Deaths Associated With Abortion Compared to Childbirth
A Review of New and Old Data and The Medical and Legal Implications

Martha Shuping
and David C. Reardon, Ph.D.,
Thomas W. Strahan, J.D.,
John M. Thorp. Jr., M.D.
Reproduced with Permission

In the landmark ruling of Roe v. Wade, the Supreme Court identified three reasons why States might seek to proscribe or regulate induced abortion: (1) "to discourage illicit sexual conduct;" (2) to exercise "the State's interest - some phrase it in terms of duty - in protecting prenatal life;" and (3) "to protect the pregnant woman, that is, to restrain her from submitting to a procedure that placed her life in serious jeopardy."1

The first reason, to discourage illicit sexual conduct, was quickly rejected by the Court, because it was not argued by the State and was also inconsistent with a ban on abortion for married couples.

The second reason, the State's interest in protecting prenatal life, was the subject of much discussion (and remains so today), but this interest was substantially limited by the Court's determination that no consensus exists as to when human life begins. The Court allowed that the State's interest in protecting a prenatal life became more compelling as the pregnancy moved into later trimesters, but this interest could not impose a burden on the woman's own health needs.2

The third reason, to protect the life and health of the woman, was upheld as a legitimate and compelling state interest. However, the Court also noted that, due to medical advances,

abortion in early pregnancy, that is, prior to the end of the first trimester, although not without its risk, is now relatively safe. Mortality rates for women undergoing early abortions, where the procedure is legal, appear to be as low as or lower than the rates for normal childbirth. Consequently, any interest of the State in protecting the woman from an inherently hazardous procedure, except when it would be equally dangerous for her to forgo it, has largely disappeared.3

Using comparative mortality rates (abortion versus childbirth) as its marker, the Court held that "in the light of present medical knowledge," the State could only proscribe or regulate abortion to protect women's health after

approximately the end of the first trimester. This is so because of the now-established medical fact . . . that until the end of the first trimester mortality in abortion may be less than mortality in normal childbirth. It follows that, from and after this point, a State may regulate the abortion procedure to the extent that the regulation reasonably relates to the preservation and protection of maternal health.4

It was probably in regard to this judicial assessment of abortion's safety relative to childbirth that Chief Justice Berger stated that he was "troubled that the Court has taken notice of various scientific and medical data in reaching its conclusion."5 In stating his concern, the Chief Justice may have been reflecting the insight that medical opinions are often reversed in light of new discoveries. If later research were to reveal that abortion is not safer than childbirth, and more particularly that the mortality rate associated with first trimester abortions is not lower than that for childbirth, the logic of Roe would require a reversal of the practical impact of the ruling.6 By establishing comparative mortality rates as the standard that determines when a State's interests become "compelling," it follows that the constitutional restrictions on the State's right to proscribe or regulate abortion would necessarily contract if and when it were found that the mortality rates of abortion were higher than mortality rates associated with childbirth.

Thirty years later, the best available evidence now contradicts the "established medical fact" relied upon in Roe. Recent analyses of large medical databases linked to death certificates have now shown that when mortality rates associated with abortion and childbirth are examined using a single uniform standard, significantly higher mortality rates are associated with abortion (see Table 1).7,8 These record linkage studies have demonstrated that pregnancy-associated deaths are actually two to four times higher for aborting women compared to delivering women.

While no state has yet attempted to regulate or proscribe abortion based on these findings, it is likely that the key "medical fact" relied upon in Roe will come under much closer judicial scrutiny in the near future. Presented with evidence that abortion, even in the first trimester, is associated with higher rates of death among women, the Supreme Court could determine that the new demonstration of facts is sufficient to establish that the State has a compelling interest in regulating or even prohibiting abortion in the first trimester.9 While such a ruling might open the way for states to enact and enforce the type of restrictions that existed prior to 1973, it would not require any change in constitutional law. Instead, such a ruling would simply apply the existing standard in Roe to the best current understanding of facts. In that event, the principle of stare decisis would be preserved and the Court could not be accused of "compromises with social and political pressures."10 In short, the Court may allow new restrictions on abortion due to a change in facts without engaging in any reinterpretation of constitutional law.

Alternatively, as signaled in Planned Parenthood v. Casey, a demonstration that key factual assumptions in Roe were actually false might justify a complete repudiation of Roe.11 However, since such a departure from stare decisis might undermine the Court's credibility,12 it is likely that the Court would overrule Roe only if its failure to do so would cost women, their families, and the nation, a "terrible price."13 For abortion opponents, that "terrible price" would be the physical and psychological damage women suffer from abortion, plus the loss of previable human lives, that would continue in states that might fail to adequately restrict abortion. For abortion supporters, that "terrible price" would be the loss of easy access to abortion, as either a tool for self-determination or as a tool for population control, eugenics, or social engineering. Faced with these two opposing viewpoints, but confronted with a change in facts sufficient to reject Roe's standard as useful for determining when the State has a compelling interest in protecting maternal health, the Court could either (a) return the task of judging the medical evidence for when abortion is contraindicated or medically justified to state legislators and state regulators, thus opening the way to a hodgepodge of different regulations in different states; (b) require the States to prohibit abortion on the grounds of an affirmative duty to protect women's health and the lives of their pre-viable children, and thereby establish a single national standard; (c) define a new, more lax medical standard for determining when the State's interest in protecting women's health is compelling; or (d) define the right to seek and perform abortions as an absolute right that is not subject to any State regulation, an option that the Court has, in the past, firmly dismissed.14 While none of these options will satisfy all parties, the latter two seem especially unlikely as these would require the Court to determine that a higher priority must be placed on protecting the abortion "liberty"15 than on the protecting women's health.

Even if new information on abortion-associated mortality were not to have any effect on the constitutional law governing abortion regulations, however, it should have an impact on the medical judgment of physicians recommending abortion. This is especially true in countries such as Great Britain, which only allow abortion when it is medically deemed to be safer than carrying a child to term.16

Clearly, the question of the comparative mortality rates of abortion and childbirth is an important legal and medical issue. Therefore, the purpose of this paper is to examine the evidence accumulated regarding these comparative mortality rates in greater detail than has previously been done. To that end, the remaining portion of this paper will review the basis for and the difficulties involved in prior efforts to compare the mortality rates of abortion and childbirth, examine the strengths and weaknesses of the new record-based studies, and examine these findings in the light of related research that provides an additional context in which to interpret these results.

Obstacles in Assessing Pregnancy Associated Deaths

On March 1, 1989, Erica Richardson, a sixteen-year-old Maryland resident, bled to death from a uterine perforation only hours after undergoing an abortion. During the next five months, two other residents of Maryland, Gladys Estanislao and Debra Gray, also died from abortion complications.17 Surprisingly, none of these women were ever granted the smallest of recognitions-becoming a statistic. The official statistics issued by Maryland public health officials showed that there were no deaths from abortion in 1989. Indeed, Maryland only reported a single abortion-related death for the entire decade of 1980 to 1989.18

Actually, there was a fourth woman who died as a result of a 1989 abortion in Maryland. In this case, Susanne Logan fell into a coma during her abortion and awoke four months later as a quadriplegic, unable to talk. She survived for three years, dying in 1992 at the age of twenty-four.19 Since Susanne's death did not occur within forty-five days of her abortion, it has not been counted in any of the official abortion mortality statistics.

These four deaths occurred in one small state. For that same year, 1989, the Abortion Surveillance Unit of the Centers for Disease Control and Prevention (CDC) reported only twelve abortion-related deaths for the entire country.20 But, as we will see, the CDC lacks any regular and systematic means of identifying abortion-related deaths.

There are numerous inherent difficulties involved in efforts to identify deaths related to pregnancy, childbirth, and abortion. First, it is obvious that the quality of analyses and conclusions can never surpass the quality of data sources. Inaccurate data will produce inaccurate conclusions.

Claims that abortion mortality rates are lower than maternal mortality rates related to childbirth are based on comparing two sets of statistics: maternal mortality rates compiled by the National Center for Health Statistics (NCHS) through its National Vital Statistics System, and the number of deaths reported to be abortion related by the CDC.21 This comparison is problematic for two general reasons. First, NCHS and CDC employ different standards and means of data collection. Second, both systems are prone to missing a large percentage of deaths associated with childbirth and abortion.

Death certificates are the primary source of data used by NCHS to compile mortality statistics through its National Vital Statistics System. In the United States, cause of death on death certificates is normally reported by the attending physician. In some cases, particularly when the cause of death is due to violent or unknown causes, medical examiners or coroners will make the final classification of causes. In either case, a recent pregnancy may not be recorded due to error or lack of knowledge on the part of the attending physician or coroner. More careful analyses in individual states reveal that fifty percent or more of death certificates for pregnant or recently pregnant women failed to note the pregnancy.22

In many cases, the physician filing the death certificate may not know about a recent birth unless told by relatives. The physician is even less likely to know about a recent abortion, since most American women obtain abortions from specialists, not their own personal physicians. Similarly, interviews with relatives are less likely to be revealing in regard to abortion than they are in regard to childbirth, or even miscarriage. The deceased may not have told relatives about her abortion. Even if they are aware of it, relatives might refrain from telling the physician completing the death certificate about the abortion simply because they would not want it noted on this public record. There is also the risk that persons involved in reporting the death may deliberately obscure the underlying cause in cases of abortion-related death. This may be done either to protect families from potential embarrassment or to avoid the implication of malpractice against the abortion providers, who in some cases may also be the attending physician who is completing the death certificate.

Additional ambiguities arise in regard to efforts to accurately identify deaths that are related to pregnancy. The International Classification of Diseases ninth revision (ICD-9) defined "maternal death" as one which occurs while a woman is pregnant or within fortytwo days of the termination of the pregnancy, regardless of the outcome (abortion, miscarriage, or delivery) and anatomical site of the pregnancy, where the death is judged to be caused by a disease related to or aggravated by the pregnancy or its management, excluding incidental deaths.23 The provision to exclude deaths deemed to be "incidental" introduces subjective judgments that are often reversed upon closer review.24 For example, it is frequently unclear what role, if any, a current or recent pregnancy may have in deaths resulting from "some cancers, stroke, asthma, liver cirrhosis, pneumonia with influenza, anorexia nervosa, and many violent deaths such as suicide, homicide, and accidents."25

Moreover, the limitation to forty-two days after the pregnancy outcome is an arbitrary one, chosen to parallel the time period used in defining infant mortality. While it is known that some deaths related to pregnancy complications may occur outside this forty-two day period, as in the case of Susanne Logan, a specific time restriction is a practical coding convenience. A forty-two day time limit would also distort comparisons if delayed deaths are more commonly associated with abortion than childbirth. More certainly, this definition of maternal death associates a far larger period of time to maternal deaths than abortion deaths. In the case of abortion, the maximum period of time covered is forty-two days. In all other cases, every death occurring at any time during the pregnancy, plus an additional fortytwo days, must be considered as possible maternal deaths. In other words, abortion related deaths cover forty-two days of a woman's life compared to 312 days (on average) for women who carry to term. This seven-fold longer time period automatically encompasses a larger number of deaths and greater exposure to errors in judgment regarding whether or not the pregnancy was a contributing cause of death, especially in cases of death due to natural causes, such as heart failure.

To further complicate matters, if a woman undergoing an abortion has an unidentified ectopic pregnancy that subsequently ruptures and causes her death, should that be counted as an abortion related death or a maternal death? The CDC researchers who compile statistics on abortion deaths have chosen to exclude deaths from ectopic pregnancy following an abortion26 even though the deaths are at least partially due to the failure of the abortion provider to verify the site of the pregnancy and the completion of the abortion.27 An additional confounding factor is that scar tissue resulting from a prior induced abortion may be associated with increased risk of subsequent ectopic pregnancies.28 All of these considerations regarding abortion and ectopic pregnancies are especially problematic since ectopic pregnancies are the leading cause of maternal death in the United States.29

Another disparity in tracking deaths associated with childbirth and abortion is related to different coding standards. Coding rule 12 of the ICD-9 requires that deaths due to medical and surgical treatment must be reported under the complication of the procedure (embolism, for example) and not under the condition for treatment (elective abortion). According to researcher Isabelle Bégin,

In effect, this makes abortion a "ghost" category under which it is impossible to code a death. Medical coders have, in fact, relayed that any attempt to code a death due to abortion under an abortion category yields a "reject message" from the computer programs provided by the National Center for Health Statistics of Washington D.C., a division of the U.S. Centers for Disease Control in Atlanta, Georgia. . . . These computer programs simply incorporate the same problematic coding rules already used throughout the world. Only a minute number of abortion-related deaths actually qualify to be declared under abortion, i.e. those for which the medical certificate of death categorically and unequivocally gives abortion as the underlying cause of death. If abortion is mentioned anywhere else on the death certificate, on the underlying cause line, the death gets coded as an accident of some kind, a sudden or unexpected death, an illness (like septicaemia-blood poisoning) or an injury, etc.30

While there are numerous three digit ICD codes for identifying specific causes of death related to pregnancy and delivery (such as 633 for ectopic pregnancy, 640 for hemorrhage in early pregnancy, 666 for postpartum hemorrhage, et cetera), there is only one code for legally induced abortion (635), and it is chiefly intended to identify when an abortion was provided in medical and billing records. As described above, the code for a legally induced abortion is not intended for use when identifying cause of death. Examination of death certificates where abortion is known to be the underlying cause of death reveals that code 635 (legal abortion) is rarely identified as the cause of death.31

In short, death certificates are not a reliable source for identifying deaths related to childbirth and are an even weaker means of identifying deaths related to abortion. In 1972, Family Planning Evaluations Division of the CDC implemented a patchwork method for identifying abortion related deaths to "complement the vital statistics activities of NCHS by identifying causes of preventable abortion deaths."32 We describe this surveillance system as a "patchwork system" because it does not rely on any regular source of data. Since no state has laws or regulations requiring doctors, emergency room personnel, or coroners to report possible abortion related deaths for investigation, the CDC system simply investigates such deaths if they happen to come to their attention through "reports from state health departments, case histories published in medical journals, anecdotal reports from state medical or hospital associations, CDC special surveys of deaths from other fertility control measures, reports to national abortion organizations, registries from the Food and Drug Administration, and reports from state maternal mortality review committees."33

The inadequacy of this system of reliance on ad hoc tips is illustrated by the fact that for the period immediately after it was implemented, the CDC identified twice the number of abortion related deaths as those reported by NCHS,34 while in the 1980s it discovered only three percent more abortion related deaths than NCHS. By contrast, a single investigative journalist examining public records, including autopsies and malpractice suits, documented thirty to thirty-nine percent more abortion related deaths than were reported by either NCHS or the CDC.35

Next Page: How Accurate is the CDC Surveillance System?
1, 2, 3, 4, 5