Abortion is one of the most common medical procedures in the United States, with one in four women having an abortion in her lifetime.1 Studies have repeatedly demonstrated the safety of abortion in both outpatient and hospital settings.2 Although most abortions (95%) are performed in outpatient facilities,1 hospital-based abortion is often the only option for women with complex medical needs.3
U.S. federal law permits abortion until viability and allows states to narrow parameters within that. In addition, individual institutions have been reported to restrict abortion more narrowly than state law.4 These restrictions can decrease the availability of abortion and result in negative health outcomes for certain patient populations.5
The Accreditation Council for Graduate Medical Education (ACGME) requires obstetrics and gynecology residency programs to provide access to abortion training,6 but research shows that many residents graduate with insufficient abortion procedural skills7 and seek additional training after residency.8 In fact, hospital policy is the most commonly reported restriction that interferes with obstetrics and gynecology resident abortion training.9
Given that teaching hospitals are critical sources of medically complex abortion care and the primary training grounds for physicians, understanding the characteristics of hospital-level abortion restrictions can help to clarify how they affect patient care and clinician training. In this mixed-methods study, we conducted a national survey of obstetrics and gynecology teaching hospitals and interviews with training program directors to document the prevalence and nature of hospital regulation of abortion.
The study was composed of two sets of qualitative interviews with a quantitative national survey conducted in between and was approved by the University of California, San Francisco Institutional Review Board (study number 12-11246, initially approved on January 31, 2014). The first set of qualitative interviews, conducted in 2014, explored obstetrics and gynecology residency and site director perceptions of hospital abortion policies and approval processes. Results from these interviews informed the development of the national survey of U.S. obstetrics and gynecology residency training program directors, conducted 2015–2016, to understand the prevalence of different policies and enforcement mechanisms. A second round of qualitative interviews was conducted in 2017 at the close of the national survey to balance the sample by residency type and to confirm sufficient thematic saturation.
We recruited interviewees from a geographically diverse mix of obstetrics and gynecology residency programs around the country and purposively sampled from Kenneth J. Ryan Residency Training Programs in Abortion and Family Planning. We aimed to include at least three residencies in each of the four regions of the United States. A trained research assistant sent emails to three to five program directors in each of the Southern, Midwestern, Eastern, and Western regions on a rolling basis to ask to interview people most knowledgeable about their hospital's abortion policies and practices, until at least three were completed in each region. We selected Ryan programs because they have an explicit commitment to abortion training with the expectation that their faculty would be most willing to share their thoughts on this sensitive topic and be willing to refer colleagues to participate. We focused on hospitals with obstetrics and gynecology residencies, assuming that a majority of residencies would likely have some sort of abortion service given the 1996 U.S. ACGME abortion training mandate.6
In 2014, we conducted a first round of in-depth 30–60-minute interviews with residency program directors, abortion training site directors, or both, at each responding institution. Interviews included eight open-ended questions about hospital abortion practice, policies, authorization, and enforcement, with several follow up questions to each asked only if the information was not volunteered spontaneously. Some examples were: “What abortion services does your hospital provide?”, “Are some abortions not allowed in your hospital? Which ones?”, “Do you ever need to get authorization to do an abortion procedure?”, and “What's that process like?” Questions were asked in the order most organic to the conversation in order to maximize rapport and allow the story to unfold in the interviewee's own words. Interviews were conducted by L.R.F., a sociologist with extensive experience conducting qualitative interviews with professionals in the field of obstetrics and gynecology.
Interviews were recorded and transcribed, and transcriptions were uploaded into Atlas.ti, a qualitative data management software program. Atlas.ti allows qualitative researchers to view all transcripts and then assign codes to segments of the texts, allowing for an organized analysis. Codes represent behaviors or concepts that give insight into the research question. Code output is then generated to group similar text segments together so that interviewees' experiences and statements can be compared and contrasted. The software program does not conduct analysis but provides a structure for researchers to analyze the data in a systematic way.
Two authors (V.B.Z. and L.R.F.) coded the transcripts together, developing consensus on codebook definitions through an iterative process of coding and reassessing jointly. The codebook definitions were finalized after four transcripts and then applied throughout the rest. The analysis was conducted with a modified grounded theoretical approach that was both iterative and adaptive; we used insights from early interviews to inform subsequent survey development (below) and further qualitative inquiry. In 2017, the second round of interviews was added to test whether the themes that arose remained salient over time and to include data from obstetrics and gynecology residencies without Ryan programs. These interviewees were drawn from a geographically balanced set of teaching programs with abortion training, referred by members of the professional networks of one the coauthors (U.L.).
We sought to survey a residency program or site director at all obstetrics and gynecology teaching hospitals about their institution's abortion policies. Contact information for all ACGME-accredited obstetrics and gynecology residency programs and program directors in the United States, excluding military programs and those in Puerto Rico, was drawn from a publicly available ACGME database. An email introducing the study was sent to all residency directors, noting its endorsement by the President of the Association of Professors of Gynecology and Obstetrics. Three to seven days later, we emailed each residency director to formally invite them to participate. In order to ensure the survey was completed by the person most knowledgeable about abortion provision, we asked residency directors to send us the contact information of the site director of the abortion training program if they themselves did not serve in this role. Emails were sent via secure Qualtrics system and linked to an online consent form explaining the study purpose, risks, and benefits. Participants checked a box indicating consent and continued to the survey. The survey included modules about Institutional Policy and Policy Enforcement, with five to 14 multiple choice questions, some of which allowed for open-ended responses. Skip patterns were preprogrammed into Qualtrics to ensure respondents flowed through the survey appropriately. After completing the survey, participants received a $25 gift card for remuneration. For prospective participants who did not respond to the survey invitation, we sent up to three additional emails. Paper surveys were sent to prospective participants who did not respond electronically; the nine paper surveys returned were entered into the Qualtrics database by the data manager. Data were collected from October 2015 to February 2016 and managed through secure Qualtrics databases.
Catholic affiliation of surveyed institutions was queried specifically, because extensive data document the significant effects of Catholic affiliation on provision of abortion care and training at these hospitals.10 Affiliation of sites was determined through a national database of Catholic-affiliated hospitals kept by MergerWatch, a nonprofit organization that tracks the growth of religious health care.11
Survey items were developed based on the most frequently occurring themes from qualitative interviews. For instance, given the high frequency with which qualitative participants said policies were based on reasons for the abortion, we asked separately about policies affecting “nonmedically indicated” abortions (sometimes called “elective,” an abortion for reasons other than maternal health, fetal health, rape, or incest) and “medically indicated” (reason of maternal or fetal health, rape, or incest). Although this demarcation of abortion types has been critiqued for reinforcing a problematic hierarchy of abortion legitimacy,12 we chose to use these categories to reflect the language used by health care providers themselves in the interviews and to most accurately quantify the policies in place.
Similarly, given the emphasis on the division between written and unwritten policies that emerged from the qualitative interviews, we asked about the existence of each. Written policies were considered those formally codified, and unwritten policies were rules set informally by hospital leaders. For enforcement of policies, we differentiated between the need to get approval (a signature or verbal agreement from another party) and the need to notify someone (communication with another party about a planned abortion, but not requiring a response to proceed.)
In order to define institutional policies that restricted abortion beyond state legal limitations, survey respondents were first queried about the gestational limits to which they were allowed to provide abortion care based on state law, and all participants were able to answer. Subsequent questions asked participants to specifically identify policies that restricted abortion provision further than their state's law. All respondents were licensed physicians and directors of an obstetrics and gynecology residency training program that provides abortion care, thus we believe these physicians were knowledgeable of the legal boundaries of the care they provide and were able to describe deviations from the state law within their institutions.
We investigated differences in response rates by residency site characteristics using a series of logistic regression models. We described site and respondent characteristics, as well as the types of policies in place, and who makes and enforces the policies. We used logistic regression to investigate differences in policy structure by U.S. region. The few missing responses in the survey data (less than 2% for any variable) were removed from the denominator for the proportions presented. All analyses were conducted with Stata 15.
For the qualitative portion, 18 obstetrics and gynecology faculty members were interviewed from 15 residency training programs throughout the United States. (four in the West, three Northeast, five Midwest, three South). The first round, consisting of 13 interviews, were completed with faculty at 10 sites; round two included five additional interviews from faculty at five sites.
Of the 231 obstetrics and gynecology programs contacted, directors from 169 (73%) completed the survey. Respondents represented all four geographic regions (Table 1). Five percent represented Catholic-affiliated programs. Forty-three percent of respondents were residency program directors, and 57% were the training directors at the particular site where most abortions were performed. The average respondent's tenure at their site was 12.5 years.
The geography of responding sites were not significantly different from nonresponding sites. Responding sites had a higher average number of residents (5.7) compared with nonresponding sites (4.4; P<.001). Non–Catholic-affiliated sites were more likely to respond (76%) than Catholic-affiliated (42%; P<.01), although only 8% of target sites were Catholic-affiliated.
A majority of survey respondents (57%) reported that their training hospital had some sort of policy that restricted abortion provision beyond what is allowable under their state's laws. Policies were established based on the reason given for the abortion and often divided into “nonmedically indicated” and “medically indicated” procedures. It was more common for policies to restrict nonmedically indicated abortions (48% of sites) than medically indicated ones (28%), and a quarter of institutions restricted both types beyond state law (Figs. 1A and 2A for types of policies.)
In the nearly half (48%) of institutions that restricted nonmedically indicated abortions, policies were specified further: a quarter of all sites (25%) prohibited these types of abortions altogether, and many restricted procedures to a specific gestational age lower than that set by state law (Fig. 1B).
There were regional differences in hospital policies limiting nonindicated abortion provision, but not for indicated abortions. Teaching hospitals in the Midwest (63%) and South (61%) were more likely to limit nonindicated abortions beyond state law than those in the Northeast (28%) (odds ratio 4.3, P<.01 for Midwest; odds ratio 4.0, P=.001 for South). Forty-four percent of surveyed institutions in the West limited nonindicated abortions. However, there were no significant differences in prevalence of policies limiting indicated abortions (34% Midwest, 28% South, 21% Northeast, 33% West).
For medically indicated abortions, hospital policies reflected a wide variety of required criteria, including maternal and fetal indications in a spectrum of severity. Institutions delineated that some indications were never allowed, others were allowable on a case-by-case basis, and some were restricted to below a certain gestational age (Fig. 2B.) Some respondents marked “unsure” or “unclear,” suggesting that some kind of policy regarding the indication existed, but it was unclear when and how it was to be invoked. Mental health and rape or incest were the indications most commonly limited by hospital policies, with fetal indications following, and maternal indications relatively less scrutinized (Fig. 2B).
Interviewees described a slew of nuanced rules surrounding medical indications. For example, one interviewee explained that the only abortions allowed in her hospital for fetal indications were for lethal anomalies, “So trisomy 21, Down's syndrome, would not qualify. Trisomy 13 and 18 would.”
Many interviewees described an opacity in the definition of various medical indications. For example, in a hospital with no written policy specifying exactly which indicated abortions were allowable, the criterion was vague and ad hoc:
I am unaware and I've asked for written guidelines on what is medically indicated and I haven't received that. They seemed to imply at one point that our Chief of Services [could decide case-by-case, but she refused that responsibility]. So they just left it…up to the physician and patient to agree on what's medically indicated.
Although many interviewees expressed that vague rules around medical indications gave them the flexibility they wanted in their practice, others were critical of individual clinicians having the power and responsibility of decision making, suggesting that the lack of definition allowed for overly subjective and hence biased decision making. For example, one physician said,
We are basically telling patients that we take care of, “We don't think your reason is good enough”…I struggled to explain to students and residents how we make those decisions. Because it seems like it's all based on our values and what we deem to be a worthy reason.
Although survey findings measured national prevalence of hospital abortion restrictions and criteria, qualitative interviews shed light on a range of ways that these abortion restrictions were understood and operationalized by physicians.
Survey respondents were asked about whether abortion policies at their institution were formally codified in writing or unwritten. About a third (31%) of hospitals had written policies limiting abortion, and another third (33%) had unwritten rules. Overall, restrictions on nonmedically indicated abortion were more likely to be written (30%) than restrictions on medically indicated ones (15%).
Unwritten rules were communicated verbally by colleagues and could be opaque. One interviewee described being told that nonmedically indicated terminations were strictly forbidden by her institution owing to a historical association with a Catholic hospital and rumor of an abortion prohibition being attached to the building bylaws (a national phenomenon13). For years, she explained, the department followed this informal policy, until a new obstetrics and gynecology department chair, who was invested in “expanding our role in the community and having an expanded capability to train residents,” questioned the policy and thereafter began providing these services:
[The] new department chairman, who had a bit of a different attitude and actually went to the greater hospital administration in his role as chair and said, you know, “What really is the institutional position on this?” And what he found out was that the whole story, that there was some kind of bylaws agreement, you know, against doing elective termination of pregnancy, was all a whole bunch of BS. It never existed in the first place.
As with rules delineating permissible indications for abortion, some respondents conveyed frustration at the opacity that unwritten policies created. However, others described how this lack of clarity gave them flexibility, often in terms of gestational limits or the nature of medical indication. For example, a respondent explained that she understood there to be an exception to her abortion policy's gestational limit when the fetus could not survive.
Well the department has—I don't know if it's written or unwritten—but an informal rule that…for lethal anomalies beyond 24 weeks I think there is some wiggle room in there; for nonlethal anomalies beyond that, there's not.
Furthermore, interviewees explained that reluctance to formalize abortion policy in writing came from a desire to maintain this flexibility over practice, especially pertaining to cases at the locally accepted margins.
Nobody wants to talk about it, nobody wants to formalize it, you know, and nobody wants to kind of write any policies. It's all sort of unspoken…the worry was that if you went to the hospital and said we insist on a policy, that you'd get a policy that was very restrictive.
Thus, although survey data quantified prevalence of formalization of policies in writing, qualitative interviews revealed that some physicians found lack of formalization to preserve their autonomy and ability to provide abortion care for certain patients with extenuating circumstances.
Just as institutional policies were variable, so was the enforcement of policies. Among hospitals with any type of abortion restriction, 57% required someone's approval to provide an abortion, and 34% required notifying someone, with others unclear on the exact enforcement policy. The parties from whom actual approval was most commonly required were maternal–fetal medicine specialists (53%), hospital ethics committees (45%), and obstetrics and gynecology department chairs (43%) (Fig. 3). In contrast, for policies requiring notification, the parties most commonly needing notification were obstetrics and gynecology department chairs (40%), maternal–fetal medicine specialists (33%), and then hospital ethics (14%) or abortion-specific (14%) committees (Fig. 4).
In institutions where nonmedically indicated terminations were allowed, interviewees sought approval for complex cases that were beyond the hospital's gestational limit. On the other hand, for medically indicated procedures, physicians sometimes needed to certify the medical necessity of the procedure using individual experts or a committee. One interviewee described the complex process:
[I] can initiate it by going to the Chief of Service, [an] MFM [maternal–fetal medicine specialist], and what she will normally do is get whoever else, like another MFM or if the person has a cardiac problem, you know, a cardiologist or a cardio-thoracic surgeon to say that this is life-threatening. And then she will take that to the Chief of Staff for the hospital. And then if he has any questions, then we'll convene the Ethics Committee to decide. But, if it's straightforward, he may just bypass everyone and say it can be done.
Although some interviewees felt that enforcement processes were convoluted or idiosyncratic, others felt approval was helpful in legitimizing the procedure in the eyes of others, and thus curbing opposition. They described approval as somewhat of a formality used to solidify support from hospital personnel and the community at large:
Because of the potential for people catching wind of it and not understanding…the direction was given to involve the Ethics Committee. 'Cause then you've just, you know, dotted all the i's and crossed the t's.
In sum, survey findings showed approval processes were prevalent and commonly fell to leadership and specialists. Interviews elucidated that their function was not only as a mechanism for policing the limits of abortion care, but also to lend legitimacy and authority to the abortions allowed within the hospital.
To understand the source of hospital abortion policies, we asked respondents to identify who they thought was responsible for establishing policies. Many interviewees had a hard time clearly identifying the origins of their institution's abortion policies; however, a range of parties were identified as likely to be responsible for setting them (Fig. 3A). Written policies were most commonly set by hospital executives and chief executive officers (57%), obstetrics and gynecology department chairs (51%), and attorneys (45%). Maternal–fetal medicine specialists and hospital boards also played a role (38% each). Interestingly, for unwritten policies, involved parties were most commonly obstetrics and gynecology department chairs (49%) and nurses and nursing administration (45%), with maternal–fetal medicine specialists and other obstetrics and gynecology clinicians playing important roles (38% each).
When survey respondents were asked why they thought parties restricted abortion beyond state law, responses were similar for written and unwritten policies and included: the desire to avoid controversy through the media, personal beliefs or comfort with abortion, desire to avoid protesters, or concern for loss of funding. Some identified pressure from politicians and personal job security as additional concerns (Fig. 3B.)
Interviewees described instances when abortion criteria would become narrower based on the political and religious beliefs of hospital funders or political leadership. One interviewee describes her hospital CEO's (chief executive officer) motivations:
Well, he's anti-abortion too, but he's mostly anti-controversy. Like he doesn't want people writing him, calling him, he doesn't want protestors outside of the hospital, he doesn't want to piss off the politicians in the state.
Policies could also change quickly with a new hire based on their personal beliefs. One interviewee described a swift and direct decree about abortion services and education in the hospital as a result of an executive's beliefs about abortion:
A CEO came in and essentially said, “I'm on a mission from God. You can't do any more electives.”…our entire Ryan [training] program and the education it was built on and everything like that, done away with.
Several interviewees lamented that policies were not grounded in medicine, but rather determined by those in power, with negative consequences for abortion access and patient outcomes. One physician assessed that the abortion policy set by his chair, dean, president, and the CEO “seems extremely arbitrary in a way that's detrimental to our community's health needs.” Another physician was similarly frustrated when her chair set a policy restricting abortion in both the hospital and offsite abortion clinic where residents trained based on his personal discomfort. She challenged, “It doesn't have to do with [state] law. Why are we at this high-risk medical center not providing this service to women in need, you know? And in particular, how is he able to dictate practice [where residents train offsite] where he doesn't have any real authority?”
Thus, survey findings quantified the perceived origins of abortion policies, and interviews fleshed out how, in some specific instances, beliefs of particular department or hospital leadership served to narrow or foreclose training and provision.
This mixed-methods study examined U.S. teaching hospitals' policies for abortion practice. We found that abortion policies existed in 57% of teaching hospitals, most prominently in the South and Midwest, and approximately one-third were codified in writing. Policies restricted nonmedically indicated abortions by gestational age and medically indicated procedures by the individual indication. Qualitative data highlighted the nuances of policies, which were sometimes complex or confusing. Rules were enforced primarily by maternal–fetal medicine specialists, obstetrics and gynecology department chairs, and hospital ethics committees, established by individuals with institutional power, and perceived to be motivated by personal beliefs and a desire to avoid controversy. Interview respondents reported that hospital leaders prioritized broad financial and political considerations when creating abortion policies which, in some cases, put physicians at odds with their mission to train future health care providers according to ACGME standards for abortion.
Both survey respondents and interviewees reported that, in all their forms, hospital abortion policies perpetuated a system of value judgments on patients' reasons for seeking abortion. Interviewees used the categories “medically indicated” and “nonmedically indicated” or “elective” abortion to explain differential allowances, a dichotomy that masks the complex social and medical situations pregnant individuals face, and perpetuates abortion stigma.12 Survey respondents corroborated that even within medical indications, abortions for some indications were more heavily scrutinized than others. This stratification of legitimacy within teaching hospital abortion policies serves to reinforce abortion stigma for both patients and medical trainees.14
One of the most notable themes in the data was the prominence and manifold effects of unclear abortion policies. Vague or unwritten policies, although difficult to navigate, gave health care providers more agency and flexibility, especially in complicated cases where patients did not fit the understood criteria for abortion in their hospitals. On the other hand, formalized rules were perceived to lend protection to clinicians and the hospital from opposing coworkers or community members who may question the legitimacy of particular abortion procedures.
Opaque policies were sometimes perpetuated by the lore of an institution but could also be influenced by those responsible for their enforcement. Although the theoretical goal of approval and enforcement mechanisms was to lend objectivity and support for a difficult clinical decision, interviewees perceived these strategies as protecting the hospital from political and financial—rather than medical, legal, or ethical—risk. In this way, hospital-level enforcement mechanisms resembled pre-Roe abortion committees.15 Many interviewees expressed discomfort with institutionally defined processes that adjudicated a legally private medical decision. Overall, in hospitals with restrictive abortion policies, power rested more subtly in the flexibility of vague policies and the sway of leaders tasked with enforcing them.
One limitation of the survey was the challenge of capturing the complexity of site-specific policies in a closed-ended format. We believe the qualitative data served as a counterbalance to that limitation, giving a sense of the nuance of policies. The survey response rate (73%) was good. Still, with greater representation of larger sites and those in the West and underrepresentation of Midwestern and religious institutions, our figures likely underestimated abortion restrictions. Finally, our findings may not be generalizable to nonteaching U.S. hospitals. We suspect, based on qualitative work, that the average nonacademic U.S. hospital has more unyielding fiscal responsibilities and political motivations, and no ACGME mandate to have abortion services as part of training, likely resulting in more abortion restrictions.16 The primary limitation of the qualitative data was in its over-representation of sites with Ryan training programs (14/18), which explicitly train residents in abortion. Nevertheless, the four interviews with non-Ryan directors yielded similar themes as those with Ryan directors.
Patients presenting to hospitals for medical services reasonably expect to get legally protected care. In religiously affiliated hospitals, patients are often unaware how institutional policies curb their legal rights to abortion.17 It is likely that patients presenting to U.S. teaching hospitals with policies that restrict abortion access beyond state law are similarly unaware of restrictions. Decreasing access to abortion has been shown to create delays in care and poor patient outcomes.1 Additionally, teaching hospital abortion policies present barriers to clinical training for obstetrics and gynecology residents.9 Understanding the nature, etiology and arbiters of these policies is the first step toward preserving hospital abortion as a vital resource for patients at the medical margins and for the education of the next generation of physicians.18