Prevalence of Self-Managed Abortion Among Women of Reproductive Age in the United States

Key Points Question What is the prevalence of self-managed abortion (SMA) among US women of reproductive age? Findings In this cross-sectional survey of 7022 women aged 18 to 49, 1.4% reported ever having attempted SMA. Using age at SMA attempt and adjusting for underreporting of abortion, it is estimated that 7.0% of US women will attempt SMA at some point in their lives. Meaning These findings suggest that SMA is occurring in the US, highlighting the need for innovative models to ensure people have access to the safest and most effective methods of SMA, particularly where facility-based care is inaccessible.


Introduction
Self-managed abortion (SMA) is any action taken to end a pregnancy outside the formal health care system and includes self-sourcing the World Health Organization (WHO)-recommended medications (ie, mifepristone and misoprostol), ingesting herbs or other drugs, and physical methods, such as insertion of foreign objects into the uterus. 1 People's reasons for SMA are varied, and include both difficulty accessing facility-based abortion care and a preference for a more private or natural abortion experience. [2][3][4] Recent evidence sheds some light on the practice of SMA in the United States, highlighting online searches for information about ways to end a pregnancy without clinical supervision, [5][6][7][8] requests for medication abortion through an online telemedicine service, 2 and preferences for SMA outside the health care system. 9 Abortion clinics and practitioners report caring for an increasing number of individuals who have attempted SMA, 10,11 and researchers speculate that the decline in the US of facility-based abortion care may, at least in part, be driven by increases in abortions occurring outside of the formal health care system not captured by current measurement efforts. 12 Global estimates suggest that 17 million less safe abortions, including use of misoprostol without information or support from a trained individual and an additional 8 million unsafe abortions using other techniques occur annually, with 97% taking place in developing countries. 13 Data on SMA in the US come primarily from studies of people accessing abortion and other primary or reproductive health care, which may miss the experiences of people unable to access facility-based care. In this research, between 2% and 13% of people accessing facility-based abortion and other health care reported having ever attempted to end a pregnancy on their own. 3,[14][15][16] In the only population-based study available, conducted among reproductive-aged women in Texas, 1.7% reported having ever self-managed an abortion. 17 To date, there are no national, population-based estimates of the prevalence of SMA or people's experience with SMA in the US, to our knowledge.
As access to legal, facility-based abortion care in the US becomes more difficult owing to increased restrictions on abortion provision, there is increasing interest in understanding and documenting SMA. In this study, we use nationally representative survey data to estimate the proportion of US women of reproductive age who have ever attempted SMA, hypothesizing that SMA experience will be more common among populations that lack access to health care or experience stigma associated with abortion; project lifetime prevalence of SMA; and among those

Methods
This cross-sectional study was prepared in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for reporting on cross-sectional studies. All study activities were approved by the University of California, San Francisco, institutional review board. Study participants provided electronic informed consent after reading about the study online through the GfK platform.

Sample
Data for this analysis were collected by the GfK Group (now Ipsos) from members of its web-based KnowledgePanel. Panel members are recruited from a probability sample of US addresses to be representative of the noninstitutionalized US population; those who agree to join the panel are regularly invited to participate in online survey activities, and if needed are provided with the technology to do so. 18 Between August 2 and 17, 2017, GfK invited 14 151 eligible panel members via email to complete a 53-item, cross-sectional survey designed by University of California, San Francisco, researchers on women's experiences and opinions related to reproductive health care.
Eligibility was defined as self-identifying as female and being aged 18 to 49 years. Reminders were sent 3 and 8 days after initial contact. Participants were reimbursed for participation through GfK's points program, in which panelists receive cash-equivalent checks ranging from $4 to $6 per month depending on their participation.

Measures
The primary outcome of interest was experience with SMA, assessed using the question, "Have you ever taken or used something on your own, without medical assistance, to try to end an unwanted pregnancy?" A longer description of SMA, developed using cognitive interviews, 19 preceded this question: "As we mentioned earlier, some women may do something on their own to try to end a pregnancy without medical assistance. For example, they may get information from the internet, a friend, or family member about pills, medicine, or herbs they can take on their own, or they may do something else to try to end the pregnancy. Have you ever taken or used something on your own, without medical assistance, to try to end an unwanted pregnancy?" Participants who responded affirmatively about SMA experience were asked a series of follow-up questions about their most recent attempt including what methods they used; in what year the attempt took place (from a drop-down menu); in what state they were living at the time (from a drop-down menu), with the option to write-in another country; whether the methods worked to end the pregnancy; whether they experienced any medical complications, defined as something requiring "treatment by a doctor or nurse"; and their reasons for "ending the pregnancy on their own instead of going to a clinic." All questions were close-ended but included an other, something else, or not sure option. For methods, efficacy, and reasons, participants could select multiple response options and write-in additional responses. If they selected herbs for method type or yes to experiencing any complications, an open-ended follow-up question probed for additional detail.
Using their self-reported age and the year of SMA attempt, we calculated their age at the time of SMA. We collapsed states into geographic regions defined by the US Census. 20 Finally, we asked about past-year and lifetime experience with abortion and the number of financial and logistical barriers experienced when accessing reproductive health care in the past 3 years.
Sociodemographic information is routinely collected by GfK for all KnowledgePanel participants. In this analysis, we used panel variables on participant's age in years, race/ethnicity, highest level of education completed, marital status, place of birth (US vs another country), parity, current state of residence, and past-year material hardship (defined as ability to come up with $2000). 21 Using responses to household income and size questions, we calculated household

Statistical Analysis
We used descriptive statistics to describe participant characteristics, generate estimates of the proportion who had ever attempted SMA, and describe experiences with SMA. For all SMA outcomes, we removed attempts that occurred outside the US and those that used emergency contraception (EC) before confirming the pregnancy as their only method of SMA, since this may have represented appropriate use of EC and not abortion.
We estimated the weighted proportion with SMA experience within subgroups of age, race/ ethnicity, and other sociodemographic characteristics and described the bivariable association between these characteristics and experience with SMA using Poisson regression. Exponentiated coefficients represent prevalence ratios (PRs).
We generated a multivariable Poisson regression model estimating factors associated with SMA.
To establish the temporal association between covariates and SMA experience, we only included covariates that are time-invariant (race/ethnicity), preceded the attempted SMA (completion of high school), or serve as a proxy for a time-invariant construct (survey language). We originally planned to use place of birth in the multivariable analysis; however, there was higher nonresponse to this question than language of the survey (5% vs 0%). Given that some participants reported SMA prior to age 18 years, we conducted a second, sensitivity analysis with the subset of participants who reported that their SMA attempt occurred at age 18 or older.
From our cross-sectional data that include people ages 18 to 49 year, we projected lifetime experience with SMA using discrete-time event history models. Using participants' reports of their age at the time of most recent SMA attempt and their current age, we applied the tfr2 command in Stata version 15.1 (StataCorp) to estimate 5-year age-specific rates of SMA and then, based on these age-specific rates, sum across the age groups to generate an estimate of lifetime prevalence. 23,24 For all analyses, we applied survey weights using the svyset command in Stata and weights provided by GfK designed to reweight the sample such that it is representative of noninstitutionalized women ages 18 to 49 years living in the US with respect to age, race/ethnicity, census region, education, household income, and language. To ensure the weighting resulted in a sample representative of our target population of women of reproductive age, we descriptively compared our study sample with the 2015 to 2017 National Survey of Family Growth (NSFG), using public use data files. 25 Given that underreporting of abortion is a common challenge in survey research, [26][27][28] we assumed that respondents underreported SMA. We adjust for underreporting by comparing reported past-year abortions in the GfK to estimates derived from surveys of abortion facilities, which are generally considered more complete.  in our sample reported a past-year abortion, which is lower than national estimates of 1.45%. 10 We therefore multiplied our overall estimates of SMA by a factor of 1.45 in order to account for underreporting of abortion.

Among
Using discrete-time event models, women's projected lifetime prevalence of SMA was 4.8% (95% CI, 3.8%-5.8%) ( Table 2). Estimated age-specific rates indicated that experience with SMA peaks in young adulthood and then declines with age (eFigure 2 in the Supplement). Adjusted for  underreporting, projected lifetime prevalence of SMA was estimated at 7.0% (95% CI, 5.5%-8.4%), (      39 suggest an urgent need to pay close attention to trends in SMA, both overall and by region, in the coming years.
One-third of participants who had ever attempted SMA indicated that their attempt was unsuccessful and they sought a facility-based abortion. This suggests that SMA attempts identified at the time of abortion care-seeking may represent only one-third of all SMA attempts; accordingly, estimates of SMA among people recruited in clinic settings, which currently range from 2% to 13% depending on the study, 4,5,15 should be interpreted as underestimates of the frequency of SMA.
Finally, our results highlight how SMA encompasses many different types of methods with vastly different and sometimes unknown 40 levels of safety and effectiveness. Consistent with prior research, 4 herbs were the most common method used. However, others took misoprostol, drugs or medications other than misoprostol, EC after confirming a pregnancy, or used a physical method, often in combination. Sample sizes were too small to evaluate whether there were statistically significant differences in safety and effectiveness by method used. However, notably, complications were not common, despite use of some methods with the potential for harm.

Limitations
There were limitations to this study. Abortion is a stigmatized and underreported health behavior. 26 We compare past-year reports of abortion to national data from clinics in an attempt to adjust for underreporting; this method assumes that people underreport their SMAs at a similar rate as facilitybased abortions. However, we cannot empirically test this assumption, leaving open the possibility that our results adjusted for underreporting are inaccurate. Approximately one-half of GfK panel members invited to complete our survey participated, and there was evidence of underrepresentation of people living in poverty in our sample compared with the NSFG. However, this suggests that we are underestimating SMA prevalence, as SMA experience was more common among people living at less than 100% of FPL. Our question assessing the methods used to SMA did not specifically list mifepristone as a response option. As a result, participants may have described use of mifepristone in the "other drug" category. Furthermore, we excluded use of EC prior to pregnancy confirmation as a method of SMA because this may have represented appropriate use of the method; however, if EC was not used as indicated, we may again be underestimating SMA.
Response rates for the open-ended prompt on complications were low; future research should consider use of a validated measure of complications or adverse events. 41 Despite cognitive work to ensure a comprehensive definition of SMA, it remains possible that some people reported medication abortion obtained from a clinician as self-managed care. Additionally, we did not ask participants whether and how they confirmed their pregnancy, so it is possible that some respondents are describing methods used while experiencing a pregnancy scare. This is not a limitation for developing accurate estimates of experience attempting SMA, but it does introduce caution into interpreting our estimates of method effectiveness.

Conclusions
There has been much attention focused on the continued decline in the US abortion rate, but that estimate only counts abortions occurring in facilities. 12 This national cross-sectional survey study provides further evidence that SMA is occurring outside of the formal health care system, and people of color, those with lower incomes, and those who face barriers to care were more likely to attempt SMA. As abortion clinics close owing to increased abortion restrictions, a reduced demand for facilitybased abortions, 39 and a growing demand for convenience, privacy, and the comfort of selfmanaged abortion 9 and self-care more broadly, 42 it is likely that SMA will become more prevalent in the US, as it is today in other countries. 13 This national estimate serves as an important baseline to track this phenomenon moving forward.