Severe maternal morbidity and maternal mortality in the United States are on the rise.1,2 Access to induced abortion may reduce these risks in some populations.3 Recent research suggests that both unintended pregnancy and chronic health conditions are risk factors for maternal morbidity.4 Women with unintended pregnancies might develop significant health problems as a consequence of not seeking out and engaging with adequate prenatal care.5Furthermore, women who begin pregnancies with pre-existing medical conditions and who lack access to medically necessary abortion may go on to have severe maternal morbidity or mortality.6
In the United States, however, access to induced abortion is not universal. Under the policy known as the Hyde Amendment, federal funding cannot be spent on abortion except in cases of life endangerment, rape, or incest. This policy may have a major effect on pregnant women who receive Medicaid, the joint state–federal program that is the largest single payer for perinatal care.7 Thirty-three states and Washington, DC, provide no Medicaid coverage of induced abortion. In 17 states, Medicaid programs choose to provide coverage of medically necessary abortion using state-only funding.8
Our objective was to estimate the association between state Medicaid programs' coverage of medically necessary abortions and in-hospital maternal mortality and hospitalizations for severe maternal morbidity among U.S. women from 2000 to 2011. We expected that state Medicaid programs' coverage of medically necessary abortions should reduce severe maternal morbidity and mortality through improved access to safe abortions for women who would otherwise experience unwanted or high-risk pregnancies. We therefore hypothesized that women who had pregnancy-related hospitalizations in states with Medicaid coverage of medically necessary abortions would have significantly lower risks of severe maternal morbidity and in-hospital maternal mortality relative to those women who had hospitalizations in states without such policies.
MATERIALS AND METHODS
We obtained a large, nationally representative sample of hospital discharge data from the Nationwide Inpatient Sample from 2000 to 2011 via the Healthcare Cost and Utilization Project administered by the Agency for Healthcare Research and Quality.9 After 2011, the Nationwide Inpatient Sample changed its sampling strategy and no longer included indicators of which states hospitalizations occurred in; therefore, we include data only through 2011. These data represent the largest source of all-payer administrative data in the United States and contain uniform data elements from nearly all U.S. states on diagnosis and procedure codes, diagnosis-related groups, severity and comorbidities, patient characteristics, and hospital characteristics.
We used International Classification of Diseases, 9th Revision, Clinical Modification diagnosis and procedure codes and diagnosis-related group codes to hierarchically identify delivery, antenatal, and postpartum hospitalizations for women ages younger than 15 years or older than 44 years in 47 states with data available in any year from 2000 to 2011 (unweighted n=10,131,942 hospitalizations).10–12 We excluded cases with missing information on the payer for the hospitalization (unweighted n=18,473; less than 1% of cases) and hospitalizations with missing data on mortality (unweighted n=2,102; less than 1% of cases). For some states and years, data on patient race were not reported for administrative reasons (eg, state hospital discharge data collection systems did not capture race or less populous states chose not to report race to protect patient privacy). Because maternal race is an important predictor of severe maternal morbidity and mortality and is associated with state of residence, it is critical to control for patient race in our analyses. Following previous research, we chose to exclude data from states and years where race was not reported.13 This resulted in excluding data for 22% (unweighted n=2,282,556) of pregnancy-related hospitalizations in our study. However, because race data are missing for administrative reasons, missingness should not be associated with state Medicaid policy status nor outcomes, and we do not expect this exclusion to introduce selection bias. Our analytic sample included data from 43 states on 7,828,788 pregnancy-related hospitalizations.
The main exposure was state Medicaid coverage of medically necessary abortions. Although all state Medicaid programs cover abortion in the cases of life endangerment, rape, or incest, medically necessary abortions are defined as those needed to protect a woman's health. Medically necessary abortion was defined by the U.S. Supreme Court as a professional judgment made by a physician “exercised in the light of all factors—physical, emotional, psychological, familial, and the woman's age—relevant to the wellbeing of the patient.”14 We used data collected by the Guttmacher Institute for each state and year indicating that 17 state Medicaid programs provided optional coverage of medically necessary abortions using state-only funding during the study time period.8 Coverage of medically necessary abortion was constant in all states across our study time period. Because we cannot measure state-level differences in Medicaid policy implementation, our results should be interpreted as the average of heterogeneous associations between Medicaid policies and outcomes.
We used the definitions published by the Centers for Disease Control and Prevention to identify severe maternal morbidity among pregnancy-related hospitalizations (Table 1).15 There is no single definition of severe maternal morbidity, and hospital discharge data may make it difficult to differentiate variability within diagnoses that might constitute severe maternal morbidity.16 Specifically, we used International Classification of Diseases, 9th Revision, Clinical Modification codes for each hospital discharge to identify the presence of any of 24 indicator conditions for severe maternal morbidity. We excluded blood transfusion from our definition of severe maternal morbidity because it is far more common than other morbidities and is collinear with other morbidities. We created a binary measure of having any severe maternal morbidity during a hospitalization.
In-hospital maternal mortality was based on a variable indicating whether or not a patient was discharged alive. If a hospital discharge for a pregnancy-related hospitalization had a value indicating that the patient died in the hospital, it was considered a maternal death.
We used measures of whether each patient was covered by Medicaid compared with another type of insurance (private insurance, self-pay, charity, or other insurance such as Medicare or military health plans). We also included a categorical measure of maternal race and ethnicity, defined as non-Hispanic white, non-Hispanic black, non-Hispanic Asian or Pacific Islander, non-Hispanic other race, or Hispanic ethnicity. We created a restricted cubic spline for maternal age at the time of hospitalization. We included a state-level measure of the annual proportion of adult obesity based on state-representative population survey data published by the Centers for Disease Control and Prevention17 to account for the correlation between state obesity rates and both health care policy and adverse maternal outcomes. We used a measure of obesity for all adults because state-representative data on obesity in pregnancy are not available for all states or all years in our study.
The unit of analysis was each hospitalization. We calculated weighted descriptive statistics of the study population overall and stratified by whether hospitalizations occurred in states with optional Medicaid coverage of medically necessary abortions. We used logistic regression to first assess whether state Medicaid coverage of medically necessary abortions was associated with the risk of severe maternal morbidity. Next, we extended the model to include women's individual Medicaid enrollment status with an interaction term between the state Medicaid abortion policy and Medicaid enrollment status. This allowed us to determine if the policy had different effects on risks between women with Medicaid coverage (among whom we would expect the policy to decrease the risk of mortality and severe maternal morbidity) and women with private insurance coverage (among whom we would expect weaker or no effects). From the regression results, we calculated the average predicted risks and 95% confidence intervals of severe maternal morbidity among women with Medicaid and private insurance coverage with hospitalizations occurring in states with and without Medicaid coverage of medically necessary abortion. We repeated the analyses with maternal mortality as the outcome. Per the Strengthening the Reporting of Observational Studies in Epidemiology Statement, statistical testing is conducted only for hypothesis tests and is not presented for descriptive data.18 All analysis used Stata's SVY commands to account for the complex sampling methodology of the Nationwide Inpatient Sample data.
We conducted a priori sensitivity analyses to examine the robustness of our results. Our primary analyses defined severe maternal morbidity as the presence of a diagnosis of one of 24 conditions during hospitalization regardless of length of stay for the hospitalization. Prior research has required a long hospital stay as a condition to define severe maternal morbidity for delivery hospitalizations, because conditions may be overrecorded in hospital discharge data.1 We replicated our main analyses restricting severe maternal morbidity as hospitalizations for which length of stay was in the top 50th percentile for vaginal or cesarean delivery and during which women did not die during the hospitalization (all antenatal and postpartum hospitalizations were included).
The University of Pittsburgh institutional review board determined that this study was exempt because it involved the analysis of only deidentified hospital discharge data.
In our Nationwide Inpatient Sample of 43 states across 12 years, 42.7% of pregnancy-related hospitalizations occurred in states with Medicaid coverage of medically necessary abortions (Table 2). Our study included a weighted 38,016,845 pregnancy-related hospitalizations, 41% of which were paid for by Medicaid. In comparison, there were 45,043,714 births in the United States over that time period,19 an estimated 45% of which were paid for by Medicaid.7 Relative to hospitalizations occurring in states without Medicaid coverage of medically necessary abortion, hospitalizations occurring in states with Medicaid coverage of medically necessary abortion tended to have slightly older patients (mean age 29 compared with 27 years) and had considerably more racial diversity (54% nonwhite patients compared with 43% nonwhite patients) (Table 3). Hospitalizations occurring in states with Medicaid coverage of medically necessary abortion were less likely than their counterparts to be paid for by Medicaid (38% compared with 44%).
Among women with Medicaid-paid hospitalizations, the unadjusted rate of severe maternal morbidity was 66.6 per 10,000 pregnancy-related hospitalizations; this rate was lower for hospitalizations occurring in states with Medicaid coverage of medically necessary abortion relative to those occurring in states without Medicaid coverage (62.4/10,000 compared with 69.3/10,000) (Table 4). The rates of the five most frequently occurring severe maternal morbidities among women with Medicaid (disseminated intravascular coagulation, heart failure during procedure, eclampsia, adult respiratory distress syndrome, and acute renal failure) were lower among hospitalizations in states with Medicaid coverage of medically necessary abortion relative to hospitalizations in states without such coverage.
Figure 1 shows the adjusted risks of severe maternal morbidity and maternal mortality by state Medicaid coverage of medically necessary abortion and stratified by whether Medicaid paid for the hospitalization. After stratifying on payer at hospitalization, we found that among hospitalizations paid by Medicaid, those occurring in states providing Medicaid coverage of medically necessary abortion had 8.5 per 10,000 (95% confidence interval [CI] 4.0–16.5/10,000) fewer cases of severe maternal morbidity (61.5/10,000, 95% CI 56.1–65.1) than those occurring in states without such coverage (70.0/10,000, 95% CI 65.1–75.8). Likewise, among private insurance–paid hospitalizations, there were 10.3 per 10,000 (95% CI 3.5–17.2/10,000) fewer cases of severe maternal morbidity in states with Medicaid coverage of medically necessary abortion (58.7/10,000, 95% CI 53.4–64.0) relative to those in states without such coverage (69.1/10,000, 95% CI 64.8–73.3). Results did not meaningfully differ in magnitude or direction when we varied the definition of severe maternal morbidity based on length of hospital stay, among women who did not die, in our sensitivity analysis. Across both Medicaid and private insurance–paid hospitalizations, state Medicaid coverage of medically necessary abortion was associated with an adjusted average of 9.5 per 10,000 fewer cases of severe maternal morbidity, which translates to an average 16% risk reduction.
Among women with Medicaid, the unadjusted rate of in-hospital mortality was 9.1 per 100,000 pregnancy-related hospitalizations, which did not differ significantly by state Medicaid abortion coverage status (Table 4). The average predicted risk of in-hospital maternal mortality was not different among Medicaid-paid hospitalizations in states with (9.2/100,000, 95% CI 7.3–11.0) or without (9.0/100,000, 95% CI 7.4–10.6) Medicaid coverage of medically necessary abortion. Likewise, there was no significant difference in in-hospital mortality among private insurance–paid hospitalizations in states with (5.6/100,000, 95% CI 4.5–6.8) or without (6.1/100,000, 95% CI 4.8–7.4) Medicaid coverage of medically necessary abortion.
State Medicaid coverage of medically necessary abortion was associated with an average 16% decreased risk of severe maternal morbidity. Women residing in states with state-funded Medicaid coverage of medically necessary abortion had lower risks of severe maternal morbidity during pregnancy-related hospitalizations compared with women who resided in states without such Medicaid coverage. This was true regardless of whether the hospitalization itself was paid for by Medicaid or private insurance. Such Medicaid coverage was not associated with risk of in-hospital maternal mortality.
There are several potential explanations for these findings. Medicaid coverage of medically necessary abortion may reduce the number of medically risky pregnancies and therefore reduce the risk of severe maternal morbidity during pregnancy-related hospitalizations. Interestingly, state Medicaid coverage of medically necessary abortion was associated with similar reductions in the risk of severe maternal morbidity among both Medicaid-paid and private insurance–paid hospitalizations. This finding, although surprising at first blush, is consistent with a sizable body of health economics research suggesting that policy adoption in one payer may have spillover effects among other payers.20–22 One likely mechanism for such spillover effects is that Medicaid payment policies may drive practice pattern norms, and physicians are likely to provide similar care to all patients regardless of payer.23 It follows that coverage of medically necessary abortion in Medicaid may affect provision of abortion services among women with both Medicaid and private insurance. This possible interpretation of our findings should be read with caution, because our study was not intended to explicitly test for spillover effects.
An alternate explanation for our findings is that there is an unmeasured population-level confounder or confounders causing lower severe maternal morbidity rates and is associated with state Medicaid coverage of medically necessary abortion. The fact that we observed reduced risk of severe maternal morbidity among all hospitalizations, regardless of payer, might suggest that there are other state-level factors associated with Medicaid coverage of medically necessary abortion that also lead to reduced severe maternal morbidity. One such factor might be geographic region. Notably, no states in the southern United States cover medically necessary abortion through Medicaid, and to the extent that geographic region is causally associated with different cultural and medical risk factors for severe maternal morbidity, this might explain our results. A second factor might be that state Medicaid coverage of medically necessary abortion is correlated with other state-level policies that facilitate access to induced abortion and might be causally associated with reduced risk of severe maternal morbidity. Specifically, we note that of the 17 states that provided Medicaid coverage of medically necessary abortion, only four had any policy to impose restrictions on access to abortion services (known as Targeted Regulation of Abortion Providers [TRAP] laws).24 This latter explanation is consistent with our hypothesis that increased access to medically necessary abortion would be associated with reduced risk of severe maternal morbidity and mortality.
However, there is also reason to suspect that state Medicaid coverage of medically necessary abortion is independently associated with reduced risk of severe maternal morbidity. In some states, coverage of medically necessary abortion was implemented by legislative or regulatory mechanisms, and in others, court decisions required coverage of medically necessary abortion in Medicaid programs. It is not immediately apparent what unmeasured state-level factor would be simultaneously associated with various legislative and court decisions and is causally related to severe maternal morbidity.
It is unlikely that our results are explained by an overrepresentation of lower risk pregnancies in states with Medicaid coverage of medically necessary abortion. On the contrary, in our study, pregnancy-related hospitalizations that occurred in states with Medicaid coverage of medically necessary abortion had populations that were older and more racially diverse, which would be expected to increase rates of adverse maternal outcome.
Given the observed reduction in the risk of severe maternal morbidity, it is surprising that we did not find an association between Medicaid coverage of medically necessary abortion and reduction in in-hospital maternal mortality. Our measure of maternal mortality is limited to in-hospital mortality. The rarity of this outcome might limit our ability to detect small differences in risk by state Medicaid policy status.
This study has limitations. First, this is an observational study and we cannot definitively establish a causal relationship between state Medicaid coverage of medically necessary abortion and outcomes. To address this limitation, we conducted sensitivity analyses to test the robustness of our findings, and our analyses have controlled for observable factors that might confound the relationship between state Medicaid policy status and maternal mortality and severe maternal morbidity. We cannot, however, rule out the possibility of unmeasured confounding. Second, the hospital discharge data we used do not allow us to follow patients over time or measure prenatal care or antepartum clinical risk factors. The lack of these data limits our ability to draw firm conclusions about the hypothesized mechanism of the association between coverage of medically necessary abortion and severe maternal morbidity. Third, we are unable to determine from hospital discharge data whether women residing in states with Medicaid coverage of medically necessary abortion did in fact have greater use of abortion services. There might be variation in women's access to abortion services even within the group of states that provide Medicaid coverage of medically necessary abortion, which would make our estimates less precise. Fourth, there is likely to be geographic variation in practices related to coding certain conditions during pregnancy-related hospitalizations; however, we have no reason to expect such variation to be systematically associated with our exposure or outcomes of interest. Strengths of our study include the use of a large, national data set that provides high external validity and our ability to draw comparisons in outcomes both by state and by the payer for the hospitalizations.
Our findings are consistent with the hypothesis that Medicaid coverage of medically necessary abortion does not harm maternal health and is associated with reduced risk of severe maternal morbidity during pregnancy-related hospitalizations. Such evidence is important as legislators and the courts consider whether state abortion restriction policies are needed to protect maternal health25 and as federal policymakers consider the health implications of the Hyde Amendment.26
1. Callaghan WM, Creanga AA, Kuklina EV. Severe maternal morbidity among delivery and postpartum hospitalizations in the United States. Obstet Gynecol 2012;120:1029–36.
2. Creanga AA, Berg CJ, Syverson C, Seed K, Bruce FC, Callaghan WM. Pregnancy-related mortality in the United States, 2006–2010. Obstet Gynecol 2015;125:5–12.
3. Papiernik E, Zeitlin J, Delmas D, Draper ES, Gadzinowski J, Künzel W, et al. Termination of pregnancy among very preterm births and its impact on very preterm mortality: results from ten European population-based cohorts in the MOSAIC study. BJOG 2008;115:361–8.
4. Moaddab A, Dildy GA, Brown HL, Bateni ZH, Belfort MA, Sangi-Haghpeykar H, et al. Health care disparity and state-specific pregnancy-related mortality in the United States, 2005–2014. Obstet Gynecol 2016;128:869–75.
5. Kost K, Lindberg L. Pregnancy intentions, maternal behaviors, and infant health: investigating relationships with new measures and propensity score analysis. Demography 2015;52:83–111.
6. Raymond EG, Grimes DA. The comparative safety of legal induced abortion and childbirth in the United States. Obstet Gynecol 2012;119:215–9.
7. Markus AR, Andres E, West KD, Garro N, Pellegrini C. Medicaid covered births, 2008 through 2010, in the context of the implementation of health reform. Womens Health Issues 2013;23:e273–80.
8. Guttmacher Instittue. State funding of abortion under Mediciad. 2016. Available at: https://www.guttmacher.org/sites/default/files/state_policy_overview_files/spib_sfam.pdf
. Retrieved May 21, 2016.
9. Healthcare Cost and Utilization Project. Overview of the National (Nationwide) Inpatient Sample (NIS). 2015. Available at: https://www.hcup-us.ahrq.gov/nisoverview.jsp
. Retrieved July 20, 2015.
10. Kuklina EV, Whiteman MK, Hillis SD, Jamieson DJ, Meikle SF, Posner SF, et al. An enhanced method for identifying obstetric deliveries: implications for estimating maternal morbidity. Matern Child Health J 2008;12:469–77.
11. Ghaji N, Boulet SL, Tepper N, Hooper WC. Trends in venous thromboembolism among pregnancy-related hospitalizations, United States, 1994–2009. Am J Obstet Gynecol 2013;209:433.e1–8.
12. Kuklina EV, Callaghan WM. Cardiomyopathy and other myocardial disorders among hospitalizations for pregnancy in the United States: 2004–2006. Obstet Gynecol 2010;115:93–100.
13. Hutcheon JA, Bodnar LM, Simhan HN. Medicaid pregnancy termination funding and racial disparities in congenital anomaly-related infant deaths. Obstet Gynecol 2015;125:163–9.
14. Doe v Bolton, 410 US 179 (1973).
15. Centers for Disease Control and Prevention. Severe maternal morbidity in the United States. 2015. Available at: http://www.cdc.gov/reproductivehealth/maternalinfanthealth/severematernalmorbidity.html
. Retrieved May 21, 2016.
16. Severe maternal morbidity: screening and review. Obstetric Care Consensus No. 5. American College of Obstetricians and Gynecologists. Obstet Gynecol 2016;128:e54–60.
17. Centers for Disease Control and Prevention (CDC). Vital signs: state-specific obesity prevalence among adults—United States, 2009. MMWR Morb Mortal Wkly Rep 2010;59:951–5.
18. Vandenbroucke JP, von Elm E, Altman DG, Gøtzsche PC, Mulrow CD, Pocock SJ, et al. Strengthening the Reporting of Observational Studies in Epidemiology (STROBE): explanation and elaboration. Epidemiology 2007;18:805–35.
20. Wang YR, Pauly MV, Lin YA. Impact of Maine's Medicaid drug formulary change on non-Medicaid markets: spillover effects of a restrictive drug formulary. Am J Manag Care 2003;9:686–96.
21. Dubay L, Kenney G. Expanding public health insurance to parents: effects on children's coverage under Medicaid. Health Serv Res 2003;38:1283–301.
22. Baicker K, Robbins JA. Medicare payments and system-level health-care use: the spillover effects of Medicare managed care. Am J Health Econ 2015;1:399–431.
23. Glied S, Zivin JG. How do doctors behave when some (but not all) of their patients are in managed care? J Health Econ 2002;21:337–53.
24. Guttmacher Institute. Targeted regulation of abortion providers. 2016. Available at: https://www.guttmacher.org/state-policy/explore/targeted-regulation-abortion-providers
. Retrieved October 20, 2016.
25. Charo RA. Whole women's victory—or not? N Engl J Med 2016;375:809–11.
26. Salganicoff A, Rosenzweig C, Sobel L. The Hyde Amendment and coverage for abortion services. Available at: http://kff.org/womens-health-policy/perspective/the-hyde-amendment-and-coverage-for-abortion-services/
. Retrieved October 25, 2016.