County-Level Variation in Low Birthweight and Preterm Birth: An Evaluation of State Medicaid Expansion under the Affordable Care Act

Abstract

Background:

Rates of low birthweight and prematurity vary two-fold across states in the US, with increased rates among states with higher concentrations of racial minorities. Medicaid expansion may serve as a mechanism to reduce geographic variation within states that expanded, by improving health and access to care for vulnerable populations.

Objective:

To identify the association of Medicaid expansion with changes in county-level geographic variation in rates of low birthweight and preterm births, overall and stratified by race/ethnicity.

Research Design:

We compared changes in the coefficient of variation (COV) and the ratio of the 80^th-to-20^th percentiles using bootstrap samples (n=1,000) of counties drawn separately for all births and for White, Black, and Hispanic births, separately.

Measures:

County-level rates of low birthweight and preterm birth.

Results:

Prior to Medicaid expansion, counties in expansion states were concentrated among quintiles with lower rates of adverse birth outcomes and counties in non-expansion states were concentrated among quintiles with higher rates. In expansion states, county-level variation, measured by the COV, declined for both outcomes among all racial/ethnic categories. In non-expansion states, geographic variation reduced for both outcomes among Hispanic births and for low birthweight among White births, but increased for both outcomes among Black births.

Conclusions:

The decrease in county-level variation in adverse birth outcomes among expansion states suggests improved equity in these states. Further reduction in geographic variation will depend largely on policies or interventions that reduce racial disparities in states that did and did not expand Medicaid.

Rates of low birthweight (birthweights <2,500 grams) and preterm births (births at <37 weeks gestation) have been declining in the United States since 2007; however, rates in the United States remain higher than most other developed nations.^1,2 It has been estimated that low birthweight and preterm births contribute to around 9,000 infant deaths every year,^3,4 with an estimated 35% of infant mortality being attributed to complications resulting from low birthweight and prematurity.⁴

Previous studies of geographic variation in adverse birth outcomes have indicated higher rates of low birthweight and preterm birth outcomes among regions with higher concentrations of low-income women and of racial and ethnic minorities.^5–7 In 2017, 13.6% of births in Mississippi were preterm and 11.6% of births were low birthweight, nearly twice the rates of preterm births (7.5%) and low birthweight (5.6%) in Vermont.² Given the relationship between low birthweight and preterm birth with infant mortality and chronic conditions across the lifespan,^8,9 geographic variation in rates of adverse birth outcomes across regions may ultimately be associated with geographic variations in other adverse health outcomes across the life course.

The 2012 landmark Supreme Court decision to allow states to opt out of the Affordable Care Act (ACA) Medicaid expansion left 24 states that did not expand as of 2014.¹⁰ Two recent studies found improvements in adverse infant outcomes among Black infants after Medicaid expansion,^11,12 which gave states the option to expand their Medicaid coverage to childless adults with income levels at or below 138% of the federal poverty level. While states were required to offer Medicaid coverage for pregnant women with incomes up to 133% of the federal poverty level prior to the ACA, this pregnancy-contingent Medicaid coverage ended 60 days post-partum. Healthcare coverage that is not contingent upon pregnancy status may improve continual access to healthcare for women of childbearing age in the preconception period.^13,14 Specifically, continuity in coverage prior to pregnancy and beyond 60 days postpartum may improve the health of the mother through increased access to preventive services, which may translate to improved infant health. Having insurance coverage prior to pregnancy may also result in earlier initiation of prenatal care services because of early determination of pregnancy, having an established clinical provider, and removal of administrative barriers associated with episodic pregnancy-based Medicaid coverage.^13,15,16 Evaluating the extent of changes in geographic variations in birth outcomes among counties within states that expanded Medicaid is important for understanding health equity among regions.

The primary objective of this study was to assess whether there were changes in county-level variation in adverse birth outcomes associated with Medicaid expansion. We expected that improved access to care and health for women in expansion states may have ultimately reduced geographic variation in adverse birth outcomes and improved health equity.

Methods

Data

Data for this study came from the National Center for Health Statistics (NCHS) Vital Statistics Birth Data Files (“Birth Data Files”), years 2011 through 2016. The Birth Data Files contain data abstracted from birth certificates for 100% of the registered births in all fifty states and the District of Columbia.¹⁷

The unit of analysis in this study was county-level rates of low birthweight and preterm birth outcomes. The Birth Data Files include births from all county and county equivalents (hereafter “counties”) in the United States, District of Columbia, and United States territories, which were identified using Federal Information Processing Standards (FIPS) codes. After excluding counties in United States territories, there were 3,145 counties represented. We modified FIPS codes for births from prior to 2015 in one county in South Dakota (FIPS=46,113) to reflect the new FIPS code for the county (FIPS=46,102).

Counties were limited to those in the contiguous United States (n=3,109) and excluded births from counties (n=372) in the 6 states that expanded Medicaid after January 1, 2014 and births from 9 independent cities from Virginia, leaving 2,728 counties. These independent cities, which were considered county equivalents, had populations less than 10,000, preventing identification in years 2014–2016 in the birth data files. County-level rates in the included counties were calculated among 19,454,243 singleton births to women aged 19+ at the time of the birth. This age group was chosen because low-income women under the age of 18 would be eligible for Children’s Health Insurance Program (CHIP) in most states if they met the income requirement.

County-level rates were calculated for births to all women, as well as among non-Hispanic White (“White”), non-Hispanic Black (“Black”), and Hispanic women, separately. Following the approach of the NCHS Vital Statistic Reports, women who indicated “Hispanic” were considered Hispanic, regardless of race. Counties that did not have at least 30 births for the race/ethnicity group in both the pre- and post-Medicaid expansion periods were excluded from the analyses for that race/ethnicity group. A limit of 30 births was chosen because this is a generally accepted threshold to obtain a sample with an approximately normal distribution, irrespective of the distribution of the population data.¹⁸ The final sample of county-level rates included 2,692 counties for births of all races/ethnicities, 2,651 counties for births to White women, 1,219 counties for births to Black women, and 1,300 counties for births to Hispanic women.

Counties were classified into Medicaid expansion categories based on whether a given state expanded Medicaid (AR, AZ, CA, CO, CT, DC, DE, IA, IL, KY, MA, MD, MN, ND, NJ, NM, NV, NY, OH, OR, RI, VT, WA, WV) on January 1, 2014.

Outcome Variables

The outcomes in this study included county-level rates of preterm births (<37 weeks gestation) and low birthweight (<2,500 grams). Rates were calculated separately for births of all races/ethnicities, for births to White women, for births to Black women, and for births to Hispanic women by Medicaid expansion status and by date prior to or post-Medicaid expansion.

Statistical Analysis

First, using data prior to expansion, we assessed the distribution of counties across quintiles based on Medicaid expansion status. Specifically, county-level rates were assigned to quintiles using all counties, and the distribution of counties across quintiles was determined within expansion and non-expansion states. To determine if there was a statistical difference in the distribution across quintiles, Mantel-Haenszel Chi-Square for Nonzero Correlation tests were conducted to assess linear association.

Next, descriptive statistics of county-level rates were calculated, stratified by racial/ethnic group, Medicaid expansion status, and pre/post the Medicaid expansion date. To measure geographic variation, we calculated the coefficient of variation (COV), defined as the standard deviation divided by the mean, and the 80th-to-20th percentile ratio for each outcome. The COV and 80^th-to-20^th percentile ratio both allow for comparisons among variables with different magnitudes and/or units.^19,20 The use of dimensionless measures that are not sensitive to scaling allows for comparison of variation among different populations or different measures. While the 90^th-to-10^th percentile ratio is more common in the geographic variation literature, we were unable to use the 10^th percentile because of undefined values when using 0% as a denominator in some samples in the bootstrap analyses (explained below). Both measures of variation were calculated, given that they can be used to assess slightly different aspects of variation. Specifically, the COV standardizes variation based on the mean while the 80^th-to-20^th percentile ratio allows for conclusions regarding the counties with the highest and lowest rates.

To assess changes in geographic variation from pre- to post-Medicaid expansion, we evaluated differences in the COVs and the 80^th-to-20^th percentile ratios among county-level rates. Formal statistical tests for changes in geographic variation among geographic regions are limited.^21–23 Analyses were modeled following methods used in previous studies.^21,24 First, bootstrap samples (n=1,000) were randomly selected with replacement for counties in all states, in non-expansion states, and in expansion states for each race/ethnicity group, separately. For example, for the analyses of Black infants among the counties in non-Medicaid expansion states, a bootstrap sample (n=1,000) of counties was selected for counties that had at least 30 Black infant births in the pre-expansion period and at least 30 Black infant births in the post-expansion period. Finally, the 95% confidence intervals of the COV and of the 80^th-to-20^th percentile ratio for each group of 1,000 samples was calculated.

Statistical significance was assumed at p<0.05. A lack of overlap in the 95% confidence intervals of a COV or 80^th-to-20^th percentile pair (i.e., 1 for the pre-expansion period and 1 for the post-expansion period) was considered to indicate a statistically significant difference. Because overlap does not indicate a rejection of the null hypothesis that the mean values of the pre-expansion and post-expansion periods do not differ, all pairs with overlapping 95% confidence intervals were further assessed to determine if a given mean was outside of the 95% confidence interval for the corresponding (i.e., opposite) pre- or post-expansion mean. All analyses were completed in SAS 9.4.

Results

Table 1 provides the distribution of counties among quintiles, stratified by expansion status using data from the period prior to expansion. Percentages in each row add to 100% within each expansion group separately. Counties in expansion states were concentrated in quintiles with lower rates of adverse outcomes, and counties in non-expansion states were concentrated in quintiles with higher rates of adverse outcomes. The tests of trend were statistically significant for both outcomes among all of the race/ethnicity categories, with the exception of low birthweight among Hispanic births. For example, among births of all races/ethnicities, 24.5% of counties in expansion states ranked in the Q1 (lowest rates) for preterm births while 12.3% of counties ranked in Q5. In non-expansion states, 16.9% of counties ranked in Q1, and 25.2% of counties ranked in Q5 (Mantel-Haenszel Chi-Square tests p-value <0.001). Among births to Black women in expansion states, only 9.4% of counties ranked in Q5 for low birthweight and 10.6% of counties ranked in Q5 for preterm birth, compared to 25.3% and 24.7% of counties in non-expansion states.

Table 1.

Tests of trend in the percent of Medicaid expansion counties in each quintile based on low birthweight and preterm birth, in the time period prior to Medicaid expansion.^a

	Non-Expansion					Expansion
	Q1 (%)	Q2 (%)	Q3 (%)	Q4 (%)	Q5 (%)	Q1 (%)	Q2 (%)	Q3 (%)	Q4 (%)	Q5 (%)	p-valueb
All Races/Ethnicities (n=2,692)
Low Birthweight	16.9	16.8	19.6	22.4	24.4	24.6	24.8	20.6	16.4	13.6	<.001
Preterm	16.9	15.9	19.2	22.8	25.2	24.5	25.8	21.5	15.9	12.3	<.001
White (n=2,651)
Low Birthweight	17.8	17.9	21.3	21.8	21.3	23.1	23.2	18.2	17.3	18.2	<.001
Preterm	17.1	17.9	19.5	23.0	22.5	24.3	23.1	20.7	15.7	16.3	<.001
Black (n=1,219)
Low Birthweight	14.1	16.8	22.0	21.8	25.3	31.7	26.5	16.1	16.3	9.4	<.001
Preterm	14.7	16.7	20.7	23.2	24.7	30.5	26.7	18.6	13.6	10.6	<.001
Hispanic (n=1,300)
Low Birthweight	19.5	20.7	20.7	20.7	18.4	20.7	18.8	19.2	19.0	22.4	.46
Preterm	20.6	18.5	17.1	19.8	24.0	19.2	22.2	24.4	20.3	13.9	.01

Low birthweight is a birthweight <2,500 grams, and preterm is a birth <37 weeks gestation. Q=Quintile

^aPrior to Medicaid expansion includes prior to January 1, 2014.

^bP-values are results of Mantel-Haenszel Chi-Square tests for Nonzero Correlation.

Table 2 provides descriptive statistics by race/ethnicity category, state expansion status, and pre-post expansion period. Rates among counties with the highest rates (80^th percentile) were between 1.45 (preterm births among all races/ethnicities in the post-expansion period) and 1.73 (low birthweight among Hispanic births in the pre-expansion period) times as high as counties with the lowest rates (20^th percentile). Average county-level rates were consistently higher for Black births in non-expansion states relative to Black births in expansion states and for Black births relative to White and Hispanic births. For example, during the post-expansion period, the rate of low birthweight for Black births in non-expansion states was 2.10 times the rate for White births in non-expansion states (11.84% versus 5.65%) and 1.22 times the rate for Black births in expansion states (11.84% versus 9.74%). Additionally, rates among counties in non-expansion states were higher for all four racial/ethnic categories relative to their counterparts in expansion states, with the exception of low birthweight among Hispanics.

Table 2.

Descriptive statistics of county-level rates of low birthweight and preterm birth by expansion status, pre-post expansion, and racial category.

	Pre expansiona				Post expansiona
	Mean(SD)	80/20 Pctl	(80th Pctl, 20th Pctl)	COV	Mean(SD)	80/20 Pctl	(20th Pctl, 80th Pctl)	COV
	Non-Expansion
All Races/Ethnicities
Number of Counties	1,608				1,608
% Low Birthweight	6.59(2.26)	1.68	(4.87,8.20)	34.36	6.62(2.18)	1.66	(4.92,8.20)	32.85
% Preterm	8.45(2.37)	1.55	(6.60,10.20)	28.06	8.42(2.34)	1.51	(6.70,10.10)	27.76
White
Number of Counties	1,571				1,571
% Low Birthweight	5.68(1.79)	1.63	(4.34,7.07)	31.50	5.65(1.68)	1.58	(4.39,6.94)	29.71
% Preterm	7.80(2.17)	1.53	(6.12,9.35)	27.79	7.69(2.17)	1.51	(6.11,9.22)	28.26
Black
Number of Counties	815				815
% Low Birthweight	11.62(3.19)	1.53	(9.15,13.96)	27.48	11.84(3.34)	1.51	(9.34,14.13)	28.18
% Preterm	11.91(3.37)	1.51	(9.45,14.23)	28.29	11.85(3.38)	1.54	(9.25,14.26)	28.53
Hispanic
Number of Counties	783				783
% Low Birthweight	5.45(2.00)	1.62	(4.14,6.73)	36.65	5.76(2.09)	1.70	(4.21,7.14)	36.36
% Preterm	7.71(2.62)	1.63	(5.83,9.51)	34.04	8.09(2.57)	1.53	(6.36,9.71)	31.80
	Expansion
All Races/Ethnicities
Number of Counties	1,084				1,084
% Low Birthweight	5.90(1.90)	1.64	(4.42,7.27)	32.15	5.91(1.83)	1.60	(4.53,7.28)	30.92
% Preterm	7.61(1.97)	1.46	(6.15,8.98)	25.83	7.63(1.84)	1.45	(6.20,8.96)	24.12
White
Number of Counties	1,080				1,080
% Low Birthweight	5.53(1.91)	1.68	(4.07,6.83)	34.50	5.48(1.80)	1.67	(4.07,6.79)	32.85
% Preterm	7.33(2.03)	1.51	(5.77,8.73)	27.74	7.29(1.91)	1.51	(5.82,8.78)	26.21
Black
Number of Counties	404				404
% Low Birthweight	9.83(3.44)	1.69	(7.26,12.28)	35.01	9.74(3.17)	1.69	(7.23,12.22)	32.49
% Preterm	10.13(3.54)	1.61	(7.69,12.40)	34.98	9.87(3.11)	1.63	(7.33,11.94)	31.48
Hispanic
Number of Counties	517				517
% Low Birthweight	5.61(2.14)	1.73	(4.04,6.98)	38.06	5.89(2.00)	1.62	(4.55,7.37)	33.97
% Preterm	7.29(2.03)	1.47	(5.91,8.70)	27.82	7.61(2.14)	1.49	(6.10,9.09)	28.14

Low birthweight is a birthweight <2,500 grams, and preterm is a birth <37 weeks gestation. COV = coefficient of variation; Pctl = percentile; SD = standard deviation

^aPre-expansion includes prior to January 1, 2014.

Table 3 provides the average COV for the bootstrap sample of counties within a given cell (see Table, Supplemental Digital Content 1, which includes associated 95% confidence intervals used to assess significance). For counties in expansion states, variation decreased for both adverse outcomes for all four racial/ethnic groups. The change in variation among counties in expansion states ranged from a 3.79% decline (low birthweight among births of all races/ethnicities) to a 10.17% decline (preterm birth among Black births). In non-expansion states, variation increased for preterm births among White infants (1.72%), preterm birth among Black infants (0.79%), and for low birthweight among Black infants (2.56%), and variation in non-expansion states declined for other racial/ethnic groups and outcomes.

Table 3.

Coefficient in variation of county-level rates of birth outcomes from before Medicaid expansion to after Medicaid expansion (n=1,000 samples per cell).^a

	Low Birthweight			Preterm
	Before	After	% Change	Before	After	% Change
All Races/Ethnicities
All States	34.02b	32.70b	−3.88	27.78b	26.96b	−2.94
Non-Expansion	34.35b	32.87b	−4.30	28.04b	27.76b	−1.01
Expansion	32.12b	30.90b	−3.79	25.81b	24.10b	−6.62
White
All States	32.73b	31.01b	−5.25	27.96b	27.63b	−1.19
Non-Expansion	31.50b	29.71b	−5.68	27.76c	28.24c	1.72
Expansion	34.47b	32.86b	−4.69	27.71b	26.20b	−5.45
Black
All States	30.64d	30.72d	0.25	31.13b	30.55b	−1.86
Non-Expansion	27.48c	28.18c	2.56	28.30c	28.52c	0.79
Expansion	34.98b	32.40b	−7.37	34.95b	31.39b	−10.17
Hispanic
All States	43.31b	41.37b	−4.50	36.79b	35.00b	−4.86
Non-Expansion	43.31b	42.21b	−2.55	37.54b	36.31b	−3.26
Expansion	43.40b	40.03b	−7.75	35.04b	32.32b	−7.76

Low birthweight is a birthweight <2,500 grams, and preterm is a birth <37 weeks gestation.

^aBootstrap samples of 2,692 (all races/ethnicities), 2,651 (White), 1,219 (Black), and 1,300 (Hispanic) counties with 30 or more births in the pre and post periods.

^bIndicates a significant reduction in variation using 1,000 bootstrap samples per cell.

^cIndicates a significant increase in variation using 1,000 bootstrap samples per cell.

^dIndicates a mean value that is outside the range of the 95% confidence interval in the corresponding pre/post period.

Table 4 provides the average 80^th-to-20^th percentile ratios for the bootstrap sample of counties within a given cell (see Table, Supplemental Digital Content 2, which includes associated 95% confidence intervals used to assess significance). Among births in expansion states, there were declines in geographic variation for both outcomes among births to all races/ethnicities and births to White women, as well as for low birthweight among Hispanic women. However, there were increases in geographic variation for low birthweight among Black births and for preterm births among Hispanic infants for counties in expansion states.

Table 4.

80^th-to-20^th percentile ratios for county-level rates of birth outcomes from before Medicaid expansion to after Medicaid expansion using bootstrapping (n=1,000 samples per cell).^a

	Low Birthweight			Preterm
	Before	After	% Change	Before	After	% Change
All Races/Ethnicities
All States	1.68b	1.66b	−1.39	1.53b	1.51b	−1.81
Non-Expansion	1.69b	1.66b	−1.43	1.54b	1.51b	−2.42
Expansion	1.64b	1.60b	−2.36	1.46b	1.45b	−0.85
White
All States	1.66b	1.63b	−1.75	1.54b	1.51b	−1.67
Non-Expansion	1.63b	1.59b	−2.58	1.52b	1.50b	−1.34
Expansion	1.68b	1.67b	−0.50	1.51b	1.51b	−0.48
Black
All States	1.60c	1.63c	1.88	1.56c	1.58c	1.18
Non-Expansion	1.52b	1.51b	−1.15	1.51c	1.53c	1.25
Expansion	1.68c	1.71c	1.62	1.61	1.61	−0.04
Hispanic
All States	1.80b	1.77b	−1.91	1.65b	1.60b	−2.70
Non-Expansion	1.80d	1.80d	0.39	1.72b	1.63b	−5.42
Expansion	1.81b	1.69b	−6.74	1.54c	1.58c	2.11

Low birthweight is a birthweight <2,500 grams, and preterm is a birth <37 weeks gestation.

^aBootstrap samples of 2,692 (all races/ethnicities), 2,651 (White), 1,219 (Black), and 1,300 (Hispanic) counties with 30 or more births in the pre and post periods.

^bIndicates a significant reduction in variation using 1,000 bootstrap samples per cell.

^cIndicates a significant increase in variation using 1,000 bootstrap samples per cell.

^dIndicates a mean value that is outside the range of the 95% confidence interval in the corresponding pre/post period.

Discussion

The analysis of changes in county-level variation in low birthweight and preterm birth after Medicaid expansion found reduced variation, measured by the COV, in preterm and low birthweight births among counties in expansion states, with less consistent reductions in non-expansion states. Overall, states that expanded Medicaid had lower rates of adverse birth outcomes compared to states that did not expand, and there were consistent disparities among Black infants relative to White infants, regardless of Medicaid expansion status and time period.

A previous evaluation of the association of Medicaid expansion with rates of adverse birth outcomes among Medicaid expansion states compared to non-expansion states found declines for Black infants in expansion states but no declines overall or for White infants.¹² The improvements among Black infants relative to White infants may explain the larger changes in variation among county-level rates of adverse birth outcomes among Black infants in our study. Our study adds to this literature by addressing the need to additionally consider variation among geographic regions when evaluating the impact of a policy. Specifically, policies may influence rates of health outcomes between regions (i.e., Medicaid expansion reduced rates of adverse birth outcomes beyond their already-lower rates) and within regions (i.e., variation in county-level rates among expansion states was reduced in terms of the COV).

There is a large body of geographic variation literature that focuses on inefficiency in healthcare delivery by identifying geographic areas with high levels of service utilization and no detectable improvement in health outcomes.²⁵ Geographic variation in adverse birth outcomes, on the other hand, represents differences in health equity among counties. Given the association between low birthweight and prematurity with adverse outcomes throughout infancy and into adulthood, understanding ways to reduce geographic variation in adverse birth outcomes has implications for health equity across the life course.^8,9

To our knowledge, there are few, if any published studies assessing changes in geographic variation in any county-level health outcome with respect to Medicaid expansion. Previous studies have used variation in the change in insurance coverage (or other similar access to care measure) across regions to draw causal conclusions regarding the impact of increases in insurance coverage on health, utilization, or other outcomes;^26,27 however, studies that have used geographic variation in a health outcome as the outcome of interest are limited. Because increases in health insurance coverage or improvements in continuity of care may have a more direct effect on other outcomes, additional studies should evaluate changes in geographic variation for other health conditions.

Previous studies have found that area-level rates of adverse birth outcomes are highly associated with concentrations of low socioeconomic groups and racial minorities.^5–7,28 Given the potential impact of continuous coverage through Medicaid expansion to improve access to healthcare among low-income women, it was hypothesized that Medicaid expansion would result in reductions in area-level variations in adverse birth outcomes among counties in expansion states if access to healthcare services and maternal health were improved for at-risk populations. Among expansion states, we found a reduction in geographic variation in county-level rates of low birthweight and preterm birth for all racial/ethnic categories as measured by the COV. However, in terms of the 80^th-to-20^th percentile ratio, geographic variation in expansion states increased for low birthweight among Black infants. A closer examination of the descriptive analysis suggests that the increased geographic variation among Black infants may be attributable to large improvements in select counties. Specifically, among counties in expansion states, there were relatively larger declines in the 80^th percentile of low birthweight births for Black infants in expansion states relative to the decline in the 20^th percentile. This relationship was not found among rates for White infants or for Hispanic infants.

Direct comparison of the findings in this study to findings from other studies is difficult, given the differences regarding the geographic unit and minimum number of births required within the defined geographic region. However, the findings in this study of substantial geographic variation in low birthweight and preterm birth has been found across studies.^5–7,28 While the magnitude of change in geographic variation was relatively small, the reductions in county-level rates of adverse birth outcomes, along with reductions in the average rates particularly for Black infants, suggest that the Medicaid expansion may have served as a potential source of improved outcomes for the most vulnerable counties.

The bootstrap analyses indicated that geographic variation among Black infants was generally similar in magnitude to the respective measure among White infants. This has not been found across studies and may be due to the fact that other studies limited geographic areas that met the minimum number of birth requirement for both White and Black infants simultaneously.^5,6,29 For the race/ethnicity subgroups of counties in this study, counties were chosen based on whether the county met the minimum number of births criteria for that race/ethnicity group alone. As such, the number of counties in the Black birth and Hispanic birth subsamples is less than the number of counties in the White birth subsample.

There are several limitations to this study. First, the Birth Data Files are administrative data and are not collected for research purposes. Because birth certificate data are collected across multiple settings and by a variety of people, there is the potential for recorded data to be misclassified or for missing data to exist disproportionately across subgroups; however, the Birth Data Files have been used in numerous studies regarding the implication of policy changes on birth outcomes.^30–32 Additionally, county-level identifiers were the lowest level of geographic granularity that was available for analysis using this data.

Gestational age was based on the obstetric/clinical estimate, which the NCHS uses in its annual reports. Prematurity estimated from the mother’s last menstrual period could provide different results. Birthweight information from the Birth Data Files has been found to have a higher validity and reliability than gestational age.^33–35 To mitigate concerns associated with gestational age, we additionally included birthweight, which is less sensitive to misclassification.⁷

The analyses of geographic variation among Black births and Hispanic births were limited to 1,219 and 1,300 counties. It should be noted, however, that this represented 95.3% (2,744,969/2,765,534) of Black births and 99.4% (4,777,922/4,805,778) of Hispanic births in the final sample of all eligible births used to calculate county-level rates in this study. Similarly, the relatively small sample sizes of other racial/ethnic minority groups (e.g., Asian) prevented additional analyses on such groups.

Finally, improved infant outcomes as a result of maternal insurance coverage in pre-pregnancy or inter-pregnancy time periods may take more time to accrue than the 3 years post-Medicaid expansion used in this study. However, removing births in the 12-months post expansion in each expansion state along with the first year of the study period to maintain 2 years of data in the pre- and post-expansion periods (see Table, Supplemental Digital Content 3 and 4, which include the COV and 80^th-to-20^th percentile values; and Table, Supplemental Digital Content 5 and 6, which include associated 95% confidence intervals) indicated largely similar findings.

Conclusions

This study found modest reductions in geographic variation of county-level rates of low birthweight and preterm birth after state Medicaid expansion for states that expanded Medicaid; however, there was weaker evidence of reductions in geographic variation among counties within states that did not expand. As low-income women of childbearing age have historically had significant issues regarding continuity of insurance coverage, improvements in continuity of insurance coverage related to Medicaid expansion under the ACA has the potential to improve birth outcomes for low-income women. If variation in low birthweight and prematurity are associated with disparities in health outcomes among regions,^36,37 then a reduction in variation in low birthweight and preterm births may translate to improved health equity by reducing infant mortality and improving long-term physical and cognitive health.