Impacts of Medicaid Expansion Before Conception on Prepregnancy Health, Pregnancy Health, and Outcomes

Abstract

Preconception health care is heralded as an essential method of improving pregnancy health and outcomes. However, access to health care for low-income US women of reproductive age has been limited because of a lack of health insurance. Expansions of Medicaid program eligibility under the Affordable Care Act (as well as prior expansions in some states) have changed this circumstance and expanded health insurance coverage for low-income women. These Medicaid expansions provide an opportunity to assess whether obtaining health insurance coverage improves prepregnancy and pregnancy health and reduces prevalence of adverse pregnancy outcomes. We tested this hypothesis using vital statistics data from 2011–2017 on singleton births to female US residents aged 15–44 years. We examined associations between preconception exposure to Medicaid expansion and measures of prepregnancy health, pregnancy health, and pregnancy outcomes using a difference-in-differences empirical approach. Increased Medicaid eligibility was not associated with improvements in prepregnancy or pregnancy health measures and did not reduce the prevalence of adverse birth outcomes (e.g., prevalence of preterm birth increased by 0.1 percentage point (95% confidence interval: −0.2, 0.3)). Increasing Medicaid eligibility alone may be insufficient to improve prepregnancy or pregnancy health and birth outcomes. Preconception programming in combination with attention to other structural determinants of pregnancy health is needed.

Abbreviation

ACA

Affordable Care Act

DID

difference-in-differences

NH

non-Hispanic

Editor’s note: An invited commentary on this article appears on page 1499, and the authors’ response appears on page 1502.

Over the past several decades, public health initiatives have sought to reduce rates of adverse pregnancy outcomes such as low birth weight and preterm delivery by improving women’s access to prenatal care (1). In the late 1980s and early 1990s, the federal government expanded the Medicaid program to cover prenatal and delivery care for low-income pregnant women (2). However, these previous expansions of Medicaid did not significantly reduce rates of adverse pregnancy outcomes (3, 4), and the United States’s high rates of adverse pregnancy outcomes, as well as stark racial and socioeconomic disparities in these outcomes, have persisted (5, 6).

This lack of progress arguably stems from the fact that prenatal care has limited clinical scope to affect pregnancy and infant health (7, 8) and is, in many cases, “too little, too late.” Efforts to mitigate risks from smoking, alcohol consumption, poor nutrition, obesity, and other chronic disease through prenatal care (which typically starts at the 6th–8th week of pregnancy) probably miss the early critical time window during which essential fetal development occurs (9, 10). Treatments given during pregnancy, such as nicotine replacement therapy for smoking cessation, antihypertensive medication for high blood pressure, and behavioral interventions for weight loss either have not been shown to be effective during pregnancy (11, 12) or may confer risks to the developing fetus (9, 13–15) and should thus begin months or years prior to conception to avoid these risks and be most effective.

Organizations (9, 10, 16) and scholars (1, 17, 18) alike now recommend identifying, managing, and treating risk factors prior to pregnancy at well-woman visits, preconception visits, and interconception visits. However, a significant barrier to the success of this strategy is the lack of health insurance among low-income women of reproductive age, particularly outside of the perinatal period (19). Prior to enactment of the Affordable Care Act (ACA) in 2010, about one-quarter of US women lacked health insurance 9 months prior to giving birth (19). At least half of these women went on to acquire Medicaid coverage during pregnancy (due to higher eligibility levels for pregnant women), but lack of preconception and interconception health-care coverage remained a problem for women of reproductive age (19).

The ACA expanded Medicaid eligibility to all nonelderly Americans with incomes less than 138% of the federal poverty level by January 1, 2014. However, a 2012 US Supreme Court ruling allowed states to opt out of Medicaid expansion, and from January 1, 2014, to the present, only 37 states plus the District of Columbia expanded Medicaid coverage (20). Prior to this expansion, eligibility for Medicaid also varied widely across states and time (21).

Our own work and that of others has shown that the ACA Medicaid expansions, as well as prior state-specific increases in Medicaid eligibility, increased health insurance coverage and access to preventive care for low-income women of reproductive age (22–24). The ACA expansions have also been linked to increased preconception enrollment in Medicaid and utilization of preventive care services (25, 26). Our objective in the current study was to test the hypothesis that increasing Medicaid eligibility for low-income, nonpregnant women of reproductive age improved indicators of prepregnancy and pregnancy health and reduced the prevalence of adverse pregnancy outcomes.

METHODS

Data and sample

We used deidentified vital statistics birth certificate data from the National Center for Health Statistics from the period 2011–2017, which included all live births in the United States (n = 27,670,554). Our analytical sample included singleton births to female US residents aged 18–44 years (see Web Figure 1, available online at https://doi.org/10.1093/aje/kwaa289). We excluded records that were missing information on length of gestation or birth weight and those that had implausible combinations of birth weight and gestational age (27). We limited our analysis to births for which conception occurred before March 26, 2017, in order to observe the full duration of pregnancy (approximately 40 weeks). In addition, as of 2011, 14 US states (see Web Table 1 for details) were still using the “unrevised” (1989) version of the US Standard Certificate of Live Birth (instead of the revised 2003 version). For these states, data on education were completely missing because the National Center for Health Statistics did not release unrevised education information after 2010, and data on key outcome variables such as prepregnancy smoking and trimester-specific smoking were not collected. Thus, we limited our analysis to the 36 US states and the District of Columbia that were using the revised birth certificate by 2011; this produced an analytical sample size of 20,783,845. (Results were similar when we used all states; see Web Table 2.)

Outcomes

We selected outcomes related to prepregnancy health based on the Centers for Disease Control and Prevention’s recommended areas of focus (28): cigarette smoking during the 3 months prior to pregnancy, prepregnancy body mass index (weight (kg)/height (m)²; <18.5, 18.5–24.9, 25.0–29.9, or ≥30.0), prepregnancy diabetes, and prepregnancy hypertension. If increased Medicaid eligibility improved these prepregnancy factors, we would also expect correlated factors during pregnancy to improve; thus, we also examined cigarette smoking in the first trimester, any cigarette smoking during pregnancy, gestational diabetes, gestational hypertension, and eclampsia. We also examined whether expanded Medicaid eligibility increased the level of prenatal care received in the first trimester, as has been suggested in prior work (29). Finally, we measured pregnancy outcomes hypothesized to be improved by increased attention to preconception health: preterm birth (<37 weeks’ gestation), early preterm birth (<34 weeks’ gestation), small size for gestational age (<10th percentile of birth weight for gestational age vs. 10th–90th percentiles), and large size for gestational age (>90th percentile of birth weight for gestational age vs. 10th–90th percentiles) (30). Gestational age was based on the National Center for Health Statistics obstetrical estimate of gestational age. Observations with missing data on outcomes were excluded from analyses of those outcomes only; no outcome was missing more than 5% of observations (see Table 1). We conducted a robustness check using a sample which excluded all observations that were missing data for any outcome.

Table 1.

Characteristics (%^a) of a Sample of Singleton US Births, Overall and by Preconception Exposure to Expansion of Medicaid Program Eligibility Under the Affordable Care Act (n = 20,783,845), 2011–2017

		Preconception Exposure to Medicaid Expansion
Variable	Total   (2010–2016)   (n = 20,783,845)	≥1 Year of Exposure   (n = 5,954,468)	<1 Year of Exposure or No Exposure   (n = 14,829,377)
Maternal Demographic Characteristics and Outcomes
Race/ethnicity
NH White	54.3	53.8	54.5
NH Black	14.6	15.0	14.5
Hispanic	23.8	23.8	23.8
NH American Indian/Alaska Native	0.8	0.8	0.8
NH Asian/Pacific Islander	6.5	6.6	6.4
Age, years
<20	27.9	25.2	29.0
20–29	29.6	29.9	29.5
30–39	39.7	42.1	38.8
≥40	2.8	2.9	2.7
Marital status
Married	60.4	60.8	60.2
Unmarried	39.6	39.2	39.8
Education
No high school diploma	14.3	12.8	14.9
High school diploma or some college	54.9	54.8	54.9
Bachelor’s degree or more	29.8	31.3	29.1
Missing data	1.1	1.1	1.1
Parity
No previous live birth	38.4	37.7	38.7
Any previous live birth	61.6	62.3	61.4
Prepregnancy health
Cigarette smoking before pregnancy
Yes	10.3	9.4	10.7
No	87.0	90.2	85.7
Missing data	2.7	0.4	3.6
Prepregnancy BMIb
Underweight (<18.5)	3.6	3.4	3.6
Normal-weight (18.5–24.9)	44.1	42.9	44.6
Overweight (25.0–29.9)	24.9	25.4	24.7
Obese (≥30.0)	24.2	25.8	23.6
Missing data	3.2	2.6	3.4
Prepregnancy diabetes
Yes	0.8	0.9	0.8
No	99.1	99.0	99.1
Missing data	0.2	0.1	0.2
Prepregnancy hypertension
Yes	1.6	1.8	1.5
No	98.3	98.2	98.3
Missing data	0.2	0.1	0.2
Pregnancy health
Prenatal care
First trimester	73.8	75.0	73.3
Late or none	23.4	22.4	23.8
Missing data	2.8	2.7	2.9
Cigarette smoking in first trimester
Yes	7.7	7.0	8.0
No	89.6	92.6	88.4
Missing data	2.7	0.4	3.6
Any cigarette smoking during pregnancy
Yes	7.9	7.2	8.2
No	89.4	92.4	88.2
Missing data	2.7	0.4	3.6
Gestational diabetes
Yes	5.5	6.1	5.2
No	94.4	93.9	94.6
Missing data	0.2	0.1	0.2
Gestational hypertension
Yes	5.1	6.1	4.8
No	94.7	93.9	95.1
Missing data	0.2	0.1	0.2
Eclampsia
Yes	0.2	0.2	0.2
No	99.7	99.7	99.6
Missing data	0.2	0.1	0.2
Pregnancy Outcomes
Gestational age at delivery, weeks
<34 (early preterm)	2.3	2.4	2.3
<37 (preterm)	9.3	9.6	9.2
≥37 (term)	90.7	90.5	90.8
Birth weight for gestational age
SGA	10.6	10.7	10.6
Normal	78.3	78.4	78.3
LGA	11.0	10.9	11.1

Abbreviations: BMI, body mass index; LGA, large for gestational age; NH, non-Hispanic; SGA, small for gestational age.

^a Percentages may not sum to 100 due to rounding.

^b Weight (kg)/height (m)².

Medicaid expansion under the ACA

Our first approach to measuring expanded Medicaid eligibility was to examine the effect of ACA Medicaid expansions that occurred in 2014 and later years—an approach which shed light on the impact of this specific policy on prepregnancy and pregnancy health and allowed us to compare our findings with previous literature (23–25, 31).

Births were assigned to expansion status if the mother’s state of residence expanded Medicaid eligibility under the ACA at any time in 2014, 2015, or 2016 (expansion = 1) or did not expand Medicaid eligibility (expansion = 0). (Web Table 1 lists states’ expansion statuses and dates.) Births were also assigned a value for postexpansion versus preexpansion: postexpansion = 1 if the estimated date of conception (birth date minus gestational age) was ≥12 months (≥365 days) after January 1, 2014, or the date on which the mother’s state of residence expanded Medicaid eligibility (for states expanding it after January 1, 2014), and postexpansion = 0 if the conception date was <12 months after Medicaid expansion or January 1, 2014.

There exists little empirical evidence with which to predict the “lag time” between obtaining health insurance coverage, utilizing health care, and improvements in prepregnancy or pregnancy health. Whereas contraceptive use to time pregnancies or treatment with antihypertensive agents to lower blood pressure (32) could be effective within a month, achieving a healthy body mass index or smoking cessation may take longer. Randomized trials of preconception lifestyle interventions suggested that smoking cessation advice given at a preconception consultation could result in moderate declines in smoking after 3 months (33) and that consultation on both smoking cessation and weight reduction among women seeking fertility treatments could lead to cessation in 30% of women and weight loss in 50% after 3–12 months (34). Accordingly, we hypothesized that beneficial effects of becoming eligible for Medicaid prior to conception would not manifest in improved prepregnancy or pregnancy health for at least 1 year. Because this 1-year lag period has never been tested, we also examined 3 other exposure periods: any preconception exposure to Medicaid expansion, exposure ≥6 months before conception, and exposure ≥24 months before conception (Web Table 3); findings did not differ substantially.

Simulated Medicaid eligibility

Our second approach took advantage of variation in Medicaid eligibility levels across states and times and other factors such as parental status both prior to and after the ACA Medicaid expansion (21). This approach assigned each birth in our data set a probability that the mother was eligible for Medicaid prior to pregnancy and then examined the association between that probability and the outcomes of interest. This approach also allowed us to estimate what would happen if, counter to fact, all women were eligible for Medicaid prior to pregnancy. This approach has been widely used and is generally accepted as valid (3, 35).

To create the simulated Medicaid eligibility measure, we selected a sample of noninstitutionalized reproductive-age (18–44 years) women who were not in the Armed Forces from the American Community Survey for the years 2008–2012 (prior to the ACA Medicaid expansion). For each woman in this sample from 2008–2012, we used her household income and composition to calculate her predicted eligibility for Medicaid in each year from 2010 to 2016 (i.e., preconception years for births occurring in 2011–2017) using Medicaid eligibility regulations for that year in her state. Then, using this sample, we defined demographic subgroups using the combination of state of residence, race/ethnicity (non-Hispanic (NH) White, NH Black, Hispanic, NH American Indian/Alaska Native, or NH Asian/Pacific Islander), age (18–44 years), marital status (married, unmarried), dependent children (yes, no), and educational attainment (less than high school, high school diploma or more); this resulted in 6,528 combinations of the variables race/ethnicity, age, marital status, dependent children, education, and state (50 states plus the District of Columbia). (Because some states did not include all possible combinations of demographic groups, we had 6,186 actual groups.) All of the information necessary to define these groups was also available on the birth certificates. Next, mean Medicaid eligibility within each demographic subgroup was calculated and measured the “simulated” probability that a woman in a given subgroup would be eligible for Medicaid in a given year from 2010 to 2016. These “simulated” probabilities were merged into the natality data by demographic subgroup and year (parity = 0 was used to approximate lack of dependent children).

The use of a constant sample of women from years prior to the Medicaid expansion to construct this measure of simulated Medicaid eligibility ensured that changes in Medicaid would not stem from changes in household income brought forth by the Medicaid expansions or from changes in sample composition or characteristics. In Web Table 4, we demonstrate that simulated eligibility was positively associated with Medicaid coverage and negatively associated with lack of health insurance among all women aged 18–44 years in the American Community Survey (2008–2017) and across subgroups defined by marital status and parity.

Statistical analyses

We calculated the unadjusted mean value (standard deviation) and frequency (%) for all outcomes, overall and by postexpansion exposure of ≥1 year versus <1 year. We then fitted a series of multivariable linear regression models (i.e., linear probability models for dichotomous outcomes) with robust standard errors clustered by state, with 1 model for each outcome. For the analysis that used a dichotomous indicator of ACA Medicaid expansion, we used a difference-in-differences (DID) model as follows:

(1)

where represents the outcome of interest for woman i in state s and year t. Our parameter of interest is , which represents the DID or percentage-point change in the outcome among births to mothers living in Medicaid expansion states compared with mothers living in nonexpansion states, where the child was conceived ≥12 months after expansion. and are vectors of conception year and state fixed effects; is a vector of individual covariates (age, race/ethnicity, parity, marital status, and educational attainment).

For the “simulated” eligibility analysis, the model was as follows:

(2)

where everything is defined as in equation 1 except for the “simulated” eligibility measure. This is the “simulated” probability that a woman in demographic group k in state s and year t is eligible for Medicaid. Our parameter of interest is , which represents the change in for each 1-unit change in simulated eligibility (i.e., a change from 0 to 1). Models with simulated eligibility as the exposure variable also included demographic subgroup × state and demographic subgroup × year fixed effects.

For the ACA expansion analysis (equation 1), we conducted separate analyses among women most likely to go from uninsured to eligible for Medicaid under the ACA expansions: women without a high school diploma, women without a college degree, and unmarried women. Within each subgroup, we also conducted an analysis including only women without a previous live birth, as childless adults were the target of the ACA expansions. In supplemental analyses using data form the American Community Survey and a similar DID approach as above, we verified that these groups did in fact experience substantially larger decreases in the uninsured rate than other groups of women (see Web Appendix and Web Table 5).

To assess the validity of the research design, we tested the parallel trend assumption underlying the DID approach, that is, that the trend in outcomes prior to expansion would be the same in expansion and nonexpansion states. We assessed this assumption for the DID analysis both graphically and empirically; see Web Figure 2 and Web Table 6 for details. To assess the robustness of our findings, we also conducted the following analyses: 1) analyses moving states that had fully expanded Medicaid to parents and childless adults prior to 2014 (where we would expect to see smaller impacts of the ACA expansions) to the nonexpansion control group; 2) complete-case analyses excluding observations that were missing data for any outcome. We performed a Holm-Bonferroni correction for multiple testing for all findings that were statistically significant at the α = 0.05 level.

RESULTS

Table 1 shows descriptive statistics for the analytical sample overall and by preconception exposure to the ACA Medicaid expansion. Approximately 54% of births in the study sample were of NH White infants, 15% were NH Black, 24% were Hispanic, and the remainder were NH American Indian/Alaska Native or NH Asian/Pacific Islander. Fourteen percent of births were to women with less than a high school education, and 55% were to women with a high school diploma or some college. About 9% of births were preterm, and 10.6% and 11.0% of babies were small for gestational age and large for gestational age, respectively.

Table 2 presents results from unadjusted and adjusted DID models examining associations between preconception exposure to the ACA Medicaid expansions of ≥1 year versus <1 year or no exposure and all prepregnancy health, pregnancy health, and pregnancy outcome measures. The β coefficients in these models can be interpreted as the percentage-point change (where β is multiplied by 100) in the outcome associated with at least 1 year of exposure to Medicaid expansion prior to conception. Overall, there was no evidence that a year of preconception exposure to expanded Medicaid coverage was associated with prepregnancy health, pregnancy health, or pregnancy outcomes. Estimates were also virtually the same with or without adjustment, which suggested a valid research design, as did most of the tests of the parallel trend assumption in Web Table 6 and Web Figure 2.

Table 2.

Association Between ≥1 Year of Preconception Exposure to the Affordable Care Act Medicaid Expansion and Pregnancy Outcomes (Linear Regression Models) Among Singleton US Births (n = 20,783,845), 2011–2017

Outcome	Minimally Adjusted Resultsa		Fully Adjusted Resultsb
	β Coefficientc	95% CI	β Coefficient	95% CI
Prepregnancy health
Any cigarette smoking before pregnancy vs. none	0.4	−0.4, 1.2	0.4	−0.4, 1.3
Prepregnancy BMId
≥25.0 vs. BMI <25.0	−0.1	−0.7, 0.4	0.0	−0.5, 0.5
≤18.0 vs. BMI >18.0	0.1	0.0, 0.2	0.1	0.0, 0.2
Prepregnancy diabetes	0.0	0.0, 0.1	0.0	0.0, 0.1
Prepregnancy hypertension	0.1	−0.1, 0.2	0.1	−0.1, 0.2
Pregnancy health
Prenatal care received in first trimester vs. late or never	0.5	−0.6, 1.7	0.6	−0.6, 1.7
Any cigarette smoking in first trimester vs. none	0.3	−0.4, 0.9	0.3	−0.4, 0.9
Any cigarette smoking during pregnancy vs. none	0.3	−0.4, 0.9	0.3	−0.4, 1.0
Gestational diabetese	0.4	0.0, 0.7	0.4	0.0, 0.8
Gestational hypertension	0.5	0.0, 0.9	0.5	0.0, 0.9
Eclampsia	0.5	−0.6, 1.7	0.0	−0.1, 0.1
Pregnancy outcome
Preterm delivery (<37 weeks) vs. ≥37 weeks	0.1	−0.2, 0.3	0.1	−0.2, 0.3
Early preterm delivery (<34 weeks) vs. ≥34 weeks	0.0	0.0, 0.1	0.0	0.0, 0.1
Birth weight for gestational agef
SGA vs. normal	0.1	−0.2, 0.3	0.1	−0.2, 0.3
LGA vs. normale	−0.2	−0.4, 0.0	−0.2	−0.4, 0.0

Abbreviations: BMI, body mass index; CI, confidence interval; LGA, large for gestational age; SGA, small for gestational age.

^a Models included fixed effects for state of maternal residence and year of conception only. Robust standard errors were used to account for clustering within state.

^b Models included fixed effects for state of maternal residence, year of conception, race/ethnicity, age, parity, marital status, and the interaction between expansion and exposure status. Robust standard errors were used to account for clustering within state.

^c All β coefficients in the table have been multiplied by 100 to aid interpretation.

^d Weight (kg)/height (m)².

^e There was evidence that the parallel trend assumption was not met (see Web Table 6).

^f For the SGA model, LGA births were not included. For the LGA model, SGA births were not included.

Table 3 shows estimates of adjusted DID associations between preconception exposure to the ACA Medicaid expansions of ≥1 year versus <1 year or no exposure and prepregnancy health, pregnancy health, and pregnancy outcome measures among subgroups most likely to become newly eligible for Medicaid prior to conception under the ACA expansions. A year of preconception exposure to Medicaid was associated with percentage-point increases ranging from 0.2 to 0.9 in births complicated by gestational diabetes and gestational hypertension among women without a high school education and unmarried women. While these estimates indicate a relatively large increase in these rare outcomes (Web Table 7), after adjustment for multiple testing, the 95% confidence intervals around these estimates included the null value (Table 3).

Table 3.

Association Between ≥1 Year of Preconception Exposure to the Affordable Care Act Medicaid Expansion and Pregnancy Outcomes (Adjusted^a Linear Regression Models) Among Singleton US Births to Women Most Likely to Benefit From the Expansion Prior to Conception, 2011–2017

Outcome	No High School Diploma (n = 2,967,957)		No High School Diploma and No Previous Children (n = 725,694)		High School Graduate or Some College (n = 11,403,490)		High School Graduate or Some College and No Previous Children (n = 4,350,047)		Unmarried (n = 8,235,960)		Unmarried and No Previous Children (n = 3,513,353)
	βb	95% CI	β	95% CI	β	95% CI	β	95% CI	β	95% CI	β	95% CI
Prepregnancy health
Any cigarette smoking before pregnancy vs. none	0.4	−0.8, 1.6	0.4	−1.1, 1.8	0.4	−0.4, 1.2	0.5	−0.5, 1.6	0.5	−0.7, 1.6	0.5	−0.8, 1.8
Prepregnancy BMIc
≥25.0 vs. BMI <25.0	0.6	0.1, 1.2	0.8	−0.4, 2.0	−0.1	−0.7, 0.5	0.1	−0.5, 0.7	0.2	−0.6, 0.9	0.2	−0.4, 0.9
≤18.0 vs. BMI >18.0	0.1	0.0, 0.3	0.4	0.0, 0.7	0.1	0.0, 0.2	0.2	0.0, 0.3	0.1	−0.1, 0.2	0.2	0.0, 0.4
Prepregnancy diabetes	0.0	−0.1, 0.1	0.1	−0.1, 0.2	0.0	0.0, 0.1	0.0	0.0, 0.1	0.0	0.0, 0.1	0.0	0.0, 0.1
Prepregnancy hypertension	0.0	−0.1, 0.2	0.0	−0.2, 0.3	0.0	−0.1, 0.2	0.1	−0.1, 0.2	0.1	−0.1, 0.2	0	−0.1, 0.2
Pregnancy health
Prenatal care received in first trimester vs. late or never	0.2	−1.7, 2.1	0.4	−1.6, 2.4	0.9	−0.2, 1.9	0.9	−0.3, 2.0	0.7	−0.7, 2.1	0.9	−0.4, 2.3
Any cigarette smoking in first trimester vs. none	0.3	−0.8, 1.4	0.3	−1.2, 1.7	0.2	−0.4, 0.9	0.2	−0.6, 1.1	0.3	−0.7, 1.3	0.2	−0.9, 1.4
Any cigarette smoking during pregnancy vs. none	0.3	−0.8, 1.4	0.2	−1.2, 1.7	0.3	−0.4, 0.9	0.3	−0.6, 1.1	0.3	−0.7, 1.4	0.3	−0.9, 1.4
Gestational diabetes	0.8d	0.3, 1.3	0.6d	0.1, 1.2	0.2	−0.1, 0.6	0.3	−0.1, 0.6	0.4d	0.1, 0.7	0.3	0.0, 0.6
Gestational hypertension	0.6d	0.2, 1.0	0.9d	0.3, 1.5	0.4	−0.1, 0.9	0.6	0.0, 1.3	0.6	0.0, 1.1	0.8d	0.1, 1.5
Eclampsia	−0.1	−0.2, 0.0	−0.1	−0.2, 0.1	0.0	−0.2, 0.1	−0.1	−0.2, 0.1	−0.1	−0.2, 0.1	−0.1	−0.2, 0.1
Pregnancy outcome
Preterm delivery (<37 weeks) vs. ≥37 weeks	0.0	−0.3, 0.4	0.3	−0.3, 0.9	0.1	−0.2, 0.3	0.1	−0.2, 0.4	0.1	−0.2, 0.4	0.1	−0.3, 0.4
Early preterm delivery (<34 weeks) vs. ≥34 weeks	0.0	−0.2, 0.2	0.1	−0.2, 0.4	0.1	0.0, 0.1	0.1	−0.1, 0.2	0.1	−0.1, 0.2	0.0	−0.1, 0.2
Birth weight for gestational agee
SGA vs. normal	0.1	−0.2, 0.4	0.6d	0.2, 0.9	0.1	−0.2, 0.4	0.2	−0.1, 0.5	0.0	−0.3, 0.4	0.2	−0.1, 0.5
LGA vs. normal	−0.2	−0.5, 0.2	0.1	−0.3, 0.6	−0.2	−0.5, 0.0	−0.2	−0.4, 0.0	−0.2	−0.5, 0.1	−0.1	−0.4, 0.1

Abbreviations: BMI, body mass index; CI, confidence interval; LGA, large for gestational age; SGA, small for gestational age.

^a Models included fixed effects for state of maternal residence, year of conception, race/ethnicity, age, parity, marital status, and the interaction between expansion and exposure status. Robust standard errors were used to account for clustering within state.

^b All β coefficients in the table have been multiplied by 100 to aid interpretation.

^c Weight (kg)/height (m)².

^d No longer statistically significantly different from the null value after Holm-Bonferroni correction for multiple testing (15 outcomes × 10 models = 150 tests).

^e For the SGA model, LGA births were not included. For the LGA model, SGA births were not included.

Table 4 shows the unadjusted and adjusted simulated eligibility estimates. The β coefficients in these models can be interpreted as the linear change or percentage-point change in the outcome associated with expanding Medicaid eligibility to all mothers. Estimates were consistently small and not different from the null value of 0.

Table 4.

Association Between Preconception Simulated Eligibility for Medicaid Coverage Under the Affordable Care Act and Prepregnancy and Pregnancy Outcomes (Minimally and Fully Adjusted Linear Regression Models) Among Singleton US Births (n = 20,783,845), 2011–2017

Outcome	Minimally Adjusted Resultsa		Fully Adjusted Resultsb
	β Coefficientc	95% CI	β Coefficient	95% CI
Prepregnancy health
Any cigarette smoking before pregnancy vs. none	0.0	−1.2, 1.1	0.0	−1.2, 1.1
Prepregnancy BMId
≥25.0 vs. BMI <25.0	−0.1	−1.2, 1.0	−0.1	−1.2, 1.0
≤18.0 vs. BMI >18.0	0.2	0.0, 0.5	0.2	0.0, 0.5
Prepregnancy diabetes	0.0	−0.1, 0.2	0.0	−0.1, 0.2
Prepregnancy hypertension	0.1	−0.2, 0.3	0.1	−0.2, 0.3
Pregnancy health
Prenatal care received in first trimester vs. late or never	1.5	−0.7, 3.7	1.5	−0.7, 3.7
Any cigarette smoking in first trimester vs. none	−0.2	−1.1, 0.7	−0.2	−1.1, 0.7
Any cigarette smoking during pregnancy vs. none	−0.2	−1.1, 0.7	−0.2	−1.1, 0.7
Gestational diabetes	0.2	−0.5, 0.9	0.2	−0.5, 0.9
Gestational hypertension	0.2	−0.8, 1.3	0.2	−0.8, 1.3
Eclampsia	0.0	−0.4, 0.3	0.0	−0.4, 0.3
Outcomes
Preterm delivery (<37 weeks) vs. ≥37 weeks	0.2	−0.2, 0.7	0.2	−0.2, 0.7
Early preterm delivery (<34 weeks) vs. ≥34 weeks	0.2	0.0, 0.4	0.2	0.0, 0.4
Birth weight for gestational agee
SGA vs. normal	0.0	−0.5, 0.5	0.0	−0.5, 0.5
LGA vs. normal	−0.3	−0.8, 0.2	−0.3	−0.8, 0.2

Abbreviations: BMI, body mass index; CI, confidence interval; LGA, large for gestational age; SGA, small for gestational age.

^a Models included fixed effects for state of maternal residence, year of conception, demographic subgroup × year of conception, and demographic subgroup × state. Robust standard errors were used to account for clustering within state.

^b Models included fixed effects for state of maternal residence, year of conception, demographic subgroup × year of conception, demographic subgroup × state, race/ethnicity, age, parity, marital status, and the interaction between expansion and exposure status. Robust standard errors were used to account for clustering within state.

^c All β coefficients in the table have been multiplied by 100 to aid interpretation.

^d Weight (kg)/height (m)².

^e For the SGA model, LGA births were not included. For the LGA model, SGA births were not included.

Our robustness analyses found that associations between Medicaid expansion and these outcomes did not differ substantially when different states were included or excluded from the analyses (Web Table 8).

DISCUSSION

We tested the hypothesis that increasing eligibility for Medicaid coverage prior to conception for women of reproductive age would translate to improved preconception health and, subsequently, improved pregnancy health and reduced adverse pregnancy outcomes. Increased Medicaid eligibility was not strongly or consistently associated with these outcomes. Among women with less than a high school education and unmarried women (those most likely to become newly insured following Medicaid expansion), we did find increases in the percentage of births complicated by gestational diabetes or hypertension. This finding may not necessarily indicate an increase in the incidence of these conditions; instead, it could reflect increased diagnoses or recording of these conditions in the medical record among newly insured (compared with uninsured) women or could reflect the role of chance, particularly because adjustment for multiple testing indicated that the 95% confidence intervals around these estimates included the null.

We propose several potential explanations for our primarily null findings. First, it is possible that not enough women who eventually went on to deliver a live infant enrolled in Medicaid following the ACA expansions, or that not enough of those newly enrolled experienced an improvement in health. However, with the large number of observations in this study, we had greater than 90% power to detect very small changes (e.g., a 0.3-percentage-point increase) in even rare outcomes like prepregnancy hypertension and in the smaller subsamples of women without a high school education and no previous children. In a recent study, Clapp et al. (25) reported that preconception Medicaid coverage among low-income women who delivered a live infant increased 8.6 percentage points following the ACA expansions. If only 4% of those women newly enrolled in Medicaid experienced an improvement in health, we would have had power to detect a change this large (0.3 percentage points) or larger in the prevalence of the outcome in our data set.

It is also possible that any positive effect of Medicaid uptake on the health of women of reproductive age could have been obscured by changes in the composition of women who gave birth—for example, if the women whose health improved after gaining insurance also delayed pregnancy or were less likely to have an unintended pregnancy. For example, women gaining health-care coverage under the Medicaid expansion increased use of prescription contraceptives by 22% (36), and our own research found a reduction in the odds of unintended pregnancy among women with government-sponsored insurance during the 2 years following implementation of the ACA contraception mandate (2013–2015) (37). In future research, investigators should further examine whether expanded preconception Medicaid coverage affected unintended pregnancy, interpregnancy interval, and/or live birth rates. Underreporting of prepregnancy chronic disease (diabetes and hypertension) and pregnancy complications (gestational diabetes, gestational hypertension, and eclampsia) on the birth certificate has been documented in validation studies (38–40). Although the prevalence of these conditions in our sample varied as expected by race/ethnicity, age, and education (Web Table 7), underreporting overall could still have resulted in biased estimates for these outcomes.

A final possible explanation is that health insurance and health care alone are insufficient to offset the fact that social and environmental risk factors for poor pregnancy health and outcomes are inequitably distributed by race/ethnicity and socioeconomic status in the United States (41–45), placing a greater burden on low-income communities and people of color that cannot be overcome simply by gaining health insurance coverage.

The current study was strengthened by the use of data on all live births taking place in the United States from 2011 to 2017. Another contribution was our use of the simulated eligibility measure, which took advantage of variability in Medicaid eligibility by state and time across the study period and which permitted estimation of the impact of expanding Medicaid eligibility to women across the income distribution. Our study also specifically examined preconception exposure to Medicaid expansion. In contrast, in a related study, Brown et al. (31) reported that Medicaid expansion during pregnancy and at the time of birth was associated with declines in disparities in preterm birth and early preterm birth between Black and White infants. It is unlikely, however, that exposure to the ACA Medicaid expansion during pregnancy or at the time of birth could impact birth outcomes when eligibility levels for pregnant women did not change.

A potential limitation of our study was the relatively short post-ACA Medicaid expansion period, but examination of simulated eligibility across the entire study period helped address this limitation. Additionally, the standard US birth certificate does not include data on women’s health insurance coverage or health-care utilization prior to pregnancy, household income, and other measures of pregnancy health such as folate/vitamin intake, alcohol use, and quality of prenatal care.

In conclusion, our findings do not indicate that increased Medicaid eligibility is associated with substantial improvements in prepregnancy health, pregnancy health, or pregnancy outcomes. Changes to Medicaid eligibility alone are likely only 1 piece of the puzzle of increasing health-care utilization, health behaviors, and health among women prior to conception. Health insurance coverage does offer an opportunity for the health-care system to promote women’s health before pregnancy and improve pregnancy outcomes by developing and implementing preconception programs. Indeed, even prior to the ACA Medicaid expansions, states had begun implementing programs such as integrating preconception- and interconception-specific health care into clinical settings, conducting provider trainings on preconception care, and integrating preconception care into health education curricula (46). Yet, social determinants of health, systemic discrimination, and environmental exposures across the life course are also strong determinants of preconception health and subsequent pregnancy outcomes (41–45), and the inequitable distribution of these exposures in our society cannot be remedied within the health-care system alone.