Medicaid Expansion Led to Reductions in Postpartum Hospitalizations

Abstract

The Affordable Care Act (ACA) Medicaid expansions increased preconception and postpartum insurance coverage among low-income birthing people, leading to greater use of outpatient care. In this study, we evaluated whether the expansions affected rates of postpartum hospitalization. Our analyses took advantage of underused longitudinal hospital data to examine hospitalizations following childbirth. We compared changes in hospitalizations among birthing people with a Medicaid-financed delivery in states that did and did not implement ACA Medicaid expansions. We found a 17 percent reduction in hospitalizations during the first 60 days postpartum associated with the Medicaid expansions, and some evidence of a smaller decrease in hospitalizations between 61 days and six months postpartum. Our findings indicate that expanding Medicaid coverage led to improved postpartum health for low-income birthing persons.

INTRODUCTION

Experts recommend increasing access to health care before pregnancy and during the postpartum period to promote the health of US childbearing parents and their infants.^1,2 More than one-third of women with pregnancy coverage through Medicaid, which covers over 40 percent of all births in the US and provides 60 days of postpartum coverage for qualifying parents, are uninsured before or during the two to six months following pregnancy.^3,4

In states that adopted the Affordable Care Act (ACA) Medicaid expansions, new eligibility rules allowed a larger share of low-income adults to qualify for Medicaid coverage prior to pregnancy and following the 60-day postpartum period. This led to significant increases in Medicaid enrollment and overall insurance coverage before and after pregnancy, as well as greater continuity of insurance coverage, among low-income parents.^5–8

Prior studies have documented greater use of outpatient care before and after pregnancy under the ACA Medicaid expansions.^6,7,9 There is also evidence of improved care during pregnancy from studies of state ACA Medicaid expansions in Ohio and Oregon, although less evidence in support of this from other studies.^5,10–14 Our study used a unique source of longitudinal administrative hospital data to examine whether these changes in access to care affected rates of inpatient hospitalization during the first six months postpartum.

STUDY DATA AND METHODS

Our analysis used 2010–2017 data on four expansion (Iowa, Maryland, New Mexico, Washington) and four non-expansion states (Florida, Georgia, Mississippi, Utah) from the Agency for Healthcare Research and Quality’s Healthcare Cost and Utilization Project State Inpatient Databases. These databases contain the universe of state inpatient hospital records.¹⁵ The study included all states with patient identifiers available during the pre- and post-periods, which allowed patients to be tracked across hospital encounters. We excluded four states with non-ACA Medicaid expansions for low-income adults, and two states with changes in Medicaid-funded deliveries unrelated to the expansions.

We defined the post-expansion period using the start date of Medicaid expansion implementation in each state, which was January 1, 2014, in all four of the study expansion states. Medicaid expansion applied to both adults without children as well as parents. However, the change in eligibility differed depending on parental status, since in most states adults without children were not eligible for Medicaid (regardless of their income level). Therefore, the change in preconception eligibility after Medicaid expansion depended on whether the birthing person had other children before becoming pregnant. In all four of the study expansion states, income eligibility for childless adults increased from 0% to 138% of federal poverty line (FPL);¹⁶ whereas parental eligibility increased from 80% to 138% of FPL in Iowa, 122% to 138% of FPL in Maryland, 85% to 138% in New Mexico, and 71% to 138% in Washington.¹⁷ We are unable to determine from our data source whether postpartum people had other children prior to delivery.

The study sample was defined as state residents aged 19 or older with a hospital delivery for which the expected primary payer was Medicaid. Delivery hospitalizations were identified using the method in Kuklina et al.¹⁸ In addition to examining effects for all patients meeting these criteria, we also stratified the sample by the race/ethnicity (non-Hispanic White, non-Hispanic Black, and Hispanic) of the birthing person.

Our primary outcomes were the occurrence of a postpartum hospitalization within 60 days of delivery discharge and between 61 days and up to six months after delivery discharge. This categorization was meant to distinguish between the first 60 days of the postpartum period, which was already covered by pregnancy-related Medicaid in all states before expansion, and the period after 60 days, which was newly covered for people gaining eligibility in expansion states. Because patient identifiers often could not be used to track patients across calendar years, we restricted the sample to patients whose outcomes could be observed within the calendar year. Those with deliveries during the first three quarters of each calendar year were analyzed for the 60-days postpartum outcome (n = 1,355,220) and those with deliveries during the first two quarters for the 61-days to six months postpartum outcome (n = 880,014).

In addition to examining overall hospitalizations, we used the International Statistical Classification of Diseases and Related Health Problems to categorize the primary diagnoses for hospitalization. We first examined descriptively the primary reasons for postpartum hospitalizations during the pre-expansion period, as grouped by body system. We then classified hospitalizations into two categories: (1) those with a primary diagnosis related to complications of pregnancy, childbirth, and the puerperium, and (2) all other diagnoses.

In unadjusted analyses, we examined average rates of postpartum hospitalization before and after expansion in expansion and non-expansion states. We also tested for statistically significant changes in the rates of hospitalization within each group of states using a mean comparison test (t-test) and evaluated significance at the 1, 5, and 10 percent levels. Two study states (Maryland and Mississippi) were missing some data years and were excluded from this unadjusted analysis.

We then used an event study model to test for pre-period differential trends in outcomes in expansion and non-expansion states, as well as to trace out the effects of ACA Medicaid expansion during each year of the post-period. This model was estimated with a series of indicators for deliveries each year during the study period interacted with state Medicaid expansion status; the 2013 year was omitted. Each estimated coefficient on these event study terms provided the average difference in relative outcomes for birthing people in expansion and non-expansion states during each year as compared to the year prior to expansion, adjusted for covariates (patient age, race, and ethnicity) and fixed effects for state and the quarter-year of delivery. A difference-in-differences model with an indicator variable for birthing people with deliveries in expansion states in 2014 and later provided an estimate of the average difference in relative outcomes for birthing people in the two groups of states during the post-expansion period. In the case of the race-specific analyses, we tested for significant differences across racial groups using a fully interacted triple difference model. For all analyses, we used linear regression models and clustered the standard errors by state using a wild cluster bootstrap procedure.¹⁹

Additional details on the data, sample, and statistical analyses are in the appendix,²⁰ which also includes results from two sensitivity analyses. In the first, we excluded Maryland and Mississippi in the difference-in-differences sample to present the adjusted results in the same sample of states that were used for the unadjusted analysis (see appendix Table A5).²⁰ In the second, we conducted the difference-in-differences analysis for the 60-day postpartum hospitalization outcome including only deliveries during the first and second quarters of the calendar year (see appendix Table A6).²⁰ We implemented this sensitivity analysis to show the results for this outcome in the same sample of deliveries that was used for the second study outcome (hospitalizations between 61 days and six months postpartum).

Limitations

This analysis had several limitations. The small number of sample states and the focus on deliveries with postpartum periods occurring within the same calendar year may limit the generalizability of the study’s findings. We were unable to control for seasonality beyond the inclusion of quarter-year fixed effects. We also did not observe any hospitalizations if they occurred outside of a given state. Finally, we are unable to rule out the presence of other confounders that may have differentially impacted postpartum hospitalizations in expansion and non-expansion states.

STUDY RESULTS

Prior to the Medicaid expansions, the five most common reasons for hospitalizations during both the first 60 days and the 61 days to six-month postpartum period for birthing people with a Medicaid-financed delivery were pregnancy or childbirth-related hospitalizations, digestive system diseases, mental disorders, genitourinary system diseases, and injury and poisoning (exhibit 1). Pregnancy or childbirth-related hospitalizations represented a much larger share of postpartum hospitalizations within the first 60 days (70%) than during the later period (8%).

Exhibit 1:

Primary diagnoses for postpartum hospitalizations among patients in 8 states with Medicaid-financed deliveries, by timing of hospitalization, 2010–13

SOURCE: Authors’ analysis using Florida, Georgia, Iowa, Maryland, Mississippi, New Mexico, Utah, and Washington State Inpatient Databases, Healthcare Cost and Utilization Project, Agency for Healthcare Research and Quality.

NOTES: Analysis of postpartum hospitalizations for patients with deliveries between 2010–2013.

Between 61 days and six months postpartum, there were approximately 50 percent fewer hospitalizations overall. During this period, diseases of the digestive system (25.9%), mental disorders (14.9%), diseases of the genitourinary system (10.8%), and injury and poisoning (8.3%) all surpassed pregnancy-related complications as the most common reasons for hospitalization.

Exhibit 2 reports the unadjusted comparison of rates of postpartum hospitalization for birthing people with a Medicaid-financed delivery before and after the ACA Medicaid expansions in states with and without expansions. We document that, after the ACA Medicaid expansions, birthing people in expansion states experienced reductions in hospitalizations within 60 days postpartum. In contrast, birthing people in non-expansion states experienced an increase in 60-day postpartum hospitalizations. Hospitalizations between 61 days and six months postpartum did not change substantially after Medicaid expansion in either group of states.

Exhibit 2:

Unadjusted rates of postpartum hospitalization among people with Medicaid-financed deliveries in 6 states before (2010–13) and after (2014–17) Medicaid expansion, by state expansion status

SOURCE: Authors’ analysis using Florida, Georgia, Iowa, New Mexico, Utah, and Washington State Inpatient Databases, Healthcare Cost and Utilization Project, Agency for Healthcare Research and Quality.

NOTES: Excluded two states in the sample with missing years of data during the study period. Statistically significant difference in hospitalization rates between 2010–2013 and 2014– 2017 denoted by an asterisk. * p < 0.10, **p < 0.05, ***p < 0.01.

Based on the event study estimates, we found no evidence of differential changes in postpartum hospitalizations between birthing people in the two groups of states during the pre-period. Starting at the time of expansion, the relative trend in hospitalizations within 60 days postpartum decreased in expansion states compared to non-expansion states (exhibit 3). Over the entire post-period, we found that Medicaid expansion decreased the rate of 60-day postpartum hospitalizations by 0.289 percentage points (p = 0.052), or a 17 percent decrease over the baseline rate (exhibit 4). Approximately 75 percent (−0.216 percentage points; p = 0.056) of this decline can be attributed to a decrease in childbirth-related hospitalizations (see appendix for more details).²⁰

Exhibit 3:

Effects of Medicaid expansions on 60-day postpartum hospitalizations after Medicaid-financed deliveries in 8 states, 2010–1

SOURCE: Authors’ analysis using Florida, Georgia, Iowa, Maryland, Mississippi, New Mexico, Utah, and Washington State Inpatient Databases, Healthcare Cost and Utilization Project, Agency for Healthcare Research and Quality.

NOTES: Graph depicts coefficient estimates and their 95% confidence intervals from the event study model described in the text.

EXHIBIT 4:

Effects of ACA Medicaid Expansions on Postpartum Hospitalizations Following Medicaid-Financed Deliveries, Overall and By Race and Ethnicity

	Within 60 Days of Delivery				61 Days to 6 Months of Delivery
	Baseline mean in expansion states (%)	Difference-in-differences estimate			Baseline mean in expansion states (%)	Difference-in-differences estimate
		Percentage-point change	(95% CI)	N		Percentage-point change	(95% CI)	N
All	1.741	−0.289*	(−0.499 to 0.004)	1,355,220	1.048	−0.085	(−0.369 to 0.326)	880,014
White, non-Hispanic	1.742	−0.251	(−0.494 to 0.117)	541,383	1.192	−0.137	(−0.603 to 0.703)	351,969
Black, non-Hispanic	2.436	−0.279	(−0.864 to 1.156)	428,628	1.194	−0.086	(−1.067 to 0.617)	278,456
Hispanic	1.483	−0.181**	(−0.306 to −0.030)	296,570	0.850	−0.042	(−0.250 to 0.429)	191,296

SOURCE: Authors’ analysis using Florida, Georgia, Iowa, Maryland, Mississippi, New Mexico, Utah, and Washington State Inpatient Databases, Healthcare Cost and Utilization Project, Agency for Healthcare Research and Quality.

NOTES: Baseline means in expansion states are calculated among deliveries that occurred between 2010 and 2013. Statistically significant estimates denoted by an asterisk.

^*p < 0.10

^**p < 0.05

^***p < 0.01.

We found some evidence of a decrease in postpartum hospitalizations between 61 days and six months. The coefficient estimates from the event study were negative following expansion, suggesting a relative decline in expansion states (see appendix).²⁰ The difference-in-differences estimate indicates a small decline in postpartum hospitalizations (−0.085 percentage points; p = 0.752), an eight percent decrease over the baseline rate (exhibit 4), but is imprecisely estimated. We note that analysis of this outcome was less well powered than the 60-day hospitalization outcome due to a smaller sample size and lower baseline prevalence.

In an additional analysis, we examined whether there was a change in 60-day postpartum hospitalizations using the same sample as for this outcome (i.e., deliveries occurring within the first half of each calendar year). The effect of Medicaid expansion on 60-day postpartum hospitalizations was similar in magnitude to the main estimate, but less precise (coefficient: −0.244; 95% CI: −0.528, 0.189). We also re-ran the adjusted analyses when excluding the two study states with data missing for certain years. The results were extremely similar. More detailed descriptions of these analyses may be found in the appendix.²⁰

Reported in exhibit 4, we examined changes associated with Medicaid expansion separately by individual’s race and ethnicity. Event study results for this analysis are reported in the appendix.²⁰ The baseline rate of postpartum hospitalization during the first 60 days was 2.4 percent among non-Hispanic Black individuals, compared to between 1.7 percent for non-Hispanic White individuals and 1.5 percent for Hispanic individuals. We estimated negative difference-in-differences coefficients in 60-day postpartum hospitalizations for all groups, suggestive of declines. The estimate for Hispanic individuals had the lowest p-value, but we were unable to rule out similarly sized changes for other racial groups. We also estimated negative coefficients in 61-day to six-month postpartum hospitalizations for all groups, although similar to the overall estimate, these were accompanied by wide confidence intervals.

DISCUSSION

Previous research suggests that while hospitalizations are approximately eight times more likely among postpartum women than non-postpartum women during the first 21 days postpartum, the difference between groups becomes much smaller later in the postpartum period.²¹ Consistent with these findings, we find that only eight percent of hospital admissions after 60 days for postpartum people with Medicaid-financed deliveries were related to childbirth, and that there were nearly twice as many overall hospitalizations during the first 60 days compared to between 61 days and six months postpartum for this population. Some postpartum hospitalizations after 60 days for other causes may still relate to pregnancy and childbirth, such as diseases of the genitourinary system,²² or if they are due to conditions that are exacerbated by pregnancy, such as mental health-related conditions. This suggests a need for long-term continuous health care among people recently giving birth.

We found a 17 percent decrease in the occurrence of 60-day postpartum hospitalizations among people with Medicaid-financed deliveries in states with ACA Medicaid expansions, as compared to those in non-expansion states. This is notable given that Medicaid pregnancy coverage only expires after 60 days. One potential explanation for this finding is that expanded Medicaid may affect postpartum outcomes by increasing coverage and access to care prior to pregnancy. For instance, some researchers have documented increased preconception health care use, or the diagnosis and treatment of chronic health conditions under expanded Medicaid.^9,23–25 In addition, greater insurance coverage before pregnancy has been linked to more timely prenatal care receipt.^10,14,26 Establishing care earlier during pregnancy may help to either avoid or better manage conditions that may otherwise lead to postpartum hospitalizations.

The event study estimates, albeit imprecise, suggest that the decrease in 60-day hospitalizations occurred as soon as 2014. Most postpartum persons in our sample during this first expansion year (i.e., those who gave birth during the first three quarters of 2014) could not have benefitted yet from any changes in preconception coverage. This indicates that expanded Medicaid may also be important for early decisions regarding postpartum care during the 60 days covered by pregnancy Medicaid. For instance, this may be the case if the Medicaid expansions increase patient knowledge about coverage during the postpartum period or affect their interactions with providers or recommended care. This is consistent with one study documenting an increase in postpartum visits before the 60-day pregnancy coverage cutoff after Medicaid expansion in Arkansas.⁷

To try to better understand the decrease in 60-day postpartum hospitalizations, we further examined changes in occurrence by the primary reason for hospitalization (childbirth-related vs. all other diagnoses). This analysis suggested declines in both types, although the change was largest for childbirth-related hospitalizations. The estimated change for this outcome represented about 75% of the size of the overall decline in postpartum hospitalizations, which is roughly in line with their representation among the primary reasons for postpartum hospitalizations during this 60-day period (see exhibit 1). Given that 60-day postpartum hospitalizations are relatively rare events (occurring for less than two percent of postpartum persons in our sample, see exhibit 4), we were underpowered to examine more specific reasons for hospitalizations. However, this analysis suggests that the ACA Medicaid expansions likely affected all types of hospitalizations.

Our analysis suggested a smaller decrease (8%) in hospitalizations between 61 days and six months postpartum, but was imprecisely estimated with a wide confidence interval. Analysis of this outcome was less well-powered due to a lower baseline prevalence (only one percent of postpartum people are hospitalized during this period, see exhibit 4) and a data limitation that required a smaller analytic sample than for the 60-day postpartum hospitalization outcome. When we used the same sample for the 60-day measure, we similarly found a wider confidence interval on our estimated decline in postpartum hospitalizations under the ACA Medicaid expansions.

Consistent with existing research documenting racial disparities in postpartum readmissions,²⁷ we found that rates of postpartum hospitalization were much higher among non-Hispanic Black individuals with Medicaid-financed deliveries, as compared to all other racial and ethnic groups. Our analysis suggested there were declines in 60-day, as well as 61-day to six-month, postpartum hospitalizations for all groups under the ACA Medicaid expansions, although the estimates were again imprecise.

As part of the 2021 American Rescue Plan (ARP), states have the option to extend pregnancy Medicaid coverage through 12 months postpartum and, as of August 2022, 22 states and Washington, D.C., have implemented this extension.²⁸ Importantly, seven states that have implemented or are planning to implement extensions do not have ACA Medicaid expansions in place. This study’s findings suggest that expanded coverage during the postpartum period in these states may lead to reductions in postpartum hospitalizations during the first six months following delivery. Not only do we find evidence of a decrease in 60-day postpartum hospitalizations, but some evidence suggesting a decrease beyond this period (61 days to six months). It is possible that effects might be even larger under the ARP postpartum extensions since, under this policy, coverage is automatic for individuals with pregnancy Medicaid and income eligibility thresholds for pregnancy coverage surpass the 138% FPL under the ACA Medicaid expansions. Future research focusing on a larger number of states, with sufficient power to examine the effects of extensions on mental health and other common causes of pregnancy-related morbidity in the later postpartum period is needed to determine the full effects of pregnancy Medicaid extensions.

CONCLUSION

This study provides the first evidence of a decrease in postpartum hospitalizations associated with expanded Medicaid. Our findings suggest that ongoing efforts to expand Medicaid may contribute to better postpartum health for US birthing people.

APPENDIX

A. Selection of SID States to Include in Analysis

Our analysis uses State Inpatient Databases from the Healthcare Cost and Utilization Project of the Agency for Healthcare Research and Quality. Our study sample included all SID states with at least one year pre- and post-data available during the 2010–2017 years with supplemental variables for revisit analyses (see Table A1),¹ which are necessary to track patients longitudinally. We excluded 3 states (Massachusetts, New York, Vermont) due to their presence of pre-ACA Medicaid or insurance expansions comparable to the ACA Medicaid expansions, as well as one state (Wisconsin) that implemented a non-ACA expansion in 2014.

This left us with a sample of 10 states (Arkansas, Florida, Georgia, Iowa, Maryland, Mississippi, Nebraska, New Mexico, Washington, and Utah) with consistent data available for the study period. Two of these states are missing some years of data during the study period. Maryland does not have data available with the supplemental variables for revisit analyses in 2010–2011, and the variables were missing for the majority of hospitalizations in 2012. For this reason, we exclude the 2012 Maryland data from the analysis. In addition, Mississippi does not have SID data available in 2012. We excluded these two states from the unadjusted analysis of postpartum hospitalizations presented in Exhibit 2.

Upon the initial sample construction, which restricted the sample to people with a delivery funded by Medicaid, we discovered that two states (Arkansas and Nebraska) had observable changes in Medicaid-financed deliveries during the study period that appeared to be unrelated to the ACA Medicaid expansions (see Figure A1). For this reason, we excluded these states from the analysis. This left us with a final sample of 8 states – 4 expansion (Iowa, Maryland, New Mexico, and Washington) and 4 non-expansion (Florida, Georgia, Mississippi, and Utah).

B. Details on Sample Construction

The main sample was defined as all deliveries with Medicaid listed as the primary payer in each state and year included in the analysis. Deliveries were identified using an algorithm and code provided by Elena Kuklina at the CDC that is described in further detail in Kuklina et al.² We limited deliveries to those where the patient was a resident in the hospitalization state. In stratified analyses, we examined individuals of non-Hispanic white, non-Hispanic Black, and Hispanic race/ethnicity. We did not separately examine non-Hispanic other race individuals due to their small presence in the overall sample (less than 7% of patients).

When examining postpartum utilization, we tracked individual patients with a non-missing linkage variable (i.e. “verified person number,” visitLink) provided by HCUP. The visitLink variable is created by HCUP using unique encrypted patient identifiers, date of birth, and sex information.

Using the visitLink variable, information on admission dates for each patient hospital event relative to an index event (DaysToEvent), and the length of stay variable (LOS), we were able to construct measures of postpartum utilization across two different observation windows (60 days following delivery discharge, and 61 days to 6 months following delivery discharge). In a very small number of cases (<1%), there were duplicate records for the same event in the HCUP files. We retained only the first record in each of these cases.

Due to changes in states coding schemes for the visitLink variable across data years, the visitLink variable cannot be used to track patients across years. See Table A2 for details for our states. For this reason, we limit our observations to patients with deliveries that occur within the first three quarters of the calendar year when examining postpartum outcomes that occur within 60 days. We limit observations to patients with deliveries that occur within the first two quarters of the calendar year when examining postpartum outcomes that occur within 6 months. This ensures that we observe the full postpartum period window for all states and years with SID data and the revisit variables available.

C. Details on Hospitalization Classification

We examined primary diagnoses for postpartum hospitalizations during the pre-period using body system indicators classified according to the International Statistical Classification of Diseases and Related Health Problems. We examined these characteristics separately for hospitalizations during the first 60 days postpartum and between 61 days and 6 months postpartum. We report the 5 most common diagnoses in Exhibit 1 in the text. The full list of diagnoses is reported in Appendix Table A3.

D. Details on Regression Model

For each outcome Y, we estimate the following difference-in-differences model:

$${\text{Y}}_{\text{isqt}}={\text{δ}}_{\text{s}}+{\text{δ}}_{\text{qt}}+\text{β}{\text{ExpansionState}}_{\text{s}}\times {\text{Post2014}}_{\text{t}}+\mathrm{\Psi }{\text{X}}_{\text{isqt}}+{\text{ε}}_{\text{isgt}},$$ Equation (1)

where i denotes the an individual patient, s denotes the state, q is the quarter, and t is year. In this model, X_isqt includes the following individual characteristics: indicators for race and ethnicity groups (non-Hispanic white, non-Hispanic Black, Hispanic, non-Hispanic other) and single year of age indicator variables. Indicator variables for each state are denoted by δs and indicators for each quarter-year are denoted by δ_qt. The variable ExpansionState_s is an indicator variable that equals 1 if the respondent lives in a state that adopted the ACA Medicaid expansion. The coefficient on β, captures the mean adjusted difference across the expansion and non-expansion states in the change in the outcome variable Y from the pre-ACA years (2010–2013) to the post- ACA years (2014–2017).

The event study version replaces the ExpansionState_s x Post2014_t term with a series of year indicators interacted with ExpansionState_s. The year prior to expansion is omitted. It is given by:

$${\text{Y}}_{\text{isqt}}={\text{δ}}_{\text{s}}+{\text{δ}}_{\text{qt}}+{\text{ExpansionState}}_{\text{s}}\times {\sum }_{\begin{array}{l}!\#\$\%\\ !\&{\text{\$}}^{\text{'}}\end{array}}^{"}"!\text{I}\left(\text{t}-{\text{t}}_{\text{s}}\text{*}=\text{y}\right)+{\text{Ψ}\mathrm{X}}_{\text{isqt}}+{\text{ε}}_{\text{isqt}}.$$ Equation (2)

The event study results for the 61-day to 6 month postpartum hospitalizations outcome are reported in Appendix Figure A2.

In addition to estimating the difference-in-differences model for the full sample, we also conducted stratified analyses for the non-Hispanic white, non-Hispanic Black, and Hispanic racial groups. To test for significant differences across racial groups in our estimated effects of the ACA Medicaid expansion, we implemented a fully interacted triple difference model that included race x age dummies, race x state dummies, and race x quarter-year dummies, in addition to separate expansion indicators for each racial group. The event study results for the stratified analyses may be found in Appendix Figures A3 and A4.

All confidence intervals and p-values are estimated using 1,000 replications under a wild cluster bootstrap procedure with a Webb weight.

E. Analysis of Postpartum Hospitalizations by Type

In addition to the overall change in 60-day postpartum hospitalizations, we also broke out 60-day postpartum hospitalizations into two categories: (1) those with primary diagnoses related to complications of pregnancy, childbirth, and the puerperium, and (2) all other diagnoses. Analyses of changes in these two categories of hospitalizations under the ACA Medicaid expansions suggested declines in both, although the change was largest for childbirth-related hospitalizations (see Appendix Table A4). The estimated coefficient for this outcome indicates a decline of 0.216 percentage points, which represents about 75% of the size of the overall estimated coefficient in Exhibit 4. This suggests that the decline in childbirth-related hospitalizations is roughly equivalent to their representation among the primary reasons for postpartum hospitalizations during this period (see Exhibit 1 – they represent 70% of 60-day postpartum hospitalizations at baseline).

F. Sensitivity to Dropping States with Missing Data Years

We included two states with some missing data years (Maryland: 2010–2012, Mississippi: 2012) in our adjusted analyses presented in Exhibits 2–4. In a sensitivity analysis, we re-ran the difference-in-differences and event study models when excluding these states. The results are very similar. See Appendix Figure A5 and Appendix Table A5.

G. Sensitivity to Dropping Third Quarter Births

The analyses for postpartum hospitalizations during the 60 days following delivery included patients giving birth during the first 3 quarters of any calendar year, while the analyses for postpartum hospitalizations use patients giving birth during the first 2 quarters of any calendar year. To determine whether our results were sensitive to this difference in sample criteria for the two outcomes, we re-ran the analysis of 60-day postpartum hospitalizations using only patients who delivered during the first 2 quarters of the calendar year (i.e. the same sample criteria used for the 60 day to 6 month postpartum outcome). We lose some precision in this analysis, but the coefficient estimates are very similar. See Appendix Table A6.

I. Appendix Figures

Figure A1.

Share of Medicaid-Financed Deliveries by State

Figure A2.

Effects of ACA Medicaid Expansions on 61-Day to 6-Month Postpartum Hospitalizations Following Medicaid-Financed Deliveries

Figure A3.

Event Studies for Stratified Analyses: Postpartum Hospitalizations within 60 Days of Delivery

Figure A4.

Event Studies for Stratified Analyses: Postpartum Hospitalizations 61 Days to 6 Months of Delivery by Race/Ethnicity

Figure A5.

Sensitivity of Event Study Estimates to Dropping MD and MS

J. Appendix Tables

Table A1.

Selection of HCUP SID States

	Years Available During Study Period	Has Pre- and Post-Data Available	Excluded Due to Non-ACA Expansions
Alaska	2015–2017	N
Arkansas	2010–2017	Y
California	2010–2011	N
Delaware	2016–2017	N
Florida	2010–2017	Y
Georgia	2010–2017	Y
Iowa	2010–2017	Y
Maryland	2012–2017	Y
Massachusetts	2010–2017	Y	Y
Mississippi	2010–2011, 2013–2017	Y
Nebraska	2010–2017	Y
New Mexico	2010–2017	Y
New York	2010–2017	Y	Y
North Carolina	2010	N
Utah	2010–2017	Y
Vermont	2011–2017	Y	Y
Washington	2010–2017	Y
Wisconsin	2013–2017	Y	Y

Source: Healthcare Cost and Utilization Project (2020)

Table A2.

Availability of Hospitalization Data and Revisit Variables by State and Year

	Data Years	VisitLink Availability	Patient Tracking Across Years
Expansion States
Iowa	2010–2017	All Years	2010–2012, 2013–2016
Maryland**	2010–2017	2013–2017	2013–2017
New Mexico	2010–2017	All Years	2015–2016
Washington	2010–2017	All Years	None
Non-Expansion States
Florida	2010–2017	All Years	All Years
Georgia	2010–2017	All Years	2010–2011, 2013–2017
Mississippi*	2010–2011, 2013–2017	All Years	2010–2011, 2013–2016
Utah	2010–2017	All Years	2010–2014, 2015–2017

Source: Healthcare Cost and Utilization Project (2020)

^*Patient tracking is available between 2016 and 2017 in MS, but the percent of visitLink values that are reported in consecutive data years appears to drop off for this state in 2016–2017 as compared to prior years. For this reason, we do not list that linkage here.

^**While visitLink is available in MD in 2012, it is missing for the majority of hospitalizations. For this reason, we do not list its 2012 availability here.

Table A3.

Diagnoses for Postpartum Hospitalizations Among Patients with Medicaid-Financed Deliveries by Timing of Hospitalization

	2010–2013		2014–2017
	<=60 days	61 days - 6 months	<=60 days	60 days - 6 months
Primary diagnosis code related to:
Infectious and parasitic diseases	1.0%	4.4%	2.0%	7.8%
Neoplasms	0.7%	2.2%	0.5%	2.3%
Endocrine, nutritional and metabolic diseases, and immunity disorders	0.5%	3.2%	0.5%	3.6%
Diseases of the blood and blood-forming organs	0.7%	3.2%	0.6%	2.8%
Mental disorders	4.8%	14.9%	5.5%	18.4%
Diseases of the nervous system and sense organs	0.8%	1.9%	0.6%	2.5%
Diseases of the circulatory system	1.7%	3.8%	1.8%	4.5%
Diseases of the respiratory system	1.3%	3.7%	1.3%	3.9%
Diseases of the digestive system	9.7%	25.9%	7.8%	22.8%
Diseases of the genitourinary system	3.3%	10.8%	2.5%	8.0%
Complications of pregnancy, childbirth, and the puerperium	70.0%	8.0%	72.1%	8.5%
Diseases of the skin and subcutaneous tissue	0.9%	3.0%	0.8%	2.7%
Diseases of the musculoskeletal system and connective tissue	0.3%	1.6%	0.3%	1.4%
Congenital anomalies	0.1%	0.1%	0.1%	0.1%
Certain conditions originating in the perinatal period	0.0%	0.0%	0.0%	0.0%
Symptoms, signs, and ill-defined conditions	1.2%	3.9%	0.8%	2.7%
Injury and poisoning	2.6%	8.3%	2.3%	7.0%
Factors influencing health status and contact with health services	0.5%	1.1%	0.6%	1.0%
N	15,059	6,998	17,019	7,734

Source: Authors’ analysis using Florida, Georgia, Iowa, Maryland, Mississippi, New Mexico, Utah, and Washington State Inpatient Databases, Healthcare Cost and Utilization Project, Agency for Healthcare Research and Quality. Analysis of postpartum hospitalizations for patients with Medicaid-financed deliveries between 2010–2013.

Appendix Table A4.

Effects of ACA Medicaid Expansions on 60-Day Postpartum Hospitalizations Following Medicaid-Financed Deliveries by Type

	Baseline mean in expansion states (%)	Difference-in-differences estimate
		Percentage-point change	(95% CI)	P-value	N
Childbirth-related	1.280	−0.216	(−0.378 to 0.004)	0.056	1,355,220
All other	0.461	−0.074	(−0.182 to 0.210)	0.450	1,355,220

Appendix Table A5.

Difference-in-Differences Estimates When Excluding MD and MS

	Within 60 Days of Delivery					60 Days to 6 Months of Delivery
	Baseline mean in expansion states (%)	Difference-in-differences estimate				Baseline mean in expansion states (%)	Difference-in-differences estimate
		Percentage-point change	(95% CI)	P-value	N		Percentage-point change	(95% CI)	P-value	N
All	1.711	−0.310	(−0.533 to 0.027)	0.090	1,176,789	1.027	−0.133	(−0.432 to 0.306)	0.772	764,861
White, non-Hispanic	1.725	−0.250	(−0.534 to 0.475)	0.43	484,439	1.160	−0.164	(−0.678 to 1.293)	0.860	314,951
Black, non-Hispanic	2.522	−0.385	(−5.221 to 2.421)	0.386	342,338	1.004	0.122	(−1.610 to 1.939)	0.788	222,858
Hispanic	1.491	−0.182	(−0.599 to −0.037)	0.034	271,918	0.853	−0.101	(−0.481 to 0.136)	0.2	175,614

Appendix Table A6.

Difference-in-Differences Estimates for Subsample of Patients Delivering During the First Two Calendar Quarters

	Within 60 Days of Delivery
	Baseline mean in expansion states (%)	Difference-in-differences estimate
		Percentage-point change	(95% CI)	P-value	N
All	1.726	−0.244	(−0.528 to 0.189)	0.394	880,014
White, non-Hispanic	1.761	−0.237	(−0.564 to 0.370)	0.464	351,969
Black, non-Hispanic	2.349	−0.259	(−1.048 to 1.243)	0.358	278,465
Hispanic	1.452	−0.103	(−0.302 to 0.208)	0.584	191,296