Medicaid expansion and infant mortality: the (questionable) impact of the Affordable Care Act

Abstract

Background

Many states expanded Medicaid eligibility under the Patient Protection and Affordable Care Act (PPACA). Medicaid expansion might impact infant mortality through improved maternal health prior to pregnancy and reduced insurance churn. Some studies suggest the PPACA had no significant impact on low birth weight or preterm birth, while others suggest that the PPACA led to a significant decrease in infant mortality.

Methods

Using a difference-in-differences estimator with fixed effects to control for differences in state characteristics and time trends we analyse three samples of births from the CDC’s linked birth/death files from 2011 to 2017 to estimate the impact of Medicaid expansion on infant mortality.

Results

We find mixed results. In our full sample, we find no statistically significant change in infant mortality associated with PPACA Medicaid expansion. However, when we restrict the sample to states who had adopted the 2003 birth certificate form and when we further exclude states with a Medicaid waiver, in both samples we see reductions in infant mortality for babies born to mothers of all races. When we stratify by race, we find infant mortality decreased for babies born to white mothers. However, this decrease is not seen for babies born to black mothers.

Conclusions

Medicaid expansion under the PPACA has an impact on infant mortality, but the results are sensitive to the sample of states included in the study. There is suggestive evidence that Medicaid expansion is not closing the infant mortality gap between black and white babies.

INTRODUCTION

The 2018 infant mortality rate in the USA of 5.9 deaths per 1000 live births was higher than many other developed countries. Among OECD countries, the USA was 33rd out of 36 (with only Mexico, Turkey and Chile having a higher rate).¹ Furthermore, there are large racial disparities in infant mortality in the USA. According to the Centers for Disease Control (CDC), babies born to African American mothers in 2017 in the USA were 2.3 times more likely to die as infants than babies born to white non-Hispanic mothers. The high infant mortality rate in the USA, as well as racial inequalities in infant mortality rates and infant health is of concern.

The relationship between various interventions and infant mortality has been extensively explored. Medical professionals and policy makers have examined the impact of interventions on infant mortality. In both the public health and economics literature, researchers have studied the relationship between insurance, specifically Medicaid, and infant mortality.^{2 3} Recently, Medicaid expanded in some states under the Patient Protection and Affordable Care Act (PPACA). Prior to the expansion, federal rules only required states to provide Medicaid to individuals who earned below a certain income and who were in certain categorical groups, for example, individuals with disabilities. With Medicaid expansion, states had the option with the help of a subsidy from the federal government, to increase Medicaid eligibility to include anyone earning less than 138% of the Federal Poverty Line (FPL). Researchers have attempted to examine the relationship between PPACA-related Medicaid expansion and infant mortality and perinatal health outcomes such as low birth weight and preterm birth.^4–7 One might expect that more generous eligibility requirements for publicly provided health insurance (Medicaid) may lead to a decrease in infant mortality as well as fewer poor perinatal health outcomes. Expansion of Medicaid would lead more women who were previously uninsured or underinsured to have affordable access to care before pregnancy. This may lead to better health for both mother and baby, as a mother may start her pregnancy with fewer or better controlled underlying conditions. Medicaid expansion reduces ‘insurance churn’ among mothers.⁸ Uncertainty about what is covered by insurance, which is exacerbated by switching insurance, may cause women to postpone care.

Bhatt and Beck-Sagué examine the difference in mean infant mortality rates before and after Medicaid expansion overall and by racial and ethnic subgroups. While they measure differences between groups and across time, their analysis does not determine the statistical significance of these differences.⁴ In a letter of response to Bhatt and Beck-Sagué, Mallinson et al raise two primary concerns: (1) variation in timing of Medicaid expansion and (2) time trends in infant mortality.⁵

Brown et al use a difference-in-differences estimator to examine the impact of Medicaid expansion on low birth weight and preterm birth, both of which can be precursors to infant mortality.⁶ They find no statistically significant impact of Medicaid expansion on low birth weight or preterm birth overall, but find a decrease in incidence in low birth weight and preterm births for black infants in states that expanded Medicaid, and therefore a decrease in perinatal health disparities for black babies.

METHODS

This paper revisits the impact of PPACA Medicaid expansion on infant mortality. Our study design and methodology address both of Mallinson et al’s concerns and employ a methodology similar to Brown et al. To address variation in timing of Medicaid expansion, we restrict our sample to states and the District of Columbia that expanded Medicaid eligibility on January 1, 2014 and compare them to states that had not expanded Medicaid as of May 1, 2019. In our difference-in-differences analysis, we include time trends to account for national changes in infant mortality. We also control for time-invariant state characteristics. Our strategy allows us to examine any impact of the PPACA Medicaid expansion on infant mortality separately from national trends or any differences that arise between states that are constant across time. Because of the large racial differences in infant mortality rates, we also study the impact of Medicaid expansion on infant mortality separately for babies born to black mothers and babies born to white mothers.

Study population

We combine CDC linked birth/infant death annual data from 2011 to 2017 with Medicaid expansion data by state-year.⁹ The linked birth/infant death records include all live births for the 50 states, the District of Columbia, Puerto Rico and Guam. Our study period begins in 2011 to include mothers who conceived after the beginning of the PPACA in March 2010 and to avoid the implementation of earlier PPACA policy changes such as the young adult provision and increased access to birth control.

We consider three samples. In the first sample, the treatment group is 24 states and the District of Columbia that expanded Medicaid on January 1, 2014. The control group is the 12 states yet to expand Medicaid as of May 1, 2019. The state groups are in table 1. States that expanded Medicaid after January 2014, Puerto Rico and Guam are excluded from our sample.

Table 1.

State groupings

Expansion states	Non-expansion states
Arizona†	Alabama†
Arkansas†	Florida
California	Georgia
Colorado	Kansas
Connecticut†	North Carolina
Delaware	Oklahoma
District of Columbia	South Carolina
Hawaii†	South Dakota
Illinois	Tennessee
Iowa	Texas
Kentucky	Wisconsin
Maryland	Wyoming
Massachusetts
Minnesota
Nevada
New Jersey†
New Mexico
New York
North Dakota
Ohio
Oregon
Rhode Island†
Vermont
Washington
West Virginia†

^†Indicates states that had not adopted the 2003 birth certificate form prior to the beginning of our study.

A new form for US birth certificates was adopted in 2003; however, it took until mid-2015 for all 50 states and territories to adopt this form.^{10 11} This new form slightly modified questions relating to race and education. Our second sample is states who had adopted the new form by January 1, 2011.¹² These states uniformly use the new form after 2011. Table 1 denotes states excluded from this second sample with asterisks. Our third sample excludes states that had a Section 1115 waiver for Medicaid expansion (Iowa and Kentucky).

In our analysis, we control for mother’s level of education, race and marital status. We use the variable ‘Mother’s Bridged Race Categories’ for the years 2011–2016 and ‘Mother’s Single Race Groups’ for 2017, due to a change in race reporting. We use the Stata command ‘expand’ to turn the demographic-group-state-year level of observation into a sample in which an observation is a birth, and associated demographic characteristics of the mother, which occurred in a treated or control state between 2011 and 2017.

Summary statistics for expansion and non-expansion states for each sample are in table 2. In table 3, we stratify summary statistics by race for our third and preferred sample.

Table 2.

Summary statistics for expansion and non-expansion states for the three samples

	(1) Full non-exp Mean	(2) Full expansion Mean	(3) Birth cert. non-exp Mean	(4) Birth cert. expansion Mean	(5) Birth cert. and waiver non-exp Mean	(6) Birth cert. and waiver exp Mean
Infant mortality rate (per 1000 live births)	6.23	5.27	6.21	5.30	6.20	5.22
8th grade or less	0.01	0.01	0.02	0.01	0.02	0.02
9–12 grade: no diploma	0.12	0.09	0.14	0.10	0.14	0.10
High school grad (GED)	0.29	0.28	0.33	0.31	0.33	0.30
Some college no degree	0.20	0.20	0.23	0.23	0.23	0.23
Associates degree	0.03	0.03	0.03	0.03	0.03	0.03
Bachelor’s degree	0.16	0.21	0.19	0.22	0.18	0.22
Master’s degree	0.04	0.07	0.04	0.08	0.04	0.08
Doctorate	0.00	0.00	0.00	0.00	0.00	0.00
Education excluded	0.16	0.09	0.02	0.00	0.02	0.00
Education unknown	0.00	0.01	0.00	0.02	0.00	0.02
Age of mother 15–19	0.08	0.05	0.08	0.05	0.08	0.05
Age of mother 20–24	0.28	0.23	0.29	0.23	0.29	0.22
Age of mother 25–29	0.31	0.30	0.31	0.30	0.31	0.30
Age of mother 30–34	0.25	0.32	0.25	0.31	0.25	0.32
Age of mother 35–39	0.08	0.10	0.07	0.11	0.07	0.11
Age of mother 40–44	0.00	0.00	0.00	0.00	0.00	0.00
Married	0.55	0.59	0.54	0.59	0.54	0.59
Unmarried	0.45	0.41	0.46	0.41	0.46	0.41
American Indian or Alaska Native	0.00	0.00	0.00	0.00	0.00	0.00
Black	0.16	0.10	0.16	0.10	0.16	0.10
White	0.82	0.86	0.83	0.86	0.82	0.86
Asian or Pacific Islander	0.02	0.04	0.02	0.04	0.02	0.04
Observations	8 905 591	3 318 206	7 550 372	2 724 464	7 464 931	2 615 357

Table 3.

Summary statistics for maternal characteristics for live births to white and black mothers (2011–2017) 2003 birth certificate states, excluding states with waivers

	(1) Black non-expansion Mean	(2) White non-expansion Mean	(3) Black expansion Mean	(4) White expansion Mean
Infant mortality rate (per 1000 live births)	11.545	5.407	11.589	4.668
Expansion state	0.000	0.000	0.534	0.524
8th grade or less	0.000	0.015	0.000	0.025
9–12 grade: no diploma	0.157	0.130	0.212	0.117
High school grad (GED)	0.460	0.323	0.430	0.290
Some college no degree	0.303	0.222	0.339	0.218
Associates degree	0.005	0.032	0.001	0.038
Bachelor’s degree	0.035	0.209	0.016	0.212
Master’s degree	0.001	0.035	0.001	0.083
Doctorate	0.000	0.000	0.000	0.001
Education excluded	0.037	0.033	0.000	0.000
Education unknown	0.001	0.000	0.001	0.016
Age of mother 15–19	0.124	0.076	0.132	0.058
Age of mother 20–24	0.439	0.272	0.449	0.217
Age of mother 25–29	0.284	0.323	0.282	0.302
Age of mother 30–34	0.137	0.255	0.130	0.308
Age of mother 35–39	0.016	0.072	0.006	0.109
Age of mother 40–44	0.000	0.002	0.000	0.005
Married	0.126	0.614	0.020	0.634
Unmarried	0.874	0.386	0.980	0.366
Observations	955 684	4 128 670	506 697	4 266 566

The level of observation is a birth by a woman living in a state which expanded Medicaid on January 1, 2014 or in a state which had not expanded Medicaid by May 2019. In column (1), 12.4% of babies were born to black mothers aged 15–19 in non-expansion states.

Statistical analysis

We estimate the following equations using STATA SE version 15 (StataCorp)

$$\begin{gathered} InfantMortalityRat{e}_{s,y,d}={\beta }_0+{\beta }_{1}Treate{d}_{s,y}+{\gamma }_s+{\gamma }_y \\ +X_{s,y,d}\delta +{\epsilon }_{s,y,d} \end{gathered}$$ (1)

where Treated is an indicator variable equal to 1 if the mother gave birth after the implementation of PPACA Medicaid expansion (January 1, 2014) in a state that expanded Medicaid. X is a set of maternal demographic controls including maternal race, level of education, age category and marital status. These factors are correlated with birth outcomes.^13–16 We include them to separately identify the impact of Medicaid expansion from these characteristics. γ_s and γ_t are state-fixed and time-fixed effects, respectively. State-fixed effects capture all details specific to the state of birth which are time invariant, for example, healthcare infrastructure, differences in racial composition, state-specific policies which might influence birth outcomes, take up of the young adult provision and any other unobservable differences between states. The time-fixed effects to control for national time trends which might impact infant mortality like access to birth control, changing attitudes towards motherhood and the decision to delay fertility because of a challenging economic environment. If there are comprehensive, country-wide efforts to reduce infant mortality which are successful, time-fixed effects allow us to separately identify any impact of the ACA’s Medicaid expansion on infant mortality from a year by year decline in national infant mortality rates. We cluster SEs at the state level to reflect that states across time are not independent observations.

Our coefficient of interest, β₁, determines if there is a statistically significant change in infant mortality in states that expanded Medicaid compared with states that did not expand Medicaid.

To examine differential impacts by race, we estimate the following equation to determine if there is a difference in impact of expansion on infants born to white and black mothers.

$$\begin{gathered} InfantMortalityRat{e}_{s,y,d}={\beta }_0+{\beta }_{1}Treated*whit{e}_{s,y} \\ +{\beta }_{2}Treated*blac{k}_{s,y}+{\gamma }_s+{\gamma }_y \\ +X_{s,y,d}*\delta +{\epsilon }_{s,y,d} \end{gathered}$$ (2)

where Treated * white is an indicator variable equal to 1 if a white mother gave birth after the implementation of Medicaid expansion in a state in our sample that expanded Medicaid and Treated * black is similarly defined for black mothers. Again, γ_s and γ_t are state-fixed effects and time-fixed effects, respectively, and X is a set of maternal demographic variable defined above.

In equation 2, our coefficients of interest are β₁ and β₂, which determine if there is a statistically significant change in infant mortality for babies born to white mothers or black mothers after January 1, 2014 in states that expanded Medicaid compared with non-expansion states. Finally, an F-test determines if the difference in changes to infant mortality between white and black mothers is statistically significant. This result suggests if Medicaid expansion is closing or exacerbating the infant mortality gap between babies born to black mothers and babies born to white mothers.

RESULTS

Column 1 of table 4 displays estimates of the coefficient β₁ from equation (1). For our full sample, we estimate a reduction in infant mortality of −0.192 per 1000 live births, but it is not statistically distinguishable from zero. In column 2, we estimate equation (2) for the full sample to separately estimate the impact of treatment on babies born to white and black mothers. Again, we find results that are not statistically distinguishable from zero. An F-test determines there is no statistically significant differential impact of Medicaid expansion on infant mortality for babies born to white and black mothers. This full sample suggests that Medicaid expansion under the ACA had no measurable impact on infant mortality for the population as a whole, for babies born to white mothers, or for babies born to black mothers.

Table 4.

Impact on infant mortality of giving birth in a medicaid expansion state after medicaid expansion

	Full sample		Birth certificate sample		Birth certificate and exclude waiver sample
	(1) All	(2) White and black	(3) All	(4) White and black	(5) All	(6) White and black
Treated	−0.192		−0.263‡		−0.271‡
	(0.116)		(0.121)		(0.122)
Treated_white		−0.205		−0.277†		−0.288†
		(0.125)		(0.139)		(0.143)
Treated_black		0.370		0.354		0.349
		(0.317)		(0.381)		(0.383)
N	12 223 793	12 223 793	10 274 832	10 274 832	10 080 284	10 080 284

^†p<0.10.

^‡p<0.05.

SEs in parentheses.

Difference-in-differences analysis including mother characteristics (race, age, and education of mother). SEs are clustered at the state level. The level of observation is a birth for a mother in our sample.

In column 3, we estimate equation (1) for the sample of states that had adopted the 2003 birth certificate by the start of the study period. We estimate a reduction of −0.263 (95% CI=−0.51 to −0.011) in the infant mortality rate for babies born to mothers of all races. Estimating equation (2) for this sample, we find that there is a reduction in the infant mortality rate of −0.277 (90% CI=−0.52 to −0.038) for babies born to white mothers in expansion states. There is no statistically significant impact of expansion for babies born to black mothers in expansion states, but the estimated coefficient is positive whereas it is negative for white mothers.

In our third sample, we include the states which had adopted the 2003 birth certificate but exclude states that had a Section 1115 waiver (Iowa and Kentucky). Table 4, column 5, estimates equation (1) for this sample. Again, we observe a reduction in the infant mortality rate of −0.271 (95% CI=−0.53 to −0.016) for mothers of all races. Column 6 estimates equation (2). Babies born to white mothers in expansion states had a reduction of −0.288 (90% CI=−0.53 to −0.042) in their infant mortality rate. However, there was not a statistically significant impact on infant mortality for babies born to black mothers in expansion states. In both the second and third sample, an F-test shows that there is no statistically significant difference between the coefficients for white and black mothers.

Our results are sensitive to the selection of sample states. In both the second (birth certificates) and third (waiver exclusion) samples, we find a statistically significant reduction in infant mortality among babies born to all women and white women in expansion states but no statistically significant difference in infant mortality rates between babies born to white mothers and babies born to black mothers in expansion states.

DISCUSSION

While the PPACA provides a compelling natural experiment, some issues need to be considered when examining the impact of Medicaid expansion on infant mortality and perinatal health outcomes. The first is selection into Medicaid expansion. The second is understanding the mechanism through which Medicaid expansion may impact infant mortality. The third is other contemporaneous policy changes that may impact insurance and healthcare access.

Bhatt and Beck-Sagué⁴ acknowledge concerns with selection into Medicaid expansion and regional heterogeneity. States that elected to expand Medicaid were states with lower infant mortality rates prior to the reform. Figure 1 illustrates a substantial difference in the level of infant mortality rates between expansion states and non-expansion states. Furthermore, many of the states that rejected Medicaid expansion are located in the Southern USA, which may have worse population health. However, our state-fixed effects help separately identify state-specific effects from Medicaid expansion.

Figure 1.

Infant mortality rates by expansion and non-expansion states: birth certificate and medicaid waiver sample.

Prior to the enactment of the PPACA, pregnant women with incomes up to 133% of the FPL were already a federally mandated group and therefore eligible for coverage at this income level in all states. While Medicaid expansion has no direct impact on insurance eligibility for low-income pregnant women during pregnancy, a potential mechanism for improved health is increased affordability of health services prior to pregnancy due to enrollment in Medicaid based solely on income. Clapp et al find that in states that expanded Medicaid, more births were covered by Medicaid after expansion. In states that did not expand Medicaid, mothers still had increased insurance coverage compared with before the PPACA. These mothers were insured through private insurance or another payer.⁷ While this suggests that there was very little difference in insurance coverage for pregnant women at the time of birth regardless of residence in an expansion or non-expansion state, there may still be large differences in out-of-pocket costs depending on insurance type.

Lower-income women have better access to health insurance prior to pregnancy in states that expanded Medicaid compared with those that did not. As a result, it is possible that women in expansion states may enter pregnancy in better health and with better controlled chronic conditions. Pregnant women may also receive prenatal care earlier if they are already insured, although existing literature has not found this to be the case.⁷ There is also less insurance churn among mothers in expansion states.⁸ This decreases the bureaucratic hurdles that a woman must face to receive care both before pregnancy and early on in pregnancy. Women are better able to maintain coverage between pregnancies in states that expanded Medicaid. Mothers who reside in non-expansion states are more likely to lose insurance coverage in the year after birth than those who are in expansion states.¹⁷

The young adult provision of the PPACA was enacted in September 2010 in all states and territories and allows adult children under the age of 26 to remain on their parents’ employer-provided insurance. This provision impacted the insurance options of some mothers before, during, and after pregnancy. It disproportionately impacts younger mothers who are from families with higher socioeconomic status.¹⁸ It also impacts fertility decisions of younger women in our sample.¹⁹ Because the implementation of this provision occurred prior to the start of our study period and state-fixed effects absorb the average young adult provision uptake by state, the young adult provision does not confound our results.

Policy changes to Medicaid eligibility for pregnant women on January 1, 2014 would confound our results. After review of Kaiser Family Foundation’s Annual Updates on Eligibility Rules, Enrollment and Renewal Procedures, and Cost-Sharing Practices in Medicaid and CHIP, from November 2013 and January 2015, the only contemporaneous policy changes occurred in Oklahoma and Virginia.^{20 21} Virginia is omitted from all samples and Oklahoma is part of the control group. Rerunning the model excluding Oklahoma does not substantively alter the results.

CONCLUSIONS

Insurance access, both before and during pregnancy, is thought to improve maternal and infant health outcomes. Prior to the PPACA, pregnant women were eligible for Medicaid at higher income levels than other groups. As such, there was limited ‘bite’ to the reform for pregnant women in expansion states. After the PPACA, low-income women in expansion states have access to health insurance regardless of their pregnancy status. This should lead to better control of underlying conditions and improved overall health, easier access to early prenatal care, and reduce uncertainty about insurance coverage as women become mothers and after they give birth.

Our results vary depending on which sample of states we use. In our full sample, we find no statistically significant impact of Medicaid expansion on infant mortality for the whole population, babies born to white women, or babies born to black women. However, when we focus on two samples of states that implemented the 2003 birth certificates and further exclude states on a Medicaid waiver, we find reductions in infant mortality for all mothers and white mothers. It is worth noting, that while not statistically significant, babies born to black mothers have an increased risk of infant mortality post-expansion in expansion states. While large SEs on our estimates for black mothers mean that we cannot say definitively that the healthcare landscape is worse for babies born to black mothers, we would be remiss not to mention that there is suggestive evidence that babies born to black and white mothers have differential risks of infant mortality. Furthermore, we find no evidence that Medicaid expansion lessens the gap in infant mortality rate between babies born to black mothers and babies born to white mothers.

In contrast to previous studies, our study suggests that changes in infant mortality rates pre-expansion and post-expansion are conditional on sample selection of Medicaid expansion states. This paper solely examines Medicaid expansion, and does not consider other provisions of the PPACA that may have impacted maternal and infant health, such as reduced-cost birth control. Our sample period starts after the implementation of many of these national policies, and thus we can isolate the impact of Medicaid expansion, while having a uniform environment for other maternal health initiatives as part of the PPACA. However, these earlier initiatives may have had impacts on maternal and infant health and should be considered in future research.

What is already known on this subject.

• Infant mortality in the USA has been declining in the past decade. Previous work examined the impact of Medicaid expansion under the Patient Protection and Affordable Care Act. Previous studies have identified a change in infant mortality rates around implementation of the PPACA. Other studies, using a difference-in-difference methodology, have found no significant change in overall incidence of low birth weight babies or preterm birth, but a decrease in racial health disparities among babies born to mothers in states that expanded Medicaid.

What this study adds.

• Controlling for differences in state characteristics and falling infant mortality rates, we use statistical analysis to determine the impact of the Patient Protection and Affordable Care Act’s Medicaid expansion on infant mortality and find no statistically significant change in infant mortality rates when we consider the full sample of expansion and non-expansion states. In our full sample, differences in infant mortality rates are predicted by maternal characteristics as well as state and year effects, not expansion status. Our results depend on the sample of states. When we restrict our sample to states with consistent birth certificate reporting and Medicaid waivers, we find a decrease in infant mortality for babies born to mothers of all races, but larger reductions for babies born to white women. This decrease is not seen for babies born to black mothers and there is no statistically significant difference between races. This is suggestive evidence that Medicaid expansion is not closing the infant mortality gap between babies born to black mothers and babies born to white mothers.

Data availability statement

Data are available in a public, open access repository.