You are here

Article Text

Article menu
Environment
Original article
Green spaces and adverse pregnancy outcomes
  1. Keren Agay-Shay1,2,3,
  2. Ammatzia Peled4,
  3. Antonia Valentín Crespo1,2,3,
  4. Chava Peretz5,
  5. Yona Amitai6,
  6. Shai Linn7,
  7. Michael Friger8,
  8. Mark J Nieuwenhuijsen1,2,3
  1. 1Centre for Research in Environmental Epidemiology (CREAL), Barcelona, Spain
  2. 2Universitat Pompeu Fabra (UPF), Barcelona, Spain
  3. 3CIBER Epidemiología y Salud Pública (CIBERESP), Barcelona, Spain
  4. 4University of Haifa, Department of Geography and Environmental Studies, Faculty of Social Sciences Haifa 31905, Israel
  5. 5Tel Aviv University, Department of Epidemiology, Faculty of Medicine, P.O.B. 39040, Ramat Aviv 69978, Israel
  6. 6Bar Ilan University, Department of Management, Ramat Gan, 52900, Israel
  7. 7Unit of Clinical Epidemiology, Rambam Medical Center, P.O. B 9602, Haifa 31096, Israel
  8. 8Ben-Gurion University of the Negev, Department of Epidemiology and Health Services Evaluation, Faculty of Health Sciences, P.O.B 653 Beersheba 84105, Israel
  1. Correspondence to Dr Keren Agay-Shay, Centre for Research in Environmental Epidemiology (CREAL), Barcelona 31905, Spain; kagayshay{at}gmail.com

Abstract

Objective The objective of this study was to evaluate the associations between proximity to green spaces and surrounding greenness and pregnancy outcomes, such as birth weight, low birth weight (LBW), very LBW (VLBW), gestational age, preterm deliveries (PTD) and very PTD (VPTD).

Methods This study was based on 39 132 singleton live births from a registry birth cohort in Tel Aviv, Israel, during 2000–2006. Surrounding greenness was defined as the average of satellite-based Normalised Difference Vegetation Index (NDVI) in 250 m buffers and proximity to major green spaces was defined as residence within a buffer of 300 m from boundaries of a major green space (5000 m2), based on data constructed from OpenStreetMap. Linear regression (for birth weight and gestational age) and logistic regressions models (for LBW, VLBW, PTD and VPTD) were used with adjustment for relevant covariates.

Results An increase in 1 interquartile range greenness was associated with a statistically significant increase in birth weight (19.2 g 95% CI 13.3 to 25.1) and decreased risk of LBW (OR 0.84, 95% CI 0.78 to 0.90). Results for VLBW were in the same direction but were not statistically significant. In general, no associations were found for gestational age, PTD and VPTD. The findings were consistent with different buffer and green space sizes and stronger associations were observed among those of lower socioeconomic status.

Conclusions This study confirms the results of a few previous studies demonstrating an association between maternal proximity to green spaces and birth weight. Further investigation is needed into the associations with VLBW and VPTD, which has never been studied before.

Keywords
  • NDVI
  • LBW
  • PTD
  • adverse pregnancy outcomes

https://doi.org/10.1136/oemed-2013-101961

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    Keywords

    What this paper adds

    • There is scant information about the impact of greenness on pregnancy outcomes. The associations between green spaces and very low birth weight and very preterm deliveries have never been previously reported.

    • This is the first study outside of USA and Europe demonstrating associations between maternal measures of greenness and birth weight, and the first study to report the association with low birth weight.

    • This study is a registry-based study restricted to routinely recorded information, and may not fully control for confounding and exposure misclassifications and, therefore, further investigations are needed to confirm our results.

    Introduction

    Adverse pregnancy outcomes are known to be associated with increased neonatal morbidity and mortality, as well as morbidity later in life.1 Adverse pregnancy outcomes such as preterm deliveries (PTD) and low birth weight (LBW) have been previously associated with environmental exposures, and most of these studies have focused on environmental nuisance.2 However, only a few studies considered the possible beneficial associations with environmental exposures, such as greenness and birth outcomes.3–8

    Green spaces are defined as ‘land that is partly or completely covered with grass, trees, shrubs, or other vegetation. Green spaces includes parks, community gardens and cemeteries’.9 The association between green spaces and health outcomes has been discussed in previous reviews, however; the underlying mechanisms of the beneficial effects of green spaces on health are not fully understood.10–13 These effects might be mediated through (1) promoting physical activity, which has a well-established positive impact on health, (2) affecting emotions (mainly by decreasing negative emotions and improving psychophysiological stress, or through improving cognitive attention restoration), (3) increasing social contacts/cohesion or (4) improving environmental conditions, such as decreasing noise and air pollution levels and moderating urban heat island effects. Recently Kihal-Talantikite et al8 proposed a detailed model on the beneficial effects of greenness on pregnancy outcomes.

    In Israel, the prevalence rate of adverse pregnancy outcomes is similar to the European rates.14 PTD was the leading cause of infant mortality during 2000–2006,15 and very LBW (VLBW) deliveries are a substantial public health burden requiring 75 000 annual days of hospitalisation in Israel.16 Therefore, the main aim of this study was to evaluate the association between proximity to green spaces and surrounding greenness and birth weight (including LBW and VLBW) and gestational age (PTD and very PTD (VPTD)) in a registry-based study. We hypothesised that having a maternal residence near green spaces and greater surrounding greenness, is inversely associated with adverse pregnancy outcomes.

    Methods

    Study population

    This study was based on a registry birth cohort that is operated by the National Birth Registry within the Department of Mother and Child Health in the Public Health Service of Israel's Ministry of Health and by the Ministry of Interior. The registry passively collects reports for over 99.9% of births in Israel from the different birthing centres in hospitals. The reports are based on birth certificates that are completed in the hospitals. Reporting of each live birth is obligatory to the Ministry of Interior and to the Ministry of Health by Israeli law. Reports include demographic, geographic, ethnic and educational data of the parents and information about the infant and the birth.17

    The Tel Aviv metropolitan district is the most densely populated area of Israel, with a population of 1.2 million residents. It is approximately 170 km2, in size, with an average population density of 6918 persons per square km.18 The district capital city is Tel Aviv (figure 1). Tel Aviv has a Mediterranean climate with mild, rainy winters and hot, dry summers. In Tel Aviv city there were 41 766 live-born infants, stillbirths and fetal deaths during the study period (2000–2006). We excluded births without information regarding gestational age and those whose address could not be geocoded. A total of 39 132 live-births met the study inclusion criteria for our analyses. In our study, PTD was defined as birth before 37 gestational weeks (n=5150) and VPTD was defined as birth before 32 gestational weeks (n=506). LBW was defined as birth weight below 2500 g (n=2561), and VLBW was defined as birth weight below 1500 g (n=310). For our main analysis, we included all singleton live births without restricting to term births4; however, we also conducted a sensitivity analysis restricted to term births (≥37 weeks of gestation).

    Figure 1
    Figure 1

    Tel Aviv region, major green spaces and normalised difference vegetation index (NDVI). OSM, OpenStreetMaps.

    Ethical approval

    The study was approved by the Ethics Committee of the University of Haifa. Further approvals from the director of Public Health Services and from the legal advisor in the Israeli Ministry of Health and the Ministry of Interior were also obtained. The investigators were provided with the births dataset only after deletion of all personal identifiers (eg, identity numbers or names).

    Measures of greenness

    Due to privacy and confidentiality concerns, maternal residential address at delivery was provided at the street name level. The geocoding algorithm was embedded into newly developed software and was integrated in the generic ArcGIS software programme (ESRI, Redlands, California, USA). The addresses were matched to the centre point of the street (number of street locations=1483). The address matching geocoding process has been described in detail elsewhere.19 We used a Geographic Information System approach to determine for each geocoded residential street address two aspects of proximity to greenness: surrounding greenness and proximity to major green spaces.3 We hypothesised that (1) immediate surrounding greenness (within a buffer of 250 m) would be more indicative of the impact of greenness on reducing psychophysiological stress and modifying the impact of air pollution, noise and temperature, while (2) proximity to major green spaces would be more suggestive of the effect of physical activity.20

    To determine the surrounding greenness, we used Normalised Difference Vegetation Index (NDVI) derived from the Landsat Enhanced Thematic Mapper Plus (ETM+) data at 30 m×30 m resolution. The Landsat ETM+ data were acquired for 11 September 2003, covering path 174 and row 038 (ie, the scope of Tel Aviv region).21 NDVI is an indicator of greenness based on land surface reflectance of visible (red) and near-infrared wavelengths. It ranges between −1 and 1 with higher numbers indicating more greenness. To compare our results with previously published studies we used similar buffer sizes that were used by others (100,3–7 2504 ,6 and 500 m4 ,6). For the main analysis, surrounding greenness was abstracted as the average of NDVI in the aerial buffer of 250 m around the street centre point of each maternal place of residence. We also evaluated as sensitivity analyses average NDVI in the aerial buffer of 100m and 500 m.

    To address proximity to major green spaces, OpenStreetMap was used. OpenStreetMap data are digital maps of street networks with various layers. We downloaded directly the ‘natural’ layer from OpenStreetMap, as shape files.22 From the ‘natural’ layer, we selected the major green spaces using two different approaches. In our main analysis we included green spaces from the ‘natural’ layer, larger than 5000 m2 based on the European indicator.23 A binary variable was used 33 to indicate whether maternal residential address was situated within an aerial buffer of 300 m from the boundaries of major green spaces. The selection of the buffer size and green space size was conducted to be in accordance with the European indicator.23 As a sensitivity analysis, we also evaluated major green spaces that were larger than the 75th percentile of the green spaces (11 177 m2) in the research area, as used previously by Dadvand et al.3

    Statistical analyses

    Main analysis

    We used multivariate linear regression models with adjustment for potential confounders to estimate the change in birth weight (grams), and gestational age at delivery (weeks) separately for a 1-IQR increase in average NDVI (IQR=0.0578) and residing within 300 m from the boundaries of major green spaces. To estimate the change in the risk of LBW, VLBW, PTD and VPTD, we used multivariate logistic regression models with the same greenness variables. We present the results from the linear regression models as the change in the mean, and the results from the logistic regression models as ORs along with their 95% CIs.

    We adjusted the models for several known risk factors that could potentially confound the association between pregnancy outcomes and greenness. For all outcomes we adjusted the models for the following covariates: infant’s gender (male/female), infant’s religion as declared by the parents (Jewish/non-Jewish), maternal age (14–20, 21–25, 26–30, 31–34, 35–40, 41–53 years), maternal marital status (married/unmarried), maternal origin (Israel/non-Israeli), year of birth and season of conception (spring (31 March–30 May), summer (31 May–22 September), autumn (23 September –6 December), winter (7 December–30 March)). For birth weight, LBW and VLBW analyses were also adjusted for gestational age. We could not consider maternal illness, alcohol, tobacco and drug use during pregnancy, because these data are not recorded on the Israeli birth certificate.14 Maternal education was defined as number of school years (5–12, 12, 13–15, 15–30). Maternal education is not required on the birth certificate that is reported to the Ministry of Health, and data was available only for approximately 40% of the total births. In a health survey conducted by the Central Bureau of Statistics during 2009, in a representative sample of the Israeli population, smoking was inversely associated with education level. Smoking rates among those with 16 years of education was 13% compared to 26% among those with less than 12 years of education.23 We used the maternal education data for a sensitivity analysis. Material deprivation was measured using the ward-based socioeconomic status rankings (SES) derived from the maternal residential geocoded address (SES rank values are 3–20). The SES was extracted from the Israeli Central Bureau of Statistics (based on the 1995 census data) and reflects various socioeconomic measures characterising the ward’s population, including income, education, apartment size, possession of appliances, and car ownership. A higher ranking indicates a higher socioeconomic status. The ranking represents, to a considerable extent, individual risk factors and possible confounders.24 SES was negatively correlated with smoking.25 In our study, maternal education was moderately correlated with ward-based SES (Spearman’s correlation r coefficient=0.52).

    Further analyses

    Stratification of analyses according to the quartile of ward-based SES

    We categorised SES into four categories, based on the distributions within the cohort. We compared the associations with greenness between the SES categories by stratifying analyses (using NDVI average in 250 m buffer around maternal residential address, as in our main analyses). Associations were expressed per a 1-IQR increase in surrounding greenness, using the same greenness contrasts used for the main analyses.

    Different greenness variables

    We evaluated the associations between measures of average NDVI in different aerial buffers (100 and 500 m), that differed from the main analysis of 250 m. We also evaluated the association between residing 300 m from major green spaces boundaries using different categorisation criteria (based on the size distribution of the green spaces in our research area).

    All births versus term births

    We limited our analyses of birth weight, LBW and VLBW to those participants with term births (gestational age at delivery ≥37 weeks) to evaluate the robustness of our findings to the exclusion of preterm births.

    Evaluation of the impact of SES using maternal education level

    Maternal education level was reported only for approximately 40% of the total births. As sensitivity analyses for the effect of the ward-based socioeconomic status ranking, we selected the births with maternal education level as a subpopulation and evaluated the associations with greenness adjusting first for SES and then additionally adjusting for maternal education.

    Evaluation of the inter-relationship between air pollution, surrounding greenness, and pregnancy outcomes

    Maternal exposure to particulate matter with aerodynamic diameter smaller than 10 μm (PM10) during the entire pregnancy was estimated using the inverse distance weighting method models that have been described and used previously.26 We repeated the main analyses by adding average maternal PM10 exposure levels during the entire pregnancy as a covariate in the models. This was done to explore the underlying mechanism of reducing air pollution as a possible mediator for the association between surrounding greenness and pregnancy outcomes.

    Results

    Descriptive statistics

    Characteristics of the total research population and infants with adverse pregnancy outcomes are presented in table 1. A larger proportion of PTD and LBW infants were from mothers born in Israel (p<0.0001), unmarried mothers (p<0.0001) and from mothers older than the total birth population (higher proportion of mother older than 40 years (p<0.0001)). PTD were more often male (p<0.0001), and LBW were more often female (p<0.0001).

    View this table:
    Table 1

    Characteristics of the study population from Tel Aviv during 2000–2006

    The 25th, 50th and 75th percentiles of the averaged NDVI in the buffer of 250 m around the centre point of the maternal residential street address were 0.023, 0.05 and 0.081, respectively. The NDVI average across alternative buffers of 100, 250 and 500 m were highly correlated with Spearman’s correlation r coefficients ranging from 0.87 to 0.96. There were 27 619 participants (70.6% of the population) who lived within 300 m from major green spaces.

    Main analysis

    A 1-IQR increase in surrounding greenness was associated with increased birth weight, and decreased risk for LBW in unadjusted and adjusted models. For VLBW, only the unadjusted models with measures of surrounding greenness were statistically significant (at p=0.05 level) (table 2). For gestational age at delivery, PTD and VPTD, only the unadjusted model for VPTD was statistically significantly associated (at p=0.05 level) with surrounding greenness.

    View this table:
    Table 2

    Unadjusted and adjusted change in mean birth weight (grams) and gestational age (weeks) and OR for LBW, VLBW, PTD and VPTD for an increase of a 1-IQR* in average NDVI in 250 m buffer around each maternal residential address (NDVI-250 m) and living within 300 m to major green spaces- (5000 m2)

    Living within 300 m from major green spaces was associated with increased birth weight in unadjusted and adjusted models, and associated with decreased risk for LBW in the adjusted models. No associations were observed for living within 300 m from major green spaces with gestational age PTD and VPTD.

    Further analyses

    Stratification of analyses according to the quartile of ward-based SES

    The results of stratified analyses according to the quartile of SES for birth weight are presented in table 3. The strongest association was for the lowest SES with a statistically significant increase in birth weight. In general, the direction of the associations for birth weight and LBW were consistent with those of the main analyses, though associations were not statistically significant for all strata. As for the main analysis, gestational age at delivery did not appear to be associated with surrounding greenness (data not presented).

    View this table:
    Table 3

    Adjusted change in mean birth weight (grams) and OR for LBW per a 1-IQR* increase in average NDVI in 250 m buffer around each maternal residential address (NDVI-250 m) stratified by quartiles of SES

    Different greenness variables

    Measures of surrounding greenness in different aerial buffers (100 and 500 m) and different categorisation of major green spaces (using the 75th percentile, 11 000 m2) were generally consistent with those of the main analyses, and did not change materially the results (see online supplementary material table S1).

    All births versus term births

    After limiting the study participants to those with term births (n=33 984), there was no notable change in findings in terms of direction, strength, and statistical significance of the associations (data not shown).

    Evaluation of the impact of SES using maternal education level

    After limiting the study participants to those with reported maternal education level (n=15 705), additionally adjusting for maternal education to the models did not change the associations with greenness, implying that maternal education was not a modifier in our study. In our study, maternal education was moderately correlated with SES (Spearman’s correlation r coefficient=0.52), and adding maternal education decreased the significance of SES in our models, implying that SES is a fairly good proxy for maternal education.

    Evaluation of the interrelationship between air pollution, surrounding greenness, and pregnancy outcomes

    After including average maternal PM10 exposure during the entire pregnancy as a covariate, the estimated regression coefficients for surrounding greenness stayed consistent with those of the main analyses (data not shown).

    Discussion

    In this large registry-based population cohort we investigated the associations between maternal measures of surrounding greenness during pregnancy and birth outcomes in their offspring. Overall, an increase in measures of surrounding greenness was associated with a statistically significant increase of birth weight and decreased risk for LBW. Results for VLBW were in the same direction but not statistically significant. In general, no associations were found between maternal measures of surrounding greenness and gestational age, PTD and VPTD.

    Few studies have investigated maternal measures of surrounding greenness and pregnancy outcomes.3–8 Our main findings of the association between greenness and increased birth weight are consistent with previous results.4 ,6 ,7 In our study, these associations were consistent using various buffer sizes (100, 250 and 500 m) and using different sizes of major green spaces (5000 and 11 000 m2). Similarly, Dadvand et al4 demonstrated consistent effects for average NDVI using the same buffer sizes. However, Laurent et al7 found the effect of greenness on the mean birth weight only for the 50 m buffer and Markevych et al6 only for the 500 m buffer.

    Our results for birth weight indicated an increase of 19.2 g (95% CI 13.3 to 25.1) per a 1-IQR increase of average NDVI (0.0578) and an increase of 18.1 g (95% CI 13.3 to 25.1) for proximity to major green spaces. These associations for increase in a 1-IQR of mean NDVI were larger than reported previously in Spain by Dadvand et al4 (38.3 g (95% CI 17.1 to 59.5), IQR=0.162 within 250 m); in Germany by Markevych et al6 (17.6 g (95% CI 0.5 to 34.6), IQR=0.101 within 500 m) or in California by Laurent et al7 (6.09 g (95% CI 3.11 to 9.06), IQR=0.109 within 50 m). Although, the change in the mean birth weight is quite small (19 g) and might not necessarily have clinical importance at an individual level, this increase could be associated with a notable benefit at the population level,27 and in our population this could decrease 100 births from being defined as LBW cases (4% decrease) and 20 births from being defined as VLBW (6% decrease).

    Our study is the first to demonstrate associations between maternal measures of surrounding greenness and LBW with a 19% decrease in risk (95% CI 11% to 28%, per a 1-IQR increase in the average NDVI), and also the first to report results for VLBW and greenness. We did not observe statistically significant results for VLBW in the adjusted models, however, the direction was as we hypothesised, and stayed consistent in all our sensitivity analyses. The lack of statistical significance might be due to the small number of cases of this outcome. One previous study has reported an association between an increase of tree-canopy cover (10%) in the surrounding maternal home address and small for gestational age (OR was 0.99, 95% CI 0.98 to 0.99).5

    In our study, we did not find any association with gestational age, PTD and VPTD, and this was consistent with others, who reported no associations with gestational age3 ,4 and PTD.5 Only Laurent et al7 found decreased risk for PTD with a 1-IQR increase in the mean NDVI within 150 m buffer, but this was not consistent for the 50 and 100 m buffers that were also analysed. The consistent results between studies for birth weight but not for gestational age need further investigation. Better understanding of the mediation mechanism involved in the association between greenness and birth weight, may shed light on this discrepancy.

    In the current study, the ward-based socioeconomic status ranking that characterised the ward’s population was added to the routinely collected covariates. The ward-based socioeconomic status ranking represents several individual risk factors and possible confounders. Statistical wards are designed a priori to include relatively homogeneous populations with respect to ethnicity, lifestyle and social class.24 The ward-based SES ranking was not found to be highly correlated with maternal measures of surrounding greenness (0.33, with NDVI in 250 m buffer) and moderately correlated to maternal education (0.52). Furthermore, unadjusted and adjusted estimates did not differ much, suggesting that residual confounding is unlikely to be a major concern. In our study, after stratifying the analyses according to the ward-based SES, we found stronger associations in the lower ward-based SES class. This was in line with previous reports on stronger associations among those with lower maternal education3 ,6 suggesting greater benefits among the more deprived population. This may be partly because groups with lower SES generally have a worse health status, less healthy behaviour, and live in areas with more environmental problems, and the combination of these factors may make them more likely to benefit from health promotion interventions compared to higher SES groups.28 ,29 Recently Kihal-Talantikite et al8 discussed this in detail. The associations with measures of surrounding greenness among the lower ward-based SES groups may present a useful avenue for public health intervention through provision of green spaces in more deprived areas.

    The specific underlying mechanisms of the associations between greenness and adverse pregnancy outcomes are not fully understood. Recently Kihal-Talantikite et al8 proposed a detailed model on the beneficial effects of greenness on pregnancy outcomes. These beneficial associations might be mediated through different pathways: promoting physical activity, affecting emotions (mainly by decreasing negative emotions and improving psychophysiological stress or through improving cognitive attention restoration), by increasing social contacts/cohesion, or by improving environmental conditions, such as decreasing noise and air pollution levels and moderating urban heat-island effect.8 ,10–13 To explore the underlying mechanism of reducing air pollution as a possible mediator for the association between surrounding greenness and pregnancy outcomes, we added the average PM10 exposure during the entire pregnancy as a covariate in the models, and no mediation effect was observed.

    In our study, we could not evaluate all the possible suggested mechanisms that mediate the associations between maternal measures of surrounding greenness and adverse pregnancy outcomes. However, some studies have been published in Israel regarding the effects of trees and greenness on the ambient environment, and recently a literature review on this issue was published.30 In the desert area of Israel, researchers found that the air temperature of a street planted with trees was 3°C lower compared to a similar street without trees. Urban gardens and urban parks also demonstrated lower temperatures compared to their environment.31 ,32 Others found that in Haifa, two similar neighbourhoods with the same SES and different vegetation cover had different air pollution levels. The neighbourhoods with greater vegetation cover had lower concentrations of particles. Differences were greater in the morning.33 In Tel Aviv, concentrations of air pollutants, noise and temperature in the city park, a paved square and a narrow street were compared.34 ,35 Results showed that the urban parks with many trees reduced temperatures by up to 4°C compared to bare roads or paved urban spaces, and had a significant effect on PM10 reduction and smaller effects on ozone, nitrogen oxides and carbon oxide reduction. Vegetation reduced noise by 3–5 dB. The effect of the green area on the environment was dependent on vegetation size, thickness and density.

    The measures of surrounding greenness used in this study were based on remote sensing data, and OpenStreetMap was used to calculate the distance to major green spaces. The application of remote sensing data on surrounding greenness enabled our study to take small-scale green spaces (eg, street trees and green verges) into account, while the OpenStreetMap data was representative of major green spaces. Mitchell et al20 demonstrated that surrounding greenness is more indicative of the impact of greenness on reducing psychophysiological stress and modifying the impact of air pollution, noise and temperature, while proximity to major green spaces is more suggestive of the effect of physical activity. In our study, the Polychoric correlation r coefficient36 between the two measures was 0.47, demonstrating that these two exposures variables represent, to some extent, different exposures to greenness.

    In our study, we used NDVI data derived from the Landsat ETM+ data at 30 m×30 m resolution. However, this type of data does not give information on the canopy structural variation and the greenness type (grass vs trees). The Enhanced Vegetation Index-EVI can give this type of information, however, the data is limited to a resolution of 250 m×250 m and therefore less suitable for the current study.21 The OpenStreetMap project is a volunteered geographic information with no cost and high availability, which makes it an interesting source of information for epidemiological spatial analyses. However, varying quality in the data is a concern often raised when it comes to using volunteered geographic information for professional applications.22 Some studies have evaluated the accuracy of this data.37 ,38 To the best of our knowledge, this type of analysis was not published for Israel or Tel-Aviv. If there are inaccuracies in the data they will cause exposure misclassification, however, the misclassification is assumed to be random (no differences between pregnant women with adverse pregnancy outcomes or without adverse pregnancy outcomes) and, therefore, is most likely to bias towards the null.

    In the current study, maternal measures of surrounding greenness was based on the address reported at the time of delivery, and this was geocoded to the centre of the street. Both may cause exposure misclassification; however, this is likely to be non-differential. Therefore, it may cause an underestimation of the true associations between greenness and adverse pregnancy outcomes as previously demonstrated for air pollutants.39 In our study, personal activity patterns, such as time spent indoors versus outdoors and time spent at work or home were not accounted for. This exposure misclassification is also assumed to be random.

    The current study was based on birth certificates and had the advantage of using a large number of birth records, thus reducing uncertainties due to random error misclassification and selection bias more common in studies with a small sample size. A limitation of this type of study was that birth record studies are typically restricted to routinely recorded information, and may not fully control for confounding from individual maternal or fetal risk factors, such as maternal passive or active smoking, alcohol consumption, infections and diabetes. A study by Ritz and Wilhelm40 indicated that for adverse pregnancy outcomes, this might not be a cause for concern. They found that for air pollution association estimates did not appear to be additionally confounded by covariates not routinely collected on birth certificates. This type of analysis was not conducted for green spaces and adverse pregnancy outcomes.

    Summary and conclusion

    This study contributes to the limited body of knowledge regarding the association between maternal measures of surrounding greenness during pregnancy and adverse pregnancy outcomes. Associations were consistent using different buffer sizes. The observed association with birth weight and not with gestational age at delivery was consistent with previews publication and needs further investigation. Our detected association between a 1-IQR increase in the average NDVI and 19% decrease in the risk for LBW requires further confirmation by other studies. The associations regarding VLBW also need further investigation in future studies. Our work is the first work reporting association with pregnancy outcomes outside of Europe and the US. This study is a registry-based study restricted to routinely recorded information and may not fully control for confounding and exposure misclassifications and, therefore, further investigations are needed to confirm our results.

    Acknowledgments

    Keren Agay-Shay gratefully acknowledges support by a post-doctoral fellowship from the Environment and Health Found, Jerusalem, Israel. We also would like to thank all employees of the Mother and Child Unit at the Ministry of Health for their data collection and to the Ministry of Interior for approving data transfer. We would like to thank the Israeli Ministry of Environmental Protection which supported Keren Agay-Shay's PhD that was the base for this work (research grant-7-2-7).

    References

    1. ↵
      1. Behrman RE,
      2. Butler AS
      . Preterm birth: causes, consequences, and prevention. National Academies Press, 2006.
    2. ↵
      1. Stillerman KP,
      2. Mattison DR,
      3. Giudice LC,
      4. et al
      . Environmental exposures and adverse pregnancy outcomes: a review of the science. Reprod Sci 2008;15:631–50.
      OpenUrlAbstract/FREE Full Text
    3. ↵
      1. Dadvand P,
      2. de Nazelle A,
      3. Figueras F,
      4. et al
      . Green space, health inequality and pregnancy. Environ Int 2012;40:110–15.
      OpenUrlCrossRefPubMedWeb of Science
    4. ↵
      1. Dadvand P,
      2. Sunyer J,
      3. Basagaña X,
      4. et al
      . Surrounding greenness and pregnancy outcomes in four Spanish birth cohorts. Environ Health Perspect 2012;120:1481.
      OpenUrlCrossRefPubMedWeb of Science
    5. ↵
      1. Donovan GH,
      2. Michael YL,
      3. Butry DT,
      4. et al
      . Urban trees and the risk of poor birth outcomes. Health Place 2011;17:390–3.
      OpenUrlCrossRefPubMedWeb of Science
    6. ↵
      1. Markevych I,
      2. Fuertes E,
      3. Tiesler CMT,
      4. et al
      . Surrounding greenness and birth weight: results from the GINIplus and LISAplus birth cohorts in Munich. Health Place 2014;26:39–46.
      OpenUrlCrossRefPubMedWeb of Science
    7. ↵
      1. Laurent O,
      2. Wu J,
      3. Li L,
      4. et al
      . Green spaces and pregnancy outcomes in Southern California.Health Place, 2013.
    8. ↵
      1. Kihal-Talantikite W,
      2. Padilla CM,
      3. Lalloué B,
      4. et al
      . Green space, social inequalities and neonatal mortality in France. BMC Pregnancy Childbirth 2013;13:191.
      OpenUrlCrossRefPubMed
    9. ↵
      EPA 2013. What is Open Space/Green Space? http://www.epa.gov/region1/eco/uep/openspace.html (accessed 10 Dec 2013).
    10. ↵
      1. Frumkin H
      . The evidence of nature and the nature of evidence. Am J Prev Med 2013;44:196–7.
      OpenUrlCrossRefPubMed
    11. ↵
      1. Lee A,
      2. Maheswaran R
      . The health benefits of urban green spaces: a review of the evidence. J Public Health 2011;33:212–22.
      OpenUrlAbstract/FREE Full Text
    12. ↵
      1. Bowler DE,
      2. Buyung-Ali LM,
      3. Knight TM,
      4. et al
      . A systematic review of evidence for the added benefits to health of exposure to natural environments. BMC Public Health 2010;10:456.
      OpenUrlCrossRefPubMed
    13. ↵
      1. Nieuwenhuijsen MJ,
      2. Kruize H,
      3. Gidlow C,
      4. et al
      . Positive Health Effects of the Natural Outdoor Environment in Typical Populations in Different Regions in Europe (PHENOTYPE). 2013.
    14. ↵
      1. Agay-Shay K
      . Congenital Malformations & the Associations with Ambient Air Pollution in Israel, 2000–2006. [PhD]. Haifa, Israel: University of Haifa, Faculty of social welfare and health sciences, School of Public Health, 2013.
    15. ↵
      1. Amitai Y
      . Infant and children mortality during 2005 in Israel- national report. Ministry of Health, Department of Mother and Child Health, 2006.
    16. ↵
      The Gertner Institute Web page. Trends in the number of livebirths of VLBW infants and multiple births 1996–2005. http://www.gertnerinst.org.il/epidemiology/woman_child/vlbw_registry/vlbw_registry_special_reports/ (accessed 10 Dec 2013).
    17. ↵
      1. Dollberg S,
      2. Haklai Z,
      3. Mimouni FB,
      4. et al
      . Birth weight standards in the live-born population in Israel. Isr Med Assoc J 2005;7:311–14.
      OpenUrlPubMedWeb of Science
    18. ↵
      1. Aharon A,
      2. Einhorn S,
      3. Ben-Tolila M,
      4. et al
      . Israel in Figures, The Statistical Abstract of Israel No. 57, 2006 The Central Bureau of Statistics, Jerusalem. 2008. http://www.cbs.gov.il/publications/isr_in_n06e.pdf (accessed 10 Apr 2014).
    19. ↵
      1. Agay-Shay K,
      2. Amitai Y,
      3. Peretz C,
      4. et al
      . Exploratory spatial data analysis of congenital malformations (CM) in Israel, 2000–2006. ISPRS Int J Geo-Info 2013;2:237–55.
      OpenUrlCrossRef
    20. ↵
      1. Mitchell R,
      2. Astell-Burt T,
      3. Richardson EA
      . A comparison of green space indicators for epidemiological research. J Epidemiol Community Health 2011;65:853–8.
      OpenUrlAbstract/FREE Full Text
    21. ↵
      USGS science for a changing world. USGC Global Visualitzation Viewer. http://glovis.usgs.gov/ (accessed 10 Apr 2014).
    22. ↵
      1. Ramm F
      . OpenStreetMap: using and enhancing the free map of the world. Cambridge: UIT Cambridge, 2011.
    23. ↵
      1. Ludlow D,
      2. Mitchell C,
      3. Webster M
      . European Common Indicators: Towards a Local Sustainability Profile. [serial online]. 2003. http://ec.europa.eu/environment/urban/pdf/local_sustainability_en.pdf (accessed 10 Apr 2014)..
    24. ↵
      Central Bureau of Statistics. Central Bureau of Statistics website. http://www.cbs.gov.il/census/census/main_mifkad08_e.html (accessed 10 Apr 2014).
    25. ↵
      Muhsen. Health Minister's report on smoking status in Israel, 2006. 2006. http://www.cancer.org.il/download/files/smokingreport2006-7.pdf (accessed 10 Apr 2014).
    26. ↵
      1. Agay-Shay K,
      2. Friger M,
      3. Linn S,
      4. et al
      . Air pollution and congenital heart defects. Environ Res 2013;124:28–34.
      OpenUrlWeb of Science
    27. ↵
      1. Rose G
      . Sick individuals and sick populations. Int J Epidemiol 2001;30:427–32.
      OpenUrlAbstract/FREE Full Text
    28. ↵
      1. Bolte G,
      2. Tamburlini G,
      3. Kohlhuber M
      . Environmental inequalities among children in Europe—evaluation of scientific evidence and policy implications. Eur J Public Health 2010;20:14–20.
      OpenUrlAbstract/FREE Full Text
    29. ↵
      1. Horev T
      . Health Inequality, Policy Analysis and Recommendations for the Israeli System, Taub Center for Social Policy Studies in Israel. [Hebrew]. 2008.
    30. ↵
      1. Haim M
      . Trees reduce air pollution. Ecology & Environment. [Hebrew] 2013;4:182.
      OpenUrl
    31. ↵
      1. Shashua-Bar L,
      2. Pearlmutter D,
      3. Erell E
      . The influence of trees and grass on outdoor thermal comfort in a hot-arid environment. Int J Climatol 2011;31:1498–506.
      OpenUrlCrossRefWeb of Science
    32. ↵
      1. Potchter O,
      2. Shashua-Bar L
      . The effect of vegetation and built-up elements on the micro-climate conditions, air quality and noise in urban open spaces—measurements in situ, simulations and guidelines for sustainable planning. 6–802. http://old.sviva.gov.il/Enviroment/Static/Binaries/mechkarim/6-802_1.pdf (accessed 10 Apr 2014).
    33. ↵
      1. Freiman MT,
      2. Hirshel N,
      3. Broday DM
      . Urban-scale variability of ambient particulate matter attributes. Atmos Environ 2006;40:5670–84.
      OpenUrlCrossRefWeb of Science
    34. ↵
      1. Shashua-Bar L,
      2. Potchter O,
      3. Bitan A,
      4. et al
      . Microclimate modelling of street tree species effects within the varied urban morphology in the Mediterranean city of Tel Aviv, Israel. Int J Climatol 2010;30:44–57.
      OpenUrlWeb of Science
    35. ↵
      1. Cohen P,
      2. Potcher O
      . Daily and seasonal air quality characteristics of urban parks in the mediterranean city of Tel Aviv. Antwerp, Belgium: CLIMAQS Workshop ‘Local Air Quality and its Interactions with Vegetation, 2010.
    36. ↵
      1. Ekström J
      . A generalized definition of the polychoric correlation coefficient. 2011.
    37. ↵
      Keßler C, de Groot RTA. Trust as a proxy measure for the quality of volunteered geographic information in the case of OpenStreetMap. In: Vandenbroucke D, Bucher B, Crompvoets J, eds. Geographic Information Science at the Heart of Europe. Lecture Notes in Geoinformation and Cartography. New York: Springer. 2013:21–37.
    38. ↵
      1. Neis P,
      2. Zielstra D,
      3. Zipf A
      . The street network evolution of crowdsourced maps: OpenStreetMap in Germany 2007–2011. Future Internet 2011;4:1–21.
      OpenUrlCrossRef
    39. ↵
      1. Navidi W,
      2. Lurmann F
      . Measurement error in air-pollution exposure assessment. J Expo Anal Environ Epidemiol 1995;5:111–24.
      OpenUrlPubMedWeb of Science
    40. ↵
      1. Ritz B,
      2. Wilhelm M
      . Ambient air pollution and adverse birth outcomes: methodologic issues in an emerging field. Basic Clin Pharmacol Toxicol 2008;102:182–90.
      OpenUrlCrossRefPubMedWeb of Science

    Supplementary materials

    • Supplementary Data

      This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

      Files in this Data Supplement:

    Footnotes

    • Contributors KA-S participated in the conception, design and planning of the study, performed data analysis, execution of research programme, interpretation of the results and manuscript writing and review of the literature. AP contributed in the conception of geocoding process, exposure assessment and writing of the manuscript. AVC performed data analysis CP contributed in the conception and writing of the manuscript YA supplied data from participating health registries, advice on data collection and critical discussion. MF contributed in the conception and writing of the manuscript. SL contributed in the manuscript review and critical discussion. MJN participated in the conception, design and planning of the study, and supervised the data analysis and revision of the manuscript. All authors saw and approved the final version of the manuscript.

    • Funding Environment and Health Fund.

    • Competing interests KA-S is supported by a post-doctoral fellowship from the Environment and Health Fund, Jerusalem, Israel.

    • Ethics approval Ethics Committee of the University of Haifa. Further approvals from the director of Public Health Services and from the legal advisor in the Israeli Ministry of Health and the Ministry of Interior were also obtained.

    • Provenance and peer review Not commissioned; externally peer reviewed.

    Linked Articles