Spatial variations in the associations of term birth weight with ambient air pollution in Georgia, USA
Introduction
Birth weight is a strong predictor of infant morbidity and mortality and an important indicator of overall infant and child health. Low birth weight (LBW), defined as live births weighing less than 2500 grams (g), is a major determinant of mortality, morbidity, and disability in infancy and childhood, and can affect the subsequent health status of individuals (WHO, 2015). World Health Organization (WHO) estimated about 30 million LBW babies born worldwide every year (WHO, 2015). LBW is also a leading cause of infant mortality in the United States (Carmichael et al., 1998, Tierney-Gumaer and Reifsnider, 2008). Thus, LBW is a significant health challenge in the U.S. and around the world and a reduction in LBW can contribute to an overall improvement in population health. A comprehensive understanding of the causes of LBW and the factors that affect birth weight can provide important information to formulate more effective health policy and preventive initiatives. These factors include biological (e.g. gene), socioeconomic (e.g. income and education), behavioral (e.g. exercise, maternal drinking and smoking), political (e.g. health policy and accessibility to prenatal care), and environmental (e.g. noise, land use, water pollution, and air pollution) variables that may have impact on the health of pregnant women and fetal development.
Numerous studies have been conducted around the world to explore the associations of birth outcomes, including birth weight, LBW, and preterm birth, with maternal exposure to air pollution (Basu et al., 2004, Bell et al., 2007, Kloog et al., 2012, Madsen et al., 2010, Twum et al., 2015, Yorifuji et al., 2015). The examined air pollutants included sulphur dioxide (SO2), carbon monoxide (CO), nitrogen dioxide (NO2), ozone (O3), and fine particulate matters (PM) with an aerodynamic diameter of less than 2.5 μm (PM2.5), and coarse PM of aerodynamic diameter of less than 10 μm (PM10). Although the previous studies provided considerable evidence to support an association between birth weight and ambient air pollution, the results are inconsistent and the causal mechanisms remain unclear (Madsen et al., 2010). For example, higher concentrations of air pollutants have been associated with a reduced birth weight or an elevated risk of LBW in some studies (Basu et al., 2004, Bell et al., 2007, Kloog et al., 2012, Twum et al., 2015, Yorifuji et al., 2015). Kloog et al. (2012) examined the effect of PM2.5 on birth weight of 634,244 births in the State of Massachusetts, USA between 2000 and 2008 and found that birth weight was negatively associated with PM2.5. In contrast, many studies did not find significant relationships between some air pollutants and birth weight or risk of LBW (Berrocal et al., 2011, Madsen et al., 2010, Vinikoor-Imler et al., 2014). Berrocal et al. (2011) applied a hierarchical model to specify the relationship between birth weight and personal exposure to PM2.5 in 14 counties in North Carolina in years 2001 and 2002, but did not find a significant effect of PM2.5 on birth weight. Moreover, some studies even found a protective effect of air pollutants on birth weight (Dadvand et al., 2013, Hao et al., 2015, Parker and Woodruff, 2008). Parker and Woodruff (2008) modelled birth weight as a function of county-level PM2.5 exposure for the births in the US during the period of 2001–2003 for the entire country and by regions, and reported that birth weight was positively related to PM2.5 in the model using the nationwide data, but both positive and negative relationships were identified for different regions. Hao et al. (2015) applied multilevel logistic regression models to evaluate the associations between LBW and PM2.5 exposure in USA, and different associations (significant negative, non-significant, and significant positive) were found in different census divisions.
The comparison of the results among the previous studies reveals that the association of birth weight with air pollution actually vary geographically. An air pollutant considered significant in affecting birth weight in one study may have no significance in other places. The spatial variation in the relationships between birth weight and air pollutants discovered among different studies may be due to the differences in study design, covariates selection, exposure estimation, and time of assessment (Parker and Woodruff, 2008, Madsen et al., 2010, Shah and Balkair, 2011). In addition, because the physical and social environment, such as pollution sources, intensity of emissions, meteorology, and pollution control technologies and policies, differ significantly over space, it is likely that the chemical composition and intrinsic toxicity of an air pollutant varies spatially, so that the impact of an air pollutant on a health outcome such as birth weight can be spatially altered (Basu et al., 2014, Bell et al., 2010, Coker et al., 2015, Darrow et al., 2011, Laurent et al., 2014). Furthermore, many contextual neighborhood and individual factors, ranging from socioeconomic status (SES), demographics, housing characteristics, behavioral factors, to accessibility to health care, usually vary significantly over space, so they may affect the susceptibilities of pregnant women to air pollution and subsequent birth outcomes, thereby complicating observed associations between birth outcomes and air pollution (Coker et al., 2015, Hajat et al., 2013, Tu et al., 2012).
Therefore, it is reasonable to expect spatial variation in the observed associations between birth weight and ambient air pollution. Only a few previous studies, however, have examined the spatial variation in associations of birth outcomes with air pollution (Berrocal et al., 2011, Coker et al., 2015, Dadvand et al., 2013, Hao et al., 2015, Parker and Woodruff, 2008). Most of them first divided the study area into different regions (e.g. countries, states, or census divisions), and then analyzed the associations by regions (Dadvand et al., 2013, Hao et al., 2015, Parker and Woodruff, 2008). For example, Parker and Woodruff (2008) discovered the regional variations in the associations of birth weight with air pollution by building different models for different US regions. The associations obtained from that study did not vary continuously over space, but abruptly changed across the boundaries of the regions, and no variation could be analyzed inside each region. Thus, the variations in the associations were not much different from the various results we can get by comparing multiple previous studies. Coker et al. (2015) applied a multivariate logistic regression model with multilevel spatially structured and unstructured random effects set in a Bayesian framework to model the spatial effects of PM2.5 on LBW in Los Angeles, and estimated the continuous spatial variation in the association of LBW with PM2.5 across census tracts, but didn't analyze other pollutants. To our best knowledge, no studies have explicitly examined the continuous spatial variation in the association of birth weight or LBW with air pollutants other than PM2.5. In addition, no published research has reported the impact of socioeconomic characteristics of the communities where the births are located on the associations of birth outcomes with air pollution.
Traditionally, the associations of birth weight or LBW with air pollution have been analyzed using conventional statistical methods, such as ordinary least squares (OLS) regression for continuous dependent variables (e.g., birth weight) or logistic regression for dichotomous dependent variables (e.g., LBW). In those studies, concentrations of air pollutant, with other confounding factors, such as socioeconomic, demographic, and behavioral factors, were typically used as the independent variables, while birth weight or LBW was the dependent variable. These methods are global statistics that analyze the average situation for the entire study area (Fotheringham et al., 2002), and they assume that relationships are stationary. As indicated in our review of the literature, this assumption is routinely violated because observed relationships are known to vary.
In recent years, a growing number of studies have applied a local spatial statistical technique called geographically weighted regression (GWR) to explore the spatial variations in relationships between health issues and risk factors (Chan et al., 2014, Gilbert and Chakraborty, 2011, Goovaerts et al., 2015, Tu et al., 2012, Zhang et al., 2012). However, no previous studies have applied this technique to analyze the associations of birth outcomes with ambient air pollution. GWR attempts to capture spatial variations by estimating regression model parameters for each individual regression point (the location of each observation). The local estimation of model parameters is obtained by weighting all neighboring observations using a distance decay function, assuming that the observations nearby have more influence on the regression point than the observations further away. In addition to the local parameter estimates, GWR also produces other local regression results, including the values of t-test on the local parameter estimates, the local R2 values, and the local residuals for each regression point. The spatial variations in the relationships between dependent and independent variables can be shown by comparing the local regression results among different regression points. Therefore, GWR may serve as a useful tool to explore the spatial variations in the associations of birth weight with air pollutants.
We have applied GWR to study the spatial variation in the associations of birth weight with socioeconomic, environmental, and behavioral factors in the State of Georgia, USA in a previous study (Tu et al., 2012). However, that study did not include air pollutants due to the lack of air pollution data, but used percentage of urban land as a proxy of air pollution instead. The current study maintains all the independent variables and methodologies used in the previous research but includes air pollutants to the analyses so that the spatial variation in the associations of birth weight with ambient air pollution can be investigated, and also includes gestational age and parity as independent variables. The primary objectives of this research are (1) to compare the differences in the relationships between birth weight and concentrations of PM2.5 and O3 analyzed by OLS and GWR; (2) to examine how the relationships between birth weight and concentrations of PM2.5 and O3 produced by GWR vary over space; and (3) to explore how the spatially varying relationships of birth weight and concentrations of PM2.5 and O3 are affected by socioeconomic and urban characteristics of communities. As GWR is an exploratory spatial data analysis (ESDA) technique, we do not intend to study the causal relationship between LBW and its risk factors or any critical values of the air pollutants that might cause LBW. Our goal is to explore how the associations between birth weight and ambient air pollution vary in the communities with different socioeconomic characteristics. We hope that the results gleaned from this study can help formulate community specific prevention and intervention policies.
Section snippets
Birth outcomes and maternal behavioral variables
Individual birth weight and maternal behavioral variables data were derived from the electronic birth certificate data (BCD) collected by the Georgia Vital Records Office in Atlanta. This study used the data in the year 2000 so that the birth weight and maternal behavioral variables could be more conveniently related to the socioeconomic variables from the US decennial Census in 2000. A detailed description of the birth data is available on the Vital Records website (URL //dph.georgia.gov/VitalRecords
Spatial variations in birth weight and its factors
The statistical summary of birth weight and socioeconomic, environmental, and behavioral factors is shown in Table 1. A great variation is observed for most of the variables, such as birth weight, parity, family median income, and percentage of urban land. Among male births, birth weight ranges from 1332 g to 7270 g with a mean of 3466 g. The lowest family median income is $4202, compared to the highest of $200,001. The lowest percentage of urban land is 0.53% while the highest is 100%. Similar
Discussion
The results obtained using OLS show that birth weight is significantly associated with all the selected socioeconomic, behavioral, and environmental factors except PM2.5 in the State of Georgia in 2000. Higher birth weight is significantly related to higher concentration of O3 and insignificantly related to higher concentration of PM2.5, which was not expected. When using GWR, the associations of birth weight with these factors do not show a statistically significant global trend, but tend to
Conclusions
This study analyzed the spatial variation in the associations of birth weight with concentrations of O3 and PM2.5 adjusted for gestational age, parity, and six other individual-level and community-level socioeconomic, behavioral, and land use factors in the State of Georgia, USA for the year 2000 using both OLS and GWR models. The results obtained using OLS suggest that birth weight has no significant correlation with concentration of PM2.5, but is significantly positively associated with
Acknowledgements
This study was partially supported by the 2015 Faculty Summer Research Grant from the College of Humanities and Social Sciences (CHSS) at Kennesaw State University (KSU).
References (45)
- et al.
Effects of fine particulate matter and its constituents on low birth weight among full-term infants in California
Environ. Res.
(2014) - et al.
Modeling spatial effects of PM2.5 on term low birth weight in Los Angeles County
Environ. Res.
(2015) - et al.
Traffic-related air pollution, preterm birth and term birth in the PIAMA birth cohort study
Environ. Res.
(2011) - et al.
Using geographically weighted regression for environmental justice analysis: cumulative cancer risks from air toxics in Florida
Soc. Sci. Res.
(2011) - et al.
Geographically-weighted regression analysis of percentage of late-stage prostate cancer diagnosis in Florida
Appl. Geogr.
(2015) - et al.
The association of PM2.5 with full term low birth weight at different spatial scales
Environ. Res.
(2014) - et al.
Sources and contents of air pollution affecting term low birth weight in Los Angeles county, California, 2001–2008
Environ. Res.
(2014) - et al.
Ambient air pollution exposure, residential mobility and term birth weight in Oslo, Norway
Environ. Res.
(2010) - et al.
Air pollution and birth outcomes: a systematic review
Environ. Int.
(2011) - et al.
Spatial variations in the associations of birth weight with socioeconomic, environmental, and behavioral factors in Georgia, USA
Appl. Geogr.
(2012)
Associations between prenatal exposure to air pollution, small for gestational age, and term low birthweight in a state-wide birth cohort
Environ. Res.
Outdoor air pollution and term low birth weight in Japan
Environ. Int.
An exploratory spatial analysis of western medical services in Republican Beijing
Appl. Geogr.
Comparing exposure metrics in the relationship between PM2.5 and birth weight in California
J. Expo. Anal. Environ. Epidemiol.
Ambient air pollution and low birth weight in Connecticut and Massachusetts
Environ. Health Perspect.
Prenatal exposure to fine particulate matter and birth weight: variations by particulate constituents and sources
Epidemiology
On the use of a PM2.5 exposure simulator to explain birthweight
Environmetrics
Geographically weighted regression — modeling spatial non-stationarity
Statistician
Cause-specific trends in neonatal mortality among black and white infants, United States, 1980–1995
Matern. Child Health J.
Geographic disparity in chronic obstructive pulmonary disease (COPD) mortality rates among the Taiwan Population
PLoS ONE
Air pollution and birth weight in northern Nevada, 1991–1999
Inhal. Toxicol.
Maternal exposure to particulate air pollution and term birth weight: a multi-country evaluation of effect and heterogeneity
Environ. Health Perspect.
Cited by (35)
Air pollution exposure and the risk of macrosomia: Identifying specific susceptible months
2023, Science of the Total EnvironmentCitation Excerpt :The results of studies on adverse health effects due to ozone O3 have been controversial. Several studies reported that exposure to O3 is linked to lower birth weight (Guo et al., 2020; Wang et al., 2021), while some other studies showed no association or an opposite association (Tu et al., 2016). The result about O3 effect is ambiguous, which may be explained by different pollutant concentrations, demographic characteristics and misclassification of pollutants.
Mother-level random effect in the association between PM<inf>2.5</inf> and fetal growth: A population-based pregnancy cohort
2022, Environmental ResearchCitation Excerpt :Our study report significant gender differences in the association between PM2.5 and TLBW. Findings were inconsistent with some studies, conducted stratified analysis by infants’ sex, that found that males are more vulnerable to decreased birth weight (Aparicio et al., 2019; Tu et al., 2016), stunting (Kurata et al., 2020) and increased risk of child asthma (Lee et al., 2018), but consistent with other studies that explored associations between in-utero pollution exposure and short- and long-term health outcomes other than birth outcomes. These studies reported that females are more vulnerable than males (Chen et al., 2017; Kim et al., 2019; Liu et al., 2020).
Maternal exposure to ambient PM<inf>2.5</inf> and term birth weight: A systematic review and meta-analysis of effect estimates
2022, Science of the Total Environment