Reduced access to abortion services in the United States was last documented by data from 1992.¹ These data showed a decrease in the number of abortion providers, more limited availability of abortion services in some geographic areas and a shift of services away from hospitals to specialized clinics. Between 1988 and 1992, the number of abortion service providers fell by an average of 65 a year, and in 1992, 30% of women lived in a county without an abortion provider.

In addition, between 1990 and 1992, the number of abortions and the abortion rate declined slightly. Analysis suggested that reduced availability of abortion services may have been responsible for part of the decline, although other factors, including a reduced rate of unintended pregnancy,² may have been more important. From 1992 to 1995, the abortion rate fell by an additional 13%, according to data compiled by the Centers for Disease Control and Prevention (CDC) from state health department reports and its own surveys.³ However, no information has been available since 1992 on changes in the number or geographic distribution of abortion providers or their possible influence on the abortion rate.

Along with the prevalence of abortion services, the development of methods of early medical (nonsurgical) abortion has the potential to affect the availability and utilization of these services. By 1995, mifepristone, in combination with a prostaglandin, was offered on a limited basis under experimental protocols up to 63 days from the woman's last menstrual period; however, this method has not been approved by the Food and Drug Administration (FDA), and its future availability is unknown. In 1996, the effectiveness of a second medical method to induce abortion early in pregnancy, methotrexate (also used with a prostaglandin), began to gain attention;⁴ methotrexate has FDA approval, but for uses that do not include abortion. The availability of these methods may have encouraged more physicians to provide abortions, may have increased the number of women seeking abortion or simply may have replaced surgical procedures.

This article provides new information on the number and geographic distribution of abortion providers as of 1996, and updates national and state-level data on the number of abortions performed and abortion rates. The availability of abortion services by state, metropolitan area and county is documented and compared with past years. Also described are trends in the types of providers, the extent to which early medical abortion is used, the frequency of dilation and extraction abortions, and the other services offered by abortion providers.

The data are derived from the 12th national survey by The Alan Guttmacher Institute (AGI) of all known abortion providers. These surveys are the only national source of information on the number and type of providers and their geographic distribution. In addition, for almost all states and for the United States as a whole, the AGI surveys are the most complete source of information on the number of abortions performed. While abortion statistics are collected by most state health departments and are compiled annually by the CDC,⁵ comparison of the 1992 CDC report⁶ with the results of the AGI survey for that year⁷ indicates that the reporting of abortions is incomplete in most states. In addition, few states publish information about abortion providers.

Methods

Beginning in July 1997, we mailed questionnaires to all hospitals, clinics and physicians' offices thought to have provided abortions during 1995 or 1996. The mailing list included all facilities surveyed by AGI in 1993, excluding those that did not perform abortions in 1992, those that informed AGI in 1993 that they no longer performed abortions and those that were known to have closed before 1995. Added to the list were the names of possible new providers, obtained from the telephone yellow pages for the entire country, Planned Parenthood affiliates, the chapters of the National Abortion and Reproductive Rights Action League, the membership directory of the National Abortion Federation and other miscellaneous sources. Our updated list included a total of 2,948 possible providers.

During the follow-up of nonrespondents and mail returns, 84 additional potential providers were identified and surveyed. Sixty-three others were identified from state health department information but were not surveyed, as health department data for these facilities could be used.

We created two versions of the questionnaire, one for hospitals and one for clinics and physicians' offices. Both asked the number of induced abortions performed at the location in 1995 and 1996;* in addition, hospitals were asked the number of inpatient and outpatient procedures performed. Nonhospital facilities were asked a number of questions about services provided, the clients, charges, sources of payment, harassment and abortion procedures used. For each of three procedures—early medical abortion, intact dilation and extraction, and other sharp or suction curettage—we asked the number performed in 1996, the number performed from January through June of 1997 and the minimum and maximum gestation limits. We also asked the providers whether they expect to offer early medical abortions within the next 12 months if mifepristone is available and if it is not available.

We sent up to four follow-up mailings. For facilities that did not respond to the mailings, we used health department data on the number of abortions from states that provide information on individual facilities. We contacted the remaining nonrespondents by telephone and asked the number of abortions they had performed; some of the facilities hesitant to cooperate required up to 20 calls before we obtained information or a final refusal.

Of the 3,032 facilities surveyed (including those identified after the initial mailing), 1,279 responded to the mailed questionnaire, 176 faxed or mailed a response after telephone follow-up, 706 provided information by telephone, 123 were determined to have closed and to have not performed abortions during 1995 or 1996, and 80 were found to be duplicates; health department data were used for 365. The remaining 303 facilities either did not respond to requests for data or had closed, moved or could not be located, and we could not confirm that no abortions had been performed at these facilities during 1995 and 1996. For 109 of them, we obtained estimates of the number of abortions performed in 1995 and 1996 from knowledgeable sources in their communities. We made our own estimates for an additional 48 facilities that we knew to have provided abortions; for almost all of them, we were able to project the number of abortions using data from previous years. Of the abortions recorded for 1996, 79% were reported by the providers, 12% were estimated by local experts, 9% came from health department data and 1% were projected from previous years.

No abortions were attributed to the remaining 146 facilities, for which no data or estimates were available; these were not counted as providers in 1995-1996. Forty-two had either moved or lost their physician, and we cannot be sure that no abortions were provided during 1995 and 1996. Among those for which data were available for 1992, a total of 10,000 abortions had been provided in that year.

It has become increasingly difficult to obtain the cooperation of some abortion providers due to fear of antiabortion harassment, even though we assure respondents that we do everything possible to preserve their confidentiality. Response to our mailed questionnaires before telephone follow-up fell from 51% in 1993 to 42% in the current survey. The initial response rate of large abortion providers was above the overall average, and that of newly established providers performing small numbers of abortions was below average. Nevertheless, with extensive follow-up and investigation, we were able to obtain data or estimates for 95% of the facilities identified as possible providers.

Some providers undoubtedly were missed because they could not be identified. These were likely to have been small providers, since facilities that perform large numbers of abortions usually advertise and are known by referral sources such as Planned Parenthood.

The number of providers missed can be estimated by surveying a random sample of physicians and hospitals not on our list of possible providers. For the current survey, we drew a probability sample of 288 such hospitals listed in a directory of hospitals.⁸ These hospitals were surveyed with the questionnaire for hospital providers and were telephoned if no response was received after three mailings. Of the sample, nine said they had provided, on average, 34 abortions in 1996, 277 said they had provided no abortions, one refused to respond and one was closed.

Projecting to the universe of unsurveyed hospitals, we estimate that we missed approximately 124 hospital providers and 4,200 hospital abortions in the provider survey. In 1993, a similar sample survey found that 159 hospitals, providing approximately 3,600 abortions, may have been missed in the 1992 study of abortion providers. For the current survey, we were unable to carry out a verification survey of physicians, but a 1993 sample survey of obstetrician-gynecologists indicated that in 1992 we may have missed approximately 1,000 physicians who collectively performed 44,000 abortions.

The results of these methodological studies indicate that the actual number of abortions in 1996 might have been 3-4% greater than the number we counted. (We did not adjust the number of abortions or providers for this estimated undercount.) The total could be even greater if facilities with large abortion caseloads were missing from our list of possible providers, but we believe it is unlikely that large providers would be missed. The undercount may be offset in part by inadvertent double counting of facilities listed under two different names.

Results

Numbers and Rates

The number of abortions performed in the United States fell sharply to 1.36 million in 1995, after reaching a peak of 1.61 million in 1990 (Table 1). The steady decline plateaued in 1996, however, when about the same number of abortions were performed (1.37 million) as in 1995. The abortion rate per 1,000 women aged 15-44 followed a similar pattern—a drop of 16% from 1990 to 1995, and then no change in 1996. The abortion rate in 1995 and 1996 was 23 per 1,000 women aged 15-44, lower than in any year since 1975 (when it was 22 per 1,000).

In 1995 and 1996, the abortion ratio was 26 abortions per 100 live births plus abortions—i.e., 26% of pregnancies (excluding miscarriages) were terminated by abortion. Including an estimate of the number of pregnancies ending in miscarriage^† decreases the percentage of pregnancies ending in abortion to 22%. The abortion ratio has fallen since 1990, when it was 28 per 100, but it has not declined as much as the abortion rate because the number of births also dropped somewhat.

Abortion numbers and rates vary widely among the states (Table 2). As is to be expected, the most populous states report the largest numbers of abortions: California (238,000), New York (168,000), Florida (94,000) and Texas (91,000). Since 1992, Florida has replaced Texas as the state with the third most abortions. Wyoming reported the fewest abortions (280), and Idaho, North Dakota and South Dakota each reported fewer than 2,000.

The highest abortion rates by state of occurrence were in Nevada (45 abortions per 1,000 female residents aged 15-44), New York (41) and New Jersey (36), and rates were above 30 per 1,000 in California and Florida as well (Table 2). The rate for the District of Columbia (155 per 1,000) was higher than that of any state; relatively high rates are characteristic of central cities generally, and the rate includes large numbers of women from outside the District who seek abortion services there.⁹ The census divisions with the highest rates are those on the East and West Coasts: the Middle Atlantic, Pacific, South Atlantic and New England states.

Five states had abortion rates below eight per 1,000: Idaho (6), Mississippi (7), South Dakota (7), West Virginia (7) and Wyoming (3). All of these states are mostly rural, with no large metropolitan areas. Among the census divisions, rates were lowest in the East South Central and West North Central states.

Between 1992 and 1996, the abortion rate declined 12% nationally, and it decreased in 43 of the 50 states (based on state of occurrence). Declines were greatest in the Pacific census division (22%), with a fall of 41% in Hawaii, 24% in Washington and 22% in California. Decreases also were especially large in Mississippi, Wyoming, Maine, and Delaware. In Mississippi and Maine, the largest abortion provider closed between 1992 and 1996, and in Delaware one of the two largest closed. The only two areas that recorded an increase in the abortion rate of more than 10%—New Jersey and the District of Columbia—had more providers in 1996 than in 1992. (Improved reporting may also have been a factor in New Jersey.)

However, abortion rates by state of occurrence should be interpreted cautiously, because they do not always reflect the extent of abortions obtained by residents, who may travel out-of-state for services. For example, in 1992, the most recent year for which such calculations have been made, the number of Wyoming residents who had abortions in other states was more than twice the number of residents who had abortions in the state. In Idaho, Missouri and West Virginia, the abortion rate among state residents was more than 40% higher than the rate based on the abortions occurring in the state. By the same token, abortion rates are inflated in the states that provide services to large numbers of out-of-state women. In 1992, the rates by state of residence were 26-48% lower than the rates by state of occurrence in the District of Columbia, Kansas and Vermont.

Service Availability

The proportion of counties with no abortion providers and the proportion of women who live in a county with no provider are indicators of the availability of abortion services. (These measures are imperfect, of course, because some women live close to a provider in a neighboring county, some providers are not readily accessible because of their high charges or their limitations on the patients they will accept, and some serve only a small number of abortion patients and rarely advertise their services.) Another potentially useful measure is the proportion of counties lacking a provider that is large enough to be likely to advertise and accept self-referred patients (which for this analysis we assume to be a facility that provided 400 or more abortions during the year).

In 1996, 86% of all U.S. counties had no identified abortion provider (Table 3), and 92% had none that performed as many as 400 abortions annually. Thirty-two percent of women of reproductive age lived in counties with no provider, and 41% lived in counties without a large provider. The proportion of counties with no provider was slightly higher than in 1992 (84%), continuing a long-term trend first seen in the late 1970s. However, the proportion of women who lived in a county with no provider of 400 or more abortions changed little between 1992 and 1996, after having declined from 1985 to 1992.

Abortion providers are much less available in nonmetropolitan than in metropolitan counties. Ninety-five percent of nonmetropolitan counties had no abortion services, and 87% of nonmetropolitan women lived in unserved counties. In 1996, only 1% of abortions (14,070 abortions) were reported in nonmetropolitan counties, where 18% of women of reproductive age lived (not shown).

More than half of metropolitan counties, however, also were unserved, and abortion services were effectively unavailable in one-third of U.S. cities. Of the country's 320 metropolitan areas,^‡ 89 had no known abortion provider and an additional 12 had providers that together reported fewer than 50 abortions in 1996 (Table 4). The states with the most unserved or underserved metropolitan areas were Texas (12), Pennsylvania (eight) and Wisconsin and Indiana (seven each).

Since 1992, however, four metropolitan areas were removed from the list of unserved cities because new clinics were established. In addition, six cities where fewer than 50 abortions had been performed in 1992 moved out of that category, three because of new facilities and three because existing providers increased their level of services. The gain in newly served cities was offset in part by a loss and decline of services in other metropolitan areas.

On the state level, a measure of the availability of services is the proportion of counties without any identified abortion provider (Table 5). In 41 states, a majority of counties had no provider, including 21 states where more than 90% of counties had none. On the other hand, all four counties in Hawaii had physicians performing abortions.

The proportion of women in a state living in a county without a provider (another measure of the availability of services) varied from 0% to 84%. In Mississippi, North Dakota and West Virginia, 80% or more of women lived in unserved counties; in contrast, the percentage of women living in unserved counties was very low in California, Massachusetts and New Jersey as well as Hawaii. The East South Central and West North Central states had the highest proportion of women living in counties without providers (65% and 57%, respectively); this proportion was lowest on the coasts, in the Pacific, New England and Middle Atlantic states.

The proportion of women aged 15-44 who lived in counties with a provider is highly correlated with abortion rates according to state of occurrence (r=0.82). (The correlation with the abortion rate by state of residence would be lower, since women from unserved states travel to service providers in other states.)

In 1996, abortion services were provided in 2,042 facilities, a decline of 14% since 1992. This represents a decrease of 85 providers per year, compared to a drop of 51 per year between 1988 and 1992. The number of providers has been declining since 1982, when it was at a high of 2,908.

Between 1992 and 1996, the number of identified providers fell in all census divisions and in all states except the District of Columbia, New Jersey and Utah, which experienced increases, and New Hampshire, North Dakota and South Dakota, where there was no change. The largest decrease occurred in California (62), which nevertheless continued to have more providers (492) than any other state. Decreases of 20 or more also occurred in New York, North Carolina and Pennsylvania. The percentage decline was greater in the East South Central states (31%) than in other census divisions (not shown). The number of providers fell by one-third or more in Tennessee (39%), Alaska (38%), New Mexico (35%) and Kansas (33%). In Tennessee, New Mexico and Kansas, the loss of providers continued declines that were evident between 1982 and 1992.

Types of Providers

An abortion provider was defined as a place where abortions are performed, whether a hospital, clinic or physician's office. If an organization offered abortion services at more than one location, each service site was counted as a provider. The number of providers is different from the number of physicians who perform abortions, because one physician could be responsible for services in several facilities, and several physicians could perform abortions in a single setting.

•Hospitals. The 2,042 abortion providers in 1996 included 703 hospitals where one or more abortions were performed during the year (Table 6). As a proportion of all providers, hospitals declined from 36% in 1992 to 34% in 1996 (not shown); in 1973, 81% of all providers had been hospitals.¹⁰ Similarly, the number of abortions performed in hospitals fell from 110,000 in 1992 to fewer than 93,000 in 1996, although the proportion of all abortions that were performed in hospitals remained at about 7%.

Abortions were performed in only 14% of short-term, general, nonfederal hospitals (16% if Catholic hospitals are excluded). These proportions are two percentage points lower than in 1992. Private (including voluntary) hospitals were somewhat more likely to offer abortion services (18%, excluding Catholic hospitals) than were public hospitals (10%, not shown).

In many of the hospitals classified as providers, only a few abortions are performed. Hospitals that allow abortions to be performed only when a woman's life or health are threatened by the continuation of her pregnancy are counted as abortion providers, even if only one abortion was performed at the facility. A majority (58%) of the hospitals that reported abortions provided fewer than 30 abortions each; these facilities together accounted for only about 4,500 procedures. In 1996, only 12 hospitals performed 1,000 or more abortions.

In keeping with the national trend toward day surgery, a large majority of hospital abortions are performed as outpatient procedures. In 1996, the proportion of inpatient abortions fell to 9%, down from 11% in 1992 (not shown). Of all abortions, fewer than 1% involved hospitalization of the woman (excluding cases where the woman may have been hospitalized for treatment of complications).

•Nonhospital facilities. Most abortions are performed in the country's 452 abortion clinics (defined as nonhospital facilities in which half or more of patient visits are for abortion services).^§ The proportion of abortions that are performed in abortion clinics increased from 60% in 1985 to 69% in 1992 and 70% in 1996, although only 22% of all providers are abortion clinics.

"Other clinics" include group practices with clinic names, surgical centers, health maintenance organizations, family planning clinics and other facilities with clinic names. (In addition, physicians' offices where 400 or more abortions were performed in 1996 were counted as "other clinics" if they did not fit our definition of abortion clinics.) The "other clinics" accounted for 20% of all abortion providers and 21% of abortions in 1996. Four out of five performed fewer than 1,000 abortions a year, compared with only one of five abortion clinics.

Physicians' offices comprise both solo and group practices where fewer than 400 abortions were performed in 1996. (However, some similar providers were counted as clinics because they had adopted clinic-like names.) Approximately 44% of physicians' offices reported 30 or fewer abortions in 1996, and the group as a whole performed only 3% of the country's abortions, down from 4% in 1992.

•Trends in numbers of providers. While the overall number of providers dropped by 14% between 1992 and 1996, the loss was greater among hospitals (18%) than among nonhospital facilities (12%), and was greater among public hospitals (23%) than among private ones (16%) (Table 7). This pattern of change was the same between 1982 and 1992, when the number of public hospital providers fell from 331 to 180. There are now only 42% as many public hospital providers as in 1982 and 53% as many private ones.

The change in the number of nonhospital providers has been much less pronounced and has occurred almost entirely among physician practices. In 1996, there were 20 fewer clinics than in 1992, but 80 more than in 1982. From 1992 to 1996, the number of abortion clinics increased by 11, while the number of other clinics providing abortions decreased by 31 (not shown). At the same time, abortion services provided by physicians' offices declined 26%. The decline was greater among physicians' practices that provided 30-390 abortions, possibly because some practices moved from the larger category to the smaller one (fewer than 30 abortions) as a consequence of falling caseloads.

Although the total number of providers for whom abortion services are an important activity has changed little, there is always some turnover between surveys: Clinics close or stop offering services, and new clinics are established. Fifty-one clinics that reported 400 or more abortions in 1992 reported none in 1996, while 98 of those reporting 400 or more abortions in 1996 began their abortion services after 1992.

We reviewed our records for any information that might explain why clinics that had provided more than 400 abortions in a year no longer provided any. Two continued to exist, but no longer offered abortion services. Twenty-seven had closed, and while we do not know why, several of them mentioned antiabortion harassment when they were surveyed in 1992 in answer to a question about factors that made it difficult to provide abortion services. In nine instances the physician retired or died, in five the physician lost his or her license, in three the clinic moved to a different county, two had difficulty finding a physician to perform abortions, one closed in bankruptcy, one burned down and no information could be found for one.

•Services offered. Abortion providers usually do far more than simply perform abortions. To obtain a fuller understanding of the types of facilities where abortions are performed, we asked nonhospital providers whether they offered general medical care, general gynecologic care and contraceptive services to nonabortion patients. Of nonhospital providers, abortion clinics were the most specialized; nevertheless, 83% offered contraceptive services independently of their abortion services, 74% provided gynecologic care and 20% provided general medical care. Almost all of the other clinics offered contraceptive services (97%) and gynecologic care (94%), and 53% provided general medical care. More than 99% of the physicians' offices provided gynecologic and contraceptive care, and 69% served general medical patients.

•Medical abortion. The advent of early medical abortion in recent years is a potentially important development. In our survey, 82 nonhospital facilities said they had performed one or more early medical abortions during 1996, and 114 said they had done so in the first half of 1997. Assuming that nonrespondents offered medical abortion services in the same proportions as respondents with similar abortion caseloads, approximately 117 clinics and doctors' practices were performing medical abortions in 1996 and 163 in 1997. The 1997 estimated number represents 12% of all nonhospital providers. An even larger percentage (44%), including most of the current providers, responded that they will probably provide medical abortions within the next 12 months if mifepristone becomes available, and 29% said they would do so even if it is not available.

The likelihood that early medical abortions were performed was strongly related to the size of the abortion caseload. Among those who reported fewer than 400 abortions in 1996, only 7% performed medical abortions in the first half of 1997, compared with 16% of those providing 400-990 abortions and 17% of providers of 1,000 or more abortions. Expectations for the future, however, are much less associated with provider size; the proportion expecting to offer medical abortion, even without mifepristone, ranges from 26% of the smallest to 31% of the largest providers.

Respondents were asked how many early medical abortions they had performed. Assuming that nonrespondents provided them at the same rate as respondents, approximately 4,200 medical abortions were performed in 1996 and 4,300 in the first half of 1997, indicating a rapid increase in the use of the method.

Respondents also were asked if they were aware of physicians in their community who did not provide surgical abortion services but who had begun to perform medical abortions. The few who answered in the affirmative were recontacted and asked the names of the physicians. In many cases, the respondents had misunderstood the question and did not in fact know of any such providers. Only one respondent named a physician whose office confirmed that medical abortions were provided, and one named a physician who denied performing any abortions. Six respondents had reason to believe physicians in their community performed medical abortions, but refused to supply the names because they felt they would be betraying a confidence if they told us. Six others said they had heard rumors but did not know the names of the physicians. Seven could not be reached.

•Dilation and extraction procedures. In view of the ongoing controversies around so-called partial-birth abortions, we wanted to shed light on the number of such abortions that were actually performed. "Partial-birth" abortion, however, is a nonmedical term that has been variably described in the popular press and in legislation, and would be difficult for respondents to interpret. On the questionnaire, therefore, we used the medically accepted term, intact dilation and extraction (D&X), as defined by the American College of Obstetricians and Gynecologists (ACOG),** which is the only procedure that approximates the various descriptions of "partial-birth" abortion.

Eight respondents reported that they had performed a total of 363 D&X abortions in 1996 and 201 during the first half of 1997, and one other physician reported using the procedure. But several of them reported that they had difficulty estimating the exact number of abortions they had performed that met all components of the definition. Some D&X abortions were undoubtedly performed in facilities that did not return our questionnaire;^††if nonrespondents used the procedure at the same rate as responding facilities, we can project that 14 providers performed a total of about 650 D&X abortions in 1996. However, projecting from such a small number of cases can result in a wide range of possible error.

Respondents were asked the minimum and maximum gestation at which they perform D&X abortions. The most common minimum gestation was 20 weeks (counting from the last menstrual period), and the most common maximum was 24 weeks. Only two of the nine respondents used the procedure before 20 weeks, one as early as 16 weeks and another 18 weeks. Similarly, only two said their maximum gestation was more than 24 weeks, one 26 weeks and one 33 weeks.^‡‡ Thus, the large majority of D&X abortions were performed at 20 to 24 weeks.

Discussion

Although the abortion rate in the United States has declined in recent years, at 23 per 1,000 women of reproductive age in 1995 and 1996, it is still high compared with other Western developed countries. For example, in 1995, the abortion rate was 16 per 1,000 in Canada, 15 per 1,000 in England and Wales, six per 1,000 in the Netherlands and 18 per 1,000 in Sweden. The 1995 rates are higher, however, in some countries of the former Soviet Union and eastern and central Europe: 35 per 1,000 in Hungary, 87 per 1,000 in Romania and 68 per 1,000 in Russia.¹¹ (The rates for Romania and Russia may be undercounts because reporting is believed to be incomplete.)

The U.S. abortion rate declined each year between 1990 (when it was 27 per 1,000) and 1995, and remained the same for 1996. A factor contributing to the decline was the aging of the "baby boomers" into their late 30s and 40s, ages when there are fewer pregnancies and abortions.¹² If the age-specific abortion rates had been the same in 1996 as in 1992, there would have been 38,000 fewer abortions in 1996 than in 1992. This amounts to 23% of the actual decline of 163,000 abortions. Thus, the changed age structure of the U.S. population accounts for some but not most of the decline in the number and rate of abortions. The abortion ratio should be affected even less by the changing age distribution, because birthrates as well as abortion rates are lower among women in their 30s and 40s.

Much of the drop in the abortion rate probably reflects a decreasing rate of unintended pregnancy, which fell among all age-groups between 1987 and 1994.¹³ The fall in the abortion rate was especially great among teenagers, some of whom began using effective, long-acting contraceptive methods that were not available before 1992. In 1995, 10% of contraceptive users aged 15-19 were using the injectable and 3% were using the implant.¹⁴

The abortion rate among teenagers was also affected by a decrease in the proportion of unintended pregnancies terminated by abortion—that is, more teenagers continued their unintended pregnancies in the 1990s than was the case in the mid-1980s. Among women aged 20 and older, in contrast, the proportion of unintended pregnancies resulting in abortion increased.¹⁵ It is unclear whether the increase in the proportion of teenagers continuing their unplanned pregnancies was a result of changes in their preferences or their reduced ability to obtain abortion services.

The reason the abortion rate stopped falling in 1996 is also unclear. One contributing factor may be declining use of the contraceptive implant, a method that received negative publicity about removal problems and product liability lawsuits. There is no way to predict whether the abortion rate will resume its fall, remain stable for some time or increase.

The fall in the U.S. abortion rate is probably not an artifact of declining survey coverage and reporting, although it has become somewhat more difficult over the years to obtain information from abortion providers. The AGI national abortion count exceeded the CDC's by 13% in 1995,¹⁶ about the same amount as in 1992 when the difference was 12%. (At this time, CDC data for 1996 are not yet available.) The CDC relies primarily on reports to state health departments, many of which do not receive reporting from all providers and some of which have voluntary reporting.¹⁷

Trends in the availability of abortion services are mixed. The number of abortion providers continued its sharp decline between 1992 and 1996, the proportion of counties with no identified provider rose slightly and the proportion of women who live in unserved counties increased as well. The decline in the number of providers was concentrated among hospitals and physicians' practices, however, where only a few abortions are performed. The high costs of small-scale abortion provision¹⁸ and possibly antiabortion sentiment or pressure may have discouraged some hospitals and physicians who had been performing abortions for their own patients but for whom abortion services were not a major commitment or source of revenue; as a consequence, they may have found it more attractive to refer their patients to specialized providers.

The number of nonhospital providers performing 400 or more abortions has not declined, however. Slightly more metropolitan areas had services in 1996 than in 1992, and the number of abortion clinics increased slightly. Thus, facilities that are focused on making abortion services available have held their own in the face of continuing harassment and declining rates of unintended pregnancy.

Nevertheless, many women still have limited access to abortion services. Almost a third of metropolitan areas have no provider, and a similar proportion of women live in unserved counties. Two states—North and South Dakota—have only one provider each. Abortion rates by state of occurrence closely reflect the availability of services within the state.

Early medical abortion has the potential to expand the availability of abortion services in areas without surgical abortion providers and reduce the need for interstate travel, but as of 1997 this potential had not been realized. A few physicians who were not formerly abortion providers are performing medical abortions for some of their private patients, but most do not want their names to be widely known. While these medical abortion services may be valuable to the few patients who have access to them, they do not represent an important expansion of service availability, and because the number of abortions performed by these physicians is small, would not impair the accuracy of our 1996 abortion data.

Early medical abortion may become more common in the future, however. (The company with the rights to the distribution of mifepristone expects to have the drug approved and on the market in 1999.¹⁹) Although medical abortion is offered by few physicians who did not already provide surgical abortions, existing providers appear to be adopting it rapidly. As many medical abortions were performed in the first half of 1997 as in all of 1996, and the number of providers increased 39%. Many other surgical providers say they will probably offer methotrexate in the future, and even more say they intend to provide medical abortions when mifepristone is available.

In the controversy concerning so-called partial-birth abortions, the factual question given most attention is the number of such abortions performed. It is impossible to give a definitive answer to this question because of the vagueness of the term, which has no medically accepted definition. However, intact D&X, as defined by ACOG, is rarely performed, accounting for only about 0.03-0.05% of all abortions in 1996. The large majority of D&X abortions were performed at 20-24 weeks of gestation.

Efforts to reduce the number of abortions can best be focused on reducing unintended pregnancy by improving contraceptive use. The fall in the abortion rate up until 1995 probably resulted in large part from more effective contraceptive use among teenagers. Women of all ages could achieve lower pregnancy rates if they used more effective methods and had better access to family planning services. That lower abortion rates are possible is clear from the examples of other developed countries.

At the same time, women who have an unwanted pregnancy need to be aware of and have access to abortion services; otherwise, they cannot freely decide whether to continue the pregnancy. At present, more than one-fourth of women who live in metropolitan areas have no abortion provider in their county large enough to be well-known in the area, and a majority of women in nonmetropolitan counties live far from any provider. More effort is needed to make services accessible to these women. Short of that, the local health care community needs to ensure that women are able to obtain referral to providers and gain assistance in making arrangements, and receive appropriate follow-up care.