Article Text

Download PDFPDF

Evaluating end of life practices in ten Brazilian paediatric and adult intensive care units
  1. Jefferson Piva1,2,3,
  2. Patrícia Lago4,
  3. Jairo Othero5,
  4. Pedro Celiny Garcia6,
  5. Renato Fiori7,
  6. Humberto Fiori7,
  7. Luiz Alexandre Borges8,
  8. Fernando S Dias9
  1. 1School of Medicine, Pontifícia Universidade Católica do Rio Grande do Sul (PUCRS), Brazil
  2. 2School of Medicine, Universidade Federal do RS (UFRGS), Brazil
  3. 3Pediatric Intensive Care Unit, Sao Lucas Hospital, PUCRS University, Brazil
  4. 4Pediatric Intensive Care Unit, Clinicas Hospital, UFRGS University, Brazil
  5. 5Intensive Care Unit, Luterano Hospital, ULBRA University, Brazil
  6. 6School of Medicine, PUCRS University, Pediatric Intensive Care Unit, Sao Lucas Hospital, PUCRS University, Brazil
  7. 7School of Medicine, PUCRS University, Neonatal Intensive Care Unit, Sao Lucas Hospital, PUCRS University, Brazil
  8. 8Intensive Care Unit, Conceiçao Hospital, Porto Alegre, Brazil
  9. 9School of Medicine, PUCRS University, Director of the Intensive Care Unit at Sao Lucas Hospital, PUCRS University, Brazil
  1. Correspondence to Dr Jefferson P Piva, Pediatric Intensive Care, Unit H, São Lucas-PUCRS, Av. Ipiranga 6690 – 5 andar, CEP – 90.610-000, Porto Alegre (RS), Brazil; jpiva{at}


Objective To evaluate the modes of death and treatment offered in the last 24 h of life to patients dying in 10 Brazilian intensive care units (ICUs) over a period of 2 years.

Design and setting Cross-sectional, multicentre, retrospective study based on medical chart review. The medical records of all patients that died in seven paediatric and three adult ICUs belonging to university and tertiary hospitals over a period of 2 years were included. Deaths in the first 24 h of admission to the ICU and brain death were excluded.

Intervention Two intensive care fellows of each ICU were trained in fulfilling a standard protocol (κ=0.9) to record demographic data and all medical management provided in the last 48 h of life. The Student t test, Mann–Whitney U test, χ2 test and RR were used for data comparison.

Measurements and main results 1053 medical charts were included (59.4% adult patients). Life support limitation was more frequent in the adult group (86% vs 43.5%; p<0.001). A ‘do not resuscitate’ order was the most common life support limitation in both groups (75% and 66%), whereas withholding/withdrawing were more frequent in the paediatric group (33.9% vs 24.9%; p=0.02). The life support limitation was rarely reported in the medical chart in both groups (52.6% and 33.7%) with scarce family involvement in the decision making process (23.0% vs 8.7%; p<0.001).

Conclusion Life support limitation decision making in Brazilian ICUs is predominantly centred on the medical perspective with scarce participation of the family, and consequently several non-coherent medical interventions are observed in patients with life support limitation.

  • End of life
  • life support limitation
  • death
  • palliative care
  • cardiopulmonary resuscitation
  • attitudes towards death
  • care of the dying patient

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


From a medical perspective, there is no reason for offering curative treatment to patients with irreversible and terminal disease. Perhaps one of the biggest challenges in this regard is the correct medical definition of ‘irreversible and terminal illness’. It has been recommended that after adequately defining the terminal and irreversible status, the intensive care staff must be encouraged to discuss any medical interventions with the patient and his/her family in order to reach a consensus about life support limitation.1–5

Several non-medical aspects may be involved in this consensus and modify the kind of treatment offered to terminal patients, such as, religious beliefs, cultural aspects, patient autonomy and legal aspects.6–12 It is well known that these factors have different magnitude in different locations, and it is not surprising that the treatment given to terminal patients varies among countries and even sometimes in the same region.6–17

The decision making process is based on a shared resolution in considering medical aspects (principles of beneficence and non maleficence), the patient's wishes and cultural values (principle of autonomy).1 2 4 In North American countries and some European countries the principle of autonomy has a strong influence in the final decision, whereas in most of Latin American countries the medical perspectives are the predominant aspects in this decision.6 15 16

Some authors consider that a decision centred on the medical aspects is synonymous with paternalism.6 10 14 We are not convinced that this interchangeable meaning can be applied in Latin America. In the last few years, especially in Brazil, the media and society have been involved in a discussion of the legal aspects and moral basis for accepting life support limitation in a terminal ill patient.17 Based on the Brazilian Penal Code, some judges declared that life support limitation has no legal support. Obviously, this ‘legalistic’ point of view does not help nor facilitates the medical care provided to terminal patients in this country. Probably, these legal concerns could explain some medical attitudes and decisions adopted in the last moments of life of terminal ill patients as well as impair the attainment of a consensual decision with the family and relatives.17–22

In 2009, the Brazilian Council of Physicians approved a new code of medical ethics that will regulate medical activities in Brazil from April 2010. This new code of medical ethics includes some specific articles related to end of life (EOL) decisions and emphasises relevant aspects such as: (a) avoiding futile interventions and stimulating provision of palliative care to patients with terminal and irreversible diseases (Chapter 1, main principles—XXII); (b) the obligatory participation of the patient (or family member) in the decision making process for EOL (Articles 22, 36 and 41); and (c) defining clearly the difference between euthanasia and palliative measures (Article 41).23 However, prior to this new code being in practice, Brazilian physicians are caring for terminal ill patients in a very confused environment and some of their decisions might not be considering the best interests of their patients.

The largest and most representative studies on EOL decisions carried out in Latin America evaluated exclusively terminal ill children.13 18 19 To our knowledge there are no large and collaborative studies performed in this area involving and comparing EOL practices in adult and paediatric intensive care units (ICUs). The aim of this study is to evaluate the modes of death and the treatment offered in the last 24 h of life to patients dying in 10 Brazilian referral ICUs (seven paediatric and three adult) over a period of 2 years.

Patients and methods

We conducted an observational, retrospective, multicentred study involving children and adults that died in 10 selected Brazilian ICUs, during a period of 24 months (January 2004 to December 2005). Seven paediatric intensive care units (PICUs) and three adult ICUs belonging to teaching hospitals or tertiary medical care centres in Brazil were selected to participate in the study. The protocol was approved by the ethics and research committees of each hospital involved in this study. Some results of the paediatric study have been published elsewhere before.18

The number of beds of the selected PICUs in the paediatric arm of study oscillates between 8 and 15, admitting an average of 350 to 650 children per year per unit with a mortality rate ranging from 5% to 12%. The selected PICUs had the following regional distribution: (a) two PICUs in the southern region: Hospital São Lucas (PUCRS University) and Hospital de Clinicas de Porto Alegre (UFRGS University); (b) two PICUs in the southeastern region: Hospital das Clinicas de São Paulo and University Hospital, both associated with USP University; and (c) three in the northeastern region: Hospital São Rafael, Hospital da Criança Santo Antônio, and Hospital Ernesto Simões Filho.

In the adult arm of the study three southern Brazilian adult ICUs were included as follows: (a) Hospital São Lucas (PUCRS University) with 35 ICU beds (850 patients admitted per year with a mortality rate of 16%); (b) Hospital Luterano (ULBRA University) with eight ICU beds (500 patients admitted per year with a mortality rate of 17%; (c) Hospital Conceição with 28 ICU beds (up to 1100 patients admitted per year with a mortality rate of 33%). Because the highest admission and mortality rates were observed in this ICU, we randomly selected 50% of the deaths that occurred in this period. In this regard, the medical charts were ordered according to the date of death and the odd numbers were selected for this study.

Patients with brain death and those dying in the first 24 h of ICU admission were excluded in both arms of the study.

All the PICUs and ICUs are referral centres for treating complex conditions (trauma, burns, oncology and cardiac surgery), and some of them have transplant programmes (liver, renal and bone marrow transplant). The PICUs and adult ICU have similar medical assistance, based on a medical coordinator, one or two (paediatric) intensivist assistants in charge, fellows in (paediatric) intensive care, and general paediatric or internal medicine residents. Medical decisions are defined during two daily rounds (early in the morning and late afternoon).

Two intensive care fellows at each PICU and ICU were trained to fill out the standard protocol with data obtained exclusively from medical charts. These fellows were considered adequately trained when achieving a concordance of 90% ((κ)=0.9) with the main researcher. In case of any doubt, they were instructed to contact the main authors (JP, PML and/or JO) by phone or internet to reach a consensus.

The following data were extracted from the selected medical charts: (a) demographic and general data (age, gender, main diagnosis, cause of ICU admission, and length of hospital and PICU stay); (b) data related to death (cause of death, number of organ failures in the last 48 h of life, complete cardiopulmonary resuscitation manoeuvres (defined as chest compressions, bag ventilation or mechanical ventilation through tracheal tube plus intravenous epinephrine administration) and life support limitation decisions reported in the medical chart. We systematically searched for any report in the medical chart regarding the participation of the family in the life support limitation decision making process; (c) data about medical management in the last 48 h of life in the PICU (vasopressor/inotropes drugs infusion, parameters of mechanical ventilation, sedative and analgesic drugs infused in the 48 h, 24 h and immediately before death). Imaging diagnostic and blood biochemistry analysis were selectively researched. Multiple organ dysfunctions were classified according to definition reported by Wilkinson et al.24

Continuous variables were expressed as means and SD, and variables with asymmetrical distribution were expressed as median and IQR (IQR 25–75%). The unpaired t test was used to compare continuous variables, whereas the Mann–Whitney U test was used to compare variables with asymmetrical distribution. Categorical variables were expressed as percentages and compared using the χ2 and the Fisher exact tests. The RR was used for measuring associated factors. An Excel (Microsoft) spreadsheet was used to collect data, which was analysed with SPSS V.11.0 software.


In this study, 1053 medical charts of adults (59.4%) and children (40.6%) that died in 10 Brazilian ICU in a period of 24 months were analysed (table 1). In the paediatric arm of the study, we initially identified 561 deaths. We excluded 61 children with diagnosis of brain death and another 36 children that died in the first 24 h of PICU. There were 36 missed medical charts (6.9%) equally distributed among the seven PICU, resulting in a final sample of 428 deceased children. In the adult arm of the study, we initially identified 710 deaths. We excluded 11 with diagnosis of brain death and an additional 74 medical charts (10.5%) were missed (equally distributed among the three adult ICUs), resulting in a final sample of 625 adult deaths.

Table 1

Characteristics of the patients that died in three Brazilian adult intensive care units (ICUs) and seven paediatric ICUs over a period of 24 months

Compared to children, the adult group had more severe diseases (≥3 organ dysfunction), higher hospital length of stay but with similar length of ICU stay (table 1). The main cause of death in both groups was shock, despite adults and children differing regarding the main cause of ICU admission (p<0.001).

Cardiopulmonary resuscitation (CPR) was not offered (‘no CPR’) to 724 patients in total (68.8%). No CPR was much more frequent in the adult ICU than in the PICU (86.1% vs 43.5%, respectively; RR 2.0 CI 1.8 to 2.2, p<0.001). In comparing adults and children not resuscitated it was observed that the length of ICU stay was similar in both groups, although the length of hospital stay and the prevalence of three or more multiple organ dysfunction system was higher in the adult group (table 2). The reasons for ICU admission among the patients not resuscitated maintained the same distribution that was observed in the whole group (table 1; cardiovascular and respiratory diseases were the main causes of ICU admission in the adult arm study while among the children not resuscitated the main cause for PICU admission was sepsis).

Table 2

Comparison between adult and paediatric patients that received no cardiopulmonary resuscitation (CPR) before death

It is interesting to note that in both groups the preference for withholding and withdrawing treatment was lower than the ‘do not resuscitate’ (DNR) orders (table 3). Another intriguing fact observed in both groups was the low rate of reported decision in the medical chart (52.6% and 33.7%) as well as the lower rate of family involvement in the decision making process. However, the option for withholding/withdrawing treatment had a slight higher frequency in the paediatric group (33.9% vs 24.9%; p=0.02) as well as the family involvement in the decision making process (23.0% vs 8.7%; p<0.001).

Table 3

Characteristics of the adults and paediatric patients not resuscitated in the 24 h prior death in 10 Brazilian ICU in a period of 24 months

Not performing CPR was much more frequent in the adult ICU than in the PICU (RR 2.0 CI 1.8 to 2.2). Analysing children and adults in whom CPR was not attempted before dying (table 4) it was observed that they had similar treatment in the 24 h before death: high rate of mechanical ventilation (97% and 92.4%), as also, approximately two thirds of them received vasoactive drugs infusion (72.7% and 66.1%) and constant analgesia/sedatives infusion (65% and 71%).

Table 4

Comparison of the medical interventions delivered in the last 24 h of life to patients that received cardiopulmonary resuscitation (CPR) and those that were not reanimated (no CPR) in three Brazilian adult and seven pediatric intensive care units

Comparing children that were and were not reanimated, we did not observe any difference in the average amount of inotropes/vasopressors (epinephrine, dopamine and dobutamine) infused in the last 24 h of life.

Analysing adults that were and were not reanimated before death, we did not observe any difference regarding the number of x-rays and laboratory exams performed in the last 24 h of life.


In this study we evaluated more than 1000 patients that died in 10 Brazilian referral ICUs and it was observed that life support limitation is a common practice in the adult intensive care units (close to 83%) as well as in the paediatric intensive care units (close to 44%), with DNR orders the preferred practice for life support limitation. As observed in previous studies performed in this area, the life support limitation decision is predominantly centred on the medical perspective with scarce involvement of the family (or the patient) in the decision making process and it is rarely reported in the medical chart.13 18 19 The omission of reporting life support limitation in the medical chart was much more frequently observed in the adult ICU than in the paediatric ICU.

In most European and North American countries, the decision making process for delivering life support limitation to patients with terminal illness is primarily based on the patient's (family) wishes and on their cultural values (autonomy principle).1 16 25 In other parts of the world, this decision is based essentially on the medical perspective.6 7 10 The decision centred on the medical perspectives has been considered as a synonym of paternalism, based on at least two (hypothetical) assumptions: (a) the physicians as natural healthcare leaders and having the specific knowledge domain could advise and appoint to the family (patients) the best moment and option for life support limitation; (b) the life support limitation adopted under the medical guidance would avoid and ‘protect’ the family against future guilty feelings related to this decision.6 7 15 26

Under the paternalistic perspective, it is implicit that physicians must define life support limitation based on medical perspectives taking into account the best interest of their patient. It is not illogical to assume that these idealistic postulations would not be reached in a great number of dying patients in different ICUs and under diverse environments.15 17 18 20 22 26 In this regard, it should be considered that external factors could compromise this decision and the best interest of the patient could be missed.

In this study, some results can't be explained (or accepted) under the paternalistic perspective because they obviously couldn't be associated with the best interest of the patients, such as maintaining the full support treatment for individuals that would not receive CPR as well as performing plain chest x-rays and laboratory exams in the last 24 h of life, that was observed in equal amounts in the reanimated group of patients. Similar findings were reported in an Australian study in which 46% of dying patients with DNR orders had diagnostic tests performed in the last 48 h prior to death. In the same study, only 27% of the palliative care performed had been reported in the medical chart.27 This kind of approach would be interpreted as a medical difficulty in accepting death as a natural evolution in a terminal ill patient.1–5

EOL care is a common problem for most medical specialties, and specifically in the intensive care field. Even now, palliative care is not taught to medical students in most Brazilian medical schools, and few Brazilian intensive care residency and fellowship programmes have a formal educational training in this area.4 13 15 18 20–22 27–29 The same difficulty has been described in other countries.30–32 In a study conducted in India, the authors emphasised the difficulty for physicians providing life support limitation to terminal ill patients preferring to ‘fight till the end’ and opting for DNR orders.32 Considering these aspects, it could be speculated that some physicians still consider death in the ICU as a medical failure and don't want to share their medical decisions with patients (or their families) and neither report them in the medical chart, as observed in our study.

When this study was carried out an intense discussion regarding legal aspects of life support limitation was growing in Brazil with some pending concerns related to the legal aspects of life support limitation.17 It is likely that these legal conflicts and fears could have contributed for some of the conflicting results above mentioned.17 27–30 33 34

It is quite clear that, independently of the country, EOL decisions should have legal support (judicial protection). However, the legal spectrum needs to reflect accurately what society and medical science consider fair and morally acceptable in that place at that time.30 31 33 34 Upon the above results, the Brazilian Critical Care Association (Associação de Medicina Intensiva Brasileira) actively introduced this discussion into civilian society and defined a former group to discuss these ethical aspects with lawyers and judges' representations, as well as encouraged medical schools to introduce the topic into the regular medical curricula.

Considering the methodology of this study (a retrospective analysis based on a chart review), it is quite probable that our results could have some bias. However, we would remark that these results were very similar to other studies conducted in the same region as well as in other countries.13 15 18 20–22 30–32

EOL decisions are one of the most difficult and stressful challenges that physicians, patients and their families have to make. It has been demonstrated that EOL practices differ around the world depending on cultural, moral, religious, legal and medical aspects.5–16 28 29 The final decision takes time and should be obtained in a consensual manner, under an atmosphere of trust, confidence and harmony. The physician should be the leader of this process guiding the patient and the family to reach a final consensus. However, when that process is influenced by external factors, disrupting the atmosphere of mutual trust and confidence, the decision will be taken under an atmosphere of conflict.1–3 28–31



  • Funding This study had financial support from the Brazilian National Research Council (CNPq # 054/2005).

  • Competing interests None.

  • Patient consent Obtained.

  • Ethics approval This study was conducted with the approval of the all the institutions involved.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Other content recommended for you