Euthanasia and Physician-Assisted Suicide: A Review of the Empirical Data From the United States

For more than a decade, there has been an intense debate about the ethics and legality of euthanasia and physician-assisted suicide (PAS) in the United States.^1-5 In June 1997, the US Supreme Court unanimously ruled that there is neither a constitutional right nor a constitutional prohibition to euthanasia or PAS.^6,7 This permitted Oregon to experiment with legalizing PAS. During this decade, most other states have consistently opposed legalization. In the weeks after the US Supreme Court decision, the Florida Supreme Court also ruled that there is no constitutional right to PAS.⁸ At least 7 state legislatures have voted to explicitly prohibit euthanasia and PAS.⁹ Indeed, a bill to legalize euthanasia or PAS has been considered by a full chamber of a state legislature in only one state, Maine, and that bill was defeated 99 to 42.¹⁰ In November 1998, 70% of the voters of Michigan resoundingly defeated a referendum to legalize PAS, while in November 2000 Maine voters also rejected legalizing PAS.¹¹

The extensive debates for and against euthanasia and PAS have made the arguments more refined, subtle, and sophisticated. Yet the essential claims—arguments based on patients' autonomy to control their own lives and beneficence in relieving excruciating pain and suffering—have remained remarkably the same since the late 19th-century debates about euthanasia.^5,12 However, the current debate has spawned significant and unprecedented empirical research, illuminating many aspects of and claims about euthanasia and PAS. This article reviews the empirical data about euthanasia and PAS in the United States regarding: (1) the public's attitudes, (2) physicians' attitudes, (3) physicians' practices and experiences, (4) nonphysician health care professionals' attitudes and practices, and (5) patients' attitudes and experiences. It will conclude with a summary of the most important question in need of additional empirical inquiry.

In this article, whenever the term euthanasia is used, voluntary active euthanasia is meant. Other forms of euthanasia, nonvoluntary or involuntary, have not been extensively advocated or studied.⁵

There have been innumerable surveys of the American public on euthanasia and PAS.^13-15 Most information derives from a few questions added to general surveys and do not probe deeply; only a few surveys have been in-depth analyses. In general, opponents and proponents of euthanasia or PAS endorse 4 conclusions from these data.

First, depending on how questions are worded and the types of choices offered, public support for euthanasia or PAS can vary widely, from about 34% to about 65% (Table 1).^13,14 In other words, some Americans are firm in their views of euthanasia and PAS, while others are more labile. The best way to understand public opinion might be by the "Rule of Thirds." Roughly, one third of Americans seem to support voluntary active euthanasia or PAS no matter what the circumstances. For instance, 29.3% of Americans support euthanasia or PAS for terminally ill patients who are not in pain but desire these interventions because they view life as meaningless. Similarly, 36.2% support euthanasia or PAS for terminally ill patients who give as their reason not wanting to be a burden on their family.¹⁶ These are the approximate one third whose support for euthanasia or PAS is not affected by the interventions, the patient's motivations, or the circumstances. Conversely, another third or so of Americans oppose euthanasia or PAS no matter what the circumstances. Almost all the surveys report the highest levels of support for euthanasia or PAS to be about 65%.^13-16 These data mean that roughly one third of Americans—the difference between 100% of the public and the 65% who support euthanasia for patients in pain—oppose euthanasia or PAS even for terminally ill patients who are experiencing unremitting pain, despite optimal management. The remaining third or so of Americans constitute the volatile public. They support euthanasia or PAS in some circumstances, usually involving extreme pain, but oppose it in other circumstances, such as for reasons of indignity or because the patient does not want to be a burden (Table 2).

Consequently, support for euthanasia or PAS is not as extensive as the reports that two thirds of Americans support these interventions make it appear. Furthermore, for few of these people is legalizing euthanasia or PAS a leading issue, the primary element that will determine their vote. In this sense, unlike abortion, euthanasia and PAS do not appear to be litmus test issues.

Second, surveys that assess trends over time indicate that the significant rise in support for euthanasia and PAS occurred in the mid 1970s, not the 1990s.¹⁴ Indeed, since the mid 1970s, support for these interventions has been constant (Table 1). Interestingly, this is similar to the trends found in the Netherlands.¹⁷ Consequently, the extensive public debates during the last decade do not appear to have shifted public opinion significantly.

Third, while medical ethicists, philosophers, lawyers, and others have spent much time debating whether euthanasia is fundamentally different from PAS and elucidating potential distinctions, the American public does not seem to make much of the distinction. Polls show that Americans support euthanasia at the same rate that they support PAS (Table 2).¹⁵ Conversely, the public does distinguish withdrawing life support or providing pain medications, even with the increased risk of respiratory depression and death from euthanasia and PAS.^14,15 Despite arguments by some philosophers suggesting that there is no moral difference,¹⁸ more than 90% of the public deem withdrawing life support as ethical, while at best 65% support euthanasia or PAS.¹⁵

Finally, certain sociodemographic characteristics consistently predict support and opposition to euthanasia or PAS.^13-15 Catholics and people who report themselves to be more religious are significantly more opposed to euthanasia or PAS. Similarly, African Americans and older individuals are significantly more opposed to euthanasia or PAS. Finally, some, but not all, surveys suggest that women are significantly more opposed to euthanasia or PAS. Interestingly, patients with terminal illnesses, such as cancer and chronic obstructive pulmonary disease, have attitudes that are almost identical to the public's.¹⁶ In other words, having a serious, life-threatening illness itself does not seem to alter attitudes toward the permissibility or opposition to euthanasia or PAS. Similarly, being a caregiver for a terminally ill patient or a recently bereaved caregiver does not seem to affect attitudes toward euthanasia or PAS.¹⁶

During the last decade, US physicians have been extensively surveyed about euthanasia and PAS.^19-48 Many of the surveys, especially the early ones, are problematic in their methods.⁵ The surveyed cohorts are narrow or biased, and the response rates are low. More important, questions are frequently worded poorly and abstractly in a confusing, emotionally laden, or biased manner. For instance, they often conflate terminating medical treatments with euthanasia or ask whether euthanasia or PAS is never ethically justified. Furthermore, many of the questions use multiple hypothetical propositions—requiring leaps of imagination by respondents—that are known to make the data unreliable. For instance, physicians are frequently asked, if euthanasia or PAS were legalized, would there be some circumstances in which they would be willing to perform euthanasia or PAS? In addition, there has been no consistency among the questions, making it difficult to compare the data across different surveys. In recent years, the surveys have addressed some of these problems, making the data more reliable, although there still appears to be the problem that physicians confound euthanasia with terminating life-sustaining treatments and euthanasia with PAS.^16,43

Surveys of physicians' attitudes have evaluated 3 issues that have not usually been clearly distinguished: (1) belief that euthanasia or PAS is ethically justifiable, (2) support for legalization of either intervention, and (3) willingness to perform either intervention (Table 3).^19-48 The more reliable surveys find that most US physicians do not view euthanasia or PAS as ethical. The major exceptions seem to ask abstractly whether these interventions might be justifiable "in some circumstances" (Table 3). More typical are surveys that report that fewer than half of physicians support euthanasia or PAS, or those in which respondents find suicide rational in some cases but believe that physicians should not assist (Table 3).

Regarding legalization, among physicians there seems to be no consistent pattern, probably because questions ask about specific legislation that varies and because respondents may not be familiar with the particular facets of the legislation. For instance, in a survey of Michigan physicians, Bachman et al³² could demonstrate most physicians (56.6%) supporting PAS only when they were forced to choose either legalization or an explicit ban; without being forced into this choice, only 38.9% supported permitting PAS. Consistently, few physicians would be willing to perform euthanasia or PAS if either were legalized (Table 3).

These data demonstrate another important factor: unlike the American public, US physicians distinguish between euthanasia and PAS. They are much more likely to support providing PAS than euthanasia.^{15,20,25,29,37,47,48} Only a few studies^19,35,37,42 have found most physicians supporting euthanasia. Therefore, unlike the American public, support for euthanasia or PAS among US physicians crucially depends on the intervention being asked about.¹⁵ This is different from Dutch physicians, who do not seem to distinguish euthanasia and PAS.⁴⁷

There are important factors associated with support for euthanasia or PAS. Like the American public, US physicians who are Catholic or religious are significantly less likely to support euthanasia or PAS.^{15,21,25,29,32,33,38,42,43,47,48} Similarly, surveys have reported certain specialties as more supportive of euthanasia or PAS than others.^{29,31,33,43,48} Surgical oncologists are more likely to support euthanasia or PAS than medical oncologists. Psychiatrists and obstetricians and gynecologists are more supportive of euthanasia or PAS, with internists, especially oncologists, less supportive. Still, others have found family or general practitioners as more supportive than internists.

Finally, at least among US oncologists, there appears to be a significant decline in support for euthanasia or PAS between the early and late 1990s.^15,44,48 Between 1994 and 1998, support for euthanasia and PAS significantly declined among oncologists in the scenario of a patient terminally ill with cancer who had unremitting pain.^15,48 Although it is hard to know precisely why this decline has occurred, 2 explanations seem reasonable. The recent focus on end-of-life care has revealed the multiplicity of interventions, besides euthanasia and PAS, that can be used to improve the quality of life of the terminally ill. Consequently, euthanasia and PAS seem less necessary and desirable to ensure good end-of-life care. Furthermore, support tends to be higher when considering euthanasia and PAS in the abstract, as a philosophical question. But as they become more real and personal and physicians may be called on to actually perform these interventions, physicians are likely to be less supportive. This may also partially explain why psychiatrists, obstetricians, surgeons, and others who rarely care for terminally ill patients are more supportive than oncologists.

Numerous studies have documented the practices of US physicians regarding euthanasia or PAS (Table 4). The precise proportion of physicians who have received such requests is unclear because there is significant variation in the reported frequencies. The different reported rates of requests for euthanasia and PAS may reflect methodological issues, such as: (1) the differences between mailed and telephone surveys; (2) the different dates of the surveys, with physicians being more willing to acknowledge performing these interventions in later years, as the debate becomes more public and accepted; (3) the different regions of the country, with those in the West having requests more frequently than those in the New England or North Central regions⁴³; and (4) the different investigators, with physicians more willing to acknowledge performing these interventions when the survey comes from investigators from the same state or a colleague in the same specialty.^{15,30,32-34,39,43,48} However, in general, it appears that oncologists have received many more requests than nononcologists. Fewer than 20% of nononcologists have received requests for PAS, while it appears that among oncologists as many as 50% have received requests for euthanasia or PAS (Table 4). This is probably because oncologists are more likely to care for dying patients than internists, surgeons, neurologists, or other physicians. Nevertheless, even among oncologists, the survey results vary considerably, suggesting residual methodological issues.

In general, physicians who have received requests have received few requests.^34,39,43,45 For instance, Meier et al⁴³ report that, overall, physicians who received requests for PAS received a median of 3 requests (range, 1-100) in their careers and a median of 4 requests (range, 1-50) for euthanasia. Carver et al⁴⁵ reported that, among neurologists who received requests, the mean number of requests for PAS was 7 and was 5 for euthanasia.

Many studies indicate that a small, but definite, proportion of US physicians have performed euthanasia or PAS, despite its being illegal. Again, the data provide conflicting evidence on the precise frequency of such interventions, with reported frequencies varying more than 6-fold even among the best studies (Table 4). As with requests, oncologists generally report having performed euthanasia or PAS more frequently. Much of this variation may be attributable to the reasons already cited, especially the differences in specialties. However, there is another methodological concern. The study by Meier et al⁴³ is the only study to have reported that more US physicians perform euthanasia than PAS. This finding contrasts with the data showing that US physicians are significantly more supportive of PAS than euthanasia.^{15,20,25,29,37,47,48} This result may be because physicians were classifying cases of terminating care as euthanasia. As reported by Emanuel et al,⁴⁹ despite careful wording, physicians frequently confound euthanasia and terminating life-sustaining treatments, and this may be more common and harder to control for in mailed rather than telephone surveys.

When US physicians have performed euthanasia or PAS, they have done so rarely. Meier et al⁴³ reported that the median number of PAS cases was 2 (range, 1-25), and the median number of euthanasia cases also 2 (range, 1-150). A recent survey of oncologists by the American Society of Clinical Oncology reported that, of those who had performed PAS, 37% had done so only once in their careers, while 18% had done so 5 or more times.⁴⁸ Similarly, among the US oncologists who had performed euthanasia, more than half had done so only once, and just 12% had done so 5 or more times.⁴⁸

Beyond the rates of requests and performance of euthanasia and PAS, what do physicians do when they receive a request and when they perform euthanasia or PAS? Back et al³⁴ reported that initially 76% of physicians increased treatment of physical symptoms, 65% treated depression and anxiety, and 24% referred the patient for a psychiatric evaluation. Similarly, Meier et al⁴³ reported that 71% of physicians responded to requests for euthanasia or PAS by increasing analgesic treatment, while 30% used fewer life-prolonging therapies and 25% prescribed antidepressants.

Regarding the actual performance of euthanasia and PAS, Meier⁴³ and Emanuel⁴⁹ and their colleagues provide similar data, at least as regards PAS (Table 5). They show that, while safeguards are adhered to overall, there are a myriad of problems. For instance, although most patients initiated the request for PAS, almost half of them did not repeat the request. Most important, both studies show that about 5% of patients were unconscious at the time of death and could not, therefore, provide concurrent consent. More than 95% of patients had severe symptoms, but according to Meier et al, only 54% had significant pain, while according to Emanuel et al, 84% of the patients with cancer who received PAS had substantial pain. In 40% to 54% of cases, the patients were getting hospice care, at least one measure of quality end-of-life care. Similarly, in many cases, patients who receive PAS had long-term relationships (>1 year) with their physicians. Finally, there are divergent data, ranging from 20% to 40%, on what proportion of patients provided with medications or a prescription ultimately does not use them. Differences in underlying disease may partially account for differences in the data between these 2 studies; Meier et al provide data on patients with many different terminal illnesses, whereas Emanuel et al interviewed oncologists and provided data on patients dying of cancer.

Two studies have examined the effect on physicians of performing euthanasia or PAS. Meier et al⁴³ and Emanuel et al⁴⁹ reported that most physicians were comfortable with having performed euthanasia or PAS. According to Meier et al, 19% of physicians were uncomfortable after performing PAS, and 12% were uncomfortable after performing euthanasia. (This lower proportion of uncomfortableness after performing euthanasia may reflect that many of these so-called euthanasia cases were actually cases of terminating life-sustaining treatments.) They also found that in similar circumstances only 1% would not comply with PAS and 7% would not comply with euthanasia. Emanuel et al reported that 25% regretted performing euthanasia or PAS and that 15% had adverse emotional reactions to performing euthanasia or PAS. At least in the cases reported by Emanuel et al, these reactions did not seem related to fear of prosecution.

Finally, there is some disagreement about failed PAS attempts. Emanuel et al⁴⁹ reported that in 15% of cases PAS failed; that is, patients were given a prescription or attempted suicide, but did not die. Ganzini et al⁵² recently reported that there had been no failed PAS attempts in Oregon since legalization. The reports from the first 2 years' experience by the Oregon Health Division, Portland, also show no failed PAS attempts.⁵³ As Nuland⁵⁴ notes, the lack of problems with PAS in these reports from Oregon contrasts with the recently reported Dutch experience, in which 7% of PAS cases had complications and in 16% it was taking "longer than expected."⁵⁵ Ultimately, in 18.4% of PAS cases in the Netherlands, physicians intervened to administer lethal medications, converting PAS cases into euthanasia.⁵³ The importance of this for the United States relates to the possibility of legalizing PAS without legalizing euthanasia, and what is to be done in the cases of failed PAS. As the data demonstrate, in the Netherlands, the accepted norm is to administer lethal medications—that is, perform euthanasia—in cases of failed PAS. This would not be permitted in the United States if euthanasia remains illegal. If the data from Emanuel et al and the Dutch investigators are correct, there may be serious dilemmas for physicians if PAS is legalized but euthanasia is not.

There have been at least 9 surveys of nonphysician health care professionals (mostly nurses) regarding euthanasia and PAS (Table 6).^38,56-64 Overall, these studies are not as rigorous in their methods as the best studies of physicians or patients. They demonstrate that about half of nonphysician health professionals support euthanasia or PAS in some circumstances, and that fewer than one third have received requests for euthanasia or PAS. Again, the type of religion and the strength of religious beliefs are associated with support for euthanasia and PAS. The data regarding performance of euthanasia or PAS by nurses vary widely, with one study showing that about 16% have participated in euthanasia or PAS, and others showing that fewer than 5% have done so (Table 6).

Although some studies have examined patients' wishes to hasten death and suicidal ideation, only a few studies^{15,16,52,53,65} have actually examined the attitudes and experiences of US patients regarding euthanasia and PAS (Table 5). Breitbart et al⁵⁰ examined patients with human immunodeficiency virus and acquired immunodeficiency disease syndrome (HIV/AIDS) in New York City; Ganzini et al⁵¹ interviewed patients with amyotrophic lateral sclerosis in Oregon; and Emanuel et al¹⁵ surveyed patients with cancer in Massachusetts. In addition, there are data reporting on the first 2 years' experience of legalized PAS in Oregon, involving some 43 cases.^53,65 There are additional data on the practices of euthanasia and PAS among patients determined to be terminally ill by their physicians.¹⁶ Four major conclusions can be drawn from these data.

First, mainly patients with cancer use euthanasia and PAS. Among the first 43 cases of PAS in Oregon, 72% of the patients had cancer.⁵³ Meier et al⁴³ report that among patients receiving PAS, 70% had cancer, while among those receiving euthanasia, only 23% had cancer. These data are comparable to the data from the Netherlands, in which 80% of euthanasia and 78% of PAS cases involved patients with cancer,⁶⁶ and from the Northern Territory, Australia, where all 7 patients who received euthanasia when it was briefly legalized had cancer.⁶⁷

Second, it appears that pain is not a major determinant of interest in or use of euthanasia or PAS (Table 5). Almost all of these studies—as well as the interviews with physicians who have administered euthanasia and PAS^34,43—have shown that pain is not a predictor of patients' interest in euthanasia or PAS. For instance, among the patients receiving PAS in Oregon, only 1 of 15 had uncontrolled pain.⁶⁵ Breitbart et al⁵⁰ reported that pain, pain intensity, and pain-related functional impairment were not associated with interest in PAS among patients with HIV/AIDS. Emanuel et al¹⁵ reported that for oncology patients, pain was not associated with personal interest in euthanasia or PAS. However, they did find that for terminally ill patients, pain was among the factors associated with personally considering euthanasia or PAS.¹⁶

Third, depression, hopelessness, and general psychological distress are consistently associated with interest in PAS and euthanasia (Table 5). Breitbart et al⁵⁰ reported that depression and hopelessness were strongly related to interest in PAS for patients with HIV/AIDS. Emanuel et al¹⁵ reported that, for oncology patients and terminally ill patients, depressive symptoms were associated with personal interest in euthanasia or PAS, such as discussing these interventions and hoarding drugs for the purpose of PAS. Ganzini et al⁵¹(p968) reported that hopelessness, but not depression, was associated with "considering taking a prescription for a medicine whose sole purpose was to end my life."

Fourth, Emanuel et al¹⁶ reported that among terminally ill patients, the extent of caregiving needs was associated with interest in euthanasia or PAS. Ganzini et al,⁵¹ however, reported that there was not an association between the burden of caring for the patients and whether caregivers supported or opposed a patient's request for PAS.

Although it is known that PAS and euthanasia occur in a small proportion of all deaths, what is not known is the precise frequency these interventions are used. In the Netherlands, 3.4% of all deaths are by euthanasia and PAS, including involuntary euthanasia.⁶⁶ In Oregon, the proportion of all deaths by PAS reported to the Oregon Health Division is 0.09%.⁵³ Such a low rate raises skepticism that not all cases of physician-assisted death are reported.⁵⁴ Emanuel et al¹⁶ have reported a rate of 0.4% among competent terminally ill US patients.

There are 6 major areas related to euthanasia and PAS in need of additional research in the United States. First, there are few data on the relationship between euthanasia or PAS and the provision of optimal end-of-life care. Are euthanasia and PAS used as truly last-ditch interventions for patients refractory to appropriate end-of-life interventions? Or are they used as substitutes for optimal end-of-life care? The American Society of Clinical Oncology survey suggested that there was a relationship between not being able to get dying patients all the care they needed and use of euthanasia and PAS.⁴⁶ This result needs confirmation. Furthermore, we need to understand what are the predictors of physicians who come to use euthanasia and PAS only after trying optimal care, vs those who use these interventions as a substitute. Is this the result of structural or financial barriers to optimal end-of-life care, or is it the result of problems on the part of physicians, such as lack of training in end-of-life care?

Second, there are divergent data on how frequently PAS fails and no data on what is done when it does fail. If, in the United States, only PAS will be legalized, what do physicians do when it fails?

Third, there is no information on the short- and long-term effects of euthanasia and PAS on the surviving family members of the patients.¹⁶ Immediately after the interventions, families may have the psychological need to be supportive of the decision and believe that the right thing was done. However, with the passage of time, they may have different views.

Fourth, there are conflicting data on the actual frequency of euthanasia and PAS. These interventions occur, but how frequently? It may be that conducting a death certificate follow-back study modeled on the Dutch studies^55,66 will be the best way to obtain accurate data on the frequency of these interventions, as well as the reasons for the interventions, the palliative measures taken, and the effects on the family.

Fifth, there are no data on the frequency of nonvoluntary euthanasia in the United States. In the Netherlands, nonvoluntary euthanasia occurs in 0.7% of all deaths.⁵⁵ The rate may be higher in the United States, given the expense and financial problems associated with end-of-life care.^68,69 Issues of coercion and of performing euthanasia on patients who are not competent are serious, and there are inadequate data on these events in the United States.

Finally, there are no data on euthanasia and PAS among children. Although death is rare among children, annually there are several thousand deaths among children with cancer and HIV/AIDS. These deaths tend to occur after significant and prolonged illnesses, and symptom management is less than optimal.⁷⁰ The American Society of Clinical Oncology survey of US oncologists suggests that there are instances of pediatric euthanasia or PAS.⁴⁸ Why these occur and how they are handled are also important and controversial issues.

Unfortunately, each of these issues is difficult to study because euthanasia and PAS are rare events, requiring screening of many physicians to identify just a few cases. Therefore, such studies will be large and expensive.

During the last decade, there has been a substantial amount of empirical research conducted on euthanasia and PAS in the United States. This empirical research has revealed many unexpected findings that have significantly affected the public debate. Such findings include: (1) Public support for euthanasia and PAS is closely linked with the reasons patients want these interventions; most of the public support the interventions only for patients in excruciating pain. (2) Yet, pain does not appear to be the primary factor motivating patients to request euthanasia and PAS; depressive symptoms, hopelessness, and other psychological factors appear to motivate patients' requests for euthanasia and PAS. Therefore, public support conflicts with the actual facts about patient interest in euthanasia and PAS. (3) Euthanasia and PAS occur, albeit at a low rate. Indeed, more than 99% of all dying Americans do not have these interventions, and even in the Netherlands, more than 96% of all decedents do not have these interventions.

Accepted for publication May 1, 2001.

Corresponding author and reprints: Ezekiel J. Emanuel, MD, PhD, Department of Clinical Bioethics, Warren G. Magnuson Clinical Center, Bldg 10, Room 1C118, National Institutes of Health, Bethesda, MD 20892-1156.