Over the past three years, several events have led policymakers, public health officials and the general public to focus renewed attention on abortion data in the United States. The information that is available on how many abortions are performed, when they take place and what methods are used has contributed to the public policy debate, but it also has proven inadequate in some instances to answer all the questions being asked.
For example, in 1995 Ohio outlawed dilation and extraction abortions, an event seen by opponents of abortion as the first victory in a national campaign to ban procedures they later dubbed "partial birth" abortions. The proposed federal "Partial-Birth Abortion Ban Act" has intensified the debate over abortion procedures, late-term abortions and, ultimately, the incidence and timing of abortions in general. Yet the debaters were often frustrated because specific data on the frequency of late-term abortions are limited, and data on the use of dilation and extraction do not exist either at the state or national level.
Moreover, at around the same time, Congress enacted a federal welfare reform law, the Personal Responsibility and Work Opportunity Reconciliation Act of 1996. Among several provisions intended to discourage out-of-wedlock births is the so-called illegitimacy bonus: Every year, for the next four years, the federal government will award $20 million each to the five states that can demonstrate the largest reduction in out-of-wedlock births and a simultaneous decrease in abortion rates. While the legislation establishes 1995 as the baseline against which reductions and increases will be measured, it does not address the limitations of abortion data collection efforts, which pose a significant challenge for accurately establishing a baseline level of abortion in many states, as well as for establishing accurate subsequent levels.
In 1996, as well, the Food and Drug Administration (FDA) took significant steps toward approving the use of medical (nonsurgical) abortion in the United States, essentially by "preapproving" the use of mifepristone, popularly known as RU 486, as an abortifacient; final approval is pending information on manufacturing and labeling. In addition, FDA cleared the way for clinical study by U.S. health care providers of a combination of two other drugs—methotrexate and misoprostol—used to induce early nonsurgical abortions.
While it remains to be seen to what extent the advent of medical (nonsurgical) abortions will actually change the provision of abortion services in the United States, it is at least possible that such abortions will be administered by health care providers who, for whatever reasons, have been reluctant to provide surgical abortions. If new providers do indeed emerge, incorporating abortion reporting by these providers into current reporting procedures will be critical both to measuring the number of abortions provided in the United States, and to monitoring the drugs' use and safety. Furthermore, because medical abortion is used primarily in the first seven weeks of pregnancy, the provision of nonsurgical abortion may lead to a shift in the timing of abortions. Documenting this shift might prove important to the abortion debate, since many individuals support early abortion but grow increasingly uncomfortable with the procedure as the pregnancy continues.
The Centers for Disease Control and Prevention (CDC), the government agency currently responsible for compiling U.S. abortion data, has been criticized by some people for its inability to answer all abortion-related inquiries—particularly, detailed questions relating to late-term abortions. However, such criticism does not consider that—in keeping with vital statistics tradition—CDC obtains its data through a voluntary federal-state partnership in which states are responsible for collecting and managing data in accordance with their own policies and systems, and submitting the information to the federal government. As a result, states ultimately determine the quality and availability of national, government-generated abortion data.
Background
History of U.S. Vital Statistics
The maintenance of vital records in the United States dates back to the 1600s, when colonies voluntarily or by law kept registers of births, deaths and marriages. This early recordkeeping was done primarily to protect individual rights; records were regarded as legal documents necessary for posterity and to ensure just administration of inheritance and other laws. During the 17th and 18th centuries, recognition of the utility of vital records as a public health tool grew, and local health boards began using death records to trace epidemics and evaluate community health.1
In the 1800s, several states and cities adopted laws governing the organization of public health agencies, and government maintenance of vital statistics emerged as an important public health function. Congress created the National Board of Health, which (in conjunction with the U.S. Bureau of the Census) was to spearhead establishment of a national vital statistics system. By 1900, the Census Bureau had developed the first standard certificates of birth and death, and in 1907 submitted the first in a series of model vital statistics bills to the states.
In 1946, responsibility for national vital statistics was transferred from the Census Bureau to the U.S. Public Health Service, which made two significant moves a decade later: It developed and issued the first standard records of marriage and divorce or annulment, and it issued the Certificate of Fetal Death (which later became the U.S. Standard Report of Fetal Death).
The National Center for Health Statistics (NCHS) was established in 1960 to collect statistics on a broad range of health topics, to conduct relevant research and analysis, and to publish vital statistics data. Nevertheless, the primary responsibility for collecting, managing and compiling vital records—records of births, deaths, fetal deaths, marriage and divorce or annulment—lies with the states in accordance with their own laws, regulations and public health agencies. They also submit data to the federal government on a contractual basis, through which the federal government shares in the cost of operating the state system.
Reporting Abortions
The move toward legalization of induced abortion in several states during the late 1960s provided an impetus for distinguishing between spontaneous and induced termination of pregnancy in reporting. As a result, some states began to collect induced abortion data separately, while others continued to record the events as fetal deaths. In 1969, with the original intent of monitoring the safety of abortion, CDC initiated a national abortion surveillance system to compile and analyze state-generated abortion statistics.2
Around the time of the landmark 1973 U.S. Supreme Court decision in Roe v. Wade, which legalized abortion in the United States, NCHS stepped up its efforts to obtain abortion data by attempting to install an abortion reporting system on par with other vital statistics data collection. In 1978, as part of that effort, NCHS introduced a standard form specifically for the reporting of induced abortion—the U.S. Standard Report of Induced Termination of Pregnancy. It was hoped that the NCHS system of collecting abortion data, which utilized micro data sets obtained by NCHS from the states on a contractual basis, would eventually replace the CDC abortion surveillance system, which relies on state-reported aggregate data.
However, NCHS was under severe financial constraint and failed to fund its abortion program adequately. This problem stymied the abortion data system's growth. At its peak, NCHS obtained abortion data from only 15 states, and the program was discontinued altogether after data year 1993.
Today, CDC's abortion surveillance system remains the sole governmental source of abortion data. The primary responsibility for recording, collecting and managing data rests with the states' vital statistics agencies, which submit data to CDC on a voluntary basis. CDC retains the federal role of issuing model legislation, forms and guidelines, as well as compiling and publishing state information; however, CDC does not share in the cost of the state data collection. Most recently, with the advent of medical abortion using such drugs as mifepristone and methotrexate, CDC led the effort to revise the U.S. Standard Report of Induced Terminations of Pregnancy to include medical abortions as a type of procedure.
Challenges to Abortion Reporting
Over time, all 50 states have wrestled with abortion reporting requirements, because, as with all abortion-related issues, reporting has met with controversy. At the heart of the issue is whether induced abortions should be regarded as reportable events paralleling births, deaths and fetal deaths, or rather as health events to be monitored as other surgeries and medical procedures are.
Additionally, some abortion rights supporters have raised concerns about the intent of abortion reporting requirements. They fear that abortion foes will use the laws to deter abortion provision, either by making reporting requirements too onerous or by allowing reported data to be used to harass service providers or women who have obtained abortions. In several states, reporting policies have been legally challenged; two cases argued before the Supreme Court have upheld reporting requirements.
When the Supreme Court heard challenges to Missouri's 1974 abortion law in Planned Parenthood of Central Missouri v. Danforth, the justices unanimously upheld the law's requirements that all health facilities and physicians report all abortions to the health departments. The Court concluded that such recordkeeping is useful to the state's interest in protecting the health of its female citizens, and that recordkeeping and reporting requirements "that are reasonably directed to the preservation of maternal health and that properly respect a patient's confidentiality and privacy are permissible."3
Sixteen years later, the Supreme Court reiterated its position in Danforth when it decided on the reporting requirement provisions of the Pennsylvania Abortion Control Act in Planned Parenthood of Southeastern Pennsylvania v. Casey. The decision stated that "[t]he collection of information with respect to actual patients is a vital element of medical research, and so it cannot be said that the requirements serve no purpose other than to make abortions more difficult."4These decisions largely affirmed states' moves to institutionalize the reporting of abortion data.
Data Completeness and Quality
While issues related to the quality of abortion data are outside the scope of this article, two studies that examined the completeness and consistency of state abortion data deserve mention. They highlight some of the limitations of abortion data, as well as indicate the potential impact of provider education and outreach, enforcement, follow-up and quality monitoring on state abortion data.
The first points to the underreporting and nonreporting that may occur in some states. The 1980 study compared Tennessee abortion data reported by providers to the Tennessee Department of Public Health with data reported for the state by The Alan Guttmacher Institute (AGI), which collects abortion data by surveying providers directly.5For 1974, the Tennessee Department of Public Health reported only half the number of abortions that AGI reported.
The authors concluded that "underreporting, or more specifically, nonreporting, by some facilities in Tennessee, has occurred because clinic and hospital administrators did not know that they were responsible for reporting abortions performed at their facilities and they have relied on physicians to do so." In subsequent years, according to the authors, department of health staff informed nonreporting clinics of the law, and by 1976 the department reported 74% of the number of abortions that AGI reported.
The second study illustrates the problems that arise both from measuring rare events and from human error: A few misrecorded abortions in Georgia dramatically altered the state's data on third-trimester abortions. The authors analyzed the accuracy of data on reported third-trimester abortions in Georgia by comparing the reported information with actual medical records for each case.6Upon reviewing 86 third-trimester induced abortions reported to the Georgia Department of Health and Human Services in 1979 and 1980, the authors found that the vast majority of the abortions were misreported. Only three procedures could be verified as actual third-trimester induced abortions; 58 of those reported were actually fetal deaths in utero, and 15 more were first- or second-trimester abortions that had been misclassified as third-trimester. The researchers concluded that the correct rate of third-trimester abortions for Georgia in 1979 and 1980 was 4.3 per 100,000 total abortions, rather than the rate of 123.1 per 100,000 abortions reported by the state's department of health.
Abortion Reporting
As of January 1998, 48 states, the city of New York* and the District of Columbia collect data on induced abortions.† The two nonreporting states, California and Oklahoma, have abortion reporting statutes on the books that are not currently in effect due to legal actions taken against related abortion statutes.
Laws
While 40 states and New York City collect abortion data as required by state statute, these laws vary. In 35 states and New York City, induced termination of pregnancy reporting is required specifically by statute (see Table 1). Overall, the laws are similar; by and large, they require every hospital or facility, or attending physician, to file a report regularly on each abortion performed, usually within a few days of the procedure or on a monthly basis. These laws mandate that abortion reports be submitted to the state department of health, state registrar or state vital statistics officer, and that the agency in turn publish the statistics on a regular basis.
Approximately half of the state laws specify that the department of health or a related agency will prescribe and provide the abortion reporting form, and several states require that the form be similar to the U.S. standard suggested by CDC. Virtually all of the statutes include a confidentiality provision—either emphasizing that the data collected are for statistical use only and may be published in aggregate only, or, at a minimum, mandating exclusion of the patient's or provider's name on the reporting form or in the published report.
Four additional states—Hawaii, New York, Rhode Island and Virginia—are legally obligated to collect abortion data under broader fetal death reporting statutes, rather than under laws specific to abortion. The Colorado vital statistics agency, meanwhile, collects abortion data in accordance with its death certification statute, which does not single out fetal death or abortion.
Regulations
Three states—Arizona, Connecticut and Washington—are obligated to collect abortion data solely by regulations issued by their state health agencies (Table 1). Regulations in all three echo the typical reporting statute. Nineteen more states have regulatory policies that accompany their abortion or fetal death reporting statutes. Such regulations typically reinforce the provisions put forth in the state statute and provide administrative guidance for the reporting system. For example, regulations might enumerate exactly what is required on the reporting form, discriminate between requirements for different types of medical facilities or elaborate on confidentiality provisions.
Voluntary Reporting
Five states and the District of Columbia collect abortion data on a voluntary basis, and their health departments provide forms and publish the data—even though no statute or regulation requires that abortions be reported (Table 1). New Jersey and West Virginia cite broad state health statutes as providing legal authority for a state health official to collect abortion-related data, while in Alaska, Maryland, New Hampshire and the District of Columbia, the health departments do not rely on legal authority.
State Data Collection
All states that collect abortion data utilize standardized forms, and most require a separate form for each procedure. The forms largely solicit the same baseline data as does the U.S. Standard Report of Induced Termination of Pregnancy: information on the facility (name or address, city and county); demographic information on the patient (her age, marital status, race, general educational level, and city, county and state of residence); medical information on the patient (date of last normal menses and number and results of previous pregnancies); information on the procedure itself (date of termination, clinical estimate of fetal gestation and method of termination‡) and the names of the attending physician and person completing the report.§
However, state forms tend to deviate from the U.S. standard in two ways. Many states do not require the same level of detail as the standard form on those items that might identify the facility, patient or attending physician—only 23 states** and New York City, for example, require the patient's residential zip code, and only 28 states†† and New York City request information identifying the attending physician. While all but three reporting areas‡‡ request information on the type of procedure used, only 17 states,§§ New York and the District of Columbia include "medical (nonsurgical)" in the list of abortion procedures.
Conversely, many states require more information than that required in the U.S. standard form. Twenty-seven states,*† for example, inquire about abortion-related complications, and several ask for additional information on the fetus, such as fetal viability, abnormality, length or weight. Nine states*‡ ask the reason for the abortion, and seven*§ request information on the woman's contraceptive history.
Six states and the District of Columbia do not use a separate form for each procedure. Colorado, New Jersey, Texas and West Virginia, which require the same basic information on each abortion as does the U.S. standard form, record abortions in logs that are submitted to the state agency on a regular basis. In Florida, Massachusetts and the District of Columbia, abortions are reported to health agencies in aggregate on a monthly or quarterly schedule.
National Data Collection
Annually, CDC contacts state vital statistics agencies to request certain data tabulations from the previous year. On a voluntary basis, states then submit aggregate data to CDC in the form of the requested tabulations, or as closely as possible, based on the state's available data. In 1995, the most recent year for which CDC data are available, the agency requested data on age of woman (younger than 15, 15, 16, 17, 18, 19, 20-24, 25-29, 30-34, 35-39, and 40 and older), weeks of gestation (less than or equal to 6 weeks, 7 weeks, 8 weeks, 9-10 weeks, 11-12 weeks, 13-15 weeks, 16-20 weeks, and 21 weeks or greater), type of procedure (suction curettage, all curettage, intrauterine saline instillation, prostaglandin instillation, hysterectomy or hysterotomy, other, unknown), race, Hispanic ethnicity, marital status, previous live births and abortions, and state of residence. As in previous years, CDC surveyed abortion providers in nonreporting states to estimate the number of abortions performed in those states.
Discussion
To a great degree, a national system for collecting data on induced termination of pregnancy is in place, and, by and large, states have moved to adopt federal standards that aim to make data complete and comparable across state lines. However, there remains considerable variability among state laws, policies, forms and systems, and this variability inevitably affects CDC's ability to determine accurately even the total number of abortions performed each year. While state reporting has improved over the years—and three states installed reporting systems for the first time in 1997—AGI reported 13% more abortions nationwide than did CDC in 1995,7 the latest year for which comparable abortion data are available.
This variability also exacts a toll on CDC's ability to answer specific questions about abortion in the United States. As demonstrated by the review of state reporting forms, there are considerable differences among states that do require abortion reporting in terms of the information they actually collect. Furthermore, for the information reported to the states, there often are problems with data completeness. For example, in CDC's 1995 state-level surveillance report, data on specific variables are missing for a number of states. To better assess the quality of state data, especially for small or sensitive groups, more research like the Georgia study is needed.
At the same time, it is important to understand that the information available to CDC is limited to the specific pieces of data that the agency requests from the states. For example, in 1995, in keeping with past years, the agency requested aggregated tabulations on nine variables, with some limited cross-tabulations. Therefore, the agency does not have access to state-collected abortion data in a record-by-record format, and it cannot then spontaneously answer questions about individual cases or new variables.
As a result of these data limitations, much of the information recently sought by decision-makers engaged in the "partial birth" abortion debate is currently out of CDC's grasp. Detailed information on late-term abortions is unavailable because the relatively small number of abortions beyond 20 weeks are aggregated into one gestational category. Data on certain procedures—including dilation and extraction, the medical procedure that most closely approximates characterizations of "partial-birth" abortion—are also unavailable because states and CDC collect data under broader categories.
Similarly, current limitations cast doubt on the federal government's ability to rely on existing data to responsibly award the "illegitimacy bonuses" authorized in the federal welfare reform law: Doing so would presumably require accurate, complete and consistent data that is comparable across the years—which simply do not now exist.
Finally, the existing abortion surveillance system poses challenges to public health officials in their quest to accurately trace the use of new, nonsurgical abortion techniques. Inclusion of the new techniques on a significant number of state forms demonstrates a sensitivity to the issue on the part of many state vital statistics officers. However, ensuring reporting by all new providers will undoubtedly require increased education and outreach efforts.
While some data limitations may be intrinsic to abortion—and no system is perfect—the quality of CDC's information is primarily compromised by the unevenness of reporting in the states. Policymakers need to assess the value they place on accurate abortion statistics and match information needs with resources. If accurate abortion data are as necessary to policymaking as recent debate suggests, steps need to be taken to bolster the existing systems. Doing so first requires further research into the limitations of the current systems and data, and a significant will to improve state-level data collection and management.