Advances in medicine have contributed not only to the ageing of the population but also to the increasing medicalisation of death. For example, in the USA, there is an increasing trend in intensive care unit (ICU) admission and invasive procedures including cardiopulmonary resuscitation (CPR) prior to death,1 and 1 in 5 deaths are preceded by an ICU admission.2 There are two major consequences of such a trend. The first is delivery of interventions that are invasive but without benefit to the patient at the time of death, and the second is the high cost associated with this trend. A survey of medical expenditures among elderly Medicare beneficiaries in the USA revealed a mean annual medical expenditure during the last year of life that is more than five times that for non-terminal years.3

The New Zealand healthcare system differs from that in the USA, in that it provides subsidised hospitalisations and pharmaceuticals. It also has a universal no-fault accident compensation system that contributes to the low rates of litigation and medical indemnity insurance premiums. We theorise that for these reasons, the way we deliver end-of-life care in our hospitals may differ from other countries.

In this study, we obtained data that might help us identify gaps in the way we deliver care at the end of life. We carried out a retrospective study looking at all hospital decedents as regards aspects of their end-of-life care during the study period. The information collected will be used to design performance improvement initiative in this area. On the basis of the dimensions of a good death as identified by Singer and company,4 we identified metrics that were relatively easy to extract from the medical records that we believed represented the following themes: pain and symptom management, avoidance of prolongation of dying and unnecessary interventions, strengthening relationship with family and friends, achieving a sense of control, and relieving burden on others. It was not the aim of the study to perform a comprehensive assessment of the dying experience in Dunedin Hospital. The readers are directed to the Quality of Death and Dying (QODD) questionnaire5 and the Liverpool Care Pathway6 for indepth measurement of processes and outcomes in end-of-life care.

METHODS

Dunedin Hospital is a 350-bed tertiary care teaching hospital located in the South Island of New Zealand. We carried out a retrospective, descriptive study using an audit of medical records. The Lower South Regional Ethics Committee approved the study. The 200 patients selected were the decedents in the hospital before 1 December 2003. Neonatal deaths were excluded. Two cases were excluded because their medical records did not contain sufficient information for analysis.

We identified measures that are reflective of some of the ideals in end-of-life care discussed above. These measures included whether a do not resuscitate (DNR) order was in place, whether the patient underwent surgery or interventions at the time of terminal hospitalisation, whether they were admitted to the ICU, and whether there was documentation in the medical records of communication with the family. A standard form (fig 1) was used to collect the data from a review of the patients’ clinical notes. In addition to these items, demographic data about the patient, and information regarding admission and ICU stay were also collected. The majority of the information came from the progress notes, nursing notes, coding sheets, and death certificate.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 The data acquisition form used in the present study.

RESULTS

Demographics and place of death

The 200 patients ranged in age (at time of death) from 16 years to 103 years (mean age of 76 years) (table 1). They all died between 18 April and 29 November 2003. Of these, 174 (87%) patients were admitted acutely and the remaining 26 (13%) died during elective admissions. The mean duration of hospital stay before death was 7.64 days (range 1–69). In all, 169 (84.5%) patients died in the ward and 38 (19%) died following an ICU admission.

Interventions

Forty (20%) patients underwent surgical procedures during their terminal admission. The most common was repair of a fractured neck of femur. Of the remaining 160, who did not have surgery, 43 (21.5%) underwent some form of invasive procedure. The most common was intubation. Therefore, a total of 83 (41.5%) patients had an invasive procedure during their terminal admission.

Communication and family

One measure of communication tracked was whether a DNR order was made. In Dunedin Hospital, a DNR order requires prior discussion with the patient and/or the family. We found that 148 (74%) patients had a DNR order in place when they died. Documentation of whether the family had been informed of the patient’s terminal status and whether there was discussion with either the patient and/or family about end-of-life care options was sought in the progress notes and nursing notes. These were documented in the medical records of 162 (81%) and 164 (82%) patients, respectively. Finally, we obtained information from the nursing notes whether family was present at the time of death. Families of 121 patients (60.5%) were present at the time of demise.

Pain

The pain status was described in the nursing notes of 140 (70%) patients. Of those in which documentation was made, 134 (96%) were pain free. Other symptoms were not sufficiently well documented in the notes to allow analysis. However, the most common that were mentioned were dry mouth or increased secretions.

Intensive care admission

A total of 38 (19%) patients were admitted to the ICU during their terminal hospitalisation. None of the patients had an ICU admission within a year prior to their terminal hospitalisation, and 31 (15.5%) patients died in the ICU. The mean age of the patients who died in the ICU was 56 years, which was younger than the average age of the decedents. The median and mean length of ICU stay for these patients was 1 day and 2.5 days, respectively. For the 31 patients who died during their ICU admission, 29 (94%) had documentation that a discussion took place between the ICU staff and the families, explaining the clinical condition.

Causes of death in hospital

Table 2 provides a breakdown of the causes of death of the patients.

DISCUSSION

There is a growing recognition that how patients die in the hospital is just as important as the more traditional quality indicators in hospital care delivery (eg, severity-adjusted morbidity and mortality). This has led to the development of tools to assess end-of-life care based on qualitative research looking at focus groups, such as older people, and patients with cancer and AIDS.

The Liverpool Care Pathway is a multiprofessional document providing an evidence-based framework for the delivery of end-of-life care.6 It was developed by the Royal Liverpool University Trust and the Marie Curie Centre, and provides guidance on different aspects of care, not only of the patient but also of the family. Emphasis is given on alleviation of symptoms at the time of death, discontinuation of inappropriate interventions, good communication, and spiritual and psychological support for both the patient and the family, including after death. It is a comprehensive template of care that assists clinicians in delivering consistent and appropriate care of dying patients across different hospital settings.

The establishment of palliative care services in hospitals and implementation of palliative care pathways brought about process changes in the way we care for patients with terminal conditions. However, a considerable fraction of hospital decedents do not have advanced stages of a chronic illness on admission and are not identified as dying during their terminal hospitalisation. This group of patients may not be benefiting from end-of-life pathways and protocols. For this reason, we decided to conduct a hospital-wide survey of outcomes that might reflect the quality of dying among hospital decedents rather than performing a comprehensive qualitative assessment of end-of-life care of a focus patient group.

In the medical literature, there is no previous hospital-wide assessment of the end-of-life care received by all hospital decedents. The tools that have undergone validation in measuring the quality of dying in the hospital require considerable resources to use in a hospital-wide survey for the purpose of identifying problem areas. We identified metrics that were relatively easy to extract from the medical records that we believed represented some of the dimensions of a good death as identified by Singer and colleagues.4 We were not able to come up with metrics that reflected the following dimensions: strengthening relationship with family and friends, achieving a sense of control and relieving burden on others. We focused instead on pain management, communication and avoidance of prolongation of dying and unnecessary interventions.

The documentation of the pain status in 70% of all the hospital decedents, and a pain-free status in 96% of the patients where the pain status is documented, are noteworthy. This may be due to the presence of an acute pain service that is widely consulted in Dunedin Hospital. However, the lack of mention of other symptoms at the time of death is a problem area we identified during this study that needs to be addressed in the design of a performance improvement initiative in improving end-of-life care.

A concerning finding was that 41.5% of patients who died underwent some form of invasive procedure. This must be taken into clinical context. There are several scenarios where interventions considered invasive might be appropriate. The patient may have been well prior to an elective surgery and then developed a complication leading ultimately to his or her demise. The surgery could also be palliative, or warranted to confirm a diagnosis.

It is clear that the dilemma lies because clinicians cannot (1) predict with 100% accuracy whether a patient will survive and (2) ascertain whether the functional status in the end, if the patient lives, is acceptable to the patient. These uncertainties initially drive how far to go in terms of interventions but over time are weighed against the response to therapy and the presence of irreversible comorbidities.

Our relatively low admission rate of decedents to the ICU (19%) is in stark contrast to the American data presented in the SUPPORT and other studies.17 In the review of the medical record of Medicare patients who died in hospital between 1985 and 1999, 38–39.8% of the patients were admitted to the ICU during their terminal hospitalisation. The sickest patients are cared for in the ICU making the ICU the busiest place in the hospital. For this reason, the ICU may not be an optimal environment for patients and families at the time of death. In addition, patients admitted to the ICU are more likely to undergo invasive interventions such as the placement of central venous catheters, mechanical ventilation and renal support therapy. Among our patients who died in the ICU, it was encouraging to find documentation of family discussion regarding the patient’s condition in 92% of cases.

A DNR order was in place for 74% of patients who died. As stated earlier, a DNR order requires prior discussion with the patient and/or the family in our hospital. During the discussion, the clinicians emphasise the severity of the patient’s clinical condition and the likelihood of a poor outcome if the patient were to have a cardiac arrest. Although documentation guarantees that the discussion took place, it is difficult to assess the quality of this discussion based on the progress notes and/or the nursing notes.

The study has several limitations and weaknesses as it is retrospective and observational. Were the data gathered valid indicators of the quality of dying? Is the rate of CPR among hospital decedents an appropriate metric to reflect end-of-life care in the hospital, given that the majority of deaths in the hospital are not due to unforeseen circumstances? If so, what is the optimal rate of CPR among hospital decedents?

As with other retrospective studies looking at processes in healthcare delivery, the study relied primarily on documentation to evaluate whether a component of care, such as communication, has been provided. The quality of that communication, and whether it achieved the outcomes it was intended for, cannot be assessed from the medical records. Were end-of-life discussions initiated as soon as the possibility of death was entertained? This is one area in healthcare where outcome metrics, such as patient and family satisfaction through survey, rather than process metrics, such as quantity and quality of communication, might be more practical to obtain on a regular basis.

It is important to consider the socio-cultural platform of end-of-life care when designing performance improvement initiatives for quality improvement. For example, studies have suggested that the availability of resources, including the number of ICU beds, may affect the care a patient receives at the end of life. A comparison of ICU availability and utilisation between Alberta, Canada, and western Massachusetts in the USA,8 revealed a utilisation that is two to three times higher in terms of ICU days per million population in western Massachusetts. This difference in ICU bed availability (24 adult ICU beds per 100 000 in western Massachusetts vs. 16 in Alberta) probably contributes significantly to the difference in the rate of ICU utilisation.

In New Zealand, during a survey of critical care resources from 2002 to 2003, there were approximately 5.3 available ICU beds per 100 000 people.9 International data indicate that many European countries have about 9 or 10 intensive care beds per 100 000 people, while France, Germany and the USA have over 20 beds per 100 000 people.10 ICU beds account for about 1.7% of hospital beds in New Zealand9; in the USA, ICU accounts for 10% of hospital beds and almost 30% of hospital budgets.11

Other factors might also be contributing to the difference between end-of-life care between New Zealand and the USA. In 1974, New Zealand abandoned a tort-based system for compensating people with personal injuries, and replaced it with a government-funded compensation system operated by the Accident Compensation Corporation (ACC).12 ACC is now the primary method in New Zealand for dealing with personal injury claims, including medical injuries, and has effectively barred all medical malpractice litigation in New Zealand. ACC provides a system of “no-fault” compensation to injured patients without the need for the patient to prove negligence on the part of their healthcare professional. Claims may be accepted on the basis of a “systems error” without attributing medical error to an individual healthcare professional.

In the USA, on the other hand, medical malpractice litigation is rife, with a tort-based medical malpractice system still in place. In such a system, the threat of liability for the healthcare professional is high and can lead to the practice of “defensive medicine”. Most relevant is where care is supplemented with additional services (for instance, testing or treatment) of marginal or no medical value, with the aim of reducing medical malpractice litigation. Indeed, studies have shown that healthcare professionals will often suggest, or comply with patients’ demands for, invasive procedures that, in their professional judgement, are unwarranted, to achieve this aim.13

In New Zealand, the use of advanced technology in medicine has been centrally controlled by the government in consultation with the clinicians, limiting its availability and perhaps also limiting the expectations of the public. New Zealanders tend not to expect access to unlimited healthcare services.14 In the USA, a recent poll by USA Today/Kaiser/ABC surveyed 1201 Americans about what care to give and when to quit at the end of life. Results revealed 40% believe it is better to keep a terminally ill person alive as long as possible regardless of cost.15 In another study, a comparison of ICU care between the USA and Japan revealed that there were considerably fewer patients 85 years and older in the ICU in the Japanese cohort than in the USA cohort (1.2% vs 4.6%, respectively) despite a higher per capita rate of individuals greater than 85 years old in Japan.16 According to the authors, this may reflect differences in cultural norms regarding the limits of care.

Finally, we theorise that distrust of the healthcare system may contribute to a difficulty in switching the goal of care from curative to palliative at the end of life. In New Zealand, doctors and nurses are repeatedly in the top five most trusted professionals.17 Several studies, on the other hand, report that distrust of the healthcare system being relatively high in the general population in the USA.18 Given this distrust of the healthcare system, plus a societal expectation of unlimited access to advances in medical technology, and a healthcare system that values patient autonomy over medical paternalism, it is not surprising that our and other US studies suggest that, compared with New Zealanders, Americans undergo more medical interventions at the end of life. All these socio-cultural factors will need to be addressed in designing performance improvement initiatives to optimise care at the end of life.

CONCLUSION

Our study suggests that majority of patients who died in Dunedin Hospital during the study period were managed adequately as regards certain aspects of their end-of-life care. Although it may not be as ideal as dying at home surrounded by family and friends, death in a hospital can be improved by good communication between the clinicians, the patient and their family, minimisation of invasive procedures unlikely to benefit the patient, and alleviation of pain and other unpleasant symptoms. These metrics should be routinely tracked by hospitals and the data regularly provided to the clinicians in order to maintain provision of ideal care at the end of life.