Sexual Orientation and Exposure to Violence Among U.S. Patients Undergoing Abortion

Abstract

OBJECTIVE:

To assess the characteristics of patients undergoing abortion in the United States according to sexual orientation and exposure to sexual and physical violence.

METHODS:

Data for this observational study come from the Guttmacher Institute’s 2014 Abortion Patient Survey, which obtained information from 8,380 individuals obtaining abortions at nonhospital facilities in the United States; 7,656 of those (91%) provided information on sexual orientation identity. We used simple logistic regression to assess differences between heterosexuals and three sexual minority groups—bisexual, lesbian, and something else—according to demographic characteristics and exposure to sexual and physical violence. Multivariate logistic regression was used to assess associations between sexual orientation and exposure to violence.

RESULTS:

Among patients undergoing abortion in 2014, 4.1% identified as bisexual (n = 316), 1.1% as something else (n = 81), and 0.4% as lesbian (n = 28); 94.4% identified as heterosexual (n = 7,231). Similar proportions of lesbian and heterosexual respondents reported a prior birth (53.6% and 58.2%, P = .62), whereas respondents who identified as something else were more likely to report having had a prior abortion (58.0% vs 43.9%, P = .01). Exposure to sexual violence was substantially and significantly higher among all three sexual minority groups compared with heterosexuals, and lesbian and bisexual respondents were also more likely than their heterosexual peers to report exposure to physical violence by the man involved in the pregnancy (33.3% and 8.7% vs 3.6%, P<.001).

CONCLUSION:

No patient should be presumed to be heterosexual. Understanding the disproportionate role of sexual violence in unintended pregnancies among sexual minorities may aid in the design of interventions and clinical guidelines that address the needs of sexual minority patients.

Sexual minority women (eg, bisexual, lesbian, and other nonheterosexual women) are more likely than their heterosexual peers to have unintended pregnancies.^1,2 Emerging research has found that sexual minority women may be nearly three times as likely as their heterosexual peers to have had abortions.^3–6 However, these findings have not been consistently documented^7–9 and are complicated by methodologic limitations. For example, existing research has often lacked heterosexual comparison groups^10–12 and when such comparisons are possible, there was not necessarily enough statistical power to detect sexual orientation group differences.^13,14 Different categorizations of abortion (eg, lifetime, previous 5 years, most recent pregnancy) and varying measurements of sexual orientation (eg, attraction, behavior, and identity) further complicate the literature. Moreover, data are limited on circumstances surrounding pregnancies including physical and sexual violence, which is likely higher among sexual minorities compared with heterosexuals.¹⁵ Finally, data are often from sources not ideal for abortion research such as retrospective chart reviews or in-person interviews at lesbian, gay, bisexual, and transgender community events.¹

To address some of these limitations, the Guttmacher Institute collected sexual orientation identity data on a national survey of patients undergoing abortion in the United States. The survey’s initial report noted a nonnegligible proportion of patients identifying as sexual minorities.¹⁶ The current study builds on that finding by examining characteristics of patients undergoing abortion by sexual orientation identity and also examines exposure to violence according to this characteristic.

MATERIALS AND METHODS

Data for this observational study come from the Guttmacher Institute’s 2014 Abortion Patient Survey. To collect these data, a random sample of nonhospital facilities across the United States was selected to administer the survey to their patients undergoing abortion. A total of 8,380 respondents completed the four-page, paper-and-pencil, self-administered questionnaire, available in English or Spanish. Data are nationally representative of patients undergoing nonhospital abortion in the United States. The survey and data collection procedures were approved by the Guttmacher Institute’s institutional review board.

In 2014, participating facilities were sampled from all known nonhospital abortion-providing facilities in the United States according to the Guttmacher Institute’s 2011 Abortion Provider Census¹⁷; the list was updated to include facilities known to have started providing abortions since 2011 and to remove those facilities known to have closed or to have stopped providing abortion care. Hospitals were excluded from the sample because of the logistic difficulties with recruitment in such facilities. (High-volume clinics affiliated with hospitals were considered to be clinics and were included in the sample.) A prior study comparing the characteristics of patients undergoing hospital abortions and those undergoing nonhospital abortions in 2008 suggests that the two populations are very similar¹⁸; in both years, only 4% of abortions were provided in hospitals.^17,19 In line with prior Abortion Patient Surveys,^20,21 we also excluded facilities with caseloads of less than 30 abortions per year because of an anticipated lack of patients undergoing abortion during the survey period.

The sample was stratified by facility annual caseload of abortions (30–399; 400–1,999; 2,000–4,999; and 5,000 or more) and by whether the facility was affiliated with national reproductive health organizations (eg, the National Abortion Federation). Within each stratum, facilities were ordered by census region and state. Facilities from each stratum were systematically sampled, and facilities with the largest caseloads were oversampled to ensure a diverse representation of facility types within the sample.

Our goal was to recruit 113 facilities; the final sample was obtained from patients at 87 facilities (77% of the original goal). The most difficult health care providers to recruit and retain were those with the smallest caseload (30–399) often as a result of small offices with limited staff.²⁰ Nonresponding facilities often specialized in other types of health care. They may not have been familiar with the Guttmacher Institute and, in turn, unwilling to participate in a study of this sensitive topic. An estimated 1,800 additional surveys would have been obtained if all facilities that were approached had agreed to participate.

Selected facilities were recruited and asked to administer the survey for a period that was inversely proportional to the probability of being selected, ranging from 2 to 12 weeks. Surveys were distributed to patients during their clinic visit when facility staff determined was most appropriate, and phone conversations with staff during the fielding process suggest that was typically while patients completed intake forms.¹⁶ The survey’s introductory language stated its purpose, indicated that the survey was voluntary and anonymous, and served as implied consent. All respondents were provided with an envelope in which to place the survey before returning it to facility staff. At the end of each week, staff mailed survey packets back to the Guttmacher Institute. Between April 2014 and June 2015, participating facilities reported providing a total of 11,024 abortions. We obtained surveys from 8,380 respondents, for a 76% response rate. Missing information on core demographic items was imputed using “hot-deck” single imputation. This method of imputation identifies the characteristics most strongly associated with each item requiring imputation and sorts the data file so that respondents similar on these characteristics are adjacent to one another. The missing value is replaced with the value from the adjacent case. For most demographic variables, only 1–2% of cases had imputed values, with the exception of income, in which 13% of the values were imputed.

Our key variable was a measure of sexual orientation according to identity adapted from the National Survey of Family Growth, a nationally representative survey of reproductive-aged adults. The survey asked: “Do you think of yourself as…” and provided three response categories: “heterosexual or straight,” “bisexual,” and “homosexual, gay, or lesbian.” We modified this item to include a fourth response category of “something else” and included a blank space for respondents to write in a response. We considered respondents in any category other than heterosexual as being a sexual minority and compared these three groups—bisexual, lesbian, and something else—with their heterosexual peers. Notably, this study measured sexual orientation identity only and did not assess other sexual orientation dimensions such as behavior or attraction. The study also did not assess gender identity (eg, cisgender, transgender). Sexual orientation was measured only at the time of the current abortion.

A total of 724 individuals, 8.6% of the full sample, did not respond to the sexual orientation item and were excluded from the analysis. This was a higher level of nonresponse than was found on most other items, and for example, 363 respondents (4.3%) did not indicate their state of residence in the question immediately preceding the sexual orientation item. This higher level of nonresponse is counter to prior research that has found respondents are as likely to answer questions about their sexual orientation as they are to respond to standard demographic items.^22,23 We compared the characteristics of responders and nonresponders in the sample to determine if, and how, they differed from those who answered the item (Appendix 1, available online at http://links.lww.com/AOG/B115). Relative to individuals who provided information about their sexual orientation identity, nonresponders were older (18.8% aged 35 years and older vs 11.1% among those who answered the question, P<.001) and a higher proportion were Latina (44.5% vs 22.4%, P<.001), born outside the United States (41% vs 13%, P<.001) had not graduated from high school (23.5% vs 7.3%, P<.001), had poverty-level incomes (64.6% vs 48.4%, P<.001), and reported a prior birth (74.0% vs 58.0%, P<.001) or abortion (51.4% vs 43.9%, P<.001). It is possible that some of these respondents deemed the sexual orientation item too sensitive to answer, whereas others may not have understood the question well enough to provide an accurate response.

Demographic characteristics include age, race and ethnicity, union status at the time the pregnancy occurred (including cohabitation), prior education, and poverty status. We also included two measures of reproductive experiences: whether the respondent reported one or more prior births or a prior abortion. (We did not assess sexual orientation identity at the time of previous births or abortions.) Union status was based on two items. The first asked “In the month you became pregnant, what was your marital status?” Response categories included “married,” “divorced,” “widowed,” “separated” (the last three categorized as “previously married”), and “never married.” A follow-up item asked, “In the month you became pregnant, were you living with your partner?” and respondents could indicate “yes” or “no.” Individuals who were not married but were living with a partner were considered to be cohabiting, and the items were assessed among all patients regardless of their partner’s gender. Poverty status was measured according to two items: number of family members the respondent lived with at the time of the abortion and total household incomes (based on a 12-category measure).

Three items adapted from the 2008 Abortion Patient Survey were used to assess exposure to sexual and physical violence. Respondents were asked “Has the man with whom you got pregnant ever hit, slapped, kicked, or otherwise physically hurt you?” “Has he ever forced you to do anything sexual when you did not want to?”, and “Is this pregnancy the result of a man forcing you to have sex when you did not want to have sex?” All three items provided response categories of “yes” and “no.” The last item also included the response category “do not know” because individuals who had been exposed to both consensual and nonconsensual sex during the relevant time period would not necessarily know which incident resulted in the current pregnancy. In our analytic sample (N = 7,656), 140 (1.8%) respondents did not answer one or more of the items about exposure to violence. A higher number of respondents did not answer items about physical (n = 103) and sexual violence (n = 101) by the man involved in the pregnancy compared with the question about forced sex (n = 79). To use as many cases as possible, we did not exclude these respondents from all analyses, but only from those that made use of these specific variables.

We rely on frequency distributions to describe the sample characteristics, including sexual orientation identity. Simple logistic regression was used to assess potential differences in demographic or violence measures across each of the sexual orientation groups. The three sexual minority groups—bisexual, lesbian, and something else—were each compared with the heterosexual group. Logistic regression was used to assess for associations between sexual minority status and exposure to physical and sexual violence after controlling for the demographic characteristics of age, union status, race and ethnicity, and poverty. To increase statistical power of the three sexual minority groups, we also generated models that collapsed them into a single category for both the bivariate and multivariate analysis; these are provided in Appendices 2 and 3, available online at http://links.lww.com/AOG/B115.

RESULTS

The vast majority of abortion respondents identified as heterosexual (94.4%, CI 93.9–94.9, n = 7,231). Sexual minority respondents were most likely to identify as bisexual (4.1%, CI 3.7–4.6, n = 316) and were nearly three times more likely to identify as something else (1.1%, CI 0.9–1.3, n = 81) than to identify as lesbian (0.4%, CI 3–5, n = 28). Of the 81 respondents who indicated a response of something else, the most common write-in response was “pansexual” (n = 12). This variable was not imputed, and the 724 individuals who did not respond to this item were excluded from the analysis.

Each of the three sexual minority groups differed from the heterosexual group on a number of demographic characteristics (Table 1). Overall, bisexual respondents were more likely to be younger than their heterosexual counterparts; nearly half were aged 20–24 years (44.9%, CI 39.5–50.5) compared with 34.7% (CI 33.6–35.8) of heterosexual respondents. Only 5.4% (CI 3.4–8.5) of bisexual respondents were 35 years of age or older, and this proportion was twice as high among heterosexuals (11.3, CI 10.6–12.1). Respondents who identified as lesbian or something else were significantly less likely than heterosexuals to be white (21.4%, CI 9.8–40.6, and 11.1%, CI 5.9–20.1 vs 41.3%, CI 40.1–42.4, respectively).

Table 1.

Characteristics and Exposure to Sexual and Physical Violence Among U.S. Patients Undergoing Nonhospital Abortion by Sexual Orientation Status, 2014

	Heterosexual		Homosexual, Gay, or Lesbian			Bisexual			Something Else
	%	95% CI	%	95% CI	P	%	95% CI	P	%	95% CI	P
Age (y)
15–19	11.9	11.2–12.7	7.1	1.7–25.0	.44	13.6	10.2–17.9	.37	7.4	3.3–15.6	.22
20–24	34.7	33.6–35.8	35.7	20.2–55.0	.91	44.9	39.5–50.5	<.001	39.5	29.4–50.6	.36
25–29	26.8	25.8–27.8	21.4	9.8–40.6	.53	27.2	22.6–32.4	.86	30.9	21.7–41.8	.41
30–34	15.3	14.5–16.1	17.9	7.5–36.8	.71	8.9	6.2–12.5	<.001	14.8	8.6–24.4	.91
35 or older	11.3	10.6–12.1	17.9	7.5–36.8	.28	5.4	3.4–8.5	<.001	7.4	3.3–15.6	.27
Race and ethnicity
Non-Hispanic black	28.0	26.9–29.0	42.9	25.9–61.6	.09	27.9	23.2–33.1	.97	35.8	26.1–46.8	.12
Non-Hispanic white	41.3	40.1–42.4	21.4	9.8–40.6	.04	43.4	38.0–48.9	.46	11.1	5.9–20.1	<.001
Non-Hispanic other	8.5	7.8–9.1	7.1	1.7–25.0	.80	7.0	4.6–10.4	.35	21.0	13.4–31.3	<.001
Hispanic	22.3	21.4–23.3	28.6	14.8–48.0	.43	21.8	17.6–26.7	.84	32.1	22.8–43.1	.04
Relationship status
Married	13.6	12.8–14.4	25.0	12.2–44.3	.09	10.8	7.8–14.7	.15	12.4	6.7–21.5	.75
Cohabiting	31.6	30.5–32.7	7.1	1.7–25.0	.02	31.7	26.7–37.0	.98	28.4	19.6–39.2	.54
Never married	46.5	45.3–47.6	46.4	28.9–64.9	.99	48.1	42.6–53.6	.58	53.1	42.2–63.7	.24
Previously married	8.4	7.7–9.0	21.4	9.8–40.6	.02	9.5	6.7–13.3	.48	6.2	2.6–14.1	.48
Education (aged 20 y and older)
Less than high school	7.0	6.4–7.7	15.4	5.8–35.0	.11	10.3	7.2–14.5	.04	16.0	9.3–26.2	<.001
High school or high school equivalency certificate	27.1	26.0–28.2	26.9	13.2–47.1	.99	33.0	27.6–38.8	.03	45.3	34.4–56.7	<.001
Some college	43.5	42.3–44.8	42.3	24.9–61.9	.90	44.7	38.9–50.6	.71	26.7	17.8–37.8	<.001
College graduate	22.4	21.4–23.4	15.4	5.8–35.0	.40	12.1	8.7–16.5	<.001	12.0	6.3–21.6	.04
% federal poverty level
Less than 100	47.8	46.6–48.9	57.1	38.4–74.1	.32	56.7	51.1–62.0	<.001	70.4	59.5–79.3	<.001
100–199	26.3	25.3–27.3	32.1	17.4–51.5	.49	23.1	18.8–28.1	.20	16.1	9.5–25.8	.04
200 or greater	25.9	24.9–27.0	10.7	3.4–28.9	.08	20.3	16.2–25.1	.02	13.6	7.7–23.0	.01
Prior births
0	41.9	40.7–43.0	46.4	28.9–64.9	.62	49.1	43.6–54.6	.01	27.2	18.6–37.9	.01
1 or more	58.2	57.0–59.3	53.6	35.1–71.1	.62	51.0	45.4–56.4	.01	72.8	62.1–81.4	.01
Prior abortions
0	56.1	54.9–57.2	57.1	38.4–74.1	.91	60.4	54.9–65.7	.13	42.0	31.7–53.0	.01
1 or more	43.9	42.8–45.1	42.9	25.9–61.6	.91	39.6	34.3–45.1	.13	58.0	47.0–68.3	.01
Physical abuse by man involved in pregnancy
Yes	3.6	3.2–4.0	33.3	17.4–54.3	<.001	8.7	6.0–12.4	<.001	7.7	3.5–16.2	.06
No	96.4	96.0–96.8	66.7	45.7–82.6	<.001	91.3	87.6–94.0	<.001	92.3	83.8–96.5	.06
Sexual abuse by man involved in the pregnancy
Yes	1.9	1.6–2.2	34.6	18.8–54.7	<.001	7.1	4.7–10.5	<.001	5.1	1.9–13.0	.04
No	98.2	97.8–98.4	65.4	45.3–81.2	<.001	92.9	89.5–95.3	<.001	94.9	87.0–98.1	.04
Pregnancy the result of forced sex
Yes	1.2	1.0–1.5	14.8	5.6–33.9	<.001	3.2	1.7–5.9	<.001	3.8	1.2 −11.2	.06
No	97.9	97.5–98.2	77.8	58.1–89.8	<.001	94.5	91.3–96.6	<.001	87.3	78.0–93.1	<.001
Do not know	0.9	0.7–1.1	7.4	1.8–25.8	<.001	2.3	1.1 −4.7	.02	8.9	4.3–17.5	<.001
Total	7,231		28			316			81

Bold indicates statistically significance at P<.05.

Compared with patients undergoing abortion who identified as heterosexual, relationship status in the month of pregnancy differed only for lesbian respondents, and they were more likely to have been previously married than their heterosexual peers (21.4, CI 9.8–40.6 vs 8.4%, CI 7.7–9.0%).

Sexual minority respondents were more socio-economically disadvantaged than their heterosexual counterparts. Among respondents aged 20 years and older, those who identified as bisexual or something else were approximately half as likely to have completed college compared with heterosexuals (12.1%, CI 8.7–16.5 and 12.0% CI 6.3–21.6, respectively, vs 22.4%, CI 21.4–23.4). More than half of bisexual patients, 56.7% (CI 51.1–62.0) and 70.4% (CI 59.5–79.3), of patients who identified as something else reported incomes of less than 100% of the federal poverty line compared with 47.8% (CI 46.6–48.9) of heterosexual-identified patients. Differences in education and income between lesbian and heterosexual patients undergoing abortion were similar to those of other sexual minority groups, but differences did not achieve statistical significance.

Differences in prior pregnancy outcomes differed between heterosexuals and each of the sexual minority groups, except lesbians. Compared with heterosexuals, bisexual respondents were slightly less likely to have had a prior birth (51.0%, CI 45.5–56.4 vs 58.2%, CI 57.0–59.3), whereas those who identified as something else were substantially more likely to have done so (72.8%, CI 62.1–81.4). Respondents who identified as something else were also more likely than those who identified as heterosexual to have had a prior abortion (58.0%, CI 47.0–68.3 vs 43.9%, CI 42.8–45.1).

All three sexual minority groups, and lesbians in particular, were more likely than heterosexuals to experience physical or sexual violence. Lesbians were nine times more likely than heterosexual respondents to report that they had been exposed to physical violence by the man involved in the pregnancy (33.3%, CI 17.4–54.3 vs 3.6%, CI 3.2–4.0, respectively), whereas bisexual patients were more than twice as likely (8.7%, CI 6.0–12.4). Similarly, lesbians were 18 times more likely than heterosexuals to report that the man involved in their pregnancy had sexually abused them (34.6%, CI 18.8–54.7 vs 1.9%, CI 1.6–2.2), whereas those who identified as bisexual and something else were three times and two times as likely (7.1%, CI 4.7–10.5 and 5.1%, CI 1.9–13.0, respectively). Finally, more than 1 in 10 lesbian respondents indicated that the pregnancy was the result of forced sex (14.8%, CI 5.6–33.9), and all three sexual minority groups were more likely than heterosexual patients to indicate that the pregnancy may have been the result of forced sex. Almost all of these associations were maintained in multivariate models that controlled for age, union status, race and ethnicity, and poverty (Table 2) with the exception that individuals who identified as something else did not differ from heterosexual respondents in exposure physical or sexual violence by the man involved in the pregnancy.

Table 2.

Odd Ratios for Logistic Regression Models Assessing Associations Between Sexual Orientation Identity and Exposure to Physical Abuse, Sexual Abuse, or Rape Among Patients Undergoing Nonhospital Abortion, 2014

Sexual Orientation Identity	Physical Abuse			Sexual Abuse			Forced Sex or Potential Rape
	OR	95% CI	P	OR	95% CI	P	OR	95% CI	P
Homosexual, gay, or lesbian	14.6	6.0–35.2	<.001	24.5	10.5–57.5	<.001	10.0	3.9–25.7	<.001
Bisexual	2.4	1.6–3.6	<.001	3.9	2.5–6.3	<.001	2.6	1.5–4.4	<.001
Something else	2.2	0.9–5.1	.08	2.7	1.0–7.7	.06	5.9	2.9–11.9	<.001
Heterosexual	1.00			1.00			1.00
Total	7,553			7,555			7,577

OR, odds ratio.

Logistic regression models control for age, union status, race and ethnicity, and poverty.

DISCUSSION

This study, using data from a national sample of patients undergoing abortion, affirms that a nonnegligible proportion of patients undergoing abortion identifies as lesbian, bisexual, or something else. The proportions of respondents endorsing a heterosexual or bisexual identity were similar to other national data sets.²⁴ A smaller proportion of respondents endorsed a lesbian identity, and this population may be underrepresented among patients undergoing abortion. Respondents may have chosen the category of “something else” to identify their sexual orientation as a result of a number of factors including unwillingness to identify with a specific label, conceptualization of a broader or more fluid sexual identity, uncertainty about their sexual orientation, or a perception of inadequate response categories. Future studies will benefit from more comprehensive sexual orientation measurement including the dimensions of sexual behavior and sexual attraction.

Exposure to physical and sexual violence was substantially higher among each of the sexual minority groups compared with their heterosexual counterparts, sometimes by a factor of 15 or more. We found that lesbian respondents had the highest levels of exposure to violence, perhaps because this population was more likely to have had sex with a man only in the context of forced sex. For some individuals, all three forms of violence may have been perpetrated in the same act of (forced) intercourse, whereas, for others, the violence could have occurred in the context of a longer term abusive relationship. Patterns among all three groups speak to the unique vulnerability of sexual minorities and corroborate emerging research documenting similar victimization across sexual orientation groups.^25,26 Understanding the association among sexual violence, sexual orientation, and risk of unintended pregnancy may aid in the design of interventions and clinical guidelines that address the needs of sexual minority patients.

Differences, or lack thereof, in the pregnancy histories of heterosexual and sexual minority patients may run counter to expectations. Although prior studies provide limited information on lifetime pregnancy histories, sexual minority individuals from older generations are less likely than heterosexuals to have had a prior birth.²⁷ This pattern applied only to individuals in our sample who identified as bisexual. Lesbians did not differ from heterosexuals in regard to their pregnancy experiences (birth or abortion), although this may have been the result of limited statistical power. Previous literature on pregnancy histories across sexual orientation groups has primarily highlighted the disproportionate risk of unintended pregnancies among sexual minorities compared with heterosexuals,^1,2 but it is unknown whether the prior pregnancies reported in this study were intended or not.

This study has several limitations. Although the sample was large, the low incidence of respondents who identified with any of the sexual minority groups assessed, and lesbians in particular, means the associations we uncovered are less reliable and may not always reflect the true incidence. However, in the absence of much, if any, other national data on these population, this study provides useful insights and areas for future research. The key variable used in the analysis—sexual orientation identity—had a higher level of nonresponse than most other items on the survey. If a majority of the individuals who did not answer this item—perhaps because it was too sensitive—belonged to any of the sexual minority groups, this would bias our findings. For example, some sexual minority individuals may not have felt that any of the response categories adequately captured their sexual identity and may have opted to not answer the question. The survey instrument included only a single measure of sexual orientation identity with limited response categories. Future studies should assess multiple dimensions of sexual orientation—including identity, behavior, and attraction—as well as gender identity and expression. Respondents currently categorized as heterosexual (based solely on their identity) would otherwise be categorized as sexual minorities if they could have endorsed sexual minority behaviors and attractions. Finally, the data were collected in 2014 and are potentially dated; however, we expect that the same associations also apply to more recent time periods.

Health care providers, including those working in abortion settings, need to be aware that a proportion of their patient population identifies as something other than heterosexual. No patient should be presumed to be heterosexual for any reason, including a pregnancy history. All pregnancies—like all patients —should be treated as unique and operating within the dynamic and interconnected circumstances of peoples’ lives, which may encompass differences in sexual orientation and exposure to violence.