Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2021 Dec 1.
Published in final edited form as: Drug Alcohol Depend. 2020 Sep 13;217:108273. doi: 10.1016/j.drugalcdep.2020.108273

Marital status, partner acknowledgment of paternity, and neighborhood influences on smoking during first pregnancy: findings across race/ethnicity in linked administrative and census data

Alexandra N Houston-Ludlam a,b,c,*, Mary Waldron a,d, Min Lian e, Alison G Cahill f, Vivia V McCutcheon a, Pamela AF Madden a, Kathleen K Bucholz a, Andrew C Heath a
PMCID: PMC8075321  NIHMSID: NIHMS1662589  PMID: 32971390

Abstract

Background:

Improving prediction of cigarette smoking during pregnancy (SDP), including differences by race/ethnicity and geography, is necessary for interventions to achieve greater and more equitable SDP reductions.

Methods:

Using individual-level data on singleton first births, 2010–2017 (N = 182,894), in a US state with high SDP rates, we predicted SDP risk as a function of reproductive partner relationship (marital status, paternity acknowledgement), maternal and residential census tract sociodemographics, and census tract five-year SDP rate.

Results:

SDP prevalence was 12.7% (white non-Hispanics, WNH), 6.8% (Black/African Americans, AA), 19.5% (Native American, NA), 4.7% (Hispanic, H), and 2.8% (Asian, AS). In WNH and AA, with similar trends in other groups, after adjustment for non-linear effects of maternal age and education and for census tract risk-factors, there was a consistent risk-ordering of SDP rates by reproductive partner relationship: married/with paternity acknowledged < unmarried/acknowledged < unmarried/unacknowledged < married/unacknowledged. Associations with census tract SDP rate, adjusted for maternal and census tract sociodemographics, were stronger for AA and H (OR 2.65–2.67) than for NA (OR = 1.91), WNH (OR = 1.75), or AS (NS). AA SDP was increased in tracts having a higher proportion of WNH residents and was reduced in comparison with WNH at every combination of age, education and partner relationship.

Conclusions:

Inattention to differences by race/ethnicity may obscure SDP risk factors. Despite marked race/ethnic differences in unmarried-partner cohabitation rates, failure to acknowledge paternity emerged as an important and consistent risk-predictor. Census-tract five-year SDP rates have heterogeneous origins, but the association of AA SDP risk with increased racial heterogeneity suggests an important influence of neighbor risk behaviors.

Keywords: American Community Survey, epidemiology methods, neighborhood effects, social determinants of health, tobacco, paternity acknowledgement, marital status

1. Introduction

In 2016, the estimated prevalence of any cigarette use in pregnancy in the United States was 7.2% (Drake et al., 2018), much higher than the Healthy People 2020 goal of reduction to 1.4% (U.S. Department of Health and Human Services). Although striking, national estimates have the potential to obscure the wide geographic variation in cigarette smoking during pregnancy (SDP) rate at the state level. In 2016, California was lowest at 1.6%, Missouri was 5th highest at 15.2%, and West Virginia was highest at 25.1% (Drake et al., 2018), all of which are likely underestimates due to underreporting. SDP has well-documented negative health effects for those who smoke and their offspring (Cnattingius, 2004; Crume, 2019; Dietz et al., 2010; Flenady et al., 2011; McEvoy and Spindel, 2017; Tager et al., 1983). Additionally, since most people who smoke during first pregnancy are young (Riaz et al., 2018), successful and permanent cessation by first pregnancy would achieve an important reduction in tobacco-related morbidity and mortality (Jha et al., 2013).

Despite a broad general literature on predictors of SDP (Boucher and Konkle, 2016; Riaz et al., 2018), the literature is underdeveloped in several key domains. There is minimal work examining cultural differences and consistency of risk factors across different race/ethnic groups (Dukes et al., 2017). While recent work, especially in the context of cessation interventions, has examined the role of the reproductive dyad, this is often limited to traditional classifications of ‘married’ or ‘unmarried and cohabiting’ versus non-cohabiting (Riaz et al., 2018), without considering absentee (‘disengaged’) versus engaged non-cohabiting reproductive partners (cf. Hummer and Hamilton, 2010). While there is increasing recognition of the importance of social and systemic determinants of health, the varying influences of neighborhood environment by race/ethnicity are often ignored. Lastly, most existing research is not equipped to examine the potential confounding of individual and neighborhood sociodemographic influences. By using either relatively small research samples or summary data at the population-level, the existing literature’s ability to achieve individual-level prediction that includes geographic influences is limited. Understanding individual risk prediction, including differences by race/ethnicity, is necessary for policy and therapeutic interventions to achieve greater and more equitable reductions in SDP for all.

To achieve a better understanding of individual risk prediction, we cross-link two forms of administrative data, individual-level state birth record data and associated census-tract American Community Survey data (GeoLytics, version 7.9). This ensures sufficiently large sample sizes to highlight important differences between Black/African Americans and white non-Hispanics and to show trends for other understudied US minority groups, including Native Americans. This report examines whether social influences associated with both residential neighborhood SDP rate and reproductive partner relationship (marital status and reproductive partner acknowledgement of paternity), after adjusting for other aspects of maternal and residential neighborhood sociodemographics, play an important role in assessing individual-level risk for smoking during first pregnancy.

2. Materials and Methods

2.1. Analysis Sample

In this population-cohort study, all non-adopted singleton first births in Missouri, for birth years 2010–17, were extracted from birth records, after agreement and data release from the Missouri Department of Health and Senior Services. Records about biological parents of adoptees could not be accessed due to Missouri state law. Maternal residential address was standardized (SmartyList, version 8.2.6, SmartyStreets) and geocoded to census tract using 2010 Census tract boundaries (GeoLytics, version 7.9). The success rate for geocoding ranged from 87.1% for white non-Hispanics (WNH), to 95.2% for Black/African Americans (AA), reflecting AA urban concentration in the state and lower success rate in geocoding rural route and similar rural addresses. The total sample, after exclusions for missing data (eTable 1), included 139,732 white non-Hispanics, 27,455 Black/African Americans, 7,973 Hispanics (H), 5,027 Asians (AS) and 2,707 Native Americans (NA). There were no exclusions for pregnancy outcome. This study received ethical approval from the Washington University Family and Population Research Center Review Board, Washington University Institutional Review Board, and Missouri Department of Health and Senior Services Ethical Review Board.

2.2. Measures

See eSupplement for details of variable coding.

2.2.1. Maternal variables

Missouri adopted the standard long-form birth certificate 2003 revision, published by the Centers for Disease Control and Prevention (CDC), in 2010. Variables extracted from birth record data included child birth year, maternal age, years of education completed, marital status, presence/absence of name of reproductive partner, number of living children, number of deceased children, race/ethnicity, and cigarette quantity smoked during the three months prior to pregnancy and during each trimester of pregnancy. Per CDC protocol, birth record report of smoking during pregnancy is obtained from medical records or physician report first, and if these sources are unavailable, then it is based on maternal self-report (National Center for Health Statistics, 1987). The presence/absence of the reproductive partner name indicates whether the partner signed an ‘affidavit of paternity’ either at the time of childbirth or, less commonly, subsequently. Births were counted as first births if maternal record indicated zero living and zero deceased children at time of index childbirth. Smoking was recoded as a binary variable indicating whether the gestational parent smoked in the second and/or third trimesters.

2.2.2. Census tract variables

The American Community Survey is an annual survey, conducted by the US Census Bureau, of 3.5 million households, assessing social, economic, housing, and demographic domains, with census tract-level data aggregated over five years to protect confidentiality (United States Census Bureau, 2017). American Community Survey 2012–16 variables were selected covering domains of census tract-level education, family structure, family income and wealth, poverty and socioeconomic distress, percent residents Black/African American, and population density as a measure of urban/rural status. Each variable was ranked, and ranks were recoded to create a seven-point ordinal scale with distribution 10%/10%/20%/20%/20%/10%/10% to satisfy state requirements for confidentiality protection. These selected variables were reduced to two correlated factors, with interfactor correlation 0.31, by unweighted principal factor analysis (Table 1): a tract socioeconomic advantage-disadvantage factor (F1) and a tract rural white – urban mixed race factor (F2). Factor scores were then ranked and recoded as quintiles to preserve confidentiality.

Table 1.

Factor structure of American Community Survey 2012–2016 Missouri census tract data.

Factor 1 Factor 2
Educational level adults 25+ −92 43
Median individual income (in 2016 $) −93 15
Median household income (in 2016 $) −91 −4
% households with investment income −70 −25
% children in poverty 80 6
% households with no vehicle 56 38
% women 18–44 without health insurance 76 −17
% in owner-occupied housing −45 −56
Labor force unemployment rate 64 6
% households receiving public assistance 90 6
% housing units vacant 72 −22
% children living with married parents −60 −46
% ever married now separated/divorced 62 36
% population African American 8 75
% population Hispanic −9 32
% agricultural workers 29 −72
% children in single parent families 51 50
% children living with unmarried parents 42 −8
% householders living with grandchildren 44 −15
Population density −28 87
Interfactor correlation 0.31
Open in a new tab

Factor loadings (x100) under an oblique PROMAX rotation.

Birth records for all non-adopted births in state, 2005–2016, not limited to singleton nor to first births, were used to calculate census tract five-year trailing aggregate rates of SDP (ctSDP). This variable was recoded as missing for any tract with fewer than 25 births over a five-year period to protect confidentiality. Births in 2010 used census tract SDP data for 2005–9; births in 2011 used 2006–10 data, and so on. Since the analysis sample only included first births, and the census tract-level SDP rates are trailing, the index pregnancy for analysis is never included in its respective trailing window. Tracts were ranked for five-year SDP rates, separately for each birth year, and ranks recoded to quintiles.

2.2.3. American Community Survey Microdata

In a separate follow-up analysis, we examined data drawn from the American Community Survey (ACS) 2013–2017 microdata (Ruggles et al., 2020). ACS variables FERTYR (whether a female participant had given birth within the last 12 months), MARST (six-level variable of marital status: Married, spouse present; Married, spouse absent; Separated; Divorced; Widowed; Never married/single), AGE, SEX, ELDCH (age of eldest child residing with participant, including step-, adopted, and biological children), and RACE, were downloaded. We used these variables to identify gestational parents aged 22–23, by race, who have given birth in previous 12 months, whose child resides with mother, by marital status, and whether they were cohabitating with their reproductive partner one year following childbirth.

2.3. Statistical Analysis

Analyses were conducted separately for each race/ethnic group because of the strong racial separation in the state, with ~50% of AA births but < 2% of WNH births occurring in census tracts falling in the top decile for percent residents AA. A multi-level logistic regression model, taking into account clustering of births within a given census tract, was fit to predict maternal SDP as a joint function of maternal birth record variables, including a quadratic term for age, census tract ACS factors, and five-year census tract SDP rate, separately for each race/ethnic group. To provide robust evidence for census tract-level effects, an additional logistic regression model was estimated using only maternal birth record variables, with comprehensive testing for inter-active effects (Imbens and Rubin, 2015) to generate a ‘socio-demographic’ risk score. See eSupplement for more details on risk score generation and eTable 2 for model summary statistics. The risk score was recoded to quintiles, and then used as a control variable in a multi-level model predicting SDP as a function of census tract variables. Given smaller numbers of births for all groups except WNH and AA, results for some analyses for other race/ethnic groups are presented only in supplementary tables. To examine assumptions regarding presence or lack of paternity acknowledgement, we compared current marital status and cohabitation with romantic partner patterns in ACS 2013–2017 microdata described above (2.2.3). All analyses were run in SAS 9.2 and figures were generated in Stata 15.1.

3. Results

3.1. Maternal predictors

Prevalence of smoking during first pregnancy was 12.7% (WNH), 6.8% (AA), 19.5% (NA), 4.7% (H), and 2.8% (AS). Comparing SDP to non-SDP births, we observe, for SDP births (eTable 3): younger age (except for AA); less frequently married and more frequently unmarried, and especially unmarried with no paternity acknowledgement by reproductive partner; and lower educational level. Residential census tract characteristics showed increasing individual SDP rates with greater socioeconomic disadvantage, greater rural/white population, and lesser proportion of AA residents (eTable 4). SDP rate in Black/African Americans was inversely related to the proportion of AA residents in the tract: comparing across quintiles for increasing proportion of Black/African American residents, SDP rate was 15%, 11.1%, 10.7%, 8.0%, and 6.1%, respectively. Fig. 1 shows, separately for white non-Hispanics and Black/African Americans, model-based predicted probabilities of maternal smoking during pregnancy, as a joint function of maternal age at childbirth, educational level, and relationship with reproductive partner (marital status and reproductive partner paternity acknowledgement), estimated under the full sociodemographic risk score model, including interaction terms. See eFig. 1 for other race/ethnic groups. Key findings are (i) expected risk differences by educational level; (ii) generally quadratic form of the relationship with age within educational levels; (iii) consistent risk differences at every age and educational level, as a function of marital status/paternity acknowledgement; and (iv) consistently lower SDP rate in AAs, particularly at lower educational levels. Table 2 shows, separately by maternal race/ethnicity, Odds Ratios from a multi-level logistic regression for SDP associations with marital status/paternity acknowledgement under a model including the quadratic effect of age and main effects of other maternal characteristics, both unadjusted and adjusted for neighborhood sociodemographics and SDP rate. Full model results unadjusted (eTable 5) and adjusted (eTable 6) for census tract ACS factor quintiles and five-year SDP rate quintiles are shown separately. In models unadjusted for census tract characteristics, there is a consistent ranking of SDP rate, in every race/ethnicity, with married, paternity acknowledged < unmarried, acknowledged < unmarried, unacknowledged < married, unacknowledged, the latter hypothesized to be cases of marital breakdown during pregnancy. Odds ratios are reduced only minimally, if at all, after adjustment for census tract characteristics. There are substantial differences by race/ethnicity in proportions of people by reproductive partner relationship type, and substantial differences in median age at first childbirth and percent with any college education for marrieds versus unmarrieds. However, median ages for unmarried/unacknowledged (WNH 21; AA 20) and unmarried/acknowledged (WNH 21, AA 20) are very similar, with great consistency for all groups except Asians (eTable 7). Differences in percent reporting any college education are also modest (unmarried/unacknowledged: WNH 36%, AA 35%; unmarried/acknowledged: WNH 44%, AA 47%). To examine whether our assumption about presence or absence of paternity acknowledgement potentially reflected partner support/abandonment effects, rather than cohabitation effects, we consulted American Community Survey 2013–2017 microdata. We identified gestational parents living with eldest child less than age 12 months, currently aged 22–23, to examine their current marital status and cohabitation with romantic partner. For WNH gestational parents, (N = 6,408) approximate cohabitation rates with spouse/romantic partner in the year following childbirth of oldest child were: married, 93%; separated, 16%; divorced, 39%; never married/single, 40%. For AA gestational parents (N = 1,434), the approximate cohabitation rates were: married, 76%; separated/divorced, 10%; never married/single, 14%. The much lower rate of cohabitation by AA compared to WNH single/never married gestational parents, but very consistent risk differences for pregnancies of unmarried with named versus unnamed partner, cautions against assuming equivalence between presence/absence of paternity acknowledgement and cohabitation/non-cohabitation during the pregnancy.

Fig. 1.

Fig. 1.

Open in a new tab

Predicted probability of smoking during pregnancy as a function of maternal age and years education completed, stratified by maternal race/ethnicity and marital status. Panels a-c: For white non-Hispanics, being married/having paternity acknowledgement is especially protective for smoking during pregnancy risk. Similar patterns for SDP risk are seen between both unmarried groups (panels a,b), with the protective effect of increasing years of education particularly apparent. Panels d-f: There is an overall reduction in predicted probability of SDP for Black/African Americans compared to white non-Hispanics. However, similar patterns emerge—increasing years education completed and being married/having paternity acknowledgement both reduce SDP risk. Means for predicted probability of SDP for each maternal age × marital status/paternity acknowledgement × education level are plotted with shaded areas representing the 95% confidence interval. Any cell size < 9 was censored from the plot.

Table 2.

Maternal sociodemographics at childbirth recorded on birth record predict smoking during pregnancy, by maternal race/ethnicity, births 2010–2017.

Maternal Race/Ethnicity white, non-Hispanic Black / African American Native American Hispanic Asian
Odds Ratio [95% Confidence Interval]
Maternal Marital Status Unadjusted
 Married, Paternity Acknowledgeda - 0.58 [0.47, 0.73] - - -
 Unmarried, Paternity Acknowledgedb 3.17 [3.03, 3.32] - 1.98 [1.50, 2.61] 2.25 [1.63, 3.09] 4.94 [3.07, 7.94]
 Unmarried, Paternity Unacknowledged 4.64 [4.39, 4.90] 1.52 [1.37, 1.68] 2.47 [1.81, 3.36] 3.16 [2.21, 4.51] 6.84 [3.91, 11.99]
 Married, Paternity Unacknowledged 5.46 [4.61, 6.46] 3.75 [2.28, 6.16] 7.83 [2.99, 20.52] 1.83 [0.42, 8.00] 23.11 [4.01, 133.23]
Maternal Marital Status Adjusted
 Married, Paternity Acknowledgeda - 0.54 [0.43, 0.68] - - -
 Unmarried, Paternity Acknowledgedb 3.15 [3.01, 3.30] - 2.00 [1.51, 2.66] 2.42 [1.75, 3.36] 5.21 [3.19, 8.50]
 Unmarried, Paternity Unacknowledged 4.62 [4.37, 4.88] 1.58 [1.42, 1.75] 2.80 [2.03, 3.84] 3.58 [2.49, 5.15] 7.84 [4.39, 14.01]
 Married, Paternity Unacknowledged 5.16 [4.35, 6.11] 2.20 [1.35, 3.59] 10.38 [4.09, 26.36] 1.80 [0.40, 8.20] 7.68 [1.49, 39.52]
Open in a new tab

Odds Ratios and 95% Confidence Intervals from multi-level logistic regressions predicting maternal smoking during pregnancy from maternal sociodemographics on the birth record, unadjusted and adjusted for ACS-derived census tract sociodemographics and census tract-level five-year smoking during pregnancy rate for birth years 2010–2017. Full results for both models presented in eTable 5 (unadjusted) and eTable 6 (adjusted).

a

Referent for white non-Hispanics, Native Americans, Hispanics, Asians.

b

Referent for Black/African Americans.

3.2. Census Tract Predictors

Fig. 2 shows the Odds Ratios for the effect of quintiles of ctSDP on individual SDP risk, adjusted for both maternal sociodemographic risk score and ACS factor scores, for WNH and AA. See eFig. 2 for results for other race/ethnic groups and eTable 8 for detailed results by race/ethnicity. In both WNH and AA, there is a progressive increase in risk with increasing five-year rate of ctSDP, with this effect more pronounced in AA (e.g. for top quintile compared to lowest quintile, WNH: OR = 1.75, 95% CI [1.57–1.94]; AA: OR = 2.67, 95% CI [2.03–3.52]). Risk differences as a function of census tract disadvantage (F1) are modest in both WNH and AA (highest quintile ORs: WNH 1.49 [1.34–1.66]; AA 1.23 [0.90–1.89]), with no significant association in AAs. Risk differences as a function of rural-white/urban-mixed race (F2) are minimal in WNH, but more pronounced in AA, with risk being highest in the most rural-white quintile and lowest in the most urban-mixed race quintile. Table 3 compares selected sociodemographic characteristics of the highest quintile for ctSDP versus pooled remaining quintiles. A more comprehensive selection of sociodemographic variables for comparison is shown in eTable 9. For AA, the high-risk quintile is characterized by a reduced proportion of AA residents (33% vs 57%). Also for AA, indicators of poverty and economic resources (e.g. median household income, unemployment rate, %homes receiving public assistance, %children living in poverty) were similar between the lowest 80% of tracts for ctSDP (low-risk) and highest 20% of census tracts for ctSDP (high-risk) tracts. However, for other maternal race/ethnicities, there were marked differences between low and high-risk tracts, with lower poverty and more economic resources in low-risk tracts compared to higher poverty and fewer economic resources in high-risk tracts.

Fig. 2.

Fig. 2.

Open in a new tab

Results from a multi-level logistic regression predicting smoking during pregnancy by five-year census tract smoking during pregnancy rates, adjusted for maternal sociodemographics and ACS-derived census tract sociodemographics for address at time of childbirth. When compared to the census tracts with lowest rates of smoking during pregnancy (0–20ile), increasing rate of census tract smoking during pregnancy increases maternal smoking during pregnancy, beyond individual and neighborhood-level influences. For white non-Hispanics, there is a modest but consistent elevation in smoking during pregnancy with increasing census tract smoking during pregnancy rate. This effect is stronger for Black/African Americans. Error bars reflect 95% confidence intervals for Odds Ratio estimates.

Table 3.

Census tract characteristics from American Community Survey data, 2012–2016, for census tracts with lowest four quintiles (0–80ile) vs highest quintile (81–100ile) five-year census tract-level smoking during pregnancy rate, by maternal race/ethnicity.

0–80ile census tract SDP rate 81–100ile census tract SDP rate
white non-Hispanic Black/ African American Native American Hispanic Asian white non-Hispanic Black/ African American Native American Hispanic Asian
N = 107374 N = 25055 N = 2042 N = 6661 N = 4657 N = 16121 N = 1281 N = 309 N = 625 N = 200
Population density (persons/km2) 846.50 1559.90 857.93 1133.95 1272.30 332.45 893.11 419.89 583.54 603.40
% Black / African American residents 8.37 56.55 17.26 16.88 13.68 5.05 32.70 6.08 8.60 8.63
% residents with Bachelor’s degree or higher 32.76 22.50 27.52 26.46 40.92 15.10 14.94 15.65 14.91 16.71
% residents with any college education 63.34 55.83 59.37 56.02 69.26 45.10 43.99 46.02 44.98 47.43
Median household income ($) 59,711 39,981 51,801 48,101 60,044 40,623 32,328 39,131 37,899 38,266
Income to Poverty ratioa 1.85 1.65 1.79 1.74 1.83 1.73 1.59 1.72 1.70 1.69
Unemployment rate (%) 5.38 11.82 6.76 6.90 5.57 7.73 11.09 7.72 8.47 7.50
% homes receiving public assistance 9.90 24.63 13.46 16.10 10.03 18.18 28.59 19.67 20.21 19.72
% children living in poverty 15.07 34.83 20.59 25.62 16.71 26.53 40.84 28.74 30.45 30.56
% women 18–44 without health insurance 14.39 22.59 18.30 22.73 14.04 22.36 24.09 23.49 24.83 22.25
Value of housing units ($) 162,386 106,184 136,302 128,799 183,725 104,808 80,510 101,552 96,601 102,601
% housing units vacant 10.64 22.08 13.27 16.15 10.46 19.21 23.80 20.30 22.25 19.52
% children living in married, two-parent households 72.91 42.90 65.69 63.35 70.98 64.89 42.13 63.78 60.41 60.18
% ever married, now separated/divorced 19.80 34.78 23.29 22.14 19.54 25.45 34.92 27.17 27.91 27.13
Open in a new tab

All statistics reported represent means.

a

The total family income divided by the poverty threshold is called the Ratio of Income to Poverty (US Census Bureau, 2019). https://www.census.gov/topics/income-poverty/poverty/guidance/poverty-measures.html.

4. Discussion

4.1. Summary

This report uses a US Midwest state, with consistently high SDP rates, lowest cigarette taxes, few enacted tobacco control policies, yet permissive policies for researcher data access, to improve characterization of two social influences on smoking during first pregnancy: residential census tract-level SDP prevalence and relationship with reproductive partner (marital status and whether paternity was acknowledged). Cross-linking birth record maternal sociodemographic data with census tract American Community Survey sociodemographic data and aggregated five-year census tract rates of smoking during pregnancy provided convergent support for an important role of social influences on smoking during pregnancy. These methods allowed improved statistical control for potentially confounding neighborhood characteristics when predicting SDP risk from individual maternal variables.

4.2. Census Tract Predictors

At the census tract level, five-year census tract SDP rates remained significantly predictive of SDP risk, even after adjustment for effects of sociodemographic characteristics of individuals at the time of childbirth and for the effects of census tract sociodemographic disadvantage and urban-rural characteristics. Census tract rate differences, after adjustment for maternal sociodemographics, likely stem from many origins, including a broad spectrum of differences in local taxes, local policy, tobacco retail outlet proximity, and availability and quality of health care (Center for Public Health Systems Science, 2016; Centers for Disease Control and Prevention, 2015; Hall et al., 2019). The association of SDP with increasing census tract rates of SDP was stronger in Black/African Americans than in white non-Hispanics, consistent with important influences of risk-behaviors of other census tract residents, though other interpretations cannot be excluded. Notably, the characteristics of the highest risk census tracts for census tract smoking during pregnancy rate were starkly different between white non-Hispanics versus Black/African Americans. For high-risk tracts, white non-Hispanics showed the expected pattern of increased socioeconomic disadvantage, while the high-risk tracts in which Black/African Americans resided were primarily characterized by an increased percent of white non-Hispanic residents and unchanged levels of socioeconomic disadvantage. This is consistent with an important social risk influence, where Black/African Americans who live in close proximity to white non-Hispanics, who have higher rates of smoking during pregnancy, are themselves more likely to smoke during pregnancy than Black/African Americans living in neighborhoods that are more segregated. Alternatively, Black/African Americans living in predominantly white neighborhoods may experience greater levels of discrimination and racism, making it harder to achieve smoking cessation by early pregnancy (Berger and Sarnyai, 2015; Gravlee, 2009; Sims et al., 2016; Tomfohr et al., 2016).

4.3. Individual Predictors

For individual risk characteristics, we found that, in addition to the anticipated risk differences in unmarried versus married people, lack of reproductive partner legal acknowledgement of paternity was associated with clear differences in likelihood of individual SDP across the distribution of maternal age and educational level. After adjustment for other maternal and for census tract risk factors, there was a clear ordering of risk, consistent across all race/ethnic groups: risk was lowest for married with paternity acknowledgement, somewhat greater for unmarried with paternity acknowledgement, substantially elevated for unmarried with no paternity acknowledgment, and highest for married with no paternity acknowledgement. Acknowledgement of paternity, occurring at or after childbirth, is likely capturing aspects of the engagement versus disengagement of the reproductive partner earlier in the pregnancy, given how this effect predicts differences in SDP risk, and is not reducible to simple presence/absence of cohabitation. Several considerations support this interpretation. The risk differences associated with presence/absence of paternity acknowledgment were observed consistently across race/ethnic groups. This persists despite marked differences in partner cohabitation rates, as inferred from American Community Survey microdata on gestational parents with oldest child less than one year old: 14% versus 40% for Black/African Americans versus white non-Hispanics living with oldest child of age less than 12 months are cohabiting with a partner. Additionally, if cohabitation was driving this effect, one might have anticipated differences in household income for cohabitating versus non-cohabitating pregnant people to explain our findings. However, our findings were robust to adjustment for census tract sociodemographics. Also consistent with our reproductive partner engagement/disengagement hypothesis, SDP risk is especially elevated for married persons with paternity not acknowledged: this effect is likely explained by breakdown of the relationship between reproductive partners, since, by law, being married at any time between conception and childbirth is recorded as “married” on the birth record in Missouri (Missouri Legislature, 2005). These findings not only support but also extend a broader research literature focused on pregnancy risk-behaviors (Kiernan and Pickett, 2006; Pickett et al., 2009; Riaz et al., 2018), with finer control for potential maternal and neighborhood sociodemographic confounders. Given high rates of co-use of cigarettes and alcohol and other substances in pregnancy, it seems plausible that similar risk-patterns would be observed for prenatal alcohol, marijuana and other illicit drug use during pregnancy (Desrosiers et al., 2016; Dukes et al., 2017; Frazer et al., 2019; Oga et al., 2019).

Conditional upon relationship with reproductive partner, white non-Hispanics and Black/African Americans were very similar in terms of age and educational level at first childbirth. However, except at the highest levels of maternal education, where overall rates of smoking during pregnancy were low, rates of SDP were also consistently lower across age, education, and relationship with reproductive partner in Black/African Americans compared to white non-Hispanics. Thus, in the context of known sociodemographic risk factors for smoking during pregnancy, these observed lower SDP rates among Black/African Americans suggest important resiliency factors that warrant further study.

4.4. Treatment and Policy Implications

Our results have clear applications for both treatment and policy interventions. Environmental and individual, including biological, influences are not independent (Bagby et al., 2019), and must be considered both jointly and uniquely, including in the context of smoking during pregnancy. Given the consistent effect of marital status/paternity acknowledgement on SDP risk, addressing both parties in the reproductive dyad is vital for improving smoking cessation and second hand smoke exposure for pregnant people and their offspring (Fergie et al., 2019; Meghea et al., 2018; Patrick, 2019; Román-Gálvez et al., 2018; Shawe et al., 2019). This distinction between reproductive partner engagement versus cohabitation may be of translational importance, since it is likely easier to ‘nudge’ (Thaler and Sunstein, 2008) unmarried fathers into greater engagement than into cohabitation. This may also aid in minimizing alcohol, marijuana, and other illicit drug use (Desrosiers et al., 2016; Frazer et al., 2019; Oga et al., 2019). Historical census tract five-year SDP rate is a variable easily constructed and shared by every state, as it is recorded on the birth record for all states. Sharing smoking during pregnancy rates at the census tract level protects confidentiality and allows consideration of the clear geographic variation in smoking during pregnancy. Once geocoded, these data are easily combined with American Community Survey data to improve prediction of maternal SDP risk, facilitating early detection of individuals at highest risk. Implementing health equity-focused policies could be an effective strategy to increase smoking cessation prior to pregnancy. Some examples include limiting tobacco retailer density, proximity to one another, and/or regulating tobacco product promotions. These policies have the added benefit of broadly promoting health equity and reducing tobacco-related health consequences (Centers for Disease Control and Prevention, 2015; Galiatsatos et al., 2020; Ribisl et al., 2017; Rodriguez et al., 2013). Consideration of neighborhood context is important when considering which interventions to implement and how to implement them—from individual to community to policy (D’Agostino et al., 2018; Gootjes et al., 2019; Hall et al., 2019; Williams and Jackson, 2005).

4.5. Application of Methods beyond Smoking During Pregnancy

In this population-based cohort study, we provide evidence for the important role of social influence on cigarette use during first pregnancy, with comprehensive control for confounding variables not previously achieved in the literature. The novel methods we present for combining administrative and publicly available US Census/American Community Survey data have important applications in addiction and health research beyond this application to smoking during pregnancy. Linking geocoded research and administrative datasets to US Census data allows use of universally assessed environmental measures available, which are available for all US-based samples. Most research programs can achieve this linkage at modest additional cost. This linkage of data allows important examination of the geographic and social contexts of many questions in addiction and health research.

4.6. Limitations

Administrative data analyses, as presented here, leave unanswered questions only addressable by appropriately targeted new data collection. Future research must parse out, for example, if risk differences among unmarried pregnant people as a function of presence/absence of paternity acknowledgement reflect partner conflict/abandonment effects, or partner support effects (Hummer and Hamilton, 2010), or other mechanisms. However, administrative data analyses do have the advantage of identifying with considerable precision, and therefore with reduced likelihood of false positive reporting (Ioannidis, 2005a, b; Ioannidis et al., 2011), phenomena requiring targeted research follow-up. Despite the potential limitations of birth record data, including likely underreporting of smoking during pregnancy and possible mis-specification in individual cases of other variables, it is unlikely that results presented herein can be dismissed as artefacts of reporting bias. The present study does not examine co-use of tobacco and other nicotine products or other substances during pregnancy, due to either unreliable reporting of data (alcohol use during pregnancy) or no report (other substances, electronic nicotine delivery systems). Future research should examine the role of social and other influences on co-substance use during pregnancy to inform treatment and policy interventions.

4.7. Conclusions

The social influences of marital status/paternity acknowledgement and census tract-level smoking during pregnancy rate prove to be consistent predictors of individual-risk of smoking during first pregnancy, even after accounting robustly for individual and neighborhood-level characteristics. The magnitude of these relationships varies by maternal race/ethnicity, notably that census tract SDP rate is an especially important risk factor for Black/African Americans, despite overall lower SDP rates. For neighborhood characteristics, we observed an expected transition from high economic resources in low SDP prevalence tracts to low economic resources in high SDP prevalence tracts for all maternal races, except for Black/African Americans: the tracts in which Black/African Americans reside have stable rates of lower economic resources across the distribution of census tract SDP rate. Lastly, we provide proof-of-concept for using publicly available US Census/American Community Survey in conjunction with administrative data. These methods can be extended to any US based administrative or primary research sample, with modest additional cost, to examine important social and neighborhood context of other questions in addiction and health research.

Supplementary Material

Supplementary Methods, Tables, Figures

Acknowledgements

Preliminary analyses for this report were presented as a poster at the 81st annual meeting of the College on Problems of Drug Dependence, San Antonio, Texas, June 15-19, 2019. Partial results for this report were presented as an ePoster at the 25th annual meeting of the Society for Research on Nicotine and Tobacco, New Orleans, Louisiana, February 20-23, 2020.

The authors thank Dejan Jovanovic and Rade Todorovic for the processing and maintenance of vital records and American Community Survey data used in this report, Stacey Marion for coordination of receipt and administration of vital records data, Denise Schmitz for providing quality control of vital records. We also thank the Missouri Department of Health and Senior Services staff for providing access to identified birth records.

The data used in this document was acquired from the Missouri Department of Health and Senior Services (DHSS). The contents of this document including data analysis, interpretation, or conclusions are solely the responsibility of the authors and do not represent the official views of DHSS.

Funding

This work was supported by the National Institutes of Health, grants TL1TR002344 (ANHL), F30DA047742 (ANHL), R24AA023487 (ACH, KKB, PAFM), R01DA044254 (ML, ACH), K07CA178331(ML), U01DA041120 (PAFM, ACH), and U10AA008401 (KKB).

Role of funding source

The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. The National Institutes of Health had no involvement in the study design; the collection, analysis, or interpretation of the data; writing of the report; or decision to submit the article for publication.

Abbreviations:

AA

Black/African American

ACS

American Community Survey

AS

Asian

CDC

Centers for Disease Control and Prevention

ctSDP

census tract-level cigarette smoking during pregnancy rate

F1

Factor 1 (from American Community Survey variables)

F2

Factor 2 (from American Community Survey variables)

H

Hispanic

NA

Native American

SDP

cigarette smoking during pregnancy

WNH

white non-Hispanic

Footnotes

Declaration of Competing Interest

The authors report no declarations of interest.

Appendix A. Supplementary data

Supplementary material related to this article can be found, in the online version, at doi:https://doi.org/10.1016/j.drugalcdep.2020.108273.

References

  1. Bagby SP, Martin D, Chung ST, Rajapakse N, 2019. From the Outside In: Biological Mechanisms Linking Social and Environmental Exposures to Chronic Disease and to Health Disparities. American Journal of Public Health 109 (S1), S56–S63. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Berger M, Sarnyai Z, 2015. More than skin deep”: stress neurobiology and mental health consequences of racial discrimination. Stress (Amsterdam, Netherlands) 18 (1), 1–10. [DOI] [PubMed] [Google Scholar]
  3. Boucher J, Konkle AT, 2016. Understanding Inequalities of Maternal Smoking-Bridging the Gap with Adapted Intervention Strategies. Int J Environ Res Public Health 13 (3). [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Center for Public Health Systems Science, 2016. Point-of-Sale Report to the Nation: Realizing the Power of States and Communities to Change the Tobacco Retail and Policy Landscape. Center for Public Health Systems Science at the Brown School at Washington University in St. Louis and the National Cancer Institute, State and Community Tobacco Control Research Initiative, St. Louis, MO. [Google Scholar]
  5. Centers for Disease Control and Prevention, 2015. In: Centers for Disease Control and Prevention, U.S. Department of Health and Human Services (Eds.), Best Practices User Guide: Health Equity in Tobacco Prevention and Control National Center for Chronic Disease Prevention and Health Promotion, Office on Smoking and Health, Atlanta: (Ed.). [Google Scholar]
  6. Cnattingius S, 2004. The epidemiology of smoking during pregnancy: Smoking prevalence, maternal characteristics, and pregnancy outcomes. Nicotine & Tobacco Research 6 (Suppl_2), S125–S140. [DOI] [PubMed] [Google Scholar]
  7. Crume T, 2019. Tobacco Use During Pregnancy. Clinical obstetrics and gynecology; 62 (1), 128–141. [DOI] [PubMed] [Google Scholar]
  8. D’Agostino EM, Patel HH, Ahmed Z, Hansen E, Sunil Mathew M, Nardi MI, Messiah SE, 2018. Impact of change in neighborhood racial/ethnic segregation on cardiovascular health in minority youth attending a park-based afterschool program. Social Science & Medicine 205, 116–129. [DOI] [PubMed] [Google Scholar]
  9. Desrosiers A, Thompson A, Divney A, Magriples U, Kershaw T, 2016. Romantic partner influences on prenatal and postnatal substance use in young couples. Journal of public health (Oxford, England) 38 (2), 300–307. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Dietz PM, England LJ, Shapiro-Mendoza CK, Tong VT, Farr SL, Callaghan WM, 2010. Infant morbidity and mortality attributable to prenatal smoking in the U.S. American journal of preventive medicine 39 (1), 45–52. [DOI] [PubMed] [Google Scholar]
  11. Drake P, Driscoll AK, Mathews TJ, 2018. Cigarette smoking during pregnancy: United States, 2016. National Center for Health Statistics, Hyattsville, MD. [PubMed] [Google Scholar]
  12. Dukes K, Tripp T, Willinger M, Odendaal H, Elliott AJ, Kinney HC, Robinson F, Petersen JM, Raffo C, Hereld D, Groenewald C, Angal J, Hankins G, Burd L, Fifer WP, Myers MM, Hoffman HJ, Sullivan L, 2017. Drinking and smoking patterns during pregnancy: Development of group-based trajectories in the Safe Passage Study. Alcohol (Fayetteville, N.Y.) 62, 49–60 [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Fergie L, Campbell KA, Coleman-Haynes T, Ussher M, Cooper S, Coleman T, 2019. Stop smoking practitioner consensus on barriers and facilitators to smoking cessation in pregnancy and how to address these: A modified Delphi survey. Addictive behaviors reports 9, 100164. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Flenady V, Koopmans L, Middleton P, Froen JF, Smith GC, Gibbons K, Coory M, Gordon A, Ellwood D, McIntyre HD, Fretts R, Ezzati M, 2011. Major risk factors for stillbirth in high-income countries: a systematic review and meta-analysis. Lancet (London, England) 377 (9774), 1331–1340. [DOI] [PubMed] [Google Scholar]
  15. Frazer Z, McConnell K, Jansson LM, 2019. Treatment for substance use disorders in pregnant women: Motivators and barriers. Drug Alcohol Depend 205, 107652. [DOI] [PubMed] [Google Scholar]
  16. Galiatsatos P, Brigham E, Krasnoff R, Rice J, Van Wyck L, Sherry M, Rand CS, Hansel NN, McCormack MC, 2020. Association between neighborhood socioeconomic status, tobacco store density and smoking status in pregnant women in an urban area. Prev Med 136, 106107. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Gootjes DV, van Dijk MR, Koster MP, Willemsen SP, Steegers EA, Steegers-Theunissen RP, 2019. Neighborhood Deprivation and the Effectiveness of Mobile Health Coaching to Improve Periconceptional Nutrition and Lifestyle in Women: Survey in a Large Urban Municipality in the Netherlands. JMIR mHealth and uHealth 7 (4), e11664. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Gravlee CC, 2009. How race becomes biology: Embodiment of social inequality. American Journal of Physical Anthropology 139 (1), 47–57. [DOI] [PubMed] [Google Scholar]
  19. Hall J, Cho HD, Guo Y, Maldonado-Molina MM, Thompson LA, Shenkman EA, Salloum RG, 2019. Association of Rates of Smoking During Pregnancy With Corporate Tobacco Sales Policies. JAMA Pediatrics 173 (3), 284–286. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Hummer RA, Hamilton ER, 2010. Race and ethnicity in fragile families. Future Child 20 (2), 113–131. [DOI] [PubMed] [Google Scholar]
  21. Imbens G, Rubin D, 2015. Causal Inference for Statistics, Social, and Biomedical Sciences: An Introduction. Cambridge University Press, New York. [Google Scholar]
  22. Ioannidis JP, 2005a. Contradicted and initially stronger effects in highly cited clinical research. Jama 294 (2), 218–228. [DOI] [PubMed] [Google Scholar]
  23. Ioannidis JP, 2005b. Why most published research findings are false. PLoS Med 2 (8), e124. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Ioannidis JP, Tarone R, McLaughlin JK, 2011. The false-positive to false-negative ratio in epidemiologic studies. Epidemiology (Cambridge, Mass) 22 (4), 450–456. [DOI] [PubMed] [Google Scholar]
  25. Jha P, Ramasundarahettige C, Landsman V, Rostron B, Thun M, Anderson RN, McAfee T, Peto R, 2013. 21st-Century Hazards of Smoking and Benefits of Cessation in the United States. New England Journal of Medicine 368 (4), 341–350. [DOI] [PubMed] [Google Scholar]
  26. Kiernan K, Pickett KE, 2006. Marital status disparities in maternal smoking during pregnancy, breastfeeding and maternal depression. Soc Sci Med 63 (2), 335–346. [DOI] [PubMed] [Google Scholar]
  27. McEvoy CT, Spindel ER, 2017. Pulmonary Effects of Maternal Smoking on the Fetus and Child: Effects on Lung Development, Respiratory Morbidities, and Life Long Lung Health. Paediatric respiratory reviews 21, 27–33. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Meghea CI, Brinzaniuc A, Sidor A, Chereches RM, Mihu D, Iuhas CI, Stamatian F, Caracostea G, Dascal MD, Foley K, Baban A, Voice TC, Blaga OM, 2018. A couples-focused intervention for smoking cessation during pregnancy: The study protocol of the QuitTogether pilot randomized controlled trial. Tob Prev Cessat 4,17. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Missouri Legislature, 2005. Missouri State Statute 193.085. Birth certificate — contents, filing, locale, duties of certain persons, time allowed, attestation. (Accessed 07/22/2020 2020). https://revisor.mo.gov/main/OneSection.aspx?section=193.085. [Google Scholar]
  30. National Center for Health Statistics, 1987. Hospitals’ and physicians’ handbook on birth registration and fetal death reporting. (Accessed 08/20/2020. https://www.cdc.gov/nchs/data/misc/hb_birth.pdf.
  31. Oga EA, Peters EN, Mark K, Trocin K, Coleman-Cowger VH, 2019. Prenatal Substance Use and Perceptions of Parent and Partner Use Using the 4P’s Plus Screener. Matern Child Health J 23 (2), 250–257. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Patrick SW, 2019. Improving Public Health Systems for Substance-Affected Pregnancies. American Journal of Public Health; 109 (1), 22–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Pickett KE, Wilkinson RG, Wakschlag LS, 2009. The psychosocial context of pregnancy smoking and quitting in the Millennium Cohort Study. J Epidemiol Community Health 63 (6), 474–480. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Riaz M, Lewis S, Naughton F, Ussher M, 2018. Predictors of smoking cessation during pregnancy: a systematic review and meta-analysis. Addiction 113 (4), 610–622. [DOI] [PubMed] [Google Scholar]
  35. Ribisl KM, D’Angelo H, Feld AL, Schleicher NC, Golden SD, Luke DA, Henriksen L, 2017. Disparities in tobacco marketing and product availability at the point of sale: Results of a national study. Prev Med 105, 381–388. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Rodriguez D, Carlos HA, Adachi-Mejia AM, Berke EM, Sargent JD, 2013. Predictors of tobacco outlet density nationwide: a geographic analysis. Tobacco Control 22 (5), 349. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Román-Gálvez RM, Amezcua-Prieto C, Olmedo-Requena R, Lewis-Mikhael Saad AM, Martmez-Galiano JM, Bueno-Cavanillas A, 2018. Partner smoking influences whether mothers quit smoking during pregnancy: a prospective cohort study. BJOG: an international journal of obstetrics and gynaecology 125 (7), 820–827. [DOI] [PubMed] [Google Scholar]
  38. Ruggles S, Flood S, Goeken R, Grover J, Meyer E, Pacas J, Sobek M, 2020. IPUMS USA: Version 10.0 [dataset] IPUMS, Minneapolis, MN. 10.18128/D010.V10.0. [DOI] [Google Scholar]
  39. Shawe J, Patel D, Joy M, Howden B, Barrett G, Stephenson J, 2019. Preparation for fatherhood: A survey of men’s preconception health knowledge and behaviour in England. PLoS One 14 (3), e0213897. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Sims M, Diez-Roux AV, Gebreab SY, Brenner A, Dubbert P, Wyatt S, Bruce M, Hickson D, Payne T, Taylor H, 2016. Perceived discrimination is associated with health behaviours among African-Americans in the Jackson Heart Study. Journal of epidemiology and community health 70 (2), 187–194. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Tager IB, Weiss ST, Munoz A, Rosner B, Speizer FE, 1983. Longitudinal Study of the Effects of Maternal Smoking on Pulmonary Function in Children. New England Journal of Medicine; 309 (12), 699–703. [DOI] [PubMed] [Google Scholar]
  42. Thaler RH, Sunstein CR, 2008. Nudge: Improving decisions about health, wealth, and happiness. Yale University Press, New Haven, CT, USA. [Google Scholar]
  43. Tomfohr LM, Pung MA, Dimsdale JE, 2016. Mediators of the relationship between race and allostatic load in African and White Americans. Health psychology 35 (4), 322–332. [DOI] [PubMed] [Google Scholar]
  44. U.S. Department of Health and Human Services, O.o.D.P.a.H.P., Healthy People 2020. https://www.healthypeople.gov/2020/topics-objectives/topic/maternal-infant-and-child-health/objectives.
  45. United States Census Bureau, 2017. American Community Survey Information Guide. https://www.census.gov/content/dam/Census/programs-surveys/acs/about/ACS_Information_Guide.pdf.
  46. Williams DR, Jackson PB, 2005. Social Sources Of Racial Disparities In Health. Health Affairs 24 (2), 325–334. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Methods, Tables, Figures
NIHMS1662589-supplement-Supplementary_Methods__Tables__Figures.docx (3.1MB, docx)

RESOURCES