Study design
The Louisiana Abortion Prenatal Study was designed to study impacts of Louisiana’s abortion restrictions [
16]. We recruited participants at three university-affiliated prenatal care facilities in Southern Louisiana that serve pregnant women who have or are eligible for Medicaid. We describe the study methods in detail elsewhere [
17]. Briefly, between June 2015 and May 2017, we recruited women at their first prenatal care visit. Participants first completed self-administered iPad surveys; they then completed in-clinic structured interviews with a research coordinator. The Institutional Review Boards at the University of California, San Francisco and The Louisiana State University Health Sciences Campus granted ethical approval for this study.
In this manuscript, we aim to estimate the proportion of women who give birth instead of have an abortion because neither federal Medicaid nor state funds covers abortion for low-income women in Louisiana. We chose Louisiana because Louisiana state Medicaid does not cover abortion [
2]. Abortion funds are a set of private organizations that seek to address limitations in insurance coverage and geographic access to abortion [
18]. To help pay for the costs of a low-income woman’s abortion, these abortion funds provide subsidies to health care facilities to cover some or all of the costs of the abortion. Some funds are large (covering thousands of abortions per year) and others are small (covering only a few abortions per year) [
19]. The local abortion fund in Southern Louisiana only covers a small portion of costs at the abortion clinics in Southern Louisiana. At the national level, the price for an abortion is more than $500 and the adjusted prices are higher in states that have more restrictive abortion policies, such as Louisiana [
20]. Average out-of-pocket costs for abortion (including abortion funds or clinic discounts) is more than $300 for first trimester and close to $600 across all gestations [
7]. Women in states where abortion for low-income women is covered by state funds pay, on average, $0 out of pocket [
7].
At the time we began the study in 2015, Louisiana had five abortion clinics [
21], three in the southern part of the state. By the time we finished recruitment in 2017, Louisiana had three abortion clinics, with two in the southern part of the state. Neither the prenatal care clinics where we recruited nor the local Planned Parenthood facilities provide abortions.
Study procedures
In each recruitment facility, a research coordinator approached all women over 18 who presented for their first prenatal care appointment during the study time period and who spoke English. During the first year of recruitment, we began recruiting Spanish-speaking women. Women who were ineligible included those who were under 18, not pregnant, receiving a noninitial prenatal visit, not English or Spanish speaking, or incarcerated. As reported previously, of eligible individuals, 86% consented to participate [
17].
Research coordinators first obtained informed consent. They then instructed participants on how to complete self-administered iPad surveys and left them to complete surveys independently. After participants completed iPad surveys, the research coordinator conducted brief in-clinic structured interviews with participants.
Measures
The primary outcome was whether Medicaid not paying for abortion was a reason a pregnant woman had not had an abortion. To assess this outcome, we asked multiple questions. As a first step towards assessing whether Medicaid was a reason for not having an abortion, the iPad survey asked, “Have you considered abortion for this pregnancy even for just one second?” In the in-clinic interview, the research coordinator repeated this question verbatim. As described previously, reporting having considered abortion for this pregnancy was consistent across modes; 94% of participants reported consistently across modes [
17]. To assess the main outcome, in in-clinic interviews, participants who reported considering abortion in the in-clinic interviews were asked: “Medicaid in Louisiana does not pay for abortion. Was Medicaid not paying for abortion part of why you have not had an abortion?” Those who responded yes were considered to have not had an abortion because Medicaid did not cover it.
As a secondary measure of the outcome, we used data from open-ended responses to questions about reason(s) for not having an abortion and the main reason for not having an abortion. In the in-clinic interview, the research coordinator asked participants who reported they had considered abortion “even for just one second” a series of questions on concrete actions they may have taken to seek an abortion. Specifically, the research coordinator asked about the following concrete actions, whether they had: called an abortion clinic, made an appointment for an abortion, and went to the state-mandated abortion counseling visit and the abortion appointment. Once a participant responded that she had not taken the next concrete action in the series of possible actions, the research coordinator asked an open-ended question about her reason(s) for not having taken that step and then asked her to specify her main reason for not having an abortion. We trained research coordinators to: document responses verbatim, use neutral probes for clarity, and obtain more detail from participants. We classified responses that included “fund”, “money”, “price”, “insurance”, “dollars”, “$”, “cost” as financial reasons for not having an abortion. We did face validity checks to ensure responses were related to financial reasons.
We used additional variables as validity checks for reporting Medicaid as a reason for not having an abortion. We asked which pregnancy outcome women preferred upon pregnancy discovery and which pregnancy outcome they preferred now (upon prenatal care entry). In the iPad survey, we asked: “Please think back to the week right after you found out you were pregnant. Please tell me which option you preferred the
week right after you found out you were pregnant. Having the baby; Adoption or having someone else raise it; Having an abortion.” Then, with the same answer options, we asked, “Next, please tell us which option you prefer
now.” We assessed pregnancy planning using the London Measure of Unplanned Pregnancy; for ease of interpretation, we categorized the scale as unplanned, ambivalent, or planned [
22]. We measured decisional certainty using the Decisional Conflict Scale, a 16-item scale used in multiple areas of health care to measure people’s certainty around different health care decisions; for ease of interpretation, we categorized the scale as high certainty, medium certainty, and low certainty [
23].
To assess whether participants who reported Medicaid not paying as a reason for not having an abortion may have proceeded to have an abortion after the interview, we asked “Are you still considering having an abortion?”, after the open-ended questions about reasons for not having an abortion.
As people sometimes report more than one reason for not having an abortion after considering one [
17], we used responses to the open-ended questions about reasons for not having an abortion that we previously coded into personal reason, interpersonal reason, healthcare/other organization interaction, and policy-related reason. Specifically, responses coded as policy-related reasons were used as a check on Medicaid-related reason. Responses could fall into more than one category.
We assessed characteristics, including age (continuous), race/ethnicity (categorical), parity (categorical), education (categorical), employment (dichotomous), public assistance receipt (dichotomous), food insecurity (dichotomous), housing insecurity (dichotomous), insurance status (categorical), relationship with man involved in the pregnancy (categorical), past-year alcohol use disorder risk (dichotomous from AUDIT-C scale, number of drinks modified from 6 to 4 [
24]), past-year drug use (dichotomous), and past-year tobacco use (dichotomous).
Births and abortions
We used published estimates of the number of births and abortions in Louisiana in 2015 as well as guidance on estimating the number of miscarriages based on birth and abortion data [
25‐
27] to estimate the number of abortions, births, and miscarriages in Louisiana in 2015.We obtained published estimates of the proportion of Louisiana births paid for by Medicaid in 2015 [
28].
Analysis
We estimated the proportion of participants who reported that they did not have an abortion because Medicaid did not pay, including 95% Confidence Intervals (CIs). We assessed whether this estimate varied if we instead used coded responses from open-ended questions. We then estimated this proportion among women with Medicaid insurance, including 95% CIs.
For validity checks, we examined associations between Medicaid not paying as a reason and pregnancy outcome preference at pregnancy discovery, pregnancy outcome preference at prenatal care entry, pregnancy planning, and decisional certainty using chi-square tests and Fisher’s exact tests.
We then estimated the proportion of women who gave birth instead of having an abortion due to Medicaid not covering abortion. We used data on the number of abortions and births to Louisiana residents in 2015 as well as guidance on estimating the number of miscarriages based on abortions and births to estimate total number of Louisiana births, miscarriages, and abortions in 2015. We used published estimates of births paid for by Louisiana Medicaid and study estimates of the proportion with Medicaid insurance who reported not having an abortion because Medicaid would not pay to estimate the number of births paid for by Medicaid that would instead be abortions if Medicaid covered abortion. We added this number to published estimates of abortions to estimate projected number of abortions in Louisiana if Medicaid covered abortion. We then calculated proportion of women who give birth instead of having an abortion because Medicaid does not cover it through the equation (Projected abortions – Actual abortions)/Projected abortions. We repeated these steps, replacing estimates of proportions of women who reported that they did not have an abortion because Medicaid did not pay with lower and upper bounds of our estimate of the proportion who reported not having an abortion due to Medicaid not paying, to get a 95% CI.
We then described characteristics of women who report not having an abortion because Medicaid did not pay. We conducted bivariate analyses using t-tests for continuous and chi-square tests or Fisher’s exact tests for dichotomous and categorical variables to identify characteristics associated with not having an abortion because Medicaid did not pay among those who had Medicaid insurance.