Abstract
Objective
To examine patterns of emergency department (ED) presentation and emergency hospital admission in the last 90 days of life for residents of New South Wales, Australia.
Methods
A retrospective audit of electronic clinical records. Descriptive statistics report patterns of ED presentation and emergency hospital admission in the last 90 days of life, and symptom drivers of ED presentation. Logistic regression identifies factors associated with low versus high rates of ED presentation and emergency admission.
Results
2869 ED presentations are included across 1730 decedents. 80% of ED visits led to admission. 92% of people had at least 1 ED presentation in the final 90 days of life, with 18% having 3 or more. 86% of people had at least 1 emergency admission, with 9.5% having 3 or more. Odds of high ED use and high admissions were increased for people with cancer and those under 70 years. ED visits were long and often involved multiple investigations, but a small number of patients required no investigations. Common symptom drivers of ED attendance were pain, breathlessness, and confusion/delirium.
Conclusions
ED presentations in the final months of life are common, and investigations are often required to assess for potentially reversible problems. Some people approaching end of life require admission to hospital via ED without the need for investigations, so may be seeking help for escalating nursing needs. Alternative models of care are needed to support escalating nursing needs at home, and funding for palliative services must keep pace with the rising demand.
Introduction
The role of an emergency department (ED) is to diagnose and treat acute and urgent illnesses and injuries to prevent death or disability. 1 The rising use of ED by people at the end of life, although this is not their primary role, has drawn growing interest from clinicians, researchers, and policymakers around the world. 2 The notion of ‘avoidable’ or ‘preventable’ was first coined by Rustein et al. 3 in 1976 to distinguish ‘amenable deaths’ and ‘preventable deaths’. However, the term potentially preventable hospitalisation (PPH) emerged in Australia and are defined under the National Healthcare Agreement 2024, covering 22 conditions across 3 categories; acute, vaccine-preventable and chronic. It is clearly stated that PPH does not mean that a patient admitted for that condition did not need to be hospitalised at the time of admission. But rather the PPH is a proxy to measure the effectiveness of timely, accessible and adequate primary care. 4
In a study conducted in Western Australia (WA), 2011, 80% of 1071 people had at least 1 ED presentation in the last year of life, and 4% visited ED on the day they died. 5 A larger study in 2015 including 45 749 deaths in New South Wales (NSW) reports that both ED use and hospital admissions rise sharply in the last 3 months of life. 6 Both these reports linked routinely collected administrative data to death registrations and were unable to interrogate clinical drivers of ED presentations at the patient-level. A more recent retrospective study of 295 patients in Queensland, Australia describes the characteristics of and care provided to older people who died within 48 hours of presentation to the ED, 7 and a small study of 35 people in regional Victoria reported that their most common presenting complaints to ED were pain, breathlessness, and fever, and that the majority of presentations were appropriate and unavoidable. 8 In the United Kingdom (UK), a large population-based study of 124,020 cancer deaths showed that repeated ED presentations are common in the last year of life, with many people attending multiple times. 9 The UK National Health Service (NHS) monitors and publishes the percentage of deaths with 3 or more emergency admissions (via ED) in the last 3 months of life. 10 This UK data allows clinicians and policy makers to identify factors associated with high ED use at the very end of life, and so encourages improvements in end-of-life care through benchmarking of services across different regions. National statistics of this kind are not routinely published in Australia.
There remains limited published data describing patterns of and reasons for ED use in the last months of life in Australia. This information gap hinders clinicians and policy makers to plan and provide health services to patients and their carers.
To address this gap, we carried out the current study with the following aims:
Describe patterns of ED presentation and emergency hospital admission in the last 90 days of life for residents of a regional area of greater Sydney, New South Wales, Australia. Identify clinical and demographic characteristics associated with higher risk of ED use and emergency admission in the last 90 days of life. Describe the incidence of different clinical problems and symptoms in people attending ED in the last 90 days of life.
Methods
Data were collected using a retrospective audit of local electronic clinical records of people who died between 1/8/2020 and 31/7/21. The start date 1/8/2020 was chosen because local ED presentation rates had returned to normal after an initial drop at the start of the COVID-19 pandemic. A 12-month period was used to capture seasonal variations in ED use.
Setting
The study took place in the Central Coast Local Heath District (CCLHD), a regional area of greater Sydney, New South Wales, Australia. The CCLHD has a population of over 346,000, and contains 2 acute public hospitals each with an ED, and 2 subacute public hospitals without an ED.
Inclusion/Exclusion Criteria
A complete list of decedents was identified through database searches at local public hospitals, community services and the local specialist palliative care service. Decedents were included if resident at a CCLHD address on the date of death. They were excluded if under 18 years old on the date of death, died by suicide, or died suddenly from trauma or with no underlying health condition. The excluded groups were felt to involve inevitable use of ED or fall outside the purview of adult palliative care services. Clinical records of all decedents were manually reviewed by a member of the audit team to screen against the inclusion/exclusion criteria above.
Data Extraction
Data were extracted from electronic patient records using a combination of automated reporting and manual review of records by the audit team. Automated reporting was used to extract; (1) key clinical and demographic variables, (2) number of ED visits in the last 90 days, (3) number of emergency hospital admissions in the last 90 days, (4) time spent in ED, (5) outcome of each ED visit, and (6) length of stay for those admitted to hospital. An emergency hospital admission was defined as an admission that resulted from an ED presentation.
Hospital systems could not easily report the primary diagnosis for each decedent, which often differed from the diagnosis in ED or the cause of death. The primary diagnoses were therefore extracted by a member of the audit team after manually reviewing the clinical records.
A subsample of clinical records was manually audited in greater depth by reviewing the clinical records of each deceased person to extract the symptoms felt to be contributing to ED presentation and investigations carried out in ED. A sample size calculation was performed to determine the number of records to manually audit. It was determined that n = 381 records would be required to estimate the proportion of patients who had higher ED use with a 5% margin of error. This same sample size (n = 381) would provide 80% power to detect a difference in proportions of 14% between 2 factors of interest, with a 5% type 1 error rate. To account for seasonal variations in patterns of ED use, sampling for the manual audit was stratified to selected in equal numbers across the 4 seasons, selected using random number generation. The audit tool for the manual audit was created in consultation with the multidisciplinary clinical research team and refined following a series of 3 pilots. In conducting the audit, researchers reviewed the patient's medical records, reading their listed patient contacts, ED assessment notes, discharge summaries, and death certificates (see Supplementary file 1 for the audit tool).
Data Analysis
Descriptive statistics are used to present clinical and demographic variables alongside details of ED presentations and emergency hospital admissions in the last 90 days of life. We report descriptive statistics of the sample using ‘n’ and ‘percentage’ for all categorical variables, and means and standard deviations for continuous variables. We did check for appropriateness and found that mean (SD) was sufficient in each case. Logistic regression models were utilised to identify clinical and demographic factors associated with low versus high rates of ED presentation, and low versus high rates of emergency hospital admissions in the last 90 days. ‘Low ED use’ was defined as 0 to 2 presentations in the last 90 days; ‘High ED use’ was as 3 + presentations in the last 90 days; ‘Low admissions’ was 0 to 2 emergency hospital admissions in the last 90 days; ‘High admissions’ was 3+ in the last 90 days. The proportion of people who had 3 or more emergency hospital admissions in their final 90 days has been used as a national proxy measure of quality of care in the UK.10,11 Effect sizes are reported as odds ratios (OR) with 95% Confidence Intervals and p-values. Univariate associations were explored for diagnosis category (cancer, cardiovascular, dementia, respiratory, other non-cancer), Cancer diagnosis (yes, no), age (18-69, 70+), gender (male, female), marital status (married, divorced/widowed/separated), country of birth (Australia, other), and preferred language (English, other). Multivariable models were then run including all factors (except diagnosis category due to collinearity with the Cancer Diagnosis). Logistic regression model fit was examined using the Hosmer-Lemeshow GOF test and Stukel's test. Statistical significance was set a priori at P < .05. All statistical analyses were programmed using SAS v9.4 (SAS Institute Inc. Cary, NC, USA).
Approvals and Ethical Issues
The study was reviewed by the Northern Sydney Local Health District (NSLHD) Human Research Ethics Committee (HREC) and considered to be a quality assurance/quality improvement activity not requiring HREC review. The study was therefore authorised by the CCLHD Research Office (Reference number: 0421-035c).
The audit included some decedents who identified as Aboriginal or Torres Strait Islander. These were included for internal quality improvement purposes but have been removed from the dataset for external publication and analyses re-run. This was done to ensure cultural safety on the advice of Aboriginal stakeholders, including a representative from the Aboriginal Health and Medical Research Council.
Results
A total of 1828 adult deaths were identified from the database searches. After applying the inclusion/exclusion criteria a total of 1785 deaths were analysed for internal quality improvement purposes. For external publication, a further 55 people were excluded who identified as Aboriginal or Torres Strait Islander, leaving n = 1730 decedents included for reporting in this article. A subgroup of 391/1785 records were manually audited in greater detail, of which 377/1730 are reported here after excluding people who identified as Aboriginal or Torres Strait Islander. The overall findings and conclusions of the study were not altered by removal of Aboriginal decedents from the dataset. Figure 1 summarises the sampling process.
Sample screening and selection flowchart.
Table 1 shows demographic and clinical characteristics of decedents, grouped by high versus low ED use and high versus low hospital admissions. Rates of ED presentation and emergency admission were high in the final 90 days of life (Figure 2). 92% of people had at least 1 ED presentation in the final 90 days, with 18% having 3 or more visits. Similarly high rates of hospital admission were found in the final 90 days, with 86% of people having at least 1 urgent admission and 9.5% having 3 or more.
(a) Frequency of ED presentation in the last 90 days of life; (b) frequency emergency hospital admissions in the last 90 days of life.
Sample Demographics by Low Versus High ED Use and Low Versus High Hospital Admissions in the Final 90 Days of Life.
Logistic regression analysis examined the factors associated with low ED use (0-2 ED visits in the final 90 days) versus high ED use (3 or more). The multivariate adjusted odds of high ED use were reduced in people with a non-cancer diagnosis (OR = 0.67, 95% CI 0.52-0.87, P = .003), and for people aged over 70 years (OR = 0.58, 95% CI 0.43-0.78, P < .001). Odds were not significantly different across gender, marital status, country of birth, or preferred language (Table 2).
Multivariate Logistic Regression, Modelling Dichotomised ED Presentations (0-2 vs 3+) and Emergency Hospital Admissions (0-2 vs 3+) in the Final 90 Days of Life. *Odds Ratio (OR) Represents the Odds of Being a High ED User (3+) or Having High Admissions (3+).
Logistic regression analysis also examined the factors associated with low admissions (0-2 emergency hospital admissions in the final 90 days) versus high admissions (3 or more). A similar pattern emerged, with the multivariate adjusted odds of having high admissions also being reduced in non-cancer (OR = 0.67, 95% CI 0.48-0.94, P = .021), and if aged over 70 (OR = 0.51, 95% CI 0.35-0.74, P < .001). Odds were not significantly different across gender, marital status, country of birth, or preferred language (Table 2).
There was a total of 2869 ED presentations in the final 90 days across the whole sample (n = 1,730, Table 3). Mean ED visit time was 8.5 hours across all presentations (8.9 hours if admitted, 6.5 hours if not admitted). ED presentations were spread evenly through days of the week and months of the year. 2309/2869 (80%) of ED visits led to an emergency hospital admission. The mean length of subsequent admission was 9.5 days.
Characteristics of ED Presentations in the Final 90 Days of Life.
The in-depth subgroup audit (n = 377) included a total of 601 ED presentations. The most prevalent problems contributing to an ED presentation in this group were pain (42%), breathlessness (29%) and confusion/delirium (27%). Investigations were commonly required, with 538/601 (90%) of ED attendances requiring blood tests, 373/601 (62%) having x-rays and 251/601 (42%) CT scans. However, 36/601 (6.0%) of ED presentations involved no investigations (Table 3).
Discussion
This study used a retrospective case notes audit and regression analysis to describe frequency, reasons and risk factors for ED use in the last months of life. It is the largest Australian study to report clinical and symptom drivers of ED use at the patient level towards the end of life. The study confirms earlier reports that ED use and subsequent hospital admission at end of life are common.9,13 In a recent study, 11 64% of people had a least 1 emergency admission, and 7% of people in their last 90 days of life had 3 or more emergency admissions. It was slightly higher in this study at 92% and 18% respectively. The characteristics associated with emergency presentations in earlier studies are not conclusive. In 1 study, 12 seriously ill older adults aged 74.3 (±6.5) years and with cancer was the most common diagnosis of ED presentation while older women aged 85 years and older were 40% less likely to make an ED visit than younger women aged less than 65 years old. 14 This study found that younger people aged under 70 years and people with cancer were most at risk of having high ED use and high rates of emergency hospital admission. Regardless of the differences in the characteristics associated with ED presentations at end of life, the findings in this study and other studies highlight the importance of integrating palliative care into ED.
A recent study 15 examined whether a multicomponent intervention to initiate palliative care in the ED influenced hospital admission rates. The intervention consisted of (1) evidence-based, multidisciplinary training, (2) simulation workshops focused on serious-illness communication, (3) clinical decision-support, and (4) audit and feedback for ED staff. The researchers found that the multicomponent intervention did not change hospital admission rates, subsequent health-care utilisation, or short-term mortality among older adults with severe, life-limiting conditions. The findings from this study provide some insights what integrated palliative care into ED should focus on given the most common reasons contributing to an ED presentation were poorly controlled symptoms such as pain, breathlessness and confusion/delirium. However, more research is warranted to determine what effective palliative-care integration in the ED should look like and how much impact it can realistically have. More importantly, further work is warranted to evaluate the existing models of care for people at the end of life, and develop new models of care to better support people in community as their care needs increase toward the end of life.
Pain, breathlessness, and confusion/delirium were the most common symptom-drivers of ED presentation in this study. ED visits were long and often involved multiple investigations, but some involved no investigations despite leading to hospital admission. These findings add the growing body of evidence that poorly controlled symptoms are common and an important driver of ED use in people toward the end of life. In the United States (US), it was reported that older adults attending ED often have high levels of palliative care need, with 94% of people having unmet physical and symptom needs. 12 An Irish study found that most ED presentations by palliative patients were often triggered by breathlessness, nausea, vomiting and uncontrolled pain. 16 These symptoms overlap with the present findings that pain, breathlessness and confusion/delirium were the 3 most common drivers of ED presentation. A small Australian study (n = 35) 16 also found significant overlap, with the most common presenting complaints to ED being pain, breathlessness and fever, but on close interrogation the authors concluded that most presentations were appropriate and unavoidable.
Even in the context of palliative care, it may be appropriate to explore reversible causes for acute or unmanageable symptoms, which may explain the high rates of investigations reported in the present study. It is likely the ED will remain the best place of safety for patients seeking investigation for potentially reversible problems, even in the final months of life. It is worth noting that 10% of all ED presentations in the present study required no bloods tests and 6% had no investigations at all. It is possible that many of these people were presenting to ED not for medical treatment, but due to increased physical care needs toward the end of life and/or lack of ability to manage confusion/delirium. Given that dementia, including Alzheimer's disease is now Australia's leading cause of death, 17 further work to understand the specific needs of this group may guide how to develop new community services to reduce ED use in the final months of life. Recent reductions in promised funding for palliative care services in New South Wales may have knock-on effect on the wider health system and contribute to continued high ED use and high admission rates at the end of life. 18
The present study identified a large number of people accessing ED multiple times in their final 3 months. 18% of decedents attended ED 3 or more times and 9.5% had 3 or more emergency admissions to hospital. Rates of emergency admission in other advanced economies are comparable, with a published rate of 7.5% for the UK as a whole. 10 More work is needed to understand if repeat attendances and admissions are avoidable or unavoidable. Further investigation can reveal if there are missed opportunities for advance care planning, and if improved community palliative care services may prevent this. A systematic review incorporating older adults mainly from the US showed that receiving specialist palliative care was associated with lower ED presentations in the last year of life. 10 A similar finding was reported in a Canadian study, which showed that specialist palliative care may reduce ED presentations in the last 90 and 30 days of life. 19
People with cancer and younger people were more likely to have high ED use and high admission rates. The reasons for this are unclear, but it is possible that younger people have a greater wish for acute medical treatments toward the end of life. Those patients who continue receiving systemic anti-cancer treatments in their final months of life are likely to continue to require ED assessment of treatment complications. Further work is needed to develop and test inventions to meet the needs of these groups out of hospital where possible.
Limitations
The present study was conducted in a regional area of New South Wales, and findings might not be generalisable to rural and metropolitan areas. The Central Coast has a lower proportion of people born overseas (16.1%), compared with Greater Sydney (38.6%), 20 which may impact the transferability of results. A retrospective audit of electronic clinical records does not capture the voice or experiences of the people who died, or the specific needs of families and carers. This study identified characteristics associated with higher risk of ED use and emergency admission in the last 90 days of life but did not assess whether each ED attendance or emergency admission was preventable or non-preventable. This would be a worthwhile area for future research.
Conclusion
ED presentations can be time-consuming and burdensome for patients so they should be avoided where possible. ED, by definition and its role, remains an essential place of safety for people who require urgent investigation and treatment of potentially reversible problems in the final months of life. The study investigated clinical and symptom-related drivers of ED use at the patient level in the last 90 days of life. Inconsistent with previous studies, it found that ED presentations and subsequent hospital admissions are common toward the end of life, particularly among people under 70 years of age and those with cancer. The variability in the characteristics associated with ED presentations at end of life, together with evidence that poorly controlled symptoms are a major contributor, highlights the need to evaluate current models of care and to develop new approached that better support people in the community as their needs increase toward the end of life. Australia, and likely other countries, may benefit from the routine collection and reporting of ED attendance rates and presenting symptoms in the final months of life, enabling benchmarking services across regions and driving improvements in end-of-life care. To ensure that people can receive care outside hospital when that is their preference, funding for palliative services, both hospital-based and community-based, must keep pace with growing demand.
Supplemental Material
sj-docx-1-pal-10.1177_08258597251413029 - Supplemental material for Emergency Department Use in the Last 90 Days of Life: A Retrospective Audit Regression Analysis
Supplemental material, sj-docx-1-pal-10.1177_08258597251413029 for Emergency Department Use in the Last 90 Days of Life: A Retrospective Audit Regression Analysis by Thomas Richard Osborne and Zoi Triandafilidis, Sarah Yeun-Sim Jeong, Stuart Szwec, Lucy Leigh, Nicholas Goodwin in Journal of Palliative Care
Footnotes
Acknowledgements
We would like to acknowledge Aboriginal Stakeholders at Nunyara Aboriginal Health Unit and the Central Coast Research Institute for their advice, advocacy and sharing of stories. We acknowledge the many colleagues who contributed to the design and conduct of this work, including Dr Cassie Curryer, Dr Vicki Tai, Dr Rosemary Stratford, Dr Jessica Macdonald, Tia Covi, Jane Mackintosh and Alison De Jong.
Ethical Considerations
The study was reviewed by the Northern Sydney Local Health District (NSLHD) Human Research Ethics Committee (HREC) and considered to be a quality assurance/quality improvement activity not requiring HREC review. The study was therefore authorised by the CCLHD Research Office (Reference number: 0421-035c)
Consent to Participate
Requirement for consent was waived by the Central Coast Local Health District Research Office due to the collection of anonymised data for quality improvement purposes (Reference number: 0421-035c)
Author Contributions Statement
Conceptualisation – TO, SJ, and NG; methodology – TO, ZT, SJ, and NG; formal analysis – TO, SJ, ZT, SS and LL; writing – original draft preparation – TO, ZT and SJ; writing – review and editing – TO, ZT, SJ, SS, LL and NG; and project administration – TO, ZT, SJ and NG. All authors have read and agreed to the published version of the manuscript.
Funding
The research was funded as part of the Medical Research Future Fund Palliative and End of Life Care Project, a joint venture between the NSW Regional Health Partners, NSW Health and the University of Newcastle. The funding body did not have any role in data analysis, interpretation, and the writing of this manuscript.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Data Availability Statement
The data presented in this study are not publicly available as participants whose data were included in the analysis have not consented to data sharing.
Supplemental Material
Supplemental material for this article is available online.
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
