eTable. The Impact of 2 State Tobacco Control Policies—Smoke-Free Legislation and Cigarette Tax Increases—On the Probability of Maternal Smoking During Pregnancy
Hawkins SS, Baum CF, Oken E, Gillman MW. Associations of Tobacco Control Policies With Birth Outcomes. JAMA Pediatr. 2014;168(11):e142365. doi:10.1001/jamapediatrics.2014.2365
It is unclear whether the benefits of tobacco control policies extend to pregnant women and infants, especially among racial/ethnic minority and low socioeconomic populations that are at highest risk for adverse birth outcomes.
To examine the associations of state cigarette taxes and the enactment of smoke-free legislation with US birth outcomes according to maternal race/ethnicity and education.
Design, Setting, and Participants
Using a quasi-experimental approach, we analyzed repeated cross sections of US natality files with 16 198 654 singleton births from 28 states and Washington, DC, between 2000 and 2010. We first used probit regression to model the associations of 2 tobacco control policies with the probability that a pregnant woman smoked (yes or no). We then used linear or probit regression to estimate the associations of the policies with birth outcomes. We also examined the association of taxes with birth outcomes across maternal race/ethnicity and education.
State cigarette taxes and smoke-free restaurant legislation.
Main Outcomes and Measures
Birth weight (in grams), low birth weight (<2500 g), preterm delivery (<37 weeks), small for gestational age (<10th percentile for gestational age and sex), and large for gestational age (>90th percentile for gestational age and sex).
White and black mothers with the least amount of education (0-11 years) had the highest prevalence of maternal smoking during pregnancy (42.4% and 20.0%, respectively) and the poorest birth outcomes, but the strongest responses to cigarette taxes. Among white mothers with a low level of education, every $1.00 increase in the cigarette tax reduced the level of smoking by 2.4 percentage points (−0.0024 [95% CI, −0.0004 to −0.0001]), and the birth weight of their infants increased by 5.41 g (95% CI, 1.92-8.89 g). Among black mothers with a low level of education, tax increases reduced smoking by 2.1 percentage points (−0.0021 [95% CI, −0.0003 to −0.0001]), and the birth weight of their infants increased by 3.98 g (95% CI, 1.91-6.04 g). Among these mothers, tax increases also reduced the risk of having low-birth-weight, preterm, and small-for-gestational-age babies, but increased the risk of having large-for-gestational-age babies. Associations were weaker among higher-educated black women and largely null among higher educated white women and other groups. We did not find evidence for an association of smoke-free restaurant legislation with birth outcomes.
Conclusions and Relevance
Increases in the cigarette tax are associated with improved health outcomes related to smoking among the highest-risk mothers and infants. Considering that US states increase cigarette taxes for reasons other than to improve birth outcomes, these findings are welcome by-products of state policies.
In utero exposure to tobacco smoke is one of the most modifiable determinants of low birth weight and related outcomes.1,2 Active smoking during pregnancy restricts fetal growth, reduces birth weight, and increases risk for preterm delivery.1 The consequences of exposure to secondhand smoke during pregnancy on fetal health are similar, albeit less pronounced.2,3 Although the prevalence of maternal smoking during pregnancy in the United States is at a historic low of 8.9%, striking racial/ethnic and educational gradients remain.4 Forty percent of white women and 16% of black women with less than a high school degree smoke during pregnancy.4 Furthermore, 37% of the US nonsmoking female population has detectable levels of secondhand smoke, with black and low-income populations having higher levels of exposure.5
Cigarette taxes and the enactment of smoke-free legislation across the United States and worldwide have resulted in population-level health benefits,6,7 but only a limited number of studies have investigated whether these improvements extend to pregnant women and their infants. Although both types of policies may improve birth outcomes, the mechanisms are likely quite different: taxes have been effective at reducing the number of smokers,8,9 whereas smoke-free legislation primarily reduces the level of exposure to secondhand smoke.6,10 Two US-based quasi-experimental studies using data from the 1990s found that increases in the cigarette tax were associated with an increase in the mean birth weight.11,12 A more recent series of before-after studies found that the introduction of smoke-free legislation at the country level13- 15 or city level16,17 was associated with reductions in the numbers of preterm,13,14,16- 18 small-for-gestational-age (SGA),13,15 and low-birth-weight infants.13 However, methodological limitations of these studies include the failure to account for other tobacco control policies,13- 18 the lack of a control group,13- 15,17 or the failure to account for secular trends.14,15 Furthermore, even though disparities in birth outcomes are well established,1,2 none of the studies examined whether policies differentially affect socially disadvantaged mothers and babies.
In a previous study, we exploited the natural experiment created by changes in tobacco control policy within and across the United States and showed that cigarette tax increases were associated with reductions in the number of white and black mothers with a low level of education who smoked during pregnancy (hereafter referred to as prenatal smoking).4 However, there was no association of smoke-free legislation with prenatal smoking. The purpose of the present study was to extend those findings to examine the associations of both policies with birth outcomes. We examined the associations of state cigarette taxes and the enactment of smoke-free legislation with birth weight, specifically on the numbers of low-birth-weight, preterm, SGA, and large-for-gestational-age (LGA) infants, according to the race/ethnicity and education of the mothers.
We obtained the natality public use microdata files for 2000 through 2004 from the National Vital Statistics System, which includes information on all births registered in the 50 US states, the District of Columbia, and New York City.19 From 2005 onward, state identifiers were no longer publicly available. For 2005 through 2010, we received approval from the National Association for Public Health Statistics and Information Systems for the microdata with state identifiers.20
We analyzed data from 28 states and Washington, DC, all of which used the 1989 Revision of the US Standard Certificate of Live Birth through at least 2007 (Alabama, Alaska, Arizona, Arkansas, Connecticut, the District of Columbia, Georgia, Hawaii, Illinois, Louisiana, Maine, Maryland, Massachusetts, Michigan, Minnesota, Mississippi, Missouri, Montana, Nevada, New Jersey, New Mexico, North Carolina, Oklahoma, Oregon, Rhode Island, Utah, Virginia, West Virginia, and Wisconsin). For those states that introduced the 2003 Revision of the birth certificate during the year, we excluded births from the month/year of introduction onward because smoking was assessed differently (Table 1). Our study was reviewed by the institutional review board at Boston College and considered exempt from requiring patient consent forms.
Among the total of 18 165 826 births, we included 16 648 751 singletons born at 30 to 44 weeks of gestation to mothers 18 to 50 years of age. We further excluded those missing information on smoking (224 632 births [1.3%]), maternal nativity (29 190 births [0.2%]), education (318 601 births [1.9%]), or birth weight (4135 births [0.02%]), as well as birth weights inconsistent with gestational age based on the method by Alexander et al21 (39 713 births [0.2%]). The final analytic sample included 16 198 654 births.
The 1989 Revision of the US Standard Certificate of Live Birth consists of 2 worksheets. A parent worksheet contains the legal and sociodemographic information on the child’s mother and father. A hospital worksheet is completed by a designated hospital representative who reports on prenatal care, labor and delivery, and neonatal conditions.
We first examined the associations of tobacco control policies with prenatal smoking. On the birth certificate, parents report on “tobacco use during pregnancy” with a yes or no response. Information on maternal smoking before pregnancy was not available.
Our main dependent variables were birth weight (continuous, in grams) and the following dichotomous outcomes: low birth weight (<2500 g), preterm delivery (before 37 weeks), SGA (<10th percentile for gestational age and sex),22 and LGA (>90th percentile for gestational age and sex).22
We linked state tobacco control policies to each birth certificate based on the month and year of delivery.4 The cigarette tax per pack of 20 cigarettes for each state was obtained from the historical compilation “The Tax Burden on Tobacco”23 and translated into real December 2010 US dollars using the national consumer price index.24 For each baby, we determined the cigarette tax (coded in dollars) 9 months prior to the month of delivery as an approximation for the month of conception.25 A 100% smoke-free legislation for workplaces and restaurants in each state was obtained from the American Nonsmokers’ Rights Foundation.26 Because all 17 states with smoke-free workplaces also had smoke-free restaurants (20 states), we used smoke-free restaurant legislation as a proxy for state smoke-free policies. For each baby, we determined whether there was smoke-free restaurant legislation in place (yes or no) 9 months prior to the month of delivery.
The birth certificate includes information on the mother’s race (including whether she was of Hispanic origin), highest level of education (0-11, 12, 13-15, or ≥16 years), age (<19, 20-24, 25-29, 30-34, or ≥35 years), marital status at the time of birth (married or not married), and country of birth (US born or foreign born). We combined self-reported race and Hispanic origin to indicate maternal race/ethnicity (white, black, Hispanic, Asian or Pacific Islander, or Native American or Alaskan native). Additional information was collected on the total number of live births (1, 2, or ≥3), the month prenatal care was initiated (first trimester, second trimester, third trimester, none, or unknown), gestational age (in weeks; calculated from the date of the last menstrual period), birth weight, and baby’s sex. We used imputed values provided in the data sets when data on the mother’s race, age, and marital status; number of live births; plurality; gestational age; and sex of the baby were missing.27
We conducted a series of differences-in-differences models, a causal inference technique, to evaluate the associations of tobacco control policies with birth outcomes. Because cigarette taxes and smoke-free restaurant legislation can influence birth outcomes directly and indirectly, via prenatal smoking, we used a 2-step modeling approach. We used the conditional mixed process estimator developed by Roodman28 to model the associations of tobacco control policies with birth outcomes, while adjusting for potential confounding factors. Using birth weight as an example, in the first model, we specified a probit regression for the probability that an expectant mother smoked; in the second model, we specified a linear regression for the birth weight of her baby, conditioned on the probability of smoking. That is, we take account of the mother’s probability of smoking during pregnancy as a factor influencing birth weight, allowing that association to vary across race/ethnicity and education categories. In analyses of dichotomous outcomes, we used a probit regression in the second equation. This framework allowed us first to estimate the associations of cigarette taxes and smoke-free legislation with smoking and then to estimate the direct associations of tobacco control policies with birth outcomes, as well as the indirect associations of policies conditioned on the probability of smoking.
We tested for the direct association of cigarette taxes with birth outcomes (ie, not through smoking itself), but finding no evidence for it, we did not include this pathway in the models. Further information on the statistical method is available in the eAppendix in the Supplement.
Based on our previous evaluation,4 the regression model that we used in the first step to estimate the association of taxes with the probability of smoking during pregnancy included an interaction between the mother’s race/ethnicity and education and the cigarette tax; an adjustment for 100% smoke-free restaurants and the sociodemographic characteristics of marital status, country of birth, number of live births, and prenatal care; and interactions between the mother’s race/ethnicity and age, state, and year. In the second step, for all birth outcomes, the models included an interaction between the mother’s race/ethnicity and education and the probability of smoking; an adjustment for 100% smoke-free restaurants, sex, and gestational age as linear and quadratic terms (except for preterm birth); and the same sociodemographic characteristics. This allowed the association of taxes with birth outcomes, via prenatal smoking, to differ by mother’s race/ethnicity and education. We included state and year fixed effects in all regression models to control for time-invariant state factors, as well as the decreasing secular trend in prenatal smoking.29 We calculated average marginal effects to evaluate the association of policy changes with the probability of smoking and, subsequently, the predicted change in the prevalence of each birth outcome using a chain-rule prediction. All analyses were conducted using Stata statistical software version 12.1SE (StataCorp), with cluster-robust standard errors and with clustering by state.
Geographic variation in adverse birth outcomes was widely evident (Table 1). Among these 28 states and Washington, DC, the proportion of babies born with low birth weight ranged from 3.8% in Alaska to 8.4% in Mississippi, 7.1% of babies were born preterm in Connecticut compared with 14.4% in Mississippi, and 6.8% of babies were born SGA in Alaska compared with 14.0% in Mississippi. Tobacco control policies varied considerably as well. During the study period, cigarette tax increases ranged from $0 in Missouri to $2.75 in Rhode Island, and 19 of 29 states introduced smoke-free restaurant policies.
In the first model, we estimated the associations of cigarette taxes and smoke-free legislation with the probability of smoking during pregnancy (eTable in the Supplement). We found that cigarette tax increases were associated with a reduced number of white mothers with a high school degree or less who smoked and a reduced number of black mothers across all educational levels who smoked, albeit with a smaller coefficient among the more highly educated. Every $1.00 increase in cigarette taxes were associated with reduced numbers of white and black mothers with 0 to 11 years of education who smoked by 2.4 and 2.1 percentage points, respectively, and by 0.8 and 0.9 percentage points, respectively, for those with 12 years of education. The educational gradient continued among black mothers, with cigarette tax increases reducing the number of black mothers with 13 to 15 years of education who smoked by 0.7 percentage points and reducing the number of black mothers with 16 or more years of education who smoked by 0.1 percentage points. In contrast, we did not find evidence for an association of smoke-free legislation with prenatal smoking.
Table 2 illustrates gradients across educational attainment for all birth outcomes. Although black mothers across all educational levels had the poorest birth outcomes, highly educated black mothers had poorer birth outcomes overall than white mothers with a low level of education. For example, 11% of black mothers with 16 or more years of education had a preterm birth compared with 17% of mothers with 0 to 11 years of education, while the corresponding proportions for white mothers were 7% and 11%, respectively.
In the second model, we estimated the direct association of smoke-free restaurant legislation with birth outcomes and the indirect association of cigarette taxes as they influence the probability of smoking (Table 2). The overall pattern of results showed that increases in cigarette taxes were associated with increased birth weight and a reduced number of adverse birth outcomes among white mothers with a low level of education and black mothers across all educational levels. Every $1.00 increase in cigarette taxes was associated with increases in birth weight by 5.4 and 4.0 g among white and black mothers with 0 to 11 years of education, respectively, as well as by 1.9 to 0.3 g for black mothers with 12 to 16 or more years of education. The pattern of results was consistent across the remaining birth outcomes. Among white and black mothers with a low level of education, every $1.00 increase in cigarette taxes was associated with reduced numbers of low-birth-weight infants by 0.08 and 0.12 percentage points, respectively, preterm infants by 0.07 and 0.08 percentage points, respectively, and SGA infants by 0.31 and 0.30 percentage points, respectively. In contrast, every $1.00 increase in cigarette taxes was associated with increased numbers of LGA infants by 0.18 and 0.10 percentage points for both white and black mothers with a low level of education, respectively. Cigarette tax increases, conditioned on the probability of smoking, also was associated with a reduced number of adverse birth outcomes among black mothers with 12 to 16 or more years of education, with a gradient of smaller coefficients with higher educational attainment. Overall, the associations were largely null among white women with a higher level of education and among women from other racial/ethnic groups. We did not find evidence for a direct association of smoke-free restaurant legislation with any birth outcomes.
Exploiting the natural experiment created by rapid changes in tobacco control policies within and across the United States, we found that cigarette tax increases resulted in higher birth weights and a reduced number of adverse birth outcomes among white and black mothers with a low level of education via reductions in the number of these women who smoked during pregnancy. In contrast, there was no direct association of smoke-free restaurant legislation with birth outcomes. These findings suggest that mothers at the highest risk for poor birth outcomes were the most responsive to cigarette tax increases. For the 28 states and Washington, DC, in this analysis, as of June 2014, taxes per pack varied from $0.17 in Missouri to $3.51 in Massachusetts.30 According to our estimates, if Missouri raised its cigarette tax to that of Massachusetts, the birth weight of infants born to Missouri mothers with less than a high school degree would increase, on average, by 18.1 g for white infants and 13.3 g for black infants.
Previously, we have shown that state cigarette taxes over the past decade were associated with reduced numbers of white and black mothers with a low level of education who smoked during pregnancy—the very populations with some of the highest levels of prenatal smoking.4 Now, examining the downstream policy effects, we find that cigarette taxes also are associated with improved birth outcomes via reductions in prenatal smoking rates among these same mothers. Prior studies used linear probability models and instrumental variable analysis and modeled smoking and birth weight in separate equations,11,12 which could inflate the apparent association of taxes with birth weight (eAppendix in the Supplement). Using the national natality files from 1989 to 1992, Evans and Ringel11 estimated that a $1.00 tax increase reduced the smoking rate by 8 percentage points and increased the mean birth weight by 21 g among all mothers. Our effect sizes were smaller by comparison, with a reduction in the smoking rate by 2 percentage points and an increase in the mean birth weight by nearly 5 g among only white and black mothers with a low level of education. Differences are also likely a result of changes over time in the sociodemographic composition of mothers who smoke during pregnancy4,25 due to the success of tobacco control policies and the strengthening of the political and social climate related to tobacco over the past 20 years.
Although secondhand smoke has detrimental effects on fetal health,2,3 we did not find any evidence for a direct association of smoke-free legislation with birth outcomes. Before-after studies using data from the past decade have shown that the introduction of smoke-free legislation at the country-level13- 15 or city-level16,17 was associated with reductions in the numbers of preterm,13,14,16- 18 SGA,13,15 and low-birth-weight infants.13 Although Cox and colleagues17 found that the introduction of smoke-free legislation in Flanders, Belgium, reduced the risk of preterm birth, there was no effect of the bans on mean birth weight, low birth weight, or SGA. Other studies14,16 have also not found effects on low birth weight. Without consistent results across birth outcomes, because their etiologies are often related, evidence for a causal interpretation of the association of smoke-free legislation with birth outcomes is not as strong. In the United States, municipalities often enact local smoke-free policies before a state passes smoke-free legislation, or local policies could be more comprehensive than those enacted at the state level.26 We were only able to examine smoke-free legislation at the state level; future research should test the potential differential effects of local and state policies.
While we found that increases in cigarette taxes were associated with reduced numbers of adverse birth outcomes overall, they also were associated with an increased proportion of babies born LGA. Our results suggest that, among white and black mothers with a low level of education, tax increases raised the mean distribution of birth weight nearly 5 g, resulting in about a third of a percentage point decrease in the number of SGA infants and a slightly smaller increase in the number of LGA infants (Table 2). Although higher birth weight among term infants is associated with better birth outcomes overall31 and a reduction in later risk for cardiovascular disease and hypertension,32 babies born LGA are at an increased risk for birth complications33 and later obesity.34 None of the studies reviewed included LGA as a birth outcome.11- 17 Current clinical guidelines are to screen all pregnant women for tobacco use and provide counseling and tobacco dependence interventions to increase quit rates.35,36 There are social and economic tradeoffs between choosing a population-based approach, such as state policies affecting all mothers, and choosing a targeted approach, such as clinic-based interventions, to improve maternal and infant health.37
There are limitations to our study, particularly the fact that prenatal smoking was self-reported. Pregnant women, especially those who are more educated and older, underreport their smoking status on the birth certificate.38 Because the women in our study had the lowest levels of prenatal smoking and were the least responsive to changes in cigarette taxes, our results for women with a low level of education would be less biased. We also could not account for other sources of secondhand smoke exposure, such as a partner smoking, because the birth certificate only has information on prenatal smoking. Although smoking is associated with spontaneous preterm labor,39 there is no distinction between spontaneous and indicated preterm deliveries on the birth certificate. We did not include information on maternal health conditions (such as preeclampsia) associated with birth outcomes in our models because they are likely on the causal pathway rather than confounders. Nevertheless, other studies13,16 that adjusted for these factors found no differences in their results.
We have shown that increases in cigarette taxes are associated with improved birth outcomes among mothers and infants who are at highest risk for adverse birth outcomes via reductions in prenatal smoking rates. Considering that, in the United States, individual states increase cigarette taxes for reasons other than to improve birth outcomes, these findings are welcome by-products of policies primarily aimed at improving the health of the smokers themselves.
Accepted for Publication: August 27, 2014.
Corresponding Author: Summer Sherburne Hawkins, PhD, Graduate School of Social Work, Boston College, 140 Commonwealth Ave, Chestnut Hill, MA 02467 (firstname.lastname@example.org).
Published Online: November 3, 2014. doi:10.1001/jamapediatrics.2014.2365.
Author Contributions: Drs Hawkins and Baum had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Hawkins.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Hawkins.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Baum.
Obtained funding: Hawkins.
Administrative, technical, or material support: Hawkins, Gillman.
Study supervision: Hawkins, Oken, Gillman.
Conflict of Interest Disclosures: None reported.
Funding/Support: The research reported in this article was supported by the Eunice Kennedy Shriver National Institute of Child Health and Human Development of the National Institutes of Health under award R00HD068506 to Dr Hawkins. Dr Oken was supported by grants P30 DK092924 and K24 HD069408.
Role of the Funder/Sponsor: The National Institutes of Health had no role in the design and conduct of the study; collection, management, analysis, or interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Disclaimer: The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Additional Contributions: We thank Rachel Boyce, BS, data coordinator at the Harvard Center for Population and Development Studies, for data management, and the National Center for Health Statistics for the natality files. Ms Boyce was funded by award R00HD068506.