Introduction

A growing number of patients die each year in hospital emergency departments (EDs). In the USA the incidence was estimated at 300,000 patients per year (0.3% of ED visits) in 1998 [1], and 379,000 in the year 2000 [2]. The situation is similar in France [3, 4] but has not yet been fully investigated. Characteristics of patients dying in emergency departments (demographic data, previous functional limitation, chronic underlying diseases, precipitating factors [5], percentages of decisions to withhold or to withdraw life-support therapies) are not well known. We previously conducted a monocentric, pilot study [4] which found that 0.8% of admitted patients died in the ED. A decision to withhold or withdraw life-sustaining treatments concerned 80% of these patients, many of whom had chronic underlying diseases and/or severe functional disabilities [6].

Decisions to limit life support have been widely studied in critical care medicine, both in the USA [7] and in Europe [8]. A multicenter survey found that 77% of deaths in US critical care units were associated with such a decision [7]. Interactions between end-of-life models and emergency department care have been explored by Chan [9], and even though the emergency department might not be the most appropriate place to give end-of-life care, the fact remains that many patients come to the ED and die there.

However, the emergency department remains a unique setting because physicians often lack crucial data concerning previous state of health and functional limitation. The absence of an ongoing long-term relationship with the patient and his family and the lack of time make a decision to limit life support particularly difficult [10, 11].

We conducted a multicenter prospective observational study to describe the characteristics of patients who died in hospital EDs. The second goal was to study the number of patients for whom a decision to limit life support was taken and to describe the processes leading to such a decision.

Patients and methods

Study design

A prospective cross-sectional survey with recruitment of every patient who died over two 2-month periods (November 2004 to December 2004 and April 2005 to May 2005) in participating emergency departments was carried out.

Setting

Six hundred EDs were identified in the French Society of Emergency Medicine database. One hundred seventy-one French departments (22.8% of all French emergency departments) agreed to participate in this study, 37 of which were university hospital EDs. Three other emergency departments were located in Belgium.

Patients

Every patient who died in a participating emergency department was included.

Endpoint

Withdrawal was defined as discontinuation of treatments that had previously been implemented, and withholding was defined as a predetermined decision not to implement therapies that would otherwise be deemed necessary: endotracheal intubation, mechanical ventilation, intravenous (IV) fluid expansion, massive transfusion (more than three red cell packs), vasopressor infusion, cardiopulmonary resuscitation, renal replacement therapy [12]. In some circumstances both decisions could be taken.

Methods

The following data were recorded: age, sex, previous functional limitation (Knaus’s classification)Footnote 1 [13], main underlying chronic diseases [14], Acute Physiology and Chronic Health Evaluation (APACHE) II score [13], main diagnosis, presence of organ failures, decisions to withhold or withdraw life-support therapy [15], and time interval between ED admission and death.

Institution of palliative care was also recorded from a predefined list: analgesia, sedation, hydration, mouth care, reposition for comfort as appropriate, emotional support of the patient and his relatives.

In each participating ED, a senior physician was responsible for collecting the data prospectively on a standardized case report form. If a decision to limit life support was taken, then additional data were collected: those involved in the decision were identified, the reasons for making such a decision were recorded from nine predefined criteria (Table 3), and the time interval between ED admission and the time to reach such a decision was recorded. Involvement of the patient or of families in the decision-making process was also recorded. Patients were coded to preserve anonymity. Each center was asked to provide the number of patients admitted to the ED during the study period.

This observational study was approved by the Nantes University Hospital’s Ethics Committee. Waiving of consent has been authorized for this study, according to French law.

Statistical analysis

Patient characteristics were described using mean and standard deviation (SD) for continuous outcomes, and numbers and percentages for qualitative outcomes. Analysis of presence/absence of withholding or withdrawing decision was performed first through univariate analyses: chi-square tests and Student’s t tests. Then, considering variables associated with p value lower than 0.2, multivariate analysis was performed using a logistic regression model. Patients with missing data were discarded from this multivariate analysis. Statistical analysis was performed by using SAS 9.1, and a p value lower than 0.05 was considered significant.

Results

Characteristics of patients who died in the EDs

A total of 2,512 patients died in the ED and were enrolled, of whom 92 (3.7%) were excluded because of lack of sufficient data. Analysis was therefore conducted on the remaining 2,420 patients. Table 1 shows the characteristics of these patients. Mean ± SD age was 77.3 ± 15.1 years, sex ratio was close to 1 (1,196 males, 49.4% versus 1,224 females, 50.6%), and men were significantly younger than women (74.1 ± 15.7 years versus 80.9 ± 13.6 years, p < 10−4). Less than half of the patients were referred to the emergency department after medical advice [general practitioner: 914 (37.7%), hospital physician (6.65%)]. Nine hundred patients (37.2%) were transported to the ED by mobile intensive care units staffed by physicians [Service Mobile d’Urgence et de Réanimation (SMUR)]. Most patients were living at home (71.8%) as opposed to living in a long-term care facility, 46% of patients had limited functional limitation, and 81.4% had chronic underlying disease. The most frequent presenting medical disorders were cardiovascular, neurological, and respiratory; 92.6% of patients had at least one organ failure (48.6% had only one organ failure, 28.4% had two, and 15.5% had three or more). One hundred seventy-nine (7.4%) patients died unexpectedly (sudden death) without previous organ failure.

Table 1 Demographic and clinical characteristics of the 2,420 patients who died in 174 emergency departments
Full size table

Patient management in the ED

Life-support therapies were instituted for 1,781 patients (73.6%): tracheal intubation (606, 25%), mechanical ventilation (547, 22.6%), IV fluid expansion (242, 10.0%), infusion of vasopressors (297, 12.3%), cardiopulmonary resuscitation (299, 12.3%), massive transfusion (16, 0.6%). Median time interval between ED admission and death was 7.5 h (first and third quartiles: 2.1 and 20.2 h, respectively). For 1,174 patients (48.5%), death occurred in the observation unit of the emergency department. The 1,246 remaining patients died in examination rooms of the ED. Palliative care was provided to 1,373 (56.7%) patients and included analgesics for 759 (55.3%), sedation (458, 33.4%), mouth care (488, 35.5%), reposition for comfort as appropriate (460, 33.5%), and emotional support of the patient (334, 24.3%) and his/her relatives (701, 51.1%).

Decisions to withhold or withdraw life-sustaining treatments

A decision to withhold or withdraw life support was taken for 1,907 patients (78.8%), thus 513 patients died without level-of-care limitation. Withholding concerned 791 patients (41.5%), withholding followed by withdrawal of life support therapy concerned 386 patients (20.2%), and withdrawal alone concerned 730 patients (38.3%) (Fig. 1). The decision to limit life-support procedures was recorded in the medical file for 330 patients (17.3%). Median time interval between ED admission and a decision to limit life support was 118 min (first and third quartiles: 48 and 410 min, respectively). There was no statistical difference between times to withhold (145 min, first and third quartiles: 48 and 410 min, respectively) or withdraw life-support therapies (117 min, first and third quartiles: 40 and 317 min, respectively).

Fig. 1
figure 1

Trial profile of 2,512 patients who died in emergency departments

Full size image

The decision was made by a single ED physician in 379 cases (19.9%) and by at least two ED physicians in 1,528 cases (80.1%); nursing staff was involved in 522 cases (27.4%). Medical advice was sought from an indoor hospital specialist but outside ED in 623 cases (32.7%) and from the general practitioner in 91 cases (4.8%). One hundred fifty-seven patients (8.2%) were considered to be able to participate in the decision-making process, but only 50 patients (31.8%) were actually involved in the decision. Among the 1,750 patients who were considered noncompetent to take part in the decision, the following reasons were put forward: acute loss of consciousness (1,337 patients, 76.4%), chronic vegetative state (309 patients, 17.7%), and sedation (104 patients, 5.9%). Families were involved in the decision-making in 1,114 cases (58.4%) or simply informed of the decision in 511 cases (26.8%). Although proposed in the patient’s chart, APACHE II score was never used, probably because it has not been deemed appropriate by emergency physicians.

The reasons for making a decision to withhold or withdraw life support were the principal acute presenting medical disorder, severity of illness, expected irreversibility of the acute disorder, and age (Table 3). Logistic regression for individual factors associated with such a decision were old age, very important functional limitation [odds ratio (OR) = 3.54 and 5,84 for Knaus category C and D, respectively], and certain medical disorders (Table 2). Physicians chose on average 3 ± 3 items (range 1–8) out of the 9 predefined criteria to justify their decisions to withhold or withdraw life-sustaining treatments (Table 3). Neither “expected quality of life unacceptably poor” nor age were used alone. The other criteria used for these patients were principal acute presenting medical disorder in 395 patients (86%), expected irreversibility of acute disorder in the first 24 h in 273 (59.8%), underlying chronic disease in 290 patients (63.6%), and previous functional limitation in 214 patients. This population had major underlying diseases; the presenting medical disorders were mainly neurological (34%), respiratory (32%), cardiac (22%), and septic (10%) Palliative care was provided to 1,148 patients (60.2%).

Table 2 Odds ratio for the composite outcome of withholding and/or withdrawing decisions based on the multivariate logistic regression model performed on 2,420 patients who died in an ED
Full size table
Table 3 Criteria used to justify limiting life-support therapies for 1,907 patients who died in emergency departments
Full size table

From an ethical point of view, the most frequent arguments for limiting life support were futility of care (1,098 patients, 57.6%), age (679 patients, 35.6%), physical pain (361 patients, 18.9%), psychological pain (211 patients, 11.1%), and a do-not-resuscitate order issued by the patient (36 patients, 1.9%) or transmitted by a relative (84 patients, 4.4%).

Discussion

This is the first large multicenter prospective study to describe the characteristics of patients who die in hospital emergency departments and describes the processes leading to a decision to withhold or withdraw life-support therapy. The inclusion criterion was patients who died in the ED and not patients who were dying. Indeed dying is a dynamic process and is not an operational definition. According to Seale [16], a person is considered to be dying when there is recognition that death is possible based on a physiological and a social context, and we are never certain when this phase will actually end with death.

We found that patients who died were elderly and had limited previous functional limitation, chronic underlying diseases, and often severe acute presenting medical disorders. Furthermore we found that almost 80% of these deaths were preceded by a decision to withdraw and or withhold life support. The characteristics of our patients are similar to those of smaller previous series [4, 6, 17, 18], the only significant difference being the higher proportion of traumatic causes of death in North America [5].

Our study has three main limitations. First, university hospitals were overrepresented, thus the case mix of patients might be different from other nonuniversity hospitals. However, we feel that this bias is counterbalanced by the fact that over one-quarter of all emergency departments located in France were involved in this study. Second, we considered only two 2-month periods (November–December and April–May), and therefore it is possible that we did not catch seasonal variations. Third, data on the number of patients admitted to the ED during the study period are incomplete because not all study centers responded to this request.

In France, the percentage of patients dying in hospital rather than at home has risen from 38% in the 1970s to 58% in the 1990s and has since remained constant [19]. This has invariably led to more people dying in ED. This proportion is higher than in other European countries included in the Eureld study (33–50%) [20]. According to Chan [9], end-of-life models have been developed for chronic illnesses and they cannot be transposed to everyday ED practice. Emergency departments are dedicated to making rapid decisions in a high-stress, fast-paced environment and for caring for unexpected illnesses or injuries. However, patients with terminal-stage chronic illnesses often come to the ED for the last hours of their life [18].

Few studies have focused on decisions to limit life-support treatments in the ED [6, 18, 21, 22]. Such decisions have, however, been widely studied in critical care medicine. The Ethicus Study, performed in European intensive care units (ICUs), found that decisions to limit life support were common and were associated with age, principal diagnosis, and length of ICU stay [23]. Furthermore, an International Consensus Conference has been organized on challenges in end-of-life care in intensive care and has proposed numerous actions to improve decision-making processes and care in this situation [24]. In our study we found such decisions to be associated with old age and very poor levels of previous functional limitation. Expected quality of life is a highly subjective notion and should not be used on its own. Finally, decisions to limit life-support treatments were almost always taken on the basis of multiple criteria, limiting the risk of arbitrariness. However, these decisions taken in a short timeframe are potentially dangerous and have to be taken with all precautions.

We found that the decision-making process often did not comply with French legislation, which considers that such a decision has to be taken in a collegial fashion [25]. Indeed, the decision was taken by an individual physician with no prior consultation with medical or nursing staff in 19.6% of cases. Nursing staff participated in the decision-making process in 27% of cases. This contrasts with a previous study in intensive care units [8] which found that nursing staff took part in 54% of decisions, and with another study which found nurse involvement for 78.8% of patients (Ethicus Study).

We found that only one-third of patients deemed capable of participating were actually involved in the decision-making process. Decisions to limit life support were rarely recorded in medical files, and families were not informed of the decision in one-third of cases. This underlines the persistent strong paternalism that exists in French doctor-to-patient relationships. Limiting life support must be a shared decision that involves physicians, nurses caring for the patient, family, and whenever possible the patient himself. This has recently been emphasized by the French Society of Critical Care Medicine [15] and the French Society of Emergency Medicine [12] and is now an integral part of French law [25].

Emergency departments have to improve the quality of care for end-of-life patients. Ethical standards must be applied by emergency medical staff in order to improve the quality of decisions concerning withholding or withdrawing life support. Likewise, informing the patient and his family remains a fundamental cornerstone which is too often overlooked. It should be emphasized that only treatment but not care is withheld or withdrawn, objectives being reoriented to favor the comfort of patient and relatives.

We found that palliative care in emergency departments was insufficient. Indeed, over 35% of patients for whom a decision to limit life-support therapy was taken did not receive palliative care. End-of-life management needs to be improved in emergency departments. Decisions to limit life support need to be collegial and recorded in the patient’s files; analgesia, sedation, and other palliative care measures need to be more widely implemented.

Conclusions

Our study describes the characteristics of patients who die in emergency departments. We have shown that decisions to limit life-support therapies were implemented for 80% of these patients. The recent arrival of end-of-life patients in emergency departments will be a lasting sociological phenomenon. Training of future ED physicians must be aimed at improving the level of care of dying patients, with particular emphasis on palliative care.