Associations between residential greenness and birth outcomes across Texas

Associations between residential greenness and birth outcomes across Texas

Environmental Research 152 (2017) 88–95 Contents lists available at ScienceDirect Environmental Research journal homepage: www.elsevier.com/locate/e...

510KB Sizes 0 Downloads 74 Views

Environmental Research 152 (2017) 88–95

Contents lists available at ScienceDirect

Environmental Research journal homepage: www.elsevier.com/locate/envres

Associations between residential greenness and birth outcomes across Texas

crossmark



Leanne Cusack , Andrew Larkin, Sue Carozza, Perry Hystad College of Public Health and Human Sciences, School of Biological and Population Health Sciences, Oregon State University, Corvallis, OR, USA

A R T I C L E I N F O

A BS T RAC T

Keywords: Greenness NDVI Birth outcomes Texas SGA Preterm birth

Background: The amount of greenness around mothers’ residences has been associated with positive birth outcomes; however, findings are inconclusive. Here we examine residential greenness and birth outcomes in a population-based birth cohort in Texas, a state with large regional variation in greenness levels, several distinct cities, and a diverse population. Methods: We used Vital Statistics data to create a birth cohort (n=3,026,603) in Texas from 2000 to 2009. Greenness exposure measures were estimated from full residential addresses across nine months of pregnancy, and each trimester specifically, using the mean of corresponding MODIS satellite 16-day normalized difference vegetation index (NDVI) surfaces at a 250 m resolution, which have not been previously used. Logistic and linear mixed models were used to determine associations with preterm birth, small for gestational age (SGA) and term birth weight, controlling for individual and neighborhood factors. Results: Unadjusted results demonstrated consistent protective effects of residential greenness on adverse birth outcomes for all of Texas and the four largest cities (Houston, San Antonio, Dallas, and Austin). However, in fully adjusted models these effects almost completely disappeared. For example, mothers with the highest ( > 0.52) compared to the lowest ( < 0.37) NDVI quartiles had a 24.4 g (95% CI: 22.7, 26.1) increase in term birth weight in unadjusted models, which was attenuated to 1.9 g (95% CI: 0.1, 3.7) in fully adjusted models. Maternal and paternal race, ethnicity and education had the largest impact on reducing associations. Trimesterspecific greenness exposures showed similar results to nine-month average exposures. Some evidence was seen for protective effects of greenness for Hispanics, mothers with low education and mothers living in low income neighborhoods. Conclusions: In this large population-based study, across multiple urban areas in Texas and diverse populations, we did not observe consistent associations between residential greenness and birth outcomes.

1. Introduction A rapidly growing body of evidence suggests residential greenness (also referred to as green space, or natural environments) is associated with a range of positive health outcomes, including improved mental health (Sugiyama et al., 2008; Thompson et al., 2012; Van den Berg et al., 2010), decreased mortality (Donovan et al., 2013; Mitchell and Popham, 2008; Takano et al., 2002; Villeneuve et al., 2012) and positive birth outcomes (Dadvand et al., 2012a, 2012b, 2012c, 2014; Donovan et al., 2011; Hystad et al., 2014). Pregnancy represents a susceptible time-period for environmental exposures (Fedulov et al., 2008; Jirtle and Skinner, 2007; Merlo et al., 2009) and birth outcomes are important population health indicators as they have substantial impacts immediately as well as over the life-course (Gray et al., 2014). Given the defined exposure period during pregnancy, birth outcomes



may be particularly useful for determining potential influences of residential greenness on health. A limited number of studies have examined associations between residential greenness and birth outcomes (Casey et al., 2016; Dadvand et al., 2012a, 2012c, 2014; Donovan et al., 2011; Ebisu et al., 2016; Hystad et al., 2014; Laurent et al., 2013; Markevych et al., 2014). A meta-analysis summarizing the literature showed that residential greenness was positively associated with birth weight within approximately 100 m buffers (Dzhambov et al., 2014), but other studies observed associations for different buffer sizes (Donovan et al., 2011; Markevych et al., 2014). Overall conclusions for preterm birth and small for gestational age vary and are less consistent than birth weight (Agay-Shay et al., 2014; Casey et al., 2016; Dadvand et al., 2012c; Ebisu et al., 2016; Grazuleviciene et al., 2015). Importantly, most studies used a static measure of greenness (typically from summer

Corresponding author. E-mail address: [email protected] (L. Cusack).

http://dx.doi.org/10.1016/j.envres.2016.10.003 Received 11 June 2016; Received in revised form 13 September 2016; Accepted 7 October 2016 Available online 14 October 2016 0013-9351/ © 2016 Elsevier Inc. All rights reserved.

Environmental Research 152 (2017) 88–95

L. Cusack et al.

and not regional (e.g. > 1 km buffer) greenness levels that may capture larger-scale influences of greenness (e.g. urban core versus suburb influences). Due too computational, time and manuscript length constraints we did not examine NDVI in multiple buffer distances.

months) and none include trimester-specific exposure estimates, which have been essential for examining other environmental exposures, such as air pollution (Lee et al., 2013). In addition, only two studies (Dadvand et al., 2012c; Ebisu et al., 2016) have examined multiple urban areas and it is unknown if residential greenness has a similar effect across multiple urban areas or if associations reflects city-specific urban form characteristics and population dynamics. Finally, while there is evidence that exposure to greenness may have a larger impact on minority and low socio-economic status (SES) populations (Dadvand et al., 2012a, 2012b, 2014), the existing literature lacks the diversity to fully examine populations that may benefit the most from greenness exposures. Here we examine associations between residential greenness and birth outcomes, including preterm birth, SGA and term birth weight, for more than three million births occurring in all urban areas of Texas from 2000 to 2009. We assess the effects of greenness throughout pregnancy by looking at trimester specific models, and examine associations for the four largest cities of Texas separately and by individual and neighborhood SES and race/ethnicity. Texas has substantial regional variation in greenness levels, several large and distinct cities, and diverse populations based on SES, race and ethnicity that allow us to fully explore how residential greenness may be associated with birth outcomes and how individual, neighborhood and city-level factors may modify potential associations.

2.3. Additional spatial exposures and neighborhood measures Additional spatial exposure measures were derived to examine traffic air pollution as a potential causal pathway and to control for potential neighborhood contextual confounding factors. To derive a spatial measure of traffic air pollution, we applied a previously developed national land use regression (LUR) model for NO2 which includes satellite NO2 estimates and eight land use variables (Novotny et al., 2001). The LUR estimates were fused with monthly groundbased monitoring data to estimate average NO2 exposures during each trimester and the entire pregnancy period (Bechle et al., 2015). We chose to include NO2 air pollution since this pollutant demonstrates fine-scale spatial variability (e.g. (Brauer et al., 2007) that is most likely influenced by residential greenness levels (Hystad et al., 2014). Neighborhood SES variables and population density were downloaded at the census tract level from the American Fact Finder website (http:// factfinder.census.gov/faces/nav/jsf/pages/index.xhtml) for 2000 and 2010 census and linked to each birth within the census tract. For example, we accessed the percent of the population that was “white alone, not Hispanic or Latino” in order to derive out percent white variable. Variables included percent Hispanic; percent White; percent adult population without a high school diploma; and median household income; percentage of the population below the poverty line; and percent unemployment. Population density was used to control for potential confounding factors associated with urban form, such as inner city cores or suburban areas.

2. Materials and methods 2.1. Birth cohort Data for all births among Texas residents between 2000 and 2009 (3,899,627 birth records) were provided by the Texas Vital Statistics program. A total of 3,413,787 (88%) birth records could be geocoded to full residential address. We restricted our analysis to mothers (n=3,233,236) living in metropolitan areas of Texas (population ≥50,000) by using the Core-Based Statistical Areas classification system. We further excluded any infants whose birth weight was not between 500 and 5000 g (n=11,994), non-singleton births (n=126,818), and those with missing covariates (n=67,821). Our final sample for the analysis of residential greenness included 3,026,603 births.

2.4. Individual covariates Individual covariates were captured from the birth certificate data provided by the Texas Vital Statistics program. Covariates were determined a-priori. Maternal as well as paternal (if available) covariates included: age; smoking (yes/no during pregnancy); education (less than high school diploma, high school diploma, some college, college degree and post graduate degree); and race/ethnicity (White, Black, Asian, Hispanic, Other). Pregnancy-related variables included method of delivery; parity (first birth or not); prenatal care received; gestational age; baby’s sex; and month and year of birth. When available, maternal and paternal characteristics were both included in subsequent analyses to better capture household SES influences.

2.2. Residential greenness exposure assessment Estimates of residential greenness were derived from MODIS satellite NDVI imagery and maternal residence at the time of birth. NDVI is an indicator of greenness based on land surface reflectance of visible and near infrared parts of spectrum (Weier and Herring, 2000). Values range from −1 to1 with the higher numbers indicating more greenness. This measure does not discern types of vegetation but represents all green vegetation. We used 16-day, 250 m resolution composite images derived from MODIS (https://lpdaac.usgs.gov/) to estimate trimester-specific and entire pregnancy length greenness exposures based on maternal residences reported at time of birth. Estimates are based on averaging composite images over the time interval of interest (e.g. trimester 1, 2, 3, and entire pregnancy) with each image weighted based on the number of estimated pregnancy days covered by the image (Eq. (1)).

2.5. Analysis We investigated the association between NDVI and three birth outcomes of interest: 1) birth weight for full term babies (≥37 weeks of gestation); 2) odds of preterm birth ( < 37 weeks of gestation); and 3) odds of being SGA (defined as the bottom decile in birth weight, stratified by gestational age, gender, and year of birth) (Groom et al., 2007; Kliegman, 2011). Logistic and linear mixed regression models were developed to examine NDVI and birth outcomes associations. Random intercepts based on metropolitan statistical areas (MSAs) were included in the statistical models to adjust for between MSA differences in birth outcomes and potential residual confounding. Fully adjusted models included all individual covariates described previously as well as neighborhood variables, NO2 air pollution concentrations and population density. Term birth weight analyses also included a categorical variable for the estimated gestational age at birth. Associations between greenness exposures and birth outcomes are presented corresponding to quartiles (defined by the distribution of NDVI across Texas, with the lowest NDVI quartile as the reference) as we observed non-linearity in most models. We first present models for

n

NDVIresidencei =

∑ j =1 NDVI for monthi *time Rangej number of days in time interval of inteerst

(1)

where time Rangej=number of estimated gestational days within MODIS NDVI composite j and time interval of interest (e.g. trimester 1, pregnancy, etc). We chose 250 m to measure greenness as we are interested in capturing local residential greenness levels outside the mother’s home 89

Environmental Research 152 (2017) 88–95

L. Cusack et al.

all urban areas (≥50,000 people) of Texas (n=3,026,603) with entire pregnancy and trimester specific exposure estimates. Next, we present models specific to the four largest urban areas in Texas (Houston, n=832,722; Dallas/Fort Worth, n=906,053; San Antonio, n=276,013; and Austin, n=217,824). Stratified models were also explored by important individual and contextual level variables, including education, race/ethnicity and neighborhood household income. We conducted a number of additional sensitivity analyses. First, incremental models were run to evaluate how the associations between residential greenness and birth outcomes changed when individual and neighborhood covariates were added to the model. Second, state-wide models were explored without MSA random effects to examine the impacts of absolute NDVI levels on birth outcomes without accounting for between urban area differences. Third, models were explored using only maternal variables for comparison with previous studies. Fourth, models were stratified by birth seasons to explore periods with the highest (Nov-Jan) and lowest (April-June) average NDVI during pregnancy. Finally, models were run on a sub-cohort (years 2000– 2005) when pre-pregnancy weight (BMI) was available. All analyses were conducted with SAS 9.4 (Cary, NC). Our study protocol was approved by Oregon State University and the Texas Department of State Health Services human subjects review boards.

314,019 births (10.38%) were classified as SGA, and the mean term birth weight was 3,322 g. Differences in birth outcomes and individual covariate information by residential greenness quartiles were seen especially for maternal and paternal race/ethnicity and neighborhood characteristics. For example, 29.7% of women in the lowest NDVI quartile were in the lowest income category ( < $29,500) compared to only 12.6% of women in the highest NDVI quartile. For maternal race/ ethnicity, 55% of women in the lowest NDVI quartile were Hispanic compared to 42% in the highest NDVI quartile. Distribution of NDVI exposures across the state of Texas and MSAs are shown in Fig. 1. Average NDVI exposure across all urban areas was 0.45 with an IQR of 0.15. As expected, greenness varied regionally, with greater mean NDVI in eastern Texas. Average NDVI exposure in each of the MSAs is as follows: Houston (mean=0.50, IQR=0.14), Austin (mean=0.48, IQR=0.12), San Antonio (mean=0.46, IQR=0.11) and Dallas (mean=0.42, IQR=0.12). NDVI exposure varied by birth season (and preceding seasons of pregnancy), with greater NDVI exposure for winter births, (Oct-Jan, mean NDVI =0.47) compared to late spring/early summer (April-July, mean NDVI=0.43). Pearson correlations between average NDVI for the entire pregnancy period and the first, second and third trimester was 0.87, 0.93 and 0.87, respectively.

3. Results

3.2. Residential greenness and birth outcomes

3.1. Descriptive statistics

Table 2 summarizes unadjusted and adjusted associations between NDVI and birth outcomes (across pregnancy and by trimester) for all urban areas in Texas. A positive dose-response relationship was observed for mean NDVI in the 9-months of pregnancy as well as for each trimester in the unadjusted models for term birth weight and SGA. The highest NDVI quartile was associated with a 24.4 g (95% CI:

Descriptive statistics of the cohort with complete covariate and exposure information (n=3,026,626), as well as participants stratified by NDVI quartiles for the entire state of Texas, are summarized in Table 1. A total of 274,331 births (9.06%) were preterm ( < 37 weeks), Table 1 Characteristics of the Texas birth cohort (N=3,026,626). Entire cohort

Residential greenness (NDVI) quartiles* Q1 ( < 0.37)

Birth outcomes Preterm birth ( < 37 weeks), n (%) 263,194 (8.61) 67,337 (8.79) Small for gestational age, n (%) 293,653 (9.60) 78,798 (10.29) Birth weight (g) at full term 3322.05 3293.15 Covariates Female sex, (%) 48.83 48.85 Nulliparous (%) 39.33 40.09 Maternal age (%) < 19 yrs 10.31 10.85 20–29 yrs 53.42 56.30 30–34 yrs 23.47 22.13 35–39 yrs 10.66 9.05 ≥ 40 yrs 2.14 1.67 Maternal education < high school diploma 25.91 27.16 High school diploma 26.68 26.31 Some college 21.53 22.24 College degree 15.98 15.48 Post Graduate degree 9.15 7.99 Maternal ethnicity Hispanic 48.16 54.77 Non-Hispanic Asian 3.66 4.47 Non-Hispanic Black 9.28 8.76 Non-Hispanic Other 1.50 1.62 Non-Hispanic White 37.40 30.38 Neighborhood income quintiles (median household income), n (%) < 29,500 19.95 29.66 29,500–39,000 19.81 19.30 39,000–50,000 19.91 16.63 50,000–68,000 19.94 16.50 > 68,000 20.38 17.90 NO2 air pollution (ppb) 10.89 11.76

Q2 (0.37–0.45)

Q3 (0.45–0.52)

Q4 ( > 0.52)

65,504 (8.56) 73,838 (9.64) 3323.69

65,149 (8.55) 71,665 (9.41) 3330.88‚

65,204 (8.52) 69,352 (9.06) 3340.58

48.83 39.55

48.87 38.86

48.78 38.81

9.72 54.05 23.81 10.44 1.98

10.40 52.19 23.93 11.22 2.27

10.26 51.14 24.01 11.94 2.65

25.39 26.62 21.43 16.53 9.27

26.19 26.86 20.77 15.99 9.50

24.88 26.91 21.67 15.93 9.87

48.08 4.49 10.12 1.83 35.48

48.26 3.36 9.46 1.49 37.43

41.53 2.35 8.78 1.05 46.30

19.75 19.46 19.39 20.07 21.32 11.24

17.74 20.15 20.86 20.22 21.01 10.87

12.61 20.34 22.75 22.97 21.31 9.71

*Q1- lowest greenness, Q4-highest greenness, OR- odds ratio, CI- Confidence Interval, NO2 - Nitrogen dioxide, NDVI- Normalized Difference Vegetation Index.

90

Environmental Research 152 (2017) 88–95

L. Cusack et al.

Fig. 1. a. Example of the spatial distribution of greenness in urban areas of Texas, with mean NDVI for 2000. b. Spatial distribution of Houston, Texas showing NDVI and highways.

relationship between term birth weight and the highest quartile of NDVI in adjusted models, though it is only significant in Austin. In Houston there is an increase in birth weight in the highest quartile, 8.4 g (95% CI: 5.1; 11.7), as compared to quartile 1 in the fully adjusted model. The odds of having a SGA infant were slightly lower in quartile 4 versus quartile 1 in each city, though none were statistically significant. The odds of having a preterm birth decreased as NDVI quartiles increased in both San Antonio and Houston, and decreased in Dallas and Austin (marginally statistically significant in fully adjusted models). Stratified models of 9 month NDVI exposure are presented in Table 4 for all urban areas of Texas (Supplemental materials Tables 1–

22.7; 26.1) increase in term birth weight and ORs of 0.92 (95% CI: 0.91;0.93) and 0.99 (95% CI: 0.97;1.00) for SGA and preterm birth. However, in fully adjusted models these associations were attenuated. For example, the highest NDVI quartile was associated with a 1.9 g (95% CI: 0.1; 3.7) increase in term birth weight and ORs of 0.99 (95% CI: 0.97;1.00) and 1.01 (95% CI: 0.99;1.02) for SGA and preterm birth. Trimester specific models demonstrated similar results across all outcomes for unadjusted and adjusted models (Table 2). The city-specific models showed heterogeneous associations between greenness and birth outcomes, although similar patterns of attenuation to the overall model were observed for each urban area (Table 3). In all cities expect for Houston, there was a negative

Table 2 Unadjusted and adjusted models for all urban areas of Texas examining greenness (NDVI) and pregnancy outcomes (n=3,026,603).

All Texas Q1 ( < 0.37) Q2 (0.37–0.45) Q3 (0.45–0.52) Q4 ( > 0.52) Trimester 1 Q1 ( < 0.37) Q2 (0.37–0.45) Q3 (0.45–0.52) Q4 ( > 0.52) Trimester 2 Q1 ( < 0.37) Q2 (0.37–0.45) Q3 (0.45–0.52) Q4 ( > 0.52) Trimester 3 Q1 ( < 0.37) Q2 (0.37–0.45) Q3 (0.45–0.52) Q4 ( > 0.52)

Preterm birth ( < 37weeks)

Small for gestational age

Term birth weight (g)

Unadjusted

Fully adjusted

Unadjusted

Fully adjusted

Unadjusted

Fully adjusted

OR (95% CI)

OR (95% CI)

OR (95% CI)

OR (95% CI)

β (95% CI)

β (95% CI)

1 1.01 (1.00,1.03) 1.00 (0.99,1.01) 0.99 (0.97,1.00)

1 1.01 (1.00,1.02) 1.00 (0.99,1.02) 1.01 (0.99,1.02)

1 0.98 (0.97,0.99)* 0.97 (0.96,0.98)* 0.92 (0.91,0.93)*

1 0.99 (0.98,1.00) 0.99 (0.98,1.00) 0.99 (0.97,1.00)

0 8.2 (6.7,9.8)* 12.2 (10.6,13.8)* 24.4 (22.7,26.1)*

0 2.7 (1.2,4.3)* 2.5 (0.9,4.2)* 1.9 (0.1,3.7)*

1 1.02 (1.00,1.03) 1.02 (1.01,1.03) 1.02 (1.01,1.03)

1 1.01 (0.99,1.02) 1.01 (1.00,1.02) 1.01 (0.99,1.02)

1 0.97 (0.96,0.98)* 0.95 (0.94,0.96)* 0.92 (0.91,0.93)*

1 1.00 (0.99,1.01) 0.99 (0.98,1.01) 0.99 (0.97,1.00)

0 8.9 (7.5,10.5)* 13.8 (12.2,15.3)* 20.9 (19.4,22.4)*

0 0.5 (−1.1,2.01) 1.7 (−0.04,3.4) 0.2 (−1.7,2.1)

1 0.99 (0.98,1.00) 0.98 (0.97,0.99) 0.97 (0.96,0.98)*

1 1.00 (0.99,1.01) 1.00 (0.99,1.02) 1.01 (0.99,1.02)

1 0.98 (0.97,0.99)* 0.98 (0.97,0.99)* 0.95 (0.94,0.96)*

1 0.99 (0.98,1.00) 1.00 (0.98,1.01) 0.99 (0.98,1.01)

0 5.4 (3.9,6.9)* 8.0 (6.5,9.6)* 15.8 (14.3,17.4)*

0 1.4 (−0.2,2.9) 1.0 (−0.7,2.7) 1.7 (−0.2,3.6)

1 0.99 (0.97,1.00) 0.99 (0.98,1.01) 0.97 (0.96,0.99)*

1 1.00 (0.99,1.02) 0.99 (0.98,1.01) 1.00 (0.99,1.02)

1 0.97 (0.96,0.98)* 0.96 (0.95,0.97)* 0.94 (0.93,0.95)*

1 0.99 (0.98,1.01) 0.99 (0.98,1.00) 0.99 (0.97,1.01)

0 7.9 (6.4,9.4)* 11.8 (10.3,13.4)* 18.5 (17.0,20.0)*

0 2.2 (0.6,3.8)* 2.8 (1.0,4.5)* 2.3 (0.4,4.2)*

*Significance level 0.05. **Fully adjusted models include: individual characteristics - mother and father’s age, method of delivery, smoking, gestational age, baby’s sex, year, month, mother and father’s education, prenatal care, parity, race/ethnicity and neighborhood characteristics – population density, NO2, household income, household education, unemployment, % below poverty, % white, % Hispanic. Multivariate linear mixed regression analysis was used to examine birth weight; logistical regression was used to examine the odds of preterm birth and the odds of being small for gestational age.

91

Environmental Research 152 (2017) 88–95

L. Cusack et al.

Table 3 Unadjusted and adjusted models for specific urban areas of Texas examining greenness (NDVI) and pregnancy outcomes for the entire length of pregnancy (n=3,026,603).

Houston Q1 ( < 0.37) Q2 (0.37–0.45) Q3 (0.45–0.52) Q4 ( > 0.52) San Antonio Q1 ( < 0.37) Q2 (0.37–0.45) Q3 (0.45–0.52) Q4 ( > 0.52) Dallas Q1 ( < 0.37) Q2 (0.37–0.45) Q3 (0.45–0.52) Q4 ( > 0.52) Austin Q1 ( < 0.37) Q2 (0.37–0.45) Q3 (0.45–0.52) Q4 ( > 0.52)

Preterm birth ( < 37weeks)

Small for gestational age

Term birth weight (g)

Unadjusted

Fully adjusted

Unadjusted

Fully adjusted

Unadjusted

Fully adjusted

OR (95% CI)

OR (95% CI)

OR (95% CI)

OR (95% CI)

β (95% CI)

β (95% CI)

1 1.02 (0.99,1.05) 1.02 (0.99,1.04) 1.02 (0.99,1.04)

1 1.03 (0.99,1.06) 1.03 (1.00,1.06) 1.04 (1.01,1.07)*

1 0.94 (0.92,0.97)* 0.91 (0.89,0.93)* 0.88 (0.86,0.90)*

1 0.99 (0.96,1.01) 0.97 (0.95,1.00) 0.98 (0.95,1.00)

0 18.8 (15.3,22.3)* 27.1 (23.8,30.4)* 37.0 (33.9,40.1)*

0 6.6 (3.1,10.2)* 6.7 (3.2,10.1)* 7.1 (3.7,10.4)*

1 1.04 (1.00,1.08) 0.99 (0.95,1.03) 0.91 (0.87,0.95)*

1 1.04 (1.00,1.09) 1.05 (1,00.1.10) 1.03 (0.97,1.08)

1 0.97 (0.93,1.01) 0.93 (0.90,0.97)* 0.84 (0.81,0.88)*

1 1.01 (0.96,1.05) 0.99 (0.95,1.03) 0.96 (0.91,1.01)

0 9.2 (3.9,14.6)* 21.3 (16.1,26.5)* 40.1 (34.4,45.7)*

0 −3.3 (−8.8,2.2) −3.6 (−9.3,2.2) −5.1 (−11.6,1.3)

1 1.02 (1.00,1.04) 1.01 (0.99,1.03) 1.00 (0.97,1.02)

1 1.00 (0.98,1.02) 0.97 (0.95,1.00) 0.98 (0.95,1.01)

1 1.00 (0.98,1.02) 1.02 (1.00,1.04) 0.96 (0.94,0.99)

1 0.98 (0.96,1.00) 1.00 (0.98,1.02) 0.98 (0.96,1.01)

0 3.60 (1.2,5.9)* 4.00 (1.5,6.5)* 11.5 (8.2,14.7)*

0 2.7 (0.4,5.1)* 1.8 (−0.8,4.5) −2.0 (−5.5,1.4)

1 0.94 (0.89,0.99)* 0.95 (0.91,1.01) 0.90 (0.85,0.95)*

1 0.96 (0.91,1.02) 0.96 (0.91,1.02) 0.94 (0.88,1.00)

1 0.94 (0.89,0.98)* 0.95 (0.90,0.99)* 0.86 (0.82,0.91)*

1 0.99 (0.94,1.05) 1.01 (0.96,1.07) 0.98 (0.93,1.04)

0 11.5 (5.0,18.0)* 9.3 (2.9,15.7)* 24.8 (18.4,31.2)*

0 −4.9 (−11.4,1.7) −9.7 (−16.3,−3.2)* −10.1 (−16.9,−3.4)*

*Significance level 0.05. **Fully adjusted models include: individual characteristics - mother and father’s age, method of delivery, smoking, gestational age, baby’s sex, year, month, mother and father’s education, prenatal care, parity, race/ethnicity and neighborhood characteristics – population density, NO2, household income, household education, unemployment, % below poverty, % white, % Hispanic. Multivariate linear regression analysis was used to examine birth weight; logistical regression was used to examine the odds of preterm birth and the odds of being small for gestational age.

differences. Additional sensitivity analyses were explored using only maternal variables for comparison with previous studies (Agay-Shay et al., 2014; Dadvand et al., 2012a; Dadvand et al., 2012b, 2014; Grazuleviciene et al., 2015; Hystad et al., 2014; Laurent et al., 2013), but results were nearly identical. Including pre-pregnancy BMI in the model showed a −0.2 g (95% CI: −1.0; −0.6) decrease in birth weight for a 0.1 increase in NDVI. Finally, models were stratified by birth seasons having the highest (Nov–Jan) and lowest (April–June) average NDVI during pregnancy. There was a 2.7 g (95% CI: −0.8; 6.3) increase in birth weight in the winter seasons and a 1.5 g (95% CI: −2.2; −5.2) decrease in birth weight in the highest quartile.

3 includes stratified models by each urban area and race/ethnicity). Large differences in the association between residential greenness and term birth weight were observed by race. Across all of Texas we see an increase by 1.1 g (95% CI: −2.3; 4.5) for Non-Hispanic Whites in the greenest quartile compared to the least green quartile, but for NonHispanic Blacks there is a decrease of 11.0 g (95% CI: −17.4; −4.6), while Hispanics had an increase of 11.1 g (95% CI: 8.3; 13.9) respectively. However, in city specific models we see neither consistent patterns nor a dose-response relationship (see Supplemental materials, Table 1). At the neighborhood level, the largest associations between greenness and term birth weight were found for mothers living in areas with the lowest education quintile (based on % less than high school education) (5.8 g, 95% CI: 2.4;9.2) and for those in the lowest income quintile (11.5 g, 95% CI: 7.1;15.8). Hispanics had a consistent increase in term birth weight while Non-Hispanic Blacks consistently had a decrease in birth weight associated with greenness when looking at individual characteristics such as education, as well as neighborhood characteristics such as income.

4. Discussion We investigated the impacts of residential greenness on birth outcomes using a cohort of approximately three million babies born in urban areas (population ≥50,000) of Texas between 2000 and 2009. Although we consistently observed strong positive impacts of greenness on birth weight and SGA in unadjusted models, most of this relationship disappeared with adjustments for individual characteristics, especially race/ethnicity. Similar results were observed for trimester specific greenness exposures and for the four largest urban areas of Texas. We observed heterogeneous associations in models stratified by mothers’ education and race/ethnicity, with consistent positive associations between Hispanics and term birth weight and negative associations for blacks. Our finding that the amount of greenness around mothers residences is not associated with birth outcomes (preterm birth, SGA and term birth weight) is contrary to the findings of most existing studies (Dadvand et al., 2012a; Dadvand et al., 2012c, 2014; Donovan et al., 2011; Hystad et al., 2014; Laurent et al., 2013). While these studies used different exposure assessment measures to capture residential greenness, they generally observed increases in term birth weight, and decreases in preterm birth and SGA, with increasing greenness levels, although results are more mixed for preterm birth and SGA. For example, in L.A. and Orange counties, USA an increase in birth weight

3.3. Sensitivity analyses Incremental models (Fig. 2) demonstrate that the inclusion of fathers and mothers race/ethnicity had the largest influence of model attenuation for the Texas model as well as city-specific models. The inclusion of NO2 air pollution did not change associations between greenness and birth outcomes, suggesting NO2 is not a pathway of influence. Additionally, the NO2 LUR model we used contained tree canopy cover within 1 km as a predictor variable, and even though this variable contributed only 4% to model prediction this could lead to over-adjustment in our models. However, models with and without NO2 exposures were virtually identical. The inclusion of the MSA random effect in the Texas model also had a substantial influence. Fully adjusted models without MSA random effect showed statistically significant associations but were driven primarily by low NDVI in the North West of Texas. Since our measure of NDVI is capturing localscale (250 m) greenness it was appropriate to control for between MSA 92

Environmental Research 152 (2017) 88–95

L. Cusack et al.

Table 4 Adjusted regression coefficients (95% confidence intervals) across Texas examining greenness (NDVI) by comparing Quartile 4 to Quartile 1*** and pregnancy outcomes for entire length of pregnancy.

Education < High school High school > High school < High school*White < High school*Black < High school*Hispanic High school*White High school*Black High school*Hispanic > High school*White > High school*Black > High school*Hispanic Race/ethnicity Non-Hispanic White Non-Hispanic Black Non-Hispanic Other Non-Hispanic Asian Hispanic Neighborhood income < $29,500 > $66,000 < $29,500 *White < $29,500 *Black < $29,500 *Hispanic > $66,000*White > $66,000*Black > $66,000*Hispanic Population density (quintiles) Lowest density Highest density

N1

Preterm birth OR (95% CI)

SGA OR (95% CI)

Term birth weight (g) β (95% CI)

790,667 762,679 1,202,600 85,902 44,863 580,183 108,130 90,052 685,735 597,228 52,352 301,085

1.00 (0.97, 1.03) 1.00 (0.97, 1.03) 1.00 (0.98, 1.03) 1.07 (0.99,1.16) 1 (0.92,1.09) 0.97 (0.94,1) 0.93 (0.88,0.98)* 0.99 (0.93,1.06) 1.04 (1,1.09) 0.98 (0.95,1.02) 1.05 (0.99,1.11) 1 (0.95,1.05)

0.98 (0.95, 1.00) 1.00 (0.97, 1.03) 0.98 (0.96, 1.00) 1 (0.94,1.08) 1.05 (0.97,1.12) 0.95 (0.93,0.98) 0.98 (0.93,1.02) 1.01 (0.96,1.07) 0.98 (0.94,1.01) 0.95 (0.92,0.98) 1.08 (1.03,1.14) 0.95 (0.91,0.99)

5.8 (2.4, 9.2)* −2.6 (−6.2,1.0) 3.1 (0.5, 5.8)* −14.7 (−25.3,−4.2)* −15.7 (−28.5,−2.9)* 11.8 (7.9,15.6)* −6.2 (−6.2,3.2) −8.4 (−17.7,0.9) 3.3 (−1.7, 8.3) 4.1 (0.5,7.8) −10.4 (−18.5, −2.4)* 15.8 (10.5,21.2)

782,355 213,546 26,711 65,210 1,233,314

0.98 1.04 1.01 0.96 1.00

(0.95, (0.99, (0.85, (0.86, (0.98,

0.97 1.05 0.95 0.94 0.96

1.1 (−2.3, 4.5) −11 (−17.4, −4.6)* 13.5 (−4.3, 31.4) 5.9 (−5.3, 17.1) 11.1 (8.3, 13.9)

566,770 533,344 51,330 68,774 465,689 343,633 36,204 107,532

1.02 0.98 0.93 1.01 1.01 0.95 1.18 0.98

(0.99, 1.06) (0.94, 1.01) (0.84,1.04) (0.93,1.09) (0.97,1.06) (0.91,1) (1.04,1.33) (0.9,1.06)

1.01) 1.09) 1.21) 1.08) 1.03)

0.98 (0.94,1.02) 1.01 (0.98, 1.05)

(0.94, (1.01, (0.84, (0.89, (0.94,

1.00) 1.10) 1.07) 1.02) 0.98)

0.96 (0.93, 0.99)* 0.98 (0.95, 1.02) 0.89 (0.81,0.98)* 1.02 (0.96,1.09) 0.95 (0.91,0.99)* 0.97 (0.92,1.01) 1.13 (1.01,1.26) 0.96 (0.9,1.03)

11.5 (7.1, 15.8)* 1.0 (−2.9, 5.1) 13.0 (0.1,26)* −11.7 (−22.6,−0.9)* 17.0 (11.8, 22.3)* 3.1 (−1.8,8.1) −13.7 (−29.1,1.7) 9.1 (0.3,18)*

1.01 (0.97, 1.04) 0.97 (0.94, 1.00)

−5.5 (−9.9,−1.2)* 3.7 (−0.4,7.8)

*Significance level 0.05. **Fully adjusted models include: individual characteristics - mother and father’s age, method of delivery, smoking, gestational age, baby’s sex, year, month, mother and father’s education, prenatal care, parity, race/ethnicity and neighborhood characteristics – population density, NO2, household income, household education, unemployment, % below poverty, % white, % Hispanic. Multivariate linear mixed regression analysis was used to examine birth weight; logistical regression was used to examine the odds of preterm birth and the odds of being small for gestational age. ***Quartile 1 – lowest greenness, Quartile 4 – highest greenness.

health associations observed in the literature can be extrapolated to other regions. Only one other study conducted in Pennsylvania, found similar results to ours. Casey et al (2016) used MODIS 250 m 16-day composite images with a 250 m radius and found no association between continuous NDVI and birth outcomes in the boroughs and townships of Pennsylvania (Casey et al., 2016). They found a protective effect of greenness in cities but it was not statistically significant. We used MODIS derived NDVI, while most studies use mean NDVI at different buffer distances calculated from the 30 m resolution from Landsat, (typically for one period in time (July)). We chose to use the 250 m MODIS 16-d composite image product to capture important temporal differences since our study spans ten years and the entire state of Texas. The computational power and time necessary for examining such a large geographical area made MODIS invaluable, and it has been used in other greenness health studies covering large geographical extents (James et al., 2016). We focused on local greenness (i.e. NDVI average for a 250 m pixel) instead of larger buffer distances since we hypothesized that the local impact of greenness is more relevant to birth outcomes (i.e. buffering of air pollution, noise and heat and psychological and psychosocial influences). Apart from the differing spatial and temporal resolution, the differences in NDVI values found between studies may also be partly due to the different bandwidth captured by Landsat (Band 3: 631–692 nm, Band 4: 772– 898 nm) and MODIS (Band 1: 620–670 nm, Band 2: 841–876 nm) satellites (Boccardo et al., 2006). However, because NDVI is a ratio, it cancels out a large portion of signal variations that can be attributed to calibration, noise and the changing irradiance conditions that accom-

of 4.8 g (95% CI: 1.93–7.72) was observed per 0.12 increase in LandSat 30 m NDVI within a 50 m buffer (Laurent et al., 2013), while in Vancouver, Canada an increase in LandSat 30 m NDVI of 0.1 within 250 m was associated with an increase in preterm birth, SGA and term birth weight of 0.95 (95% CI: 0.91;0.99), 0.97 (95% CI: 0.94;1.00), and 20.6 g (95% CI:16.5; 24.7), respectively (Hystad et al., 2014). Reductions in the odds of being SGA but not preterm birth were also observed with increases in surrounding tree cover within 50 m in Portland, USA (Donovan et al., 2011). In our current study we did not observe statistically significant associations between MODIS 250 m NDVI, both using averages for the entire nine-month pregnancy period and trimester specific averages, and birth outcomes. One potential reason for these contrary findings may be the ecological environment of Texas and differences in vegetation, trees and general green spaces compared to other study areas. Due to the size of our study we were able to examine associations for different urban areas, although these cities had fairly similar average greenness levels. We observed very consistent results in our unadjusted (strong positive dose-response association) and adjusted (no association) models. Most studies that have examined greenness and birth outcomes were limited to a single region or city (Dadvand et al., 2012a, 2014; Donovan et al., 2011; Hystad et al., 2014; Laurent et al., 2013), except for a study in four urban areas of Spain that observed consistent increases in birth weight with LandSat 30 m NDVI within 100, 250 and 500 m (Dadvand et al., 2012c). Examining the impact of greenness across multiple urban areas is an important next step to determine specific pathways and to reduce potential spatial confounding. Further studies are needed in urban areas within different ecological zones to determine if greenness and 93

Environmental Research 152 (2017) 88–95

L. Cusack et al.

lower SES and visible minority groups. Our models stratified by race/ ethnicity, education and neighborhood characteristics demonstrated heterogeneous findings. Associations varied by race/ethnicity, with protective effects for Hispanics mothers and harmful effects for black mothers. These associations also varied by different urban areas. For example, there was a decrease in birth weight for Non-Hispanic Blacks of −30.1 g (95% CI: −59.2; −1.0) in the greenest quartile area in San Antonio compared to the lowest greenness quartile, but in Austin there was an increase of 11.6 g (95% CI: −20.7; 43.9) compared to the lowest greenness quartile. While these findings suggest there are specific urban and population characteristics modifying the impacts of greenness, we did not observe consistent patterns and few studies are available to compare these findings to. A study by Dadvand et al. (2014) examined the impact of green space on birth weight in the Born in Bradford study in England, where 40% of the population was White British and 45% was of Pakistani origin. In unadjusted analyses they found a 78.4 g (95% CI: 62.3; 94.6) increase in birth weight associated with a 0.18 increase in average NDVI within 50 m. When individual covariates were added into the model, the effect size decreased to 50.1 g and when race was added to the model the effect was reduced to an 18.0 g (95% CI: 3.7;32.3) increase in birth weight. When they stratified by race the association remained for Whites but disappeared for individuals of Pakistani origins (Dadvand et al., 2014). Consistent with previous studies, we also observed the strongest beneficial effect among mothers in the lowest neighborhood income quartiles as well as those with less than a high school education; however, the effects were much smaller in magnitude compared to the previous studies. For example, within the lowest education group (less than high school education) we observed a 5.8 g (95% CI: 2.4;9.2) increase in the highest compared to the lowest income quartile across Texas. Within the lowest neighborhood income quintile ( < $29,500) we found an 11.5 g (95% CI: 7.1;15.8) increase in birth weight for the greenest quartile as compared to the least green quartile. Similarly, in Barcelona Spain, Dadvand et al. (2014) observed the largest increase in birth weight (38.9 g; 95% CI: 13.6;64.3 for a 0.18 increase in NDVI) for the most economically deprived population group. A study conducted in Israel also found the largest increase in birth weight (26.0 g; 95%CI: 6.8,45.1) per IQR was for those in the lowest quartile of SES (Agay-Shay et al., 2014). These results highlight the complex spatial patterns between race/ethnicity, SES, greenness and other spatial confounding factors that may influence birth outcomes. While our study has several important strengths (represents the largest study conducted to date (both in terms of population and geographic size), includes temporal variation in greenness exposure assessment, and captures diverse urban areas and populations) several limitations should be highlighted. First, we used the MODIS 250 m NDVI measure, which has not been used in prior studies. While MODIS has a larger pixel size compared to Landsat, we selected MODIS NDVI because of the robust quality assurance standards of MODIS composite imagery and availability of 16 day MODIS NDVI imagery over large spatial and temporal extents. Second, the NDVI measure in general represents a broad surrogate of a number of different potential exposures that could influence birth outcomes (e.g. reductions in environmental exposures, increased physical activity, psychosocial and psychological exposures). While we were able to evaluate inter-urban differences in greenness and birth outcomes associations as well as the influence of traffic air pollution, future studies should examine and test specific hypothesized causal pathways. Third, our individual demographic covariates were limited to those available through Vital Statistics records. We included fathers’ demographic information in our models, which also slightly attenuated NDVI associations, suggesting that father’s information plays an important role in capturing potential confounding factors. Lastly, information was not available on whether families moved during pregnancy and this could cause a misclassification of exposure.

Fig. 2. Incremental models of exposure to residential greenness and birth outcomes by urban areas. Models at the far right are fully adjusted models including individual covariates, neighborhood covariates, population density and NO2, 2000–2009 (n=3,026,603). * Individual covariates include: - mother and father’s age, method of delivery, smoking, gestational age, baby’s sex, year, month, mother and father’s education, prenatal care, parity. Neighborhood characteristics –household income, household education, unemployment, % below poverty, % white, % Hispanic.

pany the changing sun angles, topography, clouds/shadows and atmospheric conditions (Boccardo et al., 2006). Future work will compare city-specific results using MODIS and LandSat NDVI as well as other measures of residential greenness since deriving LandSat NDVI for the entire state of Texas over a nine-year period is not feasible. Nevertheless, the MODIS 250 m NDVI captures one measure of local residential greenness, which was strongly related to birth outcome in unadjusted models (due to confounding by race/ethnicity and education) but not associated with birth outcomes in adjusted models this large population based study. The consistent positive dose-response relationship observed for greenness and positive birth outcomes in unadjusted models deserves further attention. Incremental models demonstrated that the variable which drove the attenuation of our results was primarily mother and fathers race/ethnicity, followed by education and neighborhood median household income. In the four largest urban areas, Hispanic, Asian and Black mothers were more likely to live in neighborhoods with the lowest quartile of greenness compared to white mothers. For example, in Dallas/Fort Worth 44% and 10% of Asian mothers lived in areas with the lowest and highest greenness, compared to 25% and 28% for white mothers, respectively. These disparities suggest that widespread environmental injustice issues accompany the distribution of greenness with cities and between racial groups. Importantly, if we only include neighborhood measures of SES and ethnicity we continue to see statistically significant dose-response associations between greenness and birth outcomes (e.g. an 11.6 g (95% CI: 9.8;13.3) increase in birth weight across Texas in the greenest quartile). This suggests that confounding occurs even within neighborhoods and that studies examining residential greenness must control for individual level confounding factors. It is possible that racial and ethnic groups value greenness differently but further research is needed to explore this issue. Overall, relatively little research has examined urban greenness as an environmental injustice issue. Prior research suggests that greenness may be more beneficial to 94

Environmental Research 152 (2017) 88–95

L. Cusack et al.

Fernández Somoano, A., et al., 2012c. Surrounding greenness and pregnancy outcomes in four Spanish birth cohorts. Environ. Health Perspect. 120 (10), 1481–1487. Dadvand, P., Wright, J., Martinez, D., Basagaña, X., McEachan, R.R., Cirach, M., et al., 2014. Inequality, green spaces, and pregnant women: roles of ethnicity and individual and neighbourhood socioeconomic status. Environ. Int. 71, 101–108. Donovan, G.H., Michael, Y.L., Butry, D.T., Sullivan, A.D., Chase, J.M., 2011. Urban trees and the risk of poor birth outcomes. Health Place 17, 390–393. Donovan, G.H., Butry, D.T., Michael, Y.L., Prestemon, J.P., Liebhold, A.M., Gatziolis, D., et al., 2013. The relationship between trees and human health: evidence from the spread of the emerald ash borer. Am. J. Prev. Med. 44, 139–145. Dzhambov, A.M., Dimitrova, D.D., Dimitrakova, E.D., 2014. Association between residential greenness and birth weight: systematic review and meta-analysis. Urban For. Urban Green. 13, 621–629. Ebisu, K., Holford, T.R., Bell, M.L., 2016. Association between greenness, urbanicity, and birth weight. Sci. Total Environ. 542, 750–756. Fedulov, A.V., Leme, A., Yang, Z., Dahl, M., Lim, R., Mariani, T.J., et al., 2008. Pulmonary exposure to particles during pregnancy causes increased neonatal asthma susceptibility. Am. J. Respir. Cell Mol. Biol. 38, 57–67. Gray, S.C., Edwards, S.E., Schultz, B.D., Miranda, M.L., 2014. Assessing the impact of race, social factors and air pollution on birth outcomes: a population-based study. Environ. Health 13, 4. Grazuleviciene, R., Danileviciute, A., Dedele, A., Vencloviene, J., Andrusaityte, S., Uždanaviciute, I., et al., 2015. Surrounding greenness, proximity to city parks and pregnancy outcomes in Kaunas cohort study. Int. J. Hyg. Environ. Health 218, 358–365. Groom, K.M., Poppe, K.K., North, R.A., McCowan, L.M., 2007. Small-for-gestational-age infants classified by customized or population birthweight centiles: impact of gestational age at delivery. Am. J. Obstet. Gynecol. 197, 239, (e1-239. e5). Hystad, P., Davies, H.W., Frank, L., Van Loon, J., Gehring, U., Tamburic, L., Brauer, M., 2014. Residential greenness and birth outcomes: evaluating the influence of spatially correlated built-environment factors. Environ. Health Perspect. 122, 1095–1102. James, P., Hart, J.E., Banay, R.F., Laden, F., 2016. Exposure to greenness and mortality in a nationwide prospective cohort study of women. Environ. Health Perspect.. Jirtle, R.L., Skinner, M.K., 2007. Environmental epigenomics and disease susceptibility. Nat. Rev. Genet. 8, 253–262. Kliegman, R.M., 2011. Nelson Textbook of Pediatrics. Saunders Elsevier, Philadelphia, Pa. Laurent, O., Wu, J., Li, L., Milesi, C., 2013. Green spaces and pregnancy outcomes in Southern California. Health Place 24, 190–195. Lee, P.-C., Roberts, J.M., Catov, J.M., Talbott, E.O., Ritz, B., 2013. First trimester exposure to ambient air pollution, pregnancy complications and adverse birth outcomes in Allegheny County, PA. Matern. Child Health J. 17, 545–555. Markevych, I., Fuertes, E., Tiesler, C.M., Birk, M., Bauer, C.-P., Koletzko, S., et al., 2014. Surrounding greenness and birth weight: results from the GINIplus and LISAplus birth cohorts in Munich. Health Place 26, 39–46. Merlo, D.F., Wild, C.P., Kogevinas, M., Kyrtopoulos, S., Kleinjans, J., 2009. NewGeneris: a European study on maternal diet during pregnancy and child health. Cancer Epidemiol. Biomark. Prev. 18, 5–10. Mitchell, R., Popham, F., 2008. Effect of exposure to natural environment on health inequalities: an observational population study. Lancet 372, 1655–1660. Novotny, E.V., Bechle, M.J., Millet, D.B., Marshall, J.D., 2011. National satellite-based land-use regression: No2 in the United States. Environ. Sci. Technol. 45, 4407–4414. http://dx.doi.org/10.1021/es103578x. Sugiyama, T., Leslie, E., Giles-Corti, B., Owen, N., 2008. Associations of neighbourhood greenness with physical and mental health: do walking, social coherence and local social interaction explain the relationships? J. Epidemiol. Community Health 62, (e9–e9). Takano, T., Nakamura, K., Watanabe, M., 2002. Urban residential environments and senior citizens' longevity in megacity areas: the importance of walkable green spaces. J. Epidemiol. Community Health 56, 913–918. Thompson, C.W., Roe, J., Aspinall, P., Mitchell, R., Clow, A., Miller, D., 2012. More green space is linked to less stress in deprived communities: evidence from salivary cortisol patterns. Landsc. Urban Plan. 105, 221–229. Van den Berg, A.E., Maas, J., Verheij, R.A., Groenewegen, P.P., 2010. Green space as a buffer between stressful life events and health. Soc. Sci. Med. 70, 1203–1210. Villeneuve, P.J., Jerrett, M., Su, J.G., Burnett, R.T., Chen, H., Wheeler, A.J., et al., 2012. A cohort study relating urban green space with mortality in Ontario, Canada. Environ. Res. 115, 51–58. Weier, J., Herring, D., 2000. Measuring vegetation (NDVI and EVI): Normalized Difference Vegetation Index (NDVI).

5. Conclusions We investigated the impacts of residential greenness on birth outcomes using a cohort of approximately three million babies born in urban areas of Texas between 2000 and 2009. This is the first study to investigate the temporal changes in NDVI throughout pregnancy and to examine such a large geographical area with substantial regional variation in greenness levels, several large and distinct cities, and diverse populations based on SES, race and ethnicity. Strong associations between residential greenness and birth weight and SGA were observed in unadjusted models, but this relationship was due to confounding by individual and neighborhood characteristics, primarily race/ethnicity, education and neighborhood income. The consistent patterns of greenness and race/ethnicity and education merits further attention as a potential environmental injustice issue. In fully adjusted models there were no significant associations between residential NDVI exposure and birth outcomes. Some evidence was seen for protective effects of greenness for Hispanics, mothers with low education and mothers living in low income neighborhoods. Similar results were observed with trimester specific exposures as well as for cityspecific models of Houston, San Antonio, Dallas and Austin. Conflict of interest Authors declare that they have no conflict of interests to disclose. Funding This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. Appendix A. Supporting information Supplementary data associated with this article can be found in the online version at doi:10.1016/j.envres.2016.10.003. References Agay-Shay, K., Peled, A., Crespo, A.V., Peretz, C., Amitai, Y., Linn, S., et al., 2014. Green spaces and adverse pregnancy outcomes. Occup. Environ. Med. 71, 562–569. Bechle, M.J., Millet, D.B., Marshall, J.D., 2015. National spatiotemporal exposure surface for NO2: monthly scaling of a satellite-derived land-use regression, 2000– 2010. Environ. Sci. Technol. 49, 12297–12305. Boccardo, P., Mondino, E.B., Perez, F., Claps, P., 2006. Co-registeration and inter-sensor comparison of MODIS and Landsat 7 ETM+ Data aimed at NDVI calculation. Rev. Fr. Photogramm. Teledetect. 182, 74–79. Brauer, M., Hoek, G., Smit, H.A., De Jongste, J.C., Gerritsen, J., Postma, D.S., et al., 2007. Air pollution and development of asthma, allergy and infections in a birth cohort. Eur. Respir. J. 29, 879–888. Casey, J.A., James, P., Rudolph, K.E., Wu, C.-D., Schwartz, B.S., 2016. Greenness and birth outcomes in a range of pennsylvania communities. Int. J. Environ. Res. Public Health, 13: 311. Dadvand, P., de Nazelle, A., Figueras, F., Basagaña, X., Su, J., Amoly, E., et al., 2012a. Green space, health inequality and pregnancy. Environ. Int. 40, 110–115. Dadvand, P., Sunyer, J., Basagana, X., Ballester, F., Lertxundi, A., Fernandez-Somoano, A., et al., 2012b. Surrounding greenness and pregnancy outcomes in four Spanish birth cohorts. Environ. Health Perspect. 120, 1481–1487. http://dx.doi.org/ 10.1289/ehp.1205244. Dadvand, P., Sunyer Deu, J., Basagaña Flores, X., Ballester Díez, F., Lertxundi, A.,

95