Differences in Self-Reported and Billed Postpartum Visits Among Medicaid-Insured Individuals

This cross-sectional study investigates the association of transitions in health care coverage with disagreement in postpartum visit use in self-reports and Medicaid claims.


Introduction
In the US, approximately half of pregnancy-related deaths occur more than 1 week after delivery. 1proving access to care during the postpartum period is a key component of the US national agenda to improve maternal health. 2 During routine postpartum visits, clinicians conduct a full assessment of physical and psychological well-being, address urgent health care needs, and coordinate a transition to primary care for patients with chronic conditions. 3Since 2018, the American College of Obstetrics and Gynecology has recommended that all birthing people have initial contact with a clinician within 3 weeks of delivery and a comprehensive postpartum visit within 3 months; however, in practice, a single routine postpartum visit is typically scheduled between 4 and 6 weeks after delivery. 3In recognition of the importance of routine postpartum visits, receipt of such a visit is included in the Healthcare Effectiveness Data and Information Set quality measures, and national progress toward 100% attendance is reported annually. 4ate Medicaid programs have been at the forefront of efforts to increase postpartum care access among individuals with lower incomes, who experience higher rates of severe maternal morbidity. 5,6These efforts include recent policy changes to expand pregnancy Medicaid eligibility and reduce immigration-related barriers to Medicaid coverage during the postpartum period. 6In addition, as of November 2023, 39 state Medicaid programs had adopted the American Rescue Plan Act option to extend pregnancy Medicaid coverage from conception through 1 year after delivery (from the previous 60 days). 7Determining whether these policies are succeeding in increasing postpartum care use will require accurate measurement between states and over time; however, estimates of postpartum visit use differ by a large margin between data sources, and little research has been conducted to reconcile differences in estimates. 8ere are 2 primary data sources used to measure postpartum visit receipt among Medicaidinsured individuals: the Pregnancy Risk Assessment Monitoring System (PRAMS) and Medicaid claims.PRAMS is a survey conducted annually by the Centers for Disease Control and Prevention and state health departments to collect information on experiences before, during, and after pregnancy from state-representative samples of people with a live birth.Medicaid claims include fee-for-service and managed-care billing records for health care paid in part or full by Medicaid.Neither Medicaid claims nor PRAMS responses are a criterion standard for measuring postpartum care.Between 2016 and 2019, 88.8% of people with Medicaid after delivery reported attending a postpartum visit in PRAMS, 9 while estimates based on Medicaid claims indicated that 66.7% of Medicaid enrollees received a postpartum visit. 4dicaid claims may undercount postpartum visit use if people with a Medicaid-reimbursed delivery receive postpartum care that is not reimbursed by Medicaid.This could occur if a person who had delivered received care at a safety net clinic or paid for the visit out of pocket.For example, among people who are undocumented and immigrants who are within 5 years of establishing permanent residency, childbirth is often financed by Emergency Medicaid, which covers labor and delivery care for individuals who do not qualify for full-coverage Medicaid due to their immigration status. 10 Given that Emergency Medicaid does not pay for postpartum care, any postpartum health care received by this group of immigrants would not be observable in Medicaid claims.Postpartum care paid by private insurance would also be absent from Medicaid claims.Switching to private insurance during the postpartum period after a Medicaid-paid delivery is common, occurring after 12% to 25% of Medicaid-paid births.11,12 Conversely, PRAMS may produce an overcount of postpartum visit attendance if people who have delivered misreport other preventive care as a postpartum visit on surveys.
A previous study 13 compared Medicaid claims for postpartum visits with self-reported receipt of postpartum care in the 2011 to 2015 PRAMS survey in Wisconsin and found that agreement between data sources ranged from 42% to 82% depending on how postpartum care was identified in Medicaid claims.Our study extends that work by examining the association between insurance transitions and differences between postpartum visit receipt in PRAMS and Medicaid claims in South Carolina.We assessed whether insurance transitions from Medicaid at the time of childbirth to other insurance types during the postpartum period were associated with the degree of disagreement between data sources.

Methods
This cross-sectional study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.The study was deemed exempt from review and informed consent by Brown University Institutional Review Boards because secondary, deidentified data were used.

Data Sources
This study used South Carolina PRAMS survey data and Medicaid claims and enrollment files from 2017 to 2020.The South Carolina Revenue and Fiscal Affairs Office provided Medicaid enrollment files for all individuals who were mailed a PRAMS survey in this period; inpatient, outpatient, and physician claims for any Medicaid-paid care those individuals received in the year after childbirth; and an individual ID to link data sources.

Study Population
Most previous studies that used Medicaid claims to examine receipt of postpartum care included all people whose delivery was paid for by Medicaid. 8Therefore, to generate a study population comparable with those in previous literature, we first restricted our sample to PRAMS respondents (2041 individuals) who were enrolled in Medicaid on the date of their delivery based on Medicaid enrollment data (1189 individuals).Given that people may be enrolled in multiple insurance types simultaneously at delivery, we further restricted our study population to people for whom Medicaid was identified as the principal payer of delivery care on the child's birth certificate (851 individuals).
For individuals with multiple PRAMS responses within the study period (15 individuals), we included only the birth corresponding to the first PRAMS response (836 individuals).A flow diagram of study inclusion is shown in the eFigure in Supplement 1.

Outcomes
We measured routine postpartum visit attendance in PRAMS using the survey question "Since your new baby was born, have you had a postpartum checkup for yourself?(A postpartum checkup is the regular checkup a woman has about 4-6 weeks after she gives birth)."We identified postpartum care visits in outpatient Medicaid claims using the International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) code Z39.2, which is used in South Carolina to reimburse for postpartum visits.South Carolina is 1 of 14 states that bills for postpartum visits separately from antepartum, labor, and delivery care rather than grouping payment for these services into a single perinatal care bundle billed at the time of delivery. 14Between 2017 and 2020, fewer than 0.5% of postpartum visits were reimbursed using bundled payment codes in South Carolina. 15We restricted postpartum visits in claims to those billed within 60 days after delivery to align with the postpartum care-identification strategy used by the Centers for Medicare & Medicaid Services (CMS). 16

JAMA Network Open | Health Policy
Differences in Self-Reported and Billed Postpartum Visits Among Medicaid-Insured Individuals The study outcome was data source disagreement between PRAMS and Medicaid claims because a respondent reported a postpartum visit in PRAMS without a Medicaid claim for a visit or had a Medicaid claim for a visit without reporting a visit in PRAMS.In eTable 3 in Supplement 1, we also examined these types of data source disagreement individually.

Demographics and Postpartum Insurance Coverage
To describe our linked PRAMS and Medicaid claims sample, we examined several demographic characteristics using self-reported information from PRAMS responses.We created a combined variable for race and ethnicity with the following groups: Hispanic, non-Hispanic Black, non-Hispanic White, and multiple races or other race not previously listed.The last category includes non-Hispanic individuals whose race was listed as American Indian, Alaska Native, Chinese, Japanese, Filipino, Hawaiian, mixed race, and other race.We collapsed these responses into 1 category owing to small sample sizes.Race responses were not assessed for people who reported Hispanic ethnicity.We also examined educational attainment (no high school diploma, high school diploma, and some college or higher) and age group (Յ19, 20-24, 25-29, 30-34, and Ն35 years).
To examine whether insurance transitions from Medicaid at childbirth to another insurance type after delivery were associated with data source disagreement, we created a variable identifying respondent self-reported insurance type at the time of the PRAMS survey.8][19] Responses were classified hierarchically into 1 of 3 mutually exclusive categories: private insurance, Medicaid, or no insurance.We also used Medicaid enrollment data to identify people whose childbirth was covered by Emergency Medicaid.
We then classified respondents into 1 of 4 insurance transition types: Medicaid at delivery to Medicaid after delivery (continuous Medicaid), Medicaid at delivery to private insurance after delivery (Medicaid to private insurance), Medicaid at delivery to no insurance after delivery (Medicaid to no insurance), and Emergency Medicaid at delivery to no insurance after delivery (Emergency Medicaid to no insurance).

Statistical Analysis
We examined differences in the prevalence of postpartum visit use between data sources by calculating survey-weighted proportions of people with a postpartum visit in PRAMS and Medicaid claims overall and by insurance transition type.We performed 2-sided t tests for differences in proportions.Then, we calculated the percentage of individuals with data source disagreement in the overall sample and by insurance transition type.Finally, we used survey-weighted logistic regression models to examine the association between each insurance transition type and data source disagreement.We present results as marginal effects, which can be interpreted as the percentage point difference in data source disagreement between individuals with continuous Medicaid (the reference group) and those with each insurance transition type.Some prior studies using Medicaid claims to measure postpartum care restricted their study populations to people continuously enrolled in Medicaid during the postpartum period. 8Therefore, in eTables 2 and 4 in Supplement 1, we present results among PRAMS respondents (2041 individuals) who were enrolled in Medicaid on the date of their delivery based on Medicaid enrollment data (1189 individuals) and who were continuously enrolled in Medicaid from the day of delivery through 60 days after delivery based on Medicaid enrollment data (960 individuals).In addition, we present results restricted to the years 2017 to 2019 given that the start of the COVID-19 pandemic was associated with a reduction in postpartum care in 2020 relative to previous years. 9 conducted analyses in Stata statistical software version 17 (StataCorp).Analyses were weighted to account for the PRAMS survey design.We set a significance level of P < .05 a priori.Data were analyzed from February through October 2023.

JAMA Network Open | Health Policy
Differences in Self-Reported and Billed Postpartum Visits Among Medicaid-Insured Individuals

Prevalence of Postpartum Visit Use
Within our sample, a mean of 85.7% (95% CI, 82.1%-88.7%) of respondents reported a postpartum visit in PRAMS and a mean of 61.6% (95% CI, 56.9%-66.0%)had a Medicaid claim for a visit (Table 2).Differences in prevalence between data sources were greatest among individuals with a transition from Emergency Medicaid at childbirth to no insurance after delivery; a mean of 81.9%

Discussion
In this cross-sectional study comparing postpartum visit receipt between South Carolina PRAMS and Medicaid claims, we found a 69.7% agreement rate between data sources, with more than one-quarter of our sample reporting that they attended a postpartum visit in PRAMS without a Medicaid claim for a visit.Having a transition from Medicaid at childbirth to private insurance after delivery and being uninsured postpartum after having Emergency Medicaid as payer of childbirth were associated with a greater probability of data source disagreement compared with remaining enrolled in Medicaid after delivery.
We found that data source disagreement was greatest among people who were uninsured after childbirth if they had Emergency Medicaid at delivery.Emergency Medicaid reimburses hospitals for labor and delivery care provided to individuals who are ineligible for Medicaid due to their immigration status. 10 However, it covers only the hospitalization for childbirth without any coverage for care in the postpartum period.This finding suggests that state-level estimates of postpartum visit use derived from Medicaid claims data may systematically underestimate postpartum visit receipt by immigrants who are undocumented and recent immigrants, who may be accessing care at safety net clinics or paying out of pocket.
We also found high rates of data source disagreement among people with a transition from Medicaid at childbirth to private insurance at the time of the PRAMS survey.In our sample, 15.4% of people with Medicaid at delivery reported private insurance at the time of the PRAMS survey.This a Numbers are unweighted, and percentages are survey weighted.A total of 63 people did not respond to the PRAMS question on postpartum insurance type or attendance and were excluded from these results.
high percentage is expected given that income eligibility limits for pregnancy Medicaid coverage in South Carolina, which had not adopted the Affordable Care Act (ACA) Medicaid expansion as of June 2023, are much higher than income eligibility limits for parental Medicaid coverage.Between 2017 and 2020, South Carolina residents with incomes less than 199% of the federal poverty level were eligible for Medicaid during pregnancy through 60 days after delivery, and parents with incomes less than 67% of the federal poverty level were eligible for Medicaid coverage thereafter. 20This pattern held within our secondary sample of people who were continuously enrolled in Medicaid from delivery through 60 days after delivery but who self-reported private insurance on the PRAMS survey.These individuals may have been simultaneously enrolled in Medicaid and private insurance, in which case private insurance would typically serve as the first line of coverage for routine postpartum care. 21In contrast, we did not find significant data source disagreement among people with a transition from Medicaid at childbirth to no insurance at the time of the PRAMS survey.It is likely that individuals in this group were able to obtain a Medicaid-reimbursed postpartum visit within the typical range of 4 to 6 weeks after delivery before their pregnancy Medicaid coverage ended at 60 days after delivery.
Notably, we found high rates of data source disagreement even for people whose delivery care was paid for by Medicaid and who remained covered by Medicaid during the postpartum period.
This finding suggests that other factors beyond insurance transitions may be associated with differences in measurement between data sources.One important possibility that we did not examine in this analysis is that some respondents may misreport having received a postpartum visit in PRAMS.Individuals who have delivered may receive a variety of health care visit types during the postpartum period that may be mistaken for a comprehensive postpartum visit, such as a postpartum cesarean wound check or health care visit to address acute health concerns. 22To our knowledge, no studies have validated the PRAMS postpartum visit question; however, previous research has used electronic health records to test the validity of a variety of other PRAMS questions. 23,24is study's findings have several important health policy implications.State-level postpartum visit use estimated from Medicaid claims is used by the CMS as part of its Medicaid and Children's Health Insurance Program Scorecard. 25There is variation among states in the share of births to people who switch from Medicaid to private insurance coverage and the share of births to immigrants who are ineligible for Medicaid due to their immigration status. 11Our findings suggest that postpartum visit rates may be underestimated at a higher rate in states with greater shares of birthing persons who experience these 2 types of postpartum insurance transitions.This includes states that have not adopted the ACA Medicaid expansion, have large shares of undocumented birthing persons, and have not adopted policies to waive the 5-year waiting period for permanent residents.
Underestimating postpartum visit receipt due to Medicaid to private insurance transitions after delivery could also complicate efforts to track progress over time within a state.This is because major insurance policy changes, such as state Medicaid expansion, affect rates of postpartum insurance churn. 11Recent extensions to postpartum Medicaid through the full postpartum year may increase postpartum care receipt; however, this policy may also reduce private insurance churn, making it difficult to disentangle any real increase in postpartum visits attributable to the policy from increases

Limitations
This study has several limitations.It did not have a criterion standard data source to assess the reliability of self-reported PRAMS responses and Medicaid claims independently.Therefore, we were not able to examine other potential measurement issues, most notably the degree to which respondents may have overreported postpartum visit attendance in PRAMS.Recall bias may be greater for individuals who completed a PRAMS survey further from their delivery date.Additionally, these results may not be equally generalizable to all states due to differences in Medicaid billing practices and eligibility criteria.For example, as noted previously, South Carolina is one of few states that do not bundle payment for prenatal, delivery, and postpartum care. 14The ability to identify postpartum visits separately from other maternity care in South Carolina claims allowed us to focus our analysis on the contribution of postpartum insurance transitions to explaining measurement discrepancies between data sources.However, in states with bundled payment codes, measurement differences between PRAMS and Medicaid claims could result from global payment billing practices, postpartum insurance transitions, and other potential factors.

Conclusions
Medicaid claims data are a useful source of information on health care use during the postpartum period; however, results from this cross-sectional study suggest that estimates of postpartum visit use derived from Medicaid claims may undercount postpartum visit receipt by systematically missing care received by individuals with Emergency Medicaid for delivery and those who transition to private insurance during the postpartum period.These findings suggest that accounting for postpartum insurance transitions when conducting studies using Medicaid claims data may be associated with improved estimates of postpartum health care use and outcomes.

Table 1 .
Study Population Demographic and Insurance Characteristics Abbreviation: PRAMS, Pregnancy Risk Assessment Monitoring System.aNumbersare unweighted, and percentages are survey weighted.bThiscategory includes non-Hispanic individuals whose race was listed as American Indian, Alaska Native, Chinese, Japanese, Filipino, Hawaiian, mixed race, and other race.Respondents with a transition from Emergency Medicaid to no insurance had an increase in the probability of data source disagreement of 37.2 percentage points (95% CI, 19.6-54.8percentage

Table 2 .
Prevalence of Postpartum Visit Use ) compared with respondents with continuous Medicaid.Respondents with a transition from Medicaid to private insurance had an increase in the probability of data source disagreement of 15.8 percentage points (95% CI, 2.6-29.1 percentage points) compared with respondents with continuous Medicaid.The probability of data source disagreement among respondents with a transition from Medicaid to no insurance was not statistically significantly different compared with respondents with BAgreement is displayed for self-reported postpartum visit attendance in the Pregnancy Risk Assessment Monitoring System (PRAMS) and postpartum visits in Medicaid claims.points

Table 3 .
Data Source Disagreement by Postpartum Insurance Transition Abbreviations: NA, not applicable; PRAMS, Pregnancy Risk Assessment Monitoring System.
14 documentation of care in Medicaid claims data.Finally, people covered by Emergency Medicaid are disproportionally likely to be Hispanic.Therefore, use of Medicaid claims to measure racial and ethnic disparities in postpartum care may lead to misleading conclusions about racial and ethnic inequities in postpartum care receipt.With the release of Transformed Medicaid Statistical Information System Analytic Files by the CMS beginning in 2019, use of Medicaid claims for research on postpartum health care use and outcomes may increase.14Forresearchers planning to use Medicaid claims data to assess postpartum care use and outcomes, our findings suggest that measurement could be strengthened by stratifying results by Emergency Medicaid status and requiring continuous enrollment in Medicaid for the period of measurement of postpartum outcomes for individuals without Emergency Medicaid.

SUPPLEMENT 2. Data Sharing Statement
Study Population Demographic and Insurance Characteristics by Insurance Transition After Delivery eTable 2. Demographic and Insurance Characteristics of the Secondary Population of People Continuously Enrolled in Medicaid for 60 d After Delivery eTable 3. Factors Associated With Postpartum Visit Data Source Disagreement in Primary Study Population eTable 4. Data Source Disagreement by Postpartum Insurance Transition in Secondary Population of People Continuously Enrolled in Medicaid for 60 d After Delivery eTable 5. Prevalence of Postpartum Visit Use by Data Source Excluding 2020 Data eTable 6.Data Source Disagreement by Postpartum Insurance Transition Excluding 2020 Data