]
  10. Dixon JB, 2010. The effect of obesity on health outcomes. Mol. Cell. Endocrinol 316, 104–108. 10.1016/j.mce.2009.07.008 [DOI] [PubMed] [Google Scholar]
  11. Evenson KR, Catellier DJ, Gill K, Ondrak KS, McMurray RG, 2008. Calibration of two objective measures of physical activity for children. J. Sports Sci 26, 1557–1565. 10.1080/02640410802334196 [DOI] [PubMed] [Google Scholar]
  12. Fan Y, Khattak AJ, 2009. Does urban form matter in solo and joint activity engagement? Landsc. Urban Plan 92, 199–209. 10.1016/j.landurbplan.2009.05.006 [DOI] [Google Scholar]
  13. French SA, Sherwood NE, Mitchell NR, Fan Y, 2017. Park use is associated with less sedentary time among low-income parents and their preschool child: The NET-Works study. Prev. Med. Reports 5, 7–12. 10.1016/j.pmedr.2016.11.003 [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. French SA, Sherwood NE, Veblen-Mortenson S, Lauren Crain A, JaKa MM, Mitchell NR, Hotop AM, Berge JM, Kunin Batson AS, Truesdale K, Stevens J, Pratt C, Esposito L, 2018. Multicomponent obesity prevention intervention in low-income preschoolers: Primary and subgroup analyses of the NET-works randomized clinical trial, 2012–2017. Am. J. Public Health 108, 1695–1706. 10.2105/AJPH.2018.304696 [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Hallal PC, Andersen LB, Bull FC, Guthold R, Haskell W, Ekelund U, Alkandari JR, Bauman AE, Blair SN, Brownson RC, Craig CL, Goenka S, Heath GW, Inoue S, Kahlmeier S, Katzmarzyk PT, Kohl HW, Lambert EV, Lee IM, Leetongin G, Lobelo F, Loos RJF, Marcus B, Martin BW, Owen N, Parra DC, Pratt M, Puska P, Ogilvie D, Reis RS, Sallis JF, Sarmiento OL, Wells JC, 2012. Global physical activity levels: Surveillance progress, pitfalls, and prospects. Lancet 380, 247–257. 10.1016/S0140-6736(12)60646-1 [DOI] [PubMed] [Google Scholar]
  16. Hinkley T, Crawford D, Salmon J, Okely AD, Hesketh K, 2008. Preschool Children and Physical Activity. A Review of Correlates. Am. J. Prev. Med 34 10.1016/j.amepre.2008.02.001 [DOI] [PubMed] [Google Scholar]
  17. Kenyon Henderson K, Song Y, 2008. Can nearby open spaces substitute for the size of a property owner’s private yard? Int. J. Hous. Mark. Anal 1, 147–165. 10.1108/17538270810877772 [DOI] [Google Scholar]
  18. Land TT for P., 2018. Minneapolis, MN City Characteristics [WWW Document]. URL https://www.tpl.org/sites/default/files/city3/city3/tpl.MINN.8_15_18.pdf (accessed 3.1.19).
  19. Little R, 1986. Survey nonresponse adjustments. Int. Stat. Rev 54, 139–157. [Google Scholar]
  20. Marino JA, Fletcher EN, Whitaker RC, Anderson SE, 2012. Amount and environmental predictors of outdoor playtime at home and school: A cross-sectional analysis of a national sample of preschool-aged children attending Head Start. Heal. Place 18, 1224–1230. 10.1016/j.healthplace.2012.08.004 [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Matthews CE, 2005. Calibration of Accelerometer Output for Adults. Med. Sci. Sport. Exerc 37, S512–S522. 10.1249/01.mss.0000185659.11982.3d [DOI] [PubMed] [Google Scholar]
  22. McGill A, 2016. The Shrinking of the American Lawn [WWW Document]. Atl.
  23. MetroGIS, 2018. The Metro Regional Parcel Dataset [WWW Document]. URL https://www.metrogis.org/how-do-i-get/parcel-data.aspx (accessed 1.29.18).
  24. MetroGIS, 2014. MetroGIS Regional Parcel Dataset (Year End 2013) [WWW Document]. URL ftp://ftp.gisdata.mn.gov/pub/gdrs/data/pub/us_mn_state_metrogis/plan_regonal_parcels_2013/metadata/metadata.html (accessed 1.31.18).
  25. MnGeo, 2013. LiDAR Elevation Data for Minnesota [WWW Document]. URL http://www.mngeo.state.mn.us/chouse/elevation/lidar.html#data (accessed 1.8.18).
  26. Sallis JF, Prochaska JJ, Taylor WC, 2000. A review of correlates of physical activity of children and adolescents. Med. Sci. Sports Exerc 32, 963–975. [DOI] [PubMed] [Google Scholar]
  27. Sard B, 2001. Housing Vouchers Should Be a Major Component of Future Housing Policy for the Lowest Income Families Housing Vouchers Should Be a Major Component of Future Housing Policy for the Lowest Income Families The Characteristics of Vouchers Are Well Suited To Me. Cityscape A J. Policy Dev. Res 5, 89–110. [Google Scholar]
  28. Sherwood NE, French SA, Veblen-Mortenson S, Crain AL, Berge J, Kunin-Batson A, Mitchell N, Senso M, 2013. NET-Works: Linking families, communities and primary care to prevent obesity in preschool-age children. Contemp. Clin. Trials 36, 544–554. 10.1016/j.cct.2013.09.015 [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Spurrier NJ, Magarey AA, Golley R, Curnow F, Sawyer MG, 2008. Relationships between the home environment and physical activity and dietary patterns of preschool children: A cross-sectional study. Int. J. Behav. Nutr. Phys. Act 5, 1–12. 10.1186/1479-5868-5-31 [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Terrazas A, 2015. New Home Construction Bulks Up: Denser and Bigger [WWW Document]. Zillow Res URL https://www.zillow.com/research/bigger-homes-smaller-lots-sept-2015-10705/(accessed 9.26.18).
  31. Troiano RP, Berrigan D, Dodd KW, Mâsse LC, Tilert T, Mcdowell M, 2008. Physical activity in the United States measured by accelerometer. Med. Sci. Sports Exerc 40, 181–188. 10.1249/mss.0b013e31815a51b3 [DOI] [PubMed] [Google Scholar]
  32. Trost SG, Loprinzi PD, Moore R, Pfeiffer KA, 2011. Comparison of accelerometer cut points for predicting activity intensity in youth. Med. Sci. Sports Exerc 43, 1360–1368. 10.1249/MSS.0b013e318206476e [DOI] [PubMed] [Google Scholar]
  33. Trost SG, Mciver KL, Pate RR, 2005. Conducting accelerometer-based activity assessments in field-based research. Med. Sci. Sports Exerc 37, 531–543. 10.1249/01.mss.0000185657.86065.98 [DOI] [PubMed] [Google Scholar]
  34. Trost SG, Owen N, Bauman AE, Sallis JF, Brown W, 2002. Correlates of adults’ participation in physical activity: Review and update. Med. Sci. Sports Exerc 34, 1996–2001. [DOI] [PubMed] [Google Scholar]
  35. Uijtdewilligen L, Nauta J, Singh AS, van Mechelen W, Twisk JWR, van der Horst K, Chinapaw MJM, 2011. Determinants of physical activity and sedentary behaviour in young people: a review and quality synthesis of prospective studies. Br. J. Sports Med 45, 896–905. 10.1136/bjsports-2011-090197 [DOI] [PubMed] [Google Scholar]
  36. Valeri L, VanderWeele TJ, 2013. Mediation analysis allowing for exposure-mediator interactions and causal interpretation: Theoretical assumptions and implementation with SAS and SPSS macros. Psychol. Methods 18, 137–150. 10.1037/a0031034 [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Vanderweele TJ, 2014. A unificaiton of mediation and interaction: a four-way decomposition. Epidemiology 25, 749–761. 10.1097/EDE.0000000000000121.A [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. WHO, 2009. Global Health Risks: Mortality and burden of disease attributable to selected major risks. Bull. World Health Organ 87 10.2471/BLT.09.070565 [DOI] [Google Scholar]
  39. Xu H, Wen LM, Rissel C, 2015. Associations of Parental Influences with Physical Activity and Screen Time among Young Children : A Systematic Review 2015. [DOI] [PMC free article] [PubMed] [Google Scholar]

RESOURCES