Late Postpartum Coverage Loss Before COVID-19: Implications For Medicaid Unwinding

Introduction

Extending pregnancy-related Medicaid eligibility to one year postpartum is one of the Biden administration’s key strategies for addressing long-standing maternal health inequities in the US.¹ Before the COVID-19 pandemic, nearly one-third of prenatal Medicaid enrollees did not have consistent Medicaid coverage through the late postpartum period (exhibit 1). In response to pandemic-related stresses, the March 2021 American Rescue Plan Act included a new option for states to implement postpartum Medicaid extensions using federal matching funding.² As of March 2023, thirty-seven states and Washington, D.C., had implemented or planned to implement this op tion.² However, these state-level extensions are moot while the national “maintenance of effort” protections that have prevented states from disenrolling Medicaid beneficiaries since the onset of the COVID-19 public health emergency in March 2020 remain in effect. As these protections “unwind” in 2023, following the end of the public health emergency declaration in May, states are resuming Medicaid coverage redeterminations. In the thirteen states that have not yet adopted the American Rescue Plan Act option, this will result in a return to the prepandemic status quo of pregnancy-related Medicaid eligibility ending after sixty days postpartum.

Exhibit 1.

Percent of respondents with consistent insurance from pregnancy to the late postpartum period, by insurance type during pregnancy in 7 states, 10/2019–02/2020

Source: Authors’ analysis of data from PRAMS and the Opioid call-back survey from seven states: Kentucky, Louisiana, Massachusetts, Missouri, Pennsylvania, Utah, and West Virginia. Notes: Bar charts represent the % with consistent coverage type by prenatal coverage category.

Data on prepandemic insurance patterns in the postpartum year could inform states’ assessment of the potential impact of not adopting the American Rescue Plan Act option. Prior studies using Pregnancy Risk Assessment Monitoring System (PRAMS) data have shown that insurance loss in the early postpartum period is common among Medicaid beneficiaries; for example, during 2015–18, 22 percent of people with Medicaid-paid births were uninsured by two to four months postpartum, despite childbirth being a qualifying event for enrollment in subsidized Marketplace coverage.³ Research using data from the American Community Survey found that in 2017, 15 percent of low-income people with a birth in the past twelve months were uninsured,⁴ but the American Community Survey data do not include information on the time since birth or on prenatal coverage type. Lack of data has precluded multistate or national estimates of coverage patterns in the later postpartum period during which the American Rescue Plan Act option will apply.

Drawing on unique prepandemic multistate survey data, this article presents novel findings on longitudinal insurance patterns from pregnancy to the early (three to five months after birth) and late (nine to ten months after birth) postpartum periods. The results provide policy makers with evidence on the post–public health emergency insurance patterns that may be expected to return in states that do not adopt the American Rescue Plan Act option.

Study Data And Methods

We used follow-up survey data from the 2019 PRAMS Opioid Call-Back Survey, conducted nine to ten months postpartum in seven states: Kentucky, Louisiana, Massachusetts, Missouri, Pennsylvania, Utah, and West Virginia.⁵ The Opioid Call-Back Survey was a new one-time initiative that sampled PRAMS respondents who had a live birth between January and June 2019 and who did not opt out of being recontacted. The follow-up survey data were collected from October 2019 through April 2020, with a median response rate of 61 percent (Beatriz Salvesen von Essen, Centers for Disease Control and Prevention, personal communication, May 3, 2022). The Opioid Call-Back Survey data were linked to PRAMS, which was conducted, on average, four months postpartum, with a median response rate of 58 percent in these states.⁶ The COVID19 pandemic significantly affected postpartum insurance patterns as a result of federal requirements that prevented Medicaid disenrollment during the public health emergency.⁷ Thus, we restricted our analysis to responses before the March 2020 public health emergency declaration (90.6 percent of the survey).We also excluded people younger than age twenty because of differences in children’s Medicaid eligibility.

Individual demographic characteristics included race and ethnicity, language of survey completion, age, income, education, marital status, parity, and whether a respondent was working for pay at nine to ten months postpartum. Health characteristics included chronic (diabetes or hypertension) and behavioral (perinatal depression, anxiety, substance use, or opioid misuse) health conditions diagnosed before, during, or after pregnancy. Opioid misuse included opioids used for nonmedical purposes or obtained through nonmedical sources.

State-level policies considered in our models included state Medicaid expansion status (expansion states: Kentucky, Louisiana, Massachusetts, Pennsylvania, and West Virginia; nonexpansion states: Utah and Missouri) and policies for insuring pregnant immigrants, including the Children’s Health Insurance Program (CHIP) unborn child option (Louisiana, Massachusetts, and Missouri), which considers the fetus to be a child eligible for CHIP during pregnancy; and waiving the five-year waiting period that legal immigrants must otherwise face to be eligible for Medicaid (Massachusetts, Pennsylvania, and West Virginia). Although we categorized Utah as a nonexpansion state, Utah implemented partial expansion up to 100 percent of the federal poverty level in April 2019 and full expansion up to 138 percent of poverty in January 2020. As a result, all early postpartum data in Utah were collected during partial expansion, whereas 44 percent of the nine-to-ten-month follow-up data were collected during full Medicaid expansion. Main results excluding Utah are in online appendix A1.⁸

The primary outcomes were self-reported insurance at three points in time: during pregnancy (prenatal care payer), early postpartum (mean: 4.0 months; interquartile range: 3–5 months postpartum), and late postpartum (mean: 9.3 months; IQR: 9–10 months postpartum). Because respondents could indicate multiple insurance types, we categorized insurance hierarchically as commercial, Medicaid, or uninsured. Commercial coverage included military or commercial insurance in combination with other coverage. Uninsured included no coverage or Indian Health Service coverage only.

We examined the proportion of respondents who maintained a consistent coverage type (Medicaid, commercial, uninsured) across the three time points, based on prenatal coverage category. We then created Sankey diagrams to visualize coverage trajectories. The main Sankey diagram displays the proportion of the sample by prenatal coverage type and follows the trajectories of each prenatal coverage group through the early and late postpartum periods. Appendix A3 displays the percentage of the sample by early postpartum coverage type and follows the trajectories of each group to the late postpartum period.⁸

We estimated unadjusted survey-weighted rates of late postpartum uninsurance by state policy characteristics overall, as well as among respondents with prenatal Medicaid. We estimated state-adjusted predicted probabilities of late postpartum uninsurance by respondent demographics both overall and among respondents with prenatal Medicaid, using logistic regression models controlling for state fixed effects.

Study limitations included the relatively small sample from only seven states, which limited statistical power, exploration of the role of state policy characteristics, and generalizability. Furthermore, five of the seven sampled states were Medicaid expansion states, which have higher rates of stable Medicaid coverage compared with nonexpansion states.³ Finally, PRAMS data are self-reported, which could lead to measurement errors or biases, such as underreporting stigmatized characteristics.

Study Results

Our sample included 1,742 respondents representing a weighted total of 151,562 postpartum people. Appendix A2 describes the survey sample characteristics.⁸ Overall, we found that among those with prenatal Medicaid, only 68 percent had continuous Medicaid coverage through to the late postpartum period (exhibit 1). Among respondents who were uninsured during pregnancy, 75 percent were continuously uninsured through to the late postpartum period. People with prenatal commercial coverage reported the most stable coverage, with 85 percent maintaining commercial coverage through to the late postpartum period.

Exhibit 2 follows coverage trajectories for each prenatal coverage group through the late postpartum period. During pregnancy, 63 percent of the sample reported commercial coverage, 35 percent reported Medicaid, and 2 percent reported being uninsured. Among respondents with prenatal commercial coverage, 93 percent and 85 percent remained commercially insured in the early and late postpartum periods, respectively. Eleven percent of respondents with prenatal Medicaid became uninsured in the early postpartum period, and nearly two-thirds of those remained uninsured in the late postpartum period (64 percent of the 11 percent). Appendix A3 displays coverage trajectories from the early to late postpartum period.⁸

Exhibit 2.

Insurance continuity in the postpartum period, by insurance type during pregnancy in 7 states, 10/2019–02/2020

Source: Authors’ analysis of data from PRAMS and the Opioid call-back survey from seven states: Kentucky, Louisiana, Massachusetts, Missouri, Pennsylvania, Utah, and West Virginia. Notes: Pregnancy coverage comprises the % of the sample with private, Medicaid, or no coverage during prenatal care receipt. Early postpartum and late postpartum coverage is shown as a % of each pregnancy coverage category.

Exhibit 3 displays uninsurance rates in the late postpartum period by state policy characteristics overall and among respondents with prenatal Medicaid. The overall rate of late postpartum uninsurance was 7.9 percent. Late postpartum uninsurance was lower in Medicaid expansion states and states with coverage options for pregnant immigrants, but differences were not statistically significant. Among respondents with prenatal Medicaid, the rate of late postpartum insurance was 11.3 percent; however, Medicaid nonexpansion states had significantly higher rates of late postpartum uninsurance compared with nonexpansion states (19.8 percent compared with 9.2 percent).

Exhibit 3.

Prevalence of late postpartum uninsurance by state policy characteristics in 7 states, 10/2019–02/2020

Characteristics	Late Postpartum Uninsurance, Total Sample (N=1,742)		Late Postpartum Uninsurance, Respondents with Prenatal Medicaid (N=604)
	%	(95% CI)	%	(95% CI)
Overall Weighted Proportion	7.9	(6.1, 10.2)	11.3	(7.7, 16.2)
Medicaid Expansion Status
Non-Expansion State	10.8	(7.7, 15.1)	19.8	(11.3, 32.4)
Medicaid Expansion State	6.9	(4.8, 9.7)*	9.2**	(5.5, 14.9)
Policies for Pregnancy Coverage for Immigrants
No	10.1	(7.2, 14)	7.0	(3.6, 13.4)
Unborn Child Option for CHIP	6.5	(4.6, 9.3)	14.0	(9.3, 20.4)
5-Year Waiting Period for Medicaid Waived	6.2	(3.7, 10.3)	7.4	(3.0, 17.2)

Source: Authors’ analysis of data from PRAMS and the Opioid call-back survey from seven states: Kentucky, Louisiana, Massachusetts, Missouri, Pennsylvania, Utah, and West Virginia. Notes: Weighted proportions presented by state policy characteristics. Although we have categorized Utah as a non-expansion state, Utah implemented partial expansion in April 2019, with full Medicaid expansion implemented as of January 2020, during our study period. Chi-2 tests used to assess differences in late postpartum uninsurance by state policy characteristics.

^*p < 0.1,

^**p < 0.05,

^***p < 0. 01,

^****p < 0.001.

The state-adjusted predicted probability of late postpartum uninsurance was substantially higher among Hispanic respondents (25.6 percent) relative to non-Hispanic White respondents, and among respondents who completed the survey in Spanish (49.3 percent) relative to respondents who completed the survey in English (exhibit 4). Among respondents with prenatal Medicaid, the predicted probability of late postpartum uninsurance was highest among respondents who completed the survey in Spanish (51.4 percent), which was significantly higher relative to respondents who completed the survey in English (9.4 percent).

Exhibit 4.

State-adjusted predicted probabilities of late postpartum uninsurance by respondent demographic characteristics in 7 states, 10/2019–02/2020

Demographics	Late Postpartum Uninsurance, Total Sample (N=1,742)		Late Postpartum Uninsurance, Respondents with Prenatal Medicaid (N=604)
	%	(95% CI)	%	(95% CI)
Race and Ethnicity
Non-Hispanic White	6.2	(3.8, 8.6)	12.3	(5.8, 18.7)
Hispanic	25.6****	(15.9, 35.4)	19.7	(9.9, 29.6)
Non-Hispanic Black	4.0	(1.4, 6.7)	3.9**	(0.3, 7.6)
Non-Hispanic Other	8.8	(0.0, 18.2)	18.4	(0.0, 46.2)
Survey Language
English	6.2	(4.3, 8.1)	9.4	(5.2, 13.6)
Spanish	49.3****	(31.6, 67.1)	51.4****	(32.6, 70.1)
Age at Childbirth
20–24 Years	9.3	(3.8, 14.9)	10.7	(1.0, 20.4)
25–29 Years	7.2	(4.0, 10.5)	11.1	(5.0, 17.3)
30–34 Years	6.2	(3.1, 9.3)	7.7	(2.4, 13.0)
35 Years or Older	10.5	(4.7, 16.4)	18.9	(3.0, 34.8)
Federal Poverty Level
<138%	11.7	(7.5, 15.9)	11.4	(5.9, 16.8)
139–199%	14.0	(7.1, 21.0)	14.9	(3.9, 26.0)
200% or Above	3.6***	(1.2, 6.0)	13.2	(0.4, 25.9)
Education
High School or Less	13.4	(8.7, 18.2)	10.5	(4.1, 17.0)
More Than High School	5.5***	(3.4, 7.6)	12.1	(6.9, 17.2)
Marital Status
Married	6.5	(4.3, 8.7)	14.9	(8.1, 21.7)
Unmarried	10.7	(6.2, 15.2)	9.5	(3.9, 15.1)
Parity
Primiparous	7.4	(3.6, 11.2)	11.1	(2.2, 20.1)
Multiparous	8.2	(5.8, 10.6)	11.3	(6.6, 16.0)
Worked for Pay, Late Postpartum Period
No	11.3	(7.7, 15.0)	12.3	(6.7, 18.0)
Yes	5.2***	(2.9, 7.4)	9.9	(3.4, 16.4)
Any Chronic Conditions – Physical Health
No	8.5	(6.0, 11.1)	12.5	(7.5, 17.6)
Yes	6.2	(3.0, 9.4)	8.2	(0.9, 15.6)
Any Behavioral Health Conditions
No	9.2	(6.3, 12.1)	16.2	(9.1, 23.2)
Yes	6.3	(4.1, 9.1)	7.0**	(1.9, 12.0)

Source: Authors’ analysis of data from PRAMS and the Opioid call-back survey from seven states: Kentucky, Louisiana, Massachusetts, Missouri, Pennsylvania, Utah, and West Virginia. Notes: State-adjusted weighted adjusted predicted probabilities presented. Other category for race/ethnicity includes self-reported Alaska Native, American Indian, Asian, Pacific Islander, mixed race, or other. Physical health chronic conditions include diabetes or high blood pressure/hypertension preconception or during pregnancy. Behavioral health conditions include perinatal depression, anxiety, substance use, or opioid misuse (including opioids used for non-medical purposes or obtained through non-medical sources). Estimates truncated at zero for negative values. Statistical significance in relation to the reference category for each variable. Reference category is the first category listed for all demographic variables.

^*p < 0.1,

^**p < 0.05,

^***p < 0. 01,

^****p < 0.001.

Discussion

Using unique postpartum follow-up survey data from seven states before the COVID-19 pandemic, we found that nearly one-third of people with prenatal Medicaid did not have consistent Medicaid coverage by the late postpartum period, with only 16 percent transitioning to commercial coverage (exhibit 2). Nearly two-thirds of people who lost Medicaid in the early postpartum period remained consistently uninsured, and three-quarters of people uninsured during pregnancy did not gain coverage by nine to ten months postpartum. These findings show the potential implications for postpartum people residing in states where postpartum extensions were not implemented as the continuous coverage requirements under the public health emergency began unwinding in April 2023. Further, our results provide baseline estimates against which to measure changes after states’ adoption of the American Rescue Plan Act option.

ACA Marketplace coverage is an alternative option for people losing pregnancy-related Medicaid after birth. Adopting a child or having a baby is a qualifying event for Marketplace enrollment, and generous subsidies are available for low-income people.⁹ However, prior research has found that Marketplace eligibility was not associated with significant changes in postpartum private coverage after implementation of the Marketplaces.¹⁰ Our results are consistent with these findings: We found that only 16 percent of people with prenatal Medicaid were enrolled in commercial coverage by nine to ten months postpartum (exhibit 2). This suggests that Marketplace eligibility is not acting as a sufficient option to cover people losing pregnancy-related Medicaid, perhaps as a result of lack of awareness, support for insurance navigation, or affordability. Targeted efforts to smooth the transition to Marketplace coverage for the postpartum population arewarranted, particularly in Medicaid nonexpansion states that do not adopt pregnancy Medicaid extensions.

Adopting the pregnancy Medicaid extensions through one year after childbirth is a more direct approach to reducing postpartum uninsurance. Our results demonstrate that before the COVID19 public health emergency, which temporarily changed coverage patterns,⁷ people who lost Medicaid after birth tended to have persistent uninsurance through the late postpartum period. Extended pregnancy Medicaid would resolve this uninsurance without the administrative burden of Marketplace enrollment or redetermination of Medicaid eligibility as a low-income adult or parent. These postpartum extensions would target people with incomes below pregnancy Medicaid levels (median: 205 percent of poverty in these states as of 2019)¹¹ and above parental Medicaid levels (138 percent of poverty in the expansion states in our sample, 21 percent in Missouri, and 60 percent in Utah as of 2019).¹²

Consistent postpartum coverage could facilitate care continuity for people with chronic physical and behavioral health conditions, with implications for addressing rising rates of maternal morbidity and mortality. Although PRAMS does not include immigration status, immigrants are disproportionately represented among Spanish-speaking and Hispanic people, groups for which we found substantially high rates of postpartum uninsurance. Postpartum immigrants will likely continue to experience higher postpartum uninsurance without federal or state policy changes such as extending postpartum coverage under the CHIP unborn child pathway or waiving five-year waiting periods for Medicaid in all states. Thus, states’ choices around immigrants’ eligibility for postpartum Medicaid extensions will play an important role in their ultimate impact on reducing uninsurance rates in the year after birth.