Methods

Population

This study was a national questionnaire survey among 2100 Dutch physicians; 1300 clinical specialists (200 cardiologists, surgeons, lung specialists, intensive care unit specialists and neurologists and 300 specialists in internal medicine), 500 general practitioners and 300 nursing home physicians. The sample sizes were based on the frequency with which physicians have medical responsibility for dying patients, and on the total number of physicians working in a specific discipline. Respondents were selected according to the following criteria: (1) they were clinically active at the time they completed the questionnaire, (2) they had actively practiced the registered specialty for at least 1 year, and (3) they had to be living in The Netherlands. All addresses were taken from the professional registries of the relevant specialties. Over a 10 month period (February 2006 to November 2006) questionnaires were sent out and returned. We applied strict rules to ensure the physicians’ anonymity. Physicians were requested to return a pre-printed card with their contact details, separately from the questionnaire. In such a way we were able to identify the physicians who responded without linking physicians to the questionnaire itself. Reminders were only sent to physicians from whom we did not receive the card. The questionnaire response rate, adjusted for physicians who were untraceable, was 55% (n = 1032). In this paper we used data of 1004 physicians; 28 physicians did not fill out any question about the hypothetical cases. Of all respondents, 72% were male and 28% were female; 82% of the physicians were older than 40 years; 60% were general practitioners, 31% medical specialists and 9% nursing home physicians. Non-responders did not differ from responders in age, sex or place of residence.

Measurement instruments

Physicians received a 10 page written questionnaire that focused on experiences with and attitudes towards the Dutch Euthanasia Act. The questionnaire had been tested in a pilot study among 16 respondents, who had little or no problems with the interpretation of the questions.10 Physicians were presented six hypothetical cases of older patients in the terminal stage of a lethal disease, who severely suffered from pain and fatigue. For each case, we applied small variations in patient-characteristics that were considered not relevant (eg, patient’s age varied from 71 to 76). The attributes that systematically varied between cases were based on earlier studies on end-of-life decision-making 10 11 12: (1) the type of medication that was used, (2) the physician’s intention while using this medication, (3) the kind of patient request, (4) the patient’s life expectancy and (5) the time until death after administration of the medication (table 1). We excluded clinically unrealistic combinations such as a life expectancy of a couple of months in combination with administration of proportional morphine and a time till death after the administration of morphine of a few minutes or hours. The total number of realistic combinations of these attributes was 50. Each physician was presented three ‘standard cases’ (table 2). The definition of these standard cases was based on the legal definition of euthanasia and on large scale studies that describe the characteristics of patients who died from euthanasia (case 1), palliative or terminal sedation (case 2) and symptom alleviation (case 3).10 13 14 Physicians were further presented three cases randomly selected out of the other 47. We ensured that all 47 cases were presented in similar frequencies but gave each physician a unique combination to avoid order effects. For all cases, physicians were asked which term would, in their opinion, describe the act in the presented case best: alleviation of symptoms, palliative or terminal sedation, ending of life, euthanasia, or another term. They were further asked whether they would (probably) report the case as an unnatural death to the municipal coroner (instead of a natural death).

Statistical analyses

For all statistical analyses we weighted the data for different sampling fractions and response rates to make the results representative for all physicians in the relevant disciplines. Missing values were excluded when these comprised less than 5% of all cases assuming that this would not influence the representativeness of the data. We reported absolute frequencies and weighted percentages using SPSS V.11.0.

We determined the relative importance of factors that may be related to the assignment of labels to the 47 other, non-standard hypothetical cases. The fact that each respondent evaluated three cases was accounted for by including respondent as a random effect in a multilevel multivariable logistic regression analysis. The model also included the five attributes (table 1) that systematically varied between the cases. The outcome variable was the dichotomy ‘Euthanasia/ending of life’ versus ‘Alleviation of symptoms/palliative or terminal sedation’. This dichotomy was used because ‘euthanasia’ and ‘ending of life’ both represent unnatural deaths, whereas alleviation of symptoms and palliative or terminal sedation are considered to involve a natural death. We present unadjusted and adjusted odds ratios and overall p values. Models were fitted with the Glimmix procedure for generalised linear mixed models in SAS V.9.2. Adaptive Gaussian quadrature was used as estimation method. The 47 non-standard hypothetical cases were further analysed to test univariate associations between physicians’ labelling of cases and their willingness to report, using the χ² test for statistical significance. In total, 3012 cases were presented (1004 respondents multiplied by three cases). In this paper we used 2976 cases because we excluded 36 cases where information about labelling or reporting was missing.

Results

Physicians’ labelling of three hypothetical end-of-life cases

A full description of the three ‘standard’ cases (A, B and C) is presented in table 2. Most physicians labelled case A, where the attending physician administered a sedative followed by a muscle relaxant upon an explicit patient request to end life, as ‘euthanasia’ (87%) or ‘ending of life’ (12%). In case B, the patient received midazolam until death; the large majority of physicians labelled this case as ‘palliative or terminal sedation’ (93%). In case C, the patient received morphine by a morphine drip that was increased twice. This case was mostly labelled as ‘symptom alleviation’ (85%) but in a minority of cases (13%) as ‘palliative or terminal sedation’.

Determinants associated with labelling hypothetical end-of-life cases as ‘euthanasia/ending of life’

The type of medication was significantly associated with physicians’ labelling of end-of-life cases, also when taking into account the other attributes in a multivariate logistic regression analysis (p<0.01): 99% of all 292 cases in which muscle relaxants had been administered were labelled as ‘euthanasia/ending of life’. Table 3 further shows that 63% of all 720 cases in which disproportional morphine was administered and 11% of the 625 cases in which proportional morphine was administered were labelled as ‘euthanasia/ending of life’. Furthermore, 15% of the 316 cases in which a sedative was administered, were labelled as ‘euthanasia/ending of life’. An intention to hasten death (54% of the 1614 cases) as compared to an intention to relieve suffering (12% of the 1322 cases) also contributed to the likelihood that cases were labelled as ‘euthanasia/ending of life’ (p<0.01). Furthermore, physicians were more inclined to label cases as ‘euthanasia/ending of life’ when the patient was expected to die within several months (81% of the 291 cases and 35% of the 886 cases) as compared to a few days (32% of the 1054 cases and 21% of the 705 cases) (p<0.01). The difference in odds ratios in the univariate and multivariate analyses for a patient’s life expectancy of a couple of months in combination with the time till death could be explained by the strong association with the type of medication. A request to end life was significantly associated with the label ‘euthanasia or ending of life’ in a univariate logistic regression analysis only; 52% of all cases in which the patient requested to end life.

Physicians’ labelling and willingness to report hypothetical end-of-life cases

Physicians were much more willing to report cases that they labelled as ‘euthanasia’ (87%) or ‘ending of life’ (56%) (table 4). The willingness to report cases labelled as ‘palliative or terminal sedation’ (9.6%) and cases labelled as ‘symptom alleviation’ (2.2%) was much lower (p<0.01).

Discussion

This study on hypothetical cases shows that labelling is strongly associated with medical (the type of medication), physician (their intention) and patient (life expectancy) related factors. Our study further shows that physicians’ labelling of similar acts varied. These labels are strongly associated with physician’s intended reporting behaviour.

The large random sample and the complete anonymity of the physicians are strengths of the study. We approached clinical reality by using hypothetical cases. Such cases enable us to standardise the characteristics and to confront all physicians with similar situations. However, concise hypothetical case descriptions may lead to different interpretation of the respondents, which is a common weakness of hypothetical case studies.15 16 Yet, our pilot study showed that most physicians found the cases realistic and could answer the questions adequately, although they sometimes expressed a need for more detailed information. Further, the willingness to report the acts presented in these end-of-life cases may not be identical to actual practice: real behaviour is known to be also influenced by cultural and situational factors.17 18 Because the association with the use of a muscle relaxant was rather strong, we ran the multivariate model again without cases in which a muscle relaxant was administered. The results were virtually identical to those of the original model.

The existence of a ‘grey zone’ between euthanasia and highly dosed alleviation of pain and symptoms was first described in 1990, in the first Dutch nationwide study on end-of-life decision-making.11 It was then estimated as concerning around 2% of all deaths. Our assessment of the grey zone identifies which characteristics clearly distinguish between different end-of-life acts and which characteristics are less distinctive and consequently may contribute to the existence of a grey zone.

The use of a muscle relaxant was a strong discriminating factor related to physicians’ label of ‘euthanasia/ending of life’, as almost all cases were labelled as such, also when taking into account the other attributes. For euthanasia, the Royal Dutch Society for the Advancement of Pharmacy (KNMP) recommends the use of a barbiturate to induce a coma, followed by a muscle relaxant to induce the patient’s death. In contrast to the use of a muscle relaxant, the use of morphine, and to a lesser extent the use of sedatives, was more ambivalent in their association with labelling the act as ‘euthanasia/ending of life’. On the one hand, in 11% of the cases in which morphine was administered in proportional dosages, physicians thought that the label ‘euthanasia/ending of life’ described the act best. Further, 15% of the cases in which a sedative was administered were labelled as ‘euthanasia/ending of life’. This is a remarkable finding as these cases are generally considered as normal medical practice because of the unlikeliness that life is actually shortened when administering morphine or sedatives in proportional dosages, as is suggested in other studies.7 8 On the other hand, cases in which morphine had been administered in disproportional dosages were labelled as ‘symptom alleviation/palliative or terminal sedation’ in 37% of cases. Thus, an even greater proportion of physicians do not always perceive the use of disproportional morphine as some form of physician-assisted dying. As such, our data show that with respect to the type of medication that is administered, the grey zone is the result of difficulties in the classification of end-of-life practices in two directions.

Apart from the type of medication, the physician’s intention and patient’s life expectancy play an important role in how physicians distinguish different end-of-life acts, even after correction for the other attributes. An intention to hasten death, which is generally included in the definition of euthanasia,19 20 was significantly associated with labelling acts as ‘euthanasia/ending of life’. However, only 54% of all cases with an intention to hasten death were labelled as such. Intentions are difficult to verify.21 More is necessary than just intention and it is often argued that acts should be morally evaluated in connection with the characteristics of the actual medical act, such as the type of medication. The association of a longer life expectancy (ie, a couple of months) with labelling cases as ‘euthanasia/ending of life’ corresponds with results from another study that suggested euthanasia is often (but not always) performed at an earlier stage in the disease process than palliative or terminal sedation.22 The fact that an intention of hastening death and a longer life expectancy significantly, but not unanimously, contributed to how physicians distinguish end-of-life acts, suggests that these factors also contribute to the existence of a grey zone. The type of request and the time till death on the other hand, were not significantly associated with labelling the act as ‘euthanasia/ending of life’. Apparently, these characteristics have virtually no influence in how physicians distinguish different end-of-life acts.

In a European study performed in intensive care units of 17 countries, the authors also reported a grey zone between treatments administered to relieve pain and suffering and treatments intended to actively shorten the patient’s life.23 Physicians had the intent to shorten the dying process by administering an overdose of morphine or diazepam in 2.2% of death cases. However, the distinction between acts intended to relieve suffering and acts intended to cause death appeared to not always be clear. Apparently, physicians in many countries encounter difficulties in drawing sharp lines between their acts and labelling is part of that discussion.24

Our study clearly shows that the ascribed label is strongly associated to whether or not physicians are willing to let their acts be reviewed. Most of the physicians were willing to report cases that they labelled as ‘euthanasia’. However, physicians were less inclined to report cases that were labelled as ‘ending of life’ than cases that were labelled as ‘euthanasia’. Possibly, physicians believe that cases of ‘ending of life’ represent acts that do not fulfil the definition of euthanasia in the Euthanasia Act, or acts in which some characteristics do not correspond with the criteria of careful practice. Some physicians would also report cases labelled as ‘palliative or terminal sedation’. At first, this seems a remarkable finding, as palliative or terminal sedation is generally considered as normal medical practice. However, if sedation involves cases with a longer life expectancy where physicians prefer to use sedatives instead of muscular relaxants, these cases indeed need to be reported. Figures from The Netherlands have shown that physicians sometimes use sedatives when they have the explicit intention of hastening death, or when the patient’s life expectancy is more than one month.1 25

The grey zone, as identified in our study, reflects physicians’ varying ideas about whether or not the type of medication, the patient’s life expectancy and their intention are determinative in whether the act should be regarded as a normal medical practice or not. It is possible that the strong association between labelling and reporting is the result of the fact that physicians base their willingness to report on their labelling of their acts. It is likely, however, that physicians simultaneously consider consequences of reporting when labelling their act. Probably, such considerations are already triggered by the question whether the physician would report the act as a ‘natural’ or ‘non-natural’ death, which is morally loaded labelling as well; the grey zone is a social phenomenon interrelated with many aspects. For instance, physicians might, even unconsciously, prefer not to label acts as euthanasia and thus refrain from reporting in order to preserve some medical freedom. Accordingly, physicians may use highly dosed morphine instead of barbiturates and neuromuscular relaxants, while intending to end or to shorten the patient’s life. In The Netherlands, fear of legal consequences of reporting has been mentioned as a reason not to report by some physicians and some of them think that reporting hampers the privacy of the relationship between patient and physician.1 Further, through labelling acts within the domain of normal medical practice, it may be easier to justify their act to themselves and others. The same probably holds for countries where euthanasia is not allowed; it cannot be precluded that physicians are tempted to label the administration of drugs with the explicit intention of hastening death as alleviation of symptoms.26 27 28

In conclusion, this hypothetical case-study shows that there are three main characteristics that contribute to the existence of a grey zone between euthanasia and highly dosed alleviation of pain and symptoms: medication other than muscle relaxants, physician’s intention and patient’s life expectancy. This grey zone is expressed in the variance of how physicians label their acts. Physician’s choice to label the case as either an act within the domain of normal medical practice or not is determinative in physicians’ attitudes towards public review of their act. The existence of a grey zone will therefore impede complete societal control and transparency. As such, this study clearly highlights that a physicians’ label is an important factor for medical, ethical and judicial evaluation. Although the label that physicians ascribe to an act may be reached unconsciously, the impact of labelling warrants specific attention. For physicians, constant awareness of their own actions and their accompanying labelling is recommended. For legal experts and ethicists it is advisable to be aware of the grey area between symptom alleviation and active ending of life.

Unanswered questions and future research

Our empirical assessment of the grey zone gives rise to several questions. To what extent is this grey zone based on insufficient knowledge and to what extent is it created on purpose? How large and to which end-of-life acts is the grey zone in countries where euthanasia is not allowed or restricted? Future studies should focus on physician’s knowledge and their personal reasons for labelling acts as practices within the domain of normal medical practice or not. International studies in countries with different legislation regarding euthanasia or other medical decisions at the end of life would also increase our understanding. Further education and debates with physicians, ethicists, legal experts and politicians, should enhance the level of agreement about what is considered physician-assisted dying, and thus should be reported, and what is not.