If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Section of Acute and Critical Care Surgery, Division of General Surgery, Department of Surgery, Washington University in St. Louis School of Medicine, St. Louis, MO, USA
Cardiovascular Division, Department of Internal Medicine, Washington University in St. Louis School of Medicine, Center for Health Economics and Policy, Institute for Public Health at Washington University, St. Louis, MO, USA
Section of Acute and Critical Care Surgery, Division of General Surgery, Department of Surgery, Washington University in St. Louis School of Medicine, St. Louis, MO, USA
To explore the relationship between abortion restrictions and maternal mortality in the United States.
Study design
This was a retrospective study examining maternal mortality in the United States from 1995 to 2017. We used the Global Health Data Exchange and the Centers for Disease Control and Prevention WONDER databases to extract maternal mortality data for all 50 states for each year from 1995 to 2017. We categorized states as restrictive, neutral, or protective of abortion access according to policy information published by the Guttmacher Institute. We assessed associations between abortion restrictions and maternal mortality ratios (maternal deaths per 100,000 live births).
Results
In 1995, the mean maternal mortality ratios were similar across all groups of states (Restrictive 12.6, 95% CI 11.4-13.6; Neutral 12.2, 95% CI 10.9-13.4; Protective 10.9, 95% CI 9.6-11.9). Maternal mortality ratios increased for each group of states over time and in 2017, the mean maternal mortality ratio was higher in restrictive states than in protective states (Restrictive 28.5, 95% CI 20.7-35.1; Neutral 22.9, 95% CI 16.1-28.6; Protective 15.7, 95% CI 10.7-19.9). Regressions accounting for policy, state and year showed a statistically significant increase in maternal mortality ratios in restrictive states relative to neutral states (1.06, 95% CI 1.01-1.11) and a non-significant decrease associated with protective states (0.89, 95% CI 0.78-1.01).
Conclusions
States that restrict abortion have higher maternal mortality than states that either protect or are neutral towards abortion. Further investigation is needed to determine how abortion restrictions are associated with increased maternal mortality.
Implications
The association between abortion restrictions and maternal mortality may reflect the overall legislative priorities of individual states as restrictive states are less likely to pass proactive legislation demonstrated to improve maternal outcomes.
]. There is also growing evidence that social determinants of health, particularly systemic racism, may play an important role in maternal mortality outcomes for women of color [
]. Use of family planning services can allow women to avoid both unplanned and high-risk pregnancies, which have previously been associated with adverse pregnancy outcomes [
]. Since 1973, states have enacted over 1,200 abortion restrictions, resulting in differences in the availability of abortion services across the country [
Thompson T, Seymour J. Evaluating priorities: measuring women's and children's health and wellbeing against abortion restrictions in the states. Ibis Reproductive Health Research Report
]; however, the relationship between abortion restrictions and maternal mortality has not been examined. The objective of our study was to explore this relationship in the United States and whether any racial and ethnic disparities exist. We hypothesized that increased restriction of abortion is associated with increased maternal mortality.
2. Methods
2.1 Classification of states as restrictive, neutral, or protective toward abortion
We used policy data from the Guttmacher Institute [
] to designate individual states as restrictive, neutral, or protective towards of abortion. Status was assigned annually for years in which data was available and individual states could change categories (e.g., neutral to protective or neutral to restrictive) during our study period. We designated a state as restrictive if it met the Guttmacher Institute definition of being “hostile” to abortion [
]. A state that did not meet either set of classification criteria was categorized as neutral.
2.2 Longitudinal trends in maternal mortality ratio
We used the World Health Organization (WHO) definition of maternal death: “the death of a woman while pregnant or within 42 days of termination of pregnancy, irrespective of the duration and site of the pregnancy, from any cause related to or aggravated by the pregnancy or its management but not from accidental or incidental causes” [
Global Burden of Disease Collaborative Network. Global Burden of Disease Study 2017 (GBD 2017) Results. Seattle, United States: Institute for Health Metrics and Evaluation (IHME), 2018. Available from http://ghdx.healthdata.org/gbd-results-tool.
] (GHDx), a publicly available database managed by the Institute for Health Metrics and Evaluation that includes global vital statistics. State-level data for MMR were abstracted from the GHDx for the years 1995 to 2017 for all maternal disorders fitting the WHO definition of maternal death (Codes A.6.1.1 through A.6.1.7, A.6.1.9, and A.6.1.10). Mean MMRs were calculated for each category of states (restrictive, neutral, and protective) and weighted by total births in each state using CDC birth data (available from 1995 to 2017).
The CDC WONDER database was queried for maternal deaths, based on the same WHO definition of maternal death (Codes O-00 through O-95, O-98, O-99, and A-34). We summed maternal mortality for each category of states and calculated MMRs using CDC birth data for the years 1999 to 2017.
The GHDx does not report information on race, but the CDC WONDER provides a racial breakdown of mortality if enough years are sampled to avoid suppression of low frequency data as defined by state-level events with nine or fewer deaths due to privacy limitations. Given published evidence of disparities in health outcomes by race and ethnicity, we planned to assess for differences in MMRs by these factors. The number of abortion restrictions greatly increased in the latter half of the time frame [
]. We divided the data into two nearly equal time periods of 1999 to 2008 (period 1) and 2009 to 2017 (period 2) to assess more recent trends in MMR. We calculated MMRs from the CDC WONDER database for each category of states based on their restrictive, neutral, or protective status in 2017, stratified by racial data.
2.3 Statistical analysis
First, we plotted the mean MMRs from GHDx for the years 1995 to 2017, weighted by each state's live birth data to assess for MMR trends in each group of states. We similarly plotted MMRs from CDC WONDER for the years 1999 to 2017 for each set of states. This approach allows for examination of raw trends between each group of states and demonstrates the overall differences in outcomes between states; however, it includes both within-states and between-state differences and thus does not eliminate potential confounding due to inherent differences between these groups of states. To mitigate potential confounding and isolate the effect of within-state shifts in policy, we used negative binomial regression to explore the relationship between abortion restrictions and maternal mortality in the GHDx database. We used a “time-lagged policy” model as policy changes are unlikely to result in changes in maternal mortality rates during the calendar year in which they are enacted. In this model, the abortion policy classification for a given state was compared to the MMR in that state three years later. We ran an additional model that included policy, state, and year in order to account for likely confounding effects of these variables. We ran a similar analysis using the CDC WONDER dataset. In the CDC WONDER model, only policy and year were included as factors, as state-level data are not available due to privacy restrictions. These analyses were performed using the MASS library in the R statistical package [
We completed a sensitivity analysis of the unweighted GHDx data to ensure MMRs from more populous states did not outweigh MMRs from less populous states. Next, we performed four sensitivity analyses with negative binomial regression of the GHDx dataset. First, we repeated analyses after removing data from the state of Texas as prior literature suggests the MMRs may be overestimated in that state [
]. Second, we included Medicaid expansion under the Affordable Care Act as a potential confounding covariate as increased insurance coverage may reduce maternal mortality [
]. Next, we included the addition of the pregnancy checkbox to a state's death certificate as a potential confounder as individual states adapted the pregnancy checkbox at different years throughout the included time frame. Finally, we repeated our analyses using a static rather than year-to-year definition of the categories, classifying all states based on their category in 2017.
Statistical analyses and graphing were performed in R 3.5.1 [
]. A two-tailed P-value of less than 0.05 was considered statistically significant. Because the data were publicly available and de-identified, the Washington University in St. Louis Human Research Protection Office considered this to not be human subjects research and waived the requirement for informed consent.
3. Results
3.1 Classification of states
In 1995, 13 states (26%) were restrictive, 32 states (64%) were neutral, and 5 states (10%) were protective (Fig. 1A). Based on 2017 policy data, 29 states (58%) were restrictive, 12 states (24%) were neutral, and 9 states (18%) were protective (Fig. 1B). An analysis of year-to-year policy shows the number of states restricting abortion increased by more than 100% from 2000 to 2017 (13 vs. 29). The number of states neutral to abortion decreased (32 vs 12). Five states were protective across the entire time frame, and four states changed from neutral to protective (eTable 1).
Fig. 1Classification of abortion restriction for individual states in 1995 and 2017, based on policy information from the Guttmacher Institute
3.2 Longitudinal estimates of maternal mortality 1995-2017
3.2.1 Global Health Data Exchange
In 1995, the mean weighted MMRs were similar for all groups of states (Fig. 2A) (Restrictive 12.6, 95% CI 11.4-13.6; Neutral 12.2, 95% CI 10.9-13.4; Protective 10.9, 95% CI 9.6-11.9). In 2009 and subsequent years, the 95% confidence intervals of restrictive and protective states did not overlap (eTable 2). In 2017, the weighted mean MMR in restrictive states was almost 70% higher than that in protective states (Restrictive 28.5, 95% CI 20.7-35.1; Protective 15.7, 95% CI 10.7-19.9). In neutral states, the mean MMR (Neutral 22.9, 95% CI 16.1-28.6) was between those for restrictive and protective states.
Fig. 2Longitudinal trends in maternal mortality ratios by abortion restriction, 1995-2017
2A – Plot of weighted mean maternal mortality ratio (deaths per 100,000 live births) through 42 days from termination of pregnancy for restrictive (orange), neutral (tan), and protective (blue) states from 1995 to 2017 from Global Health Data Exchange. Shaded areas represent 95% confidence interval (CI).
2B - Plot of maternal mortality ratio (deaths per 100,000 live births) through 42 days from termination of pregnancy for restrictive (orange), neutral (tan), and protective (blue) states from 1999 to 2017 from CDC WONDER.
CDC WONDER data (Fig. 2B) reveal absolute MMRs began to diverge beginning in 2009 between the categories of states. Prior to 2009, there was some overlap between the categories. In 2017, absolute MMRs were highest in restrictive states, followed by neutral states and lowest in protective states.
3.3 Association between abortion restriction and MMR
Using CDC WONDER data, we modeled maternal mortality in protective and restrictive states, relative to neutral states, as a function of policy and year. In this model, restrictive states have a significantly higher MMR and protective states have a significantly lower MMR compared to neutral states (Restrictive 1.24, 95% CI 1.06-1.50; Protective 0.71, 95% CI 0.59-0.85). To most closely mirror this analysis in the GHDx dataset, we first modeled maternal mortality as a function of policy and year. Again, there was a significantly higher MMR in restrictive states and a significantly lower MMR in protective states (Restrictive 1.22, 95% CI 1.17-1.27; Protective 0.84, 95% CI 0.79-0.89). Finally, we accounted for state-level effects by including state, year, and policy in our model. This again showed a significant increase in MMR in restrictive states (Restrictive 1.06; 95% CI 1.01-1.11). There was a decrease in MMR in protective states (Protective 0.89; 95% CI 0.78-1.01), although this result was not statistically significant (Table 1).
Table 1Modeled maternal mortality
Model
Classification
Relative maternal mortality ratio
95% CI
p value
Policy + Year (CDC WONDER)
Restrictive
1.24
1.06-1.50
0.008
Protective
0.71
0.59-0.85
<0.001
Policy + Year (GHDx)
Restrictive
1.22
1.17-1.27
<0.001
Protective
0.84
0.79-0.89
<0.001
Policy + Year + State (GHDx)
Restrictive
1.06
1.01-1.11
0.03
Protective
0.89
0.78-1.01
0.06
Negative binomial regression of modeled maternal mortality for restrictive and protective states as compared to neutral states.
3.4 Racial disparities 1999-2008 vs 2009-2017 by abortion restriction (CDC WONDER)
Black women had the highest MMRs in all groups of states in period 1 (1999-2008) and period 2 (2009-2017) (Fig. 3). In the restrictive states, MMR increased across all racial groups between period 1 and period 2 (Hispanic 8.5 vs 16.2, white 8.9 vs 20.4, black 28.2 vs 47.2, Asian 11.5 vs 14.6, and Native American 13.4 vs 37.9). In the neutral states, MMR increased across all racial groups between period 1 and period 2 (Hispanic 11.9 vs 14.9, white 7.0 vs 13.0, black 35.3 vs 48.7, Asian 8.6 vs 12.2, Native American 10.2 vs 35.8). In the protective states, MMR either decreased or remained stable across all racial groups between period 1 and period 2 (Hispanic 10.0 vs 5.2, white 8.9 vs 9.2, black 30.2 vs 26.3, Asian 10.1 vs 7.8, and Native American 23.6 vs 18.2).
Fig. 3Racial Disparities 1999-2008 vs 2009-2017 by abortion restriction (CDC WONDER)
Trend in MMR, by race, for the grouped years of 1999-2008 (period 1) and 2009-2017 (period 2) for each group of states: restrictive (orange), neutral (tan), and protective (blue).
The unweighted analysis of the GHDx data was still significant, though with greater spread in confidence intervals as smaller states had a greater impact. Negative binomial regression sensitivity analyses were largely similar to our main results (eTable 3). Removing Texas did not alter the results. Similarly, analyses accounting for Medicaid expansion continued to demonstrate a small but statistically significant increase in MMR in restrictive states. Medicaid expansion in protective states was associated with a decreased MMR, though this reduction was not statistically significant. There was a significant association between not having a checkbox and decreased MMR (0.93; 95% CI = 0.89- 0.98); however, in this model there was still a significant increase in MMR in restrictive states and a non-significant trend towards decreased MMR in protective states. Finally, analysis of the data using a static policy definition to categorize states, as opposed to the year-to-year definition used in our primary analysis, revealed a similar trend.
4. Discussion
This study demonstrates an association between abortion restrictions and increased MMRs. MMRs were similar across the three groups throughout the 1990s, but in the mid-2000s, began to diverge. It is notable that this corresponds with states enacting 2.8-fold more abortion restrictions in 2009 through 2017 than in 1999 through 2008 [
]. We also noted evidence of a dose-response relationship in which restrictive states had the largest increase in MMR over time, neutral states had a moderate increase, and protective states had the smallest increase.
Our findings may reflect broader issues impacting access to reproductive healthcare and maternal outcomes. There are likely a number of healthcare policies at the state level that impact overall health care access which could contribute to the association demonstrated in our analyses. Previous literature demonstrates that states that protect abortion access also implement policies that reduce maternal mortality while states with more abortion restrictions have fewer policies that support women's well-being [
Thompson T, Seymour J. Evaluating priorities: measuring women's and children's health and wellbeing against abortion restrictions in the states. Ibis Reproductive Health Research Report
]. States that pass abortion restrictions may not prioritize other policies to improve maternal outcomes, such as Medicaid expansion. While our findings are still significant when accounting for Medicaid expansion, both access to abortion services and the broader state-level care milieu likely play a role in maternal outcomes. States that restrict abortion may have broader hostility towards women's health. A research report from Ibis Reproductive Health published in 2017 concluded that states with more abortion restrictions had poorer health and well-being outcomes for women and children [
Thompson T, Seymour J. Evaluating priorities: measuring women's and children's health and wellbeing against abortion restrictions in the states. Ibis Reproductive Health Research Report
]. This report pointed out that “restrictions are often disproportionately felt by populations that may derive the greatest benefit from supportive policies.” Given the association between higher MMR and restricting abortion, restrictive states may benefit from targeted interventions to combat their rising maternal mortality.
This study does not allow us to draw any conclusions regarding direct causality, and models that restricted the effect of abortion policy to the within-state change over time were associated with a modest increase in maternal mortality ratios. However, there are several mechanisms by which abortion restrictions could be associated with higher maternal mortality. One possibility is that women who have high-risk pregnancies and wish to terminate for medical reasons may be less able to do so in states in which abortion is restricted. Thus, women with chronic medical conditions may be more likely to continue high-risk pregnancies, increasing their risk of poor obstetrical outcomes [
]; however, a more recent study found that states with a highly restrictive legislative environment towards abortion provision did have a lower abortion rate [
]. It is possible that broadly, abortion restrictions may lower the abortion rate and that certain populations high-risk populations may be disproportionately harmed by restrictions. A second possibility is that when abortion clinics close due to abortion restrictions, as seen in Texas following Texas House Bill 2 [
], there is “collateral damage” to sexual and reproductive healthcare services provided by abortion clinics. Access to contraception can decrease unintended pregnancies [
] and this may be an important area for future research.
Previous literature has shown that women of color have higher maternal mortality than white women, and suggest that structural racism may impact maternal health outcomes [
]. The associations between abortion restrictions and MMRs in our analysis are particularly strong for women of color. The longstanding history and pervasive impact of structural racism permeates all fields of medicine, but are particularly apparent in maternal health outcomes [
]. Women of color may be disproportionately affected by abortion restrictions because they already experience higher structural barriers to healthcare.
This study has important limitations. In general, there are inherent limitations to the use of longitudinal data for maternal mortality due to changes in reporting standards for maternal mortality. This is primarily due to the addition of the “pregnancy check box” to death certificates in 2003. Individual states adapted the pregnancy check box at different times between 2003 and 2017 which limits the examination of cross-sectional data [
]. We completed a sensitivity analysis to account for the year an individual state adapted the pregnancy checkbox and our results were largely unchanged. Additionally, since 2003, the California definition of maternal mortality has included death within one year of pregnancy. In our analysis, the MMR in California, a protective state, is likely inflated by the expanded definition of maternal mortality, which would bias our results to the null. A second limitation is that states were grouped by the number of enacted policies rather than focusing on individual restrictions. We did so to correspond with the Guttmacher Institute classification of abortion rights. Additionally, there may be a cumulative burden due to layering of multiple restrictions rather than the any given restriction creating a barrier [
]. Third, because our data were from de-identified public datasets, we did not have access to data on comorbidities or other factors that might explain our findings. Most importantly, the correlations we report do not prove causation. Future studies are needed to determine the mechanisms by which abortion access, or other associated policies, contribute to positive and negative changes in MMRs with specific attention to minority populations
We found an association between restricted abortion restrictions and increased maternal mortality, particularly for black and Native American women. Increased legal restrictions on abortion may be a part of a broader set of legislative and policy priorities that affect patients’ access to needed reproductive health services. Rising maternal mortality in the US is likely multifactorial, but the impact of decreasing access to comprehensive reproductive healthcare and over regulation of abortion care should be examined along with broader health policy, infrastructure, insurance, and environmental factors.
Declaration of Competing Interest
Dr. Eisenberg has the following disclosures: Medicines 360 (PI on Phase 3 trial of LNG-IUS, Scientific advisory board), Merck (served as trainer for Nexplanon® contraceptive implant), FemaSys (Consultant and Scientific advisory board), ACI Clinical (Chair of DSMB for Phase 3 contraceptive trial). Dr. Joynt Maddox receives research support from the National Heart, Lung, and Blood Institute (R01HL143421), National Institute on Aging (R01AG060935), and Commonwealth Fund, and previously did contract work for the US Department of Health and Human Services. The remaining authors have no reported conflicts of interest.
Funding
Research reported in this publication was supported by the National Center for Advancing Translational Sciences of the National Institutes of Health under award number UL1TR002345.The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Thompson T, Seymour J. Evaluating priorities: measuring women's and children's health and wellbeing against abortion restrictions in the states. Ibis Reproductive Health Research Report
Global Burden of Disease Collaborative Network. Global Burden of Disease Study 2017 (GBD 2017) Results. Seattle, United States: Institute for Health Metrics and Evaluation (IHME), 2018. Available from http://ghdx.healthdata.org/gbd-results-tool.
00090-1&id=gr1.jpg){kind=link}
00090-1&id=gr2.jpg){kind=link}
00090-1&id=gr3.jpg){kind=link}