Linking Air Pollution Data and Adverse Birth Outcomes: Environmental Public Health Tracking in New York State Jessica M. Brown, MPH, RN; Gerald Harris, PhD; Cristian Pantea, MS; Syni-An Hwang, PhD; Thomas O. Talbot, MSPH rrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrrr
Background: Studies investigating associations between ambient air pollution and fetal growth and gestational duration have reported inconclusive findings. Objectives: The study goal was to use the Environmental Public Health Tracking Network to describe the association between exposure to particulate matter (PM2.5 ) and ozone and term low birth weight (TLBW) in New York State. Methods: Birth data for the years 2001-2006 were linked to Census data and hierarchical Bayesian modeled air pollution data. Daily 8-hour maximums for ozone and daily average PM2.5 estimates were averaged by trimester and exposure quartiles. The Environmental Public Health Tracking Academic Center for Excellence at Rutgers University partnered with New York and several other states to create a statistical program that uses logistic regression to determine the association between air pollution exposure and TLBW. Results: There were no consistent dose-response relationships between the pollutants and TLBW. Ozone exposure was associated with a higher risk of TLBW only in the first trimester, but these results were not statistically significant. Exposure to the third quartile of ozone for the full gestational period had negative associations with TLBW (odds ratio = 0.86; 95% confidence interval, 0.81-0.92). Conclusion: Collaboration within the Environmental Public Health Tracking Network to share methods and data for research proved feasible and efficient in assessing the relationship of air pollutants to adverse birth outcomes. This study finds little evidence to support positive associations between exposure to ozone or PM2.5 and TLBW in New York State.
J Public Health Management Practice, 2015, 21(2 Supp), S68–S74 C 2015 Wolters Kluwer Health, Inc. All rights reserved. Copyright
KEY WORDS: air pollution, Environmental Public Health Tracking,
preterm birth, term low birth weight Studies conducted to investigate the association between air pollution and measures of fetal growth and gestational duration have yielded conflicting results.1-8 However, these studies have differences in study design, exposure assessment, and measurement metrics that make it difficult to compare results. Several studies have reported a possible association between low birth weight and an increase in PM10 exposure.1-3 However, the trimester with the greatest association differed among the 3 studies. In contrast, a study conducted by Madsen et al4 did not find significant associations between PM2.5 exposure and low birth weight and a study by Maisonet et al5 did not show statistically significant associations between PM10 and low birth weight in the northeastern United States. Studies examining the effect of ozone on low birth weight have
Author Affiliations: Bureau of Occupational and Environmental Epidemiology, New York State Department of Health, Albany, New York (Mss Brown and Mr Pantea, Dr Hwang, and Mr Talbot); and Robert Wood Johnson Medical School, Rutgers University, Piscataway (Dr Harris). This study was supported by the Centers for Disease Control and Prevention Environmental Public Health Tracking Program grant #U38EH000942 and by contract #200-2010-37441. The authors thank Dr Daniel Wartenberg of Rutgers University for helping conceptualize this project and facilitate partnership between Rutgers and New York State Health Department (NYSDOH) to the authors also thank Mr Steven Forand of the NYSDOH for help with the study design and acquisition of the data and Dr Tabassum Insaf of the NYSDOH in preparing the manuscript. The authors declare that they have no conflict of financial or personal interest. Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s Web site (http://www.JPHMP.com). Correspondence: Thomas O. Talbot, MSPH, Center for Environmental Health, Bureau of Environmental and Occupational Epidemiology, New York State Department of Health, Empire State Plaza, Corning Tower, Room 1203, Albany, NY 12237 ([email protected]). DOI: 10.1097/PHH.0000000000000171
S68 Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
EPHT Air Pollution Study
reported no significant association.3,6,7 Recommendations for future studies have included an improved method of estimating exposure at a large population level by increasing spatial resolutions of measure, inclusions of entire populations, and controlling for maternal smoking.9,10 The Centers for Disease Control and Prevention (CDC), along with state health departments and academic institutions, has long recognized the need for more accurate, finer-scale (geographically and temporally) health and environmental data, along with the tools, to study the potential associations between air pollution and adverse birth outcomes. However, academic institutions have had difficulty gaining access to health records at the appropriate geographic and temporal scales due to confidentiality concerns of the data stewards. Through CDC’s Environmental Public Health Tracking (EPHT) Network, health departments are encouraged to work closely with universities and other health and environmental agencies to conduct surveillance and research. Indeed, one of the goals of the EPHT Network is to extend traditional health surveillance by jointly tracking environmental hazards, exposures, and health effects potentially related to exposure to environmental hazards.11 The principal objective of this study was to use the EPHT Network and resources to describe the association between exposure to particulate matter (PM2.5 ) or ozone and term low birth weight (TLBW) in New York State (NYS) for the years 2001-2006 while controlling for known risk factors. Air pollution estimates used in this study, created from both monitor and modeled data on a 12-km2 grid, allow for more complete coverage both temporally and geographically than monitor data alone. NYS birth data also allow control for maternal smoking. A secondary objective of the project was to develop a framework for linking fine-scale temporal and geographic health data with air pollution and sociodemographic data. The goal was to develop an analytic data set at the appropriate level of detail for epidemiologic analyses while minimizing the risk of disclosure of confidential health information before sharing data with study partners for epidemiologic analyses.
● Methods Birth data Both the NYS Department of Health Vital Statistics Unit and NYS Department of Health Institutional Review Board approved the study and the sharing of a linked analytic data set with Rutgers University. NYS birth records were obtained from the vital records division
❘ S69
of NYS Department of Health. The study population included singleton live births between September 1, 2001, and December 31, 2006, that occurred in NYS excluding the 5 counties that make up New York City (NYC): New York, Kings, Bronx, Richmond, and Queens. Approximately 47% of the NYS births were born to NYC residents. Births from NYC were excluded for several reasons. The NYS Department of Health is not the data steward for NYC birth records. The NYC Department of Health and Mental Hygiene maintains its own vital statistics and has its own EPHT program. More importantly, excluding NYC births from the analysis allowed for improved assessment of differences in geographic areas of the state by eliminating the overwhelming population in a singular exposure category from NYC. There were 829 545 births in NYS during the study period. Births to out-of-state or NYC residents were excluded. Only singleton births were included in the study population. Births with unspecified maternal address at the time of birth were excluded. Births with a gestational age less than 17 weeks or greater than 47 weeks and births with a missing birth weight or gestational age or weight recorded as less than 100 g were also excluded as being implausible records. After these exclusions, the final sample size consisted of 480 430 births of which 9782 (2.04%) were TLBW.
Air data The US Environmental Protection Agency (EPA) created air pollution data in support of the CDC’s National EPHT Network using a hierarchical Bayesian space time modeling (HBM) system.12 These data provide daily estimates of ambient levels of ozone (8-hour maximum parts per billion [ppb]) and average micrograms per cubic meter (μg/m3 ) airborne particulates (PM2.5 ) on a 12-km2 grid scale in the eastern United States for the years 2001-2006. There were 1001 grid cells in our study area. The HBM uses Community Multiscale Air Quality data and EPA monitor data from its Air Quality System. The hierarchical Bayesian model uses monitor data from the entire geographic domain to produce estimates for the study area. The monitor data are weighted more heavily in areas near where monitors exist.12 There were approximately 14 PM2.5 monitoring stations and 30 ozone monitoring stations in the study area operating during the study period, although the number did vary slightly over the course of the study period. About three-fourth of both the PM2.5 and ozone monitors were situated in urban areas as defined by Rural-Urban Commuting Area codes.13 PM2.5 monitors are operational throughout the year. However, there were about 28% fewer daily ozone measurements taken during the colder months (NovemberMarch) when ozone levels are lower. The data were
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
S70 ❘ Journal of Public Health Management and Practice retrieved from EPA’s Air Quality Data from the CDC National EPHT Network.
We also used 2000 Census data to allow for consideration of neighborhood factors, such as household and family income; percentage of poverty and percentage of poverty in the population younger than 6 years; percentage of all minorities and percentage of white non-Hispanic population; percent unemployed; percentage with education beyond a bachelor’s degree; and percentage of population with less than high school education.
health information (Supplemental Digital Appendix 1, available at http://links.lww.com/JPHMP/A115). The data elements eliminated included birth day, street address, zip code, block group number, zip code tabulation area, and air pollution grid ID number. The final de-identified data were shared with Rutgers University. Rutgers University researchers used these shared data to develop the statistical models described in this article. Rutgers University also used similar “linkand-strip” data sharing methods to obtain data from other state EPHT programs to study the association of PM2.5 with full-term birth weight at different spatial scales.15
Data linkage
Statistical analysis
The birth data were automatically geocoded to street address by the NYSDOH Vital Statistics Unit using MapMarker (version 22). Addresses that were not geocoded automatically were corrected for misspellings of streets and missing zip codes using online resources. Where street level geocoding was not available, the grid cell containing the population-weighted centroid of that birth record’s zip code was used. Each birth was then geographically linked to a 12-km2 grid cell. EPHT indicator measures, daily 8-hour maximums for ozone and daily average PM2.5 estimates, were averaged by trimester for each 12-km2 grid cell of the HBM data to calculate individual-level, trimester-specific exposure based on mother’s residence. Births were geographically linked to Census data.
Rutgers University used SAS 9.2 to create logistic regression models to assess the relationship between TLBW and PM2.5 and ozone by estimating odds ratios (ORs) and 95% confidence intervals (CIs). TLBW was defined as a gestational age of 37 weeks or more and 40 weeks or less, with birth weight 2500 g or less. The EPHT program uses an algorithm, developed by the National Center for Health Statistics, to eliminate implausible combinations of gestational age and birth weight when calculating gestational age–based indicators. For births where the calculated gestational age lies outside the predicted parameters of the birth weight, the clinical estimate is used instead.16 Trimester divisions were determined using the last menstrual period variable from birth records and defined as 1 to 13, 14 to 26, and 27 weeks and on for the third trimester; these definitions have been applied elsewhere.17,18 The SAS program created by Rutgers University partners ranked exposure to PM2.5 and ozone into quartiles of their distribution(≤25th, >25th-50th, >50th75th, >75th). The subjects with exposure in the first quartile were used as the reference category. The range of values for the exposure categories are shown in Table 2. Sensitivity analyses were conducted by running the model with both pollutants to test stability of one pollutant model, as well as without ethnicity as a covariate to see if including those missing ethnicity (n = 33 671; 7.98%) would change results. Data were also stratified by month of conception and rural versus urban commuting areas. Because none of the census variables included were found to be significant in bivariate analysis, they were dropped from the final model.
Census data
Possible confounders/covariates Covariates potentially related to birth outcomes and exposures were included in bivariate analyses. The following covariates were obtained from birth certificate data and controlled for in all analyses: maternal age, maternal race, maternal education, adequacy of prenatal care Kessner index,14 and parity were entered as ordinal variables; infant sex, maternal ethnicity, and maternal smoking during pregnancy were entered as dichotomous variables; and pregnancy complications (eclampsia, chronic hypertension, gestational hypertension, gestational diabetes, premature rupture of membranes, and placental abruption) were totaled as a dichotomous (any or none) and a discrete variable (0, 1, 2, or greater). ZIP codes from the birth records were used to assign rural or urban location utilizing Rural-Urban Commuting Area codes.
Data sharing protocol Once the health, air pollution, and Census data were linked and air pollution metrics were created for each trimester of pregnancy, we stripped the fine-scale geographic and temporal data elements from the linked data set to reduce the risk of disclosure of confidential
● Results Vital records provided 528 312 singleton birth records with complete gestational periods occurring during 2001-2006 in NYS excluding NYC. After retaining only those records that had acceptable calculated or
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
EPHT Air Pollution Study
clinically estimated gestational ages, as well as excluding births to out-of-state residents, and observations with missing location information, birth weight, or date of birth, 480 430 (91%) records remained. Remaining data included 9782 (2.04%) TLBW births. The prevalence of TLBW and preterm birth was higher for mothers younger than 20 years, mothers with less than a high school education, Kessner prenatal care index rated lower than “good,” African American mothers, Hispanic mothers, those who smoked during pregnancy, and those with a separate pregnancy complication (Table 1). If the birth was a second birth, parity was protective for TLBW. Average ozone and PM2.5 concentrations during the full gestational period were the same, 11.0 μg/m3 and 38.7 ppb, respectively, for both the reference and TLBW births.
❘ S71
Table 2 shows the risks of TLBW associated with increases in quartile ranges in exposure to trimesteraveraged PM2.5 and ozone. There was a negative association between ozone and TLBW in the third trimester. A positive dose-response relationship was seen only with first-trimester exposure to ozone but was not statistically significant. Other results were inconclusive. Results from sensitivity analyses, stratifying by month of conception and rural or urban maternal residence, did not show any effect modification by these variables. Sensitivity analysis conducted without ethnicity as a covariate added an additional 33 671 records (7.98%) but did not improve precision. Results were not attenuated by including both pollutants in the same model. To test whether using an average as the exposure metric was resulting in a loss of precision, the
TABLE 1 ● Characteristics of Study Population by Birth Category qqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqq
Characteristics
Reference Births,a n (%)
Sample size 421 763 (87.79) Maternal age, y
Background: Studies investigating associations between ambient air pollution and fetal growth and gestational duration have reported inconclusive findings. Objectives: The study goal was to use the Environmental Public Health Tracking Network to describe the association between exposure to particulate matter (PM2.5 ) and ozone and term low birth weight (TLBW) in New York State. Methods: Birth data for the years 2001-2006 were linked to Census data and hierarchical Bayesian modeled air pollution data. Daily 8-hour maximums for ozone and daily average PM2.5 estimates were averaged by trimester and exposure quartiles. The Environmental Public Health Tracking Academic Center for Excellence at Rutgers University partnered with New York and several other states to create a statistical program that uses logistic regression to determine the association between air pollution exposure and TLBW. Results: There were no consistent dose-response relationships between the pollutants and TLBW. Ozone exposure was associated with a higher risk of TLBW only in the first trimester, but these results were not statistically significant. Exposure to the third quartile of ozone for the full gestational period had negative associations with TLBW (odds ratio = 0.86; 95% confidence interval, 0.81-0.92). Conclusion: Collaboration within the Environmental Public Health Tracking Network to share methods and data for research proved feasible and efficient in assessing the relationship of air pollutants to adverse birth outcomes. This study finds little evidence to support positive associations between exposure to ozone or PM2.5 and TLBW in New York State.
J Public Health Management Practice, 2015, 21(2 Supp), S68–S74 C 2015 Wolters Kluwer Health, Inc. All rights reserved. Copyright
KEY WORDS: air pollution, Environmental Public Health Tracking,
preterm birth, term low birth weight Studies conducted to investigate the association between air pollution and measures of fetal growth and gestational duration have yielded conflicting results.1-8 However, these studies have differences in study design, exposure assessment, and measurement metrics that make it difficult to compare results. Several studies have reported a possible association between low birth weight and an increase in PM10 exposure.1-3 However, the trimester with the greatest association differed among the 3 studies. In contrast, a study conducted by Madsen et al4 did not find significant associations between PM2.5 exposure and low birth weight and a study by Maisonet et al5 did not show statistically significant associations between PM10 and low birth weight in the northeastern United States. Studies examining the effect of ozone on low birth weight have
Author Affiliations: Bureau of Occupational and Environmental Epidemiology, New York State Department of Health, Albany, New York (Mss Brown and Mr Pantea, Dr Hwang, and Mr Talbot); and Robert Wood Johnson Medical School, Rutgers University, Piscataway (Dr Harris). This study was supported by the Centers for Disease Control and Prevention Environmental Public Health Tracking Program grant #U38EH000942 and by contract #200-2010-37441. The authors thank Dr Daniel Wartenberg of Rutgers University for helping conceptualize this project and facilitate partnership between Rutgers and New York State Health Department (NYSDOH) to the authors also thank Mr Steven Forand of the NYSDOH for help with the study design and acquisition of the data and Dr Tabassum Insaf of the NYSDOH in preparing the manuscript. The authors declare that they have no conflict of financial or personal interest. Supplemental digital content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s Web site (http://www.JPHMP.com). Correspondence: Thomas O. Talbot, MSPH, Center for Environmental Health, Bureau of Environmental and Occupational Epidemiology, New York State Department of Health, Empire State Plaza, Corning Tower, Room 1203, Albany, NY 12237 ([email protected]). DOI: 10.1097/PHH.0000000000000171
S68 Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
EPHT Air Pollution Study
reported no significant association.3,6,7 Recommendations for future studies have included an improved method of estimating exposure at a large population level by increasing spatial resolutions of measure, inclusions of entire populations, and controlling for maternal smoking.9,10 The Centers for Disease Control and Prevention (CDC), along with state health departments and academic institutions, has long recognized the need for more accurate, finer-scale (geographically and temporally) health and environmental data, along with the tools, to study the potential associations between air pollution and adverse birth outcomes. However, academic institutions have had difficulty gaining access to health records at the appropriate geographic and temporal scales due to confidentiality concerns of the data stewards. Through CDC’s Environmental Public Health Tracking (EPHT) Network, health departments are encouraged to work closely with universities and other health and environmental agencies to conduct surveillance and research. Indeed, one of the goals of the EPHT Network is to extend traditional health surveillance by jointly tracking environmental hazards, exposures, and health effects potentially related to exposure to environmental hazards.11 The principal objective of this study was to use the EPHT Network and resources to describe the association between exposure to particulate matter (PM2.5 ) or ozone and term low birth weight (TLBW) in New York State (NYS) for the years 2001-2006 while controlling for known risk factors. Air pollution estimates used in this study, created from both monitor and modeled data on a 12-km2 grid, allow for more complete coverage both temporally and geographically than monitor data alone. NYS birth data also allow control for maternal smoking. A secondary objective of the project was to develop a framework for linking fine-scale temporal and geographic health data with air pollution and sociodemographic data. The goal was to develop an analytic data set at the appropriate level of detail for epidemiologic analyses while minimizing the risk of disclosure of confidential health information before sharing data with study partners for epidemiologic analyses.
● Methods Birth data Both the NYS Department of Health Vital Statistics Unit and NYS Department of Health Institutional Review Board approved the study and the sharing of a linked analytic data set with Rutgers University. NYS birth records were obtained from the vital records division
❘ S69
of NYS Department of Health. The study population included singleton live births between September 1, 2001, and December 31, 2006, that occurred in NYS excluding the 5 counties that make up New York City (NYC): New York, Kings, Bronx, Richmond, and Queens. Approximately 47% of the NYS births were born to NYC residents. Births from NYC were excluded for several reasons. The NYS Department of Health is not the data steward for NYC birth records. The NYC Department of Health and Mental Hygiene maintains its own vital statistics and has its own EPHT program. More importantly, excluding NYC births from the analysis allowed for improved assessment of differences in geographic areas of the state by eliminating the overwhelming population in a singular exposure category from NYC. There were 829 545 births in NYS during the study period. Births to out-of-state or NYC residents were excluded. Only singleton births were included in the study population. Births with unspecified maternal address at the time of birth were excluded. Births with a gestational age less than 17 weeks or greater than 47 weeks and births with a missing birth weight or gestational age or weight recorded as less than 100 g were also excluded as being implausible records. After these exclusions, the final sample size consisted of 480 430 births of which 9782 (2.04%) were TLBW.
Air data The US Environmental Protection Agency (EPA) created air pollution data in support of the CDC’s National EPHT Network using a hierarchical Bayesian space time modeling (HBM) system.12 These data provide daily estimates of ambient levels of ozone (8-hour maximum parts per billion [ppb]) and average micrograms per cubic meter (μg/m3 ) airborne particulates (PM2.5 ) on a 12-km2 grid scale in the eastern United States for the years 2001-2006. There were 1001 grid cells in our study area. The HBM uses Community Multiscale Air Quality data and EPA monitor data from its Air Quality System. The hierarchical Bayesian model uses monitor data from the entire geographic domain to produce estimates for the study area. The monitor data are weighted more heavily in areas near where monitors exist.12 There were approximately 14 PM2.5 monitoring stations and 30 ozone monitoring stations in the study area operating during the study period, although the number did vary slightly over the course of the study period. About three-fourth of both the PM2.5 and ozone monitors were situated in urban areas as defined by Rural-Urban Commuting Area codes.13 PM2.5 monitors are operational throughout the year. However, there were about 28% fewer daily ozone measurements taken during the colder months (NovemberMarch) when ozone levels are lower. The data were
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
S70 ❘ Journal of Public Health Management and Practice retrieved from EPA’s Air Quality Data from the CDC National EPHT Network.
We also used 2000 Census data to allow for consideration of neighborhood factors, such as household and family income; percentage of poverty and percentage of poverty in the population younger than 6 years; percentage of all minorities and percentage of white non-Hispanic population; percent unemployed; percentage with education beyond a bachelor’s degree; and percentage of population with less than high school education.
health information (Supplemental Digital Appendix 1, available at http://links.lww.com/JPHMP/A115). The data elements eliminated included birth day, street address, zip code, block group number, zip code tabulation area, and air pollution grid ID number. The final de-identified data were shared with Rutgers University. Rutgers University researchers used these shared data to develop the statistical models described in this article. Rutgers University also used similar “linkand-strip” data sharing methods to obtain data from other state EPHT programs to study the association of PM2.5 with full-term birth weight at different spatial scales.15
Data linkage
Statistical analysis
The birth data were automatically geocoded to street address by the NYSDOH Vital Statistics Unit using MapMarker (version 22). Addresses that were not geocoded automatically were corrected for misspellings of streets and missing zip codes using online resources. Where street level geocoding was not available, the grid cell containing the population-weighted centroid of that birth record’s zip code was used. Each birth was then geographically linked to a 12-km2 grid cell. EPHT indicator measures, daily 8-hour maximums for ozone and daily average PM2.5 estimates, were averaged by trimester for each 12-km2 grid cell of the HBM data to calculate individual-level, trimester-specific exposure based on mother’s residence. Births were geographically linked to Census data.
Rutgers University used SAS 9.2 to create logistic regression models to assess the relationship between TLBW and PM2.5 and ozone by estimating odds ratios (ORs) and 95% confidence intervals (CIs). TLBW was defined as a gestational age of 37 weeks or more and 40 weeks or less, with birth weight 2500 g or less. The EPHT program uses an algorithm, developed by the National Center for Health Statistics, to eliminate implausible combinations of gestational age and birth weight when calculating gestational age–based indicators. For births where the calculated gestational age lies outside the predicted parameters of the birth weight, the clinical estimate is used instead.16 Trimester divisions were determined using the last menstrual period variable from birth records and defined as 1 to 13, 14 to 26, and 27 weeks and on for the third trimester; these definitions have been applied elsewhere.17,18 The SAS program created by Rutgers University partners ranked exposure to PM2.5 and ozone into quartiles of their distribution(≤25th, >25th-50th, >50th75th, >75th). The subjects with exposure in the first quartile were used as the reference category. The range of values for the exposure categories are shown in Table 2. Sensitivity analyses were conducted by running the model with both pollutants to test stability of one pollutant model, as well as without ethnicity as a covariate to see if including those missing ethnicity (n = 33 671; 7.98%) would change results. Data were also stratified by month of conception and rural versus urban commuting areas. Because none of the census variables included were found to be significant in bivariate analysis, they were dropped from the final model.
Census data
Possible confounders/covariates Covariates potentially related to birth outcomes and exposures were included in bivariate analyses. The following covariates were obtained from birth certificate data and controlled for in all analyses: maternal age, maternal race, maternal education, adequacy of prenatal care Kessner index,14 and parity were entered as ordinal variables; infant sex, maternal ethnicity, and maternal smoking during pregnancy were entered as dichotomous variables; and pregnancy complications (eclampsia, chronic hypertension, gestational hypertension, gestational diabetes, premature rupture of membranes, and placental abruption) were totaled as a dichotomous (any or none) and a discrete variable (0, 1, 2, or greater). ZIP codes from the birth records were used to assign rural or urban location utilizing Rural-Urban Commuting Area codes.
Data sharing protocol Once the health, air pollution, and Census data were linked and air pollution metrics were created for each trimester of pregnancy, we stripped the fine-scale geographic and temporal data elements from the linked data set to reduce the risk of disclosure of confidential
● Results Vital records provided 528 312 singleton birth records with complete gestational periods occurring during 2001-2006 in NYS excluding NYC. After retaining only those records that had acceptable calculated or
Copyright © 2015 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
EPHT Air Pollution Study
clinically estimated gestational ages, as well as excluding births to out-of-state residents, and observations with missing location information, birth weight, or date of birth, 480 430 (91%) records remained. Remaining data included 9782 (2.04%) TLBW births. The prevalence of TLBW and preterm birth was higher for mothers younger than 20 years, mothers with less than a high school education, Kessner prenatal care index rated lower than “good,” African American mothers, Hispanic mothers, those who smoked during pregnancy, and those with a separate pregnancy complication (Table 1). If the birth was a second birth, parity was protective for TLBW. Average ozone and PM2.5 concentrations during the full gestational period were the same, 11.0 μg/m3 and 38.7 ppb, respectively, for both the reference and TLBW births.
❘ S71
Table 2 shows the risks of TLBW associated with increases in quartile ranges in exposure to trimesteraveraged PM2.5 and ozone. There was a negative association between ozone and TLBW in the third trimester. A positive dose-response relationship was seen only with first-trimester exposure to ozone but was not statistically significant. Other results were inconclusive. Results from sensitivity analyses, stratifying by month of conception and rural or urban maternal residence, did not show any effect modification by these variables. Sensitivity analysis conducted without ethnicity as a covariate added an additional 33 671 records (7.98%) but did not improve precision. Results were not attenuated by including both pollutants in the same model. To test whether using an average as the exposure metric was resulting in a loss of precision, the
TABLE 1 ● Characteristics of Study Population by Birth Category qqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqqq
Characteristics
Reference Births,a n (%)
Sample size 421 763 (87.79) Maternal age, y
