Transfer from residential aged care to emergency departments: an analysis of patient outcomes


  • G. Arendts,

    Corresponding author
    1. Sydney School of Public Health, University of Sydney, Sydney, New South Wales
    2. Centre for Clinical Research in Emergency Medicine, Western Australian Institute for Medical Research, Perth, Western Australia
    3. Discipline of Emergency Medicine, University of Western Australia, Royal Perth Hospital, Perth, Western Australia
      Glenn Arendts, Emergency Medicine, Royal Perth Hospital, GPO Box X2213, Perth, WA 6001, Australia. Email:
    Search for more papers by this author
  • C. Dickson,

    1. Central Hospitals Network, South Eastern Sydney and Illawarra Area Health Service, Sydney, New South Wales, Australia
    Search for more papers by this author
  • K. Howard,

    1. Sydney School of Public Health, University of Sydney, Sydney, New South Wales
    Search for more papers by this author
  • S. Quine

    1. Sydney School of Public Health, University of Sydney, Sydney, New South Wales
    Search for more papers by this author

  • Funding: None.

  • Conflict of interest: None.

Glenn Arendts, Emergency Medicine, Royal Perth Hospital, GPO Box X2213, Perth, WA 6001, Australia. Email:


Background: In order to design optimal systems to meet the acute healthcare needs of the frail elderly living in residential care, good clinical information is essential. The aims of this study were to analyse the casemix and outcomes of patients transferred from residential aged care facilities to public hospital emergency departments in New South Wales.

Methods: Individual patient data from six hospital emergency departments and inpatient wards were obtained from merged databases and analysed using descriptive and comparative statistics.

Results: Outcomes in 4680 patient transfers over a 12-month period in 2006–2007 were analysed. Transfers occur mostly in high-acuity patients, with approximately three of every four transfers admitted; one in every 12 dying; and admitted patients undergoing an average of 2.4 interventions or procedures during each hospital stay. Several variables are associated with prolonged length of emergency department stay including triage urgency, type of hospital and transfers occurring in winter or out of hours.

Conclusions: Patients transferred from aged care facilities to emergency departments are predominantly high-acuity patients with a substantial likelihood of hospitalisation, intervention and death. Nevertheless, scope exists for some episodes of acute care, in both discharged and admitted patients, to be provided outside a hospital setting.


Over 150 000 people aged 65 years or over live permanently in residential aged care facilities (RACF) in Australia, 70% in high level care.1 People residing in RACF commonly suffer from chronic incurable disease, with symptom relief a prominent therapeutic goal.2 Notwithstanding this, there is often cause for provision of medical care via hospital emergency departments (ED) for this population. Such care may be compromised by myriad factors.3,4 Patients are typically frail, chronically ill, on multiple medications and present with symptoms that are atypical or unable to be communicated due to cognitive impairment. They are invariably assessed by doctors who have never seen them before, have incomplete information about their prior history, and whose training emphasises acute illness and injury management. They are frequently seen in an uncomfortable overcrowded environment within a system that emphasises rapid throughput with little time allowed for adequate assessment of their complex needs. When they receive therapy, they may respond unpredictably due to poor physiological reserve or drug interactions. While these problems apply to many elderly patients to varying degrees, several authors have noted the unique difficulties in the transition between RACF and the acute care setting.5,6

In order to address these shortfalls in care of patients from RACF in the acute care setting, and to design effective alternative models of care, it is important to have a clear understanding of the volume, casemix and outcomes of presentations to ED from RACF. Until lately, previous Australian analyses have been limited by being single hospital studies with less than 600 ED presentations, only one of which described inpatient care in a subsequent publication.7–9 Recently, a retrospective study of ambulance transfers to Perth ED compared casemix in elderly people from RACF with those not living in facilities, finding the casemix and mortality rate in patients from RACF differed from the broader elderly population.10 This study otherwise contained limited clinical information. Internationally, a large Canadian study examined health service utilisation by RACF patients and included data from almost 5000 patients over 2 years but also reported limited clinical or outcome data.11

The aim of our study was therefore to describe the clinical profile and outcomes of patients presenting to ED from RACF in one Australian state (New South Wales (NSW)). Additionally, we sought to identify factors associated with patient outcomes such as prolonged ED stay; discharge from ED back to the RACF without admission; and death. Improving the understanding of the interaction between RACF and ED will enable clinicians and policy-makers to better respond to the challenge of acute care provision for this complex population.

Materials and methods

An analysis of consecutive individual de-identified patient presentations to six ED in NSW was conducted. Ethics approval for the study was sought and determined to be within the scope of routine data usage due to the de-identified nature of all data given to the researchers (P. Fink, pers. comm., 2007, South Eastern Sydney and Illawarra Area Health Service Central Network Ethics Committee).

We selected six public hospital ED from three area health services as broadly representative of ED in NSW. We chose to look only at ED designated level 4 or above with some access to continuous medical services.12 Two tertiary ED in Sydney (one inner and one outer metropolitan), two urban district ED, one rural base hospital and one other rural hospital were included. Four of the six hospitals are accredited for advanced training by the Australasian College for Emergency Medicine.

Consecutive presentations to the six ED, over a 12-month period from 1 July 2006, by people aged 65 years and over and living in a RACF were identified using the computerised patient management system EDIS (iSOFT, NSW, Australia). We initially identified patients where any of the following criteria were found in EDIS:

  • 1Address and telephone number matched those of an accredited RACF when cross-linked to published details13,14
  • 2Source of referral in EDIS coded as RACF
  • 3Free text search within either the triage or presenting problem fields included the words nursing home, hostel, aged care facility or any recognised abbreviations for these, for example, N/H, RACF
  • 4Discharge destination at separation coded as RACF.

The presence of (1) plus at least one other criterion was taken to be evidence of RACF habitation for inclusion. Where patients were identified from (1) alone, or at least two other criteria without (1), the presentation was further assessed to determine eligibility for case inclusion. Data cleaning was undertaken to remove duplicate or invalid entries.

Because some facilities contain a mixture of high level and low level beds, the data were stratified for analysis and where relevant is reported in two groups, exclusively high level and combined high/low level. Facilities with exclusively low level and/or self-care beds were excluded from the study, as were facilities with more than 5% of their beds being respite beds.

Variables of interest for analysis are detailed in Box 1. For patients who were not admitted from ED (discharged or died within ED) we used EDIS exclusively for outcome information. For patients admitted, the ED episode of care was linked to inpatient data by a unique episode identifier, and outcome data extracted from the Health Information Exchange (iSOFT, Sydney, New South Wales, Australia) system.

Box 1 Variables analysed for each individual patient presentation to emergency departments.

  • 1ED variables
    • • ED type
    • • Mode of arrival to ED
    • • Source of referral to ED
    • • Arrival date and time (in or out of office hours 08:00–18:00 Monday–Friday excluding public holidays)
    • • Triage category
    • • Presenting symptom/s
    • • ED length of stay
    • • ED waiting time before seen by doctor
    • • Provisional diagnosis at separation from ED
    • • Urgency, Disposition and Age-related Groups weight
    • • Destination on departure from ED
  • 2Demographic variables
    • • Age
    • • Sex
    • • Country of birth
    • • Indigenous status
    • • Preferred language
    • • Need for interpreter service
    • • Health insurance status
    • • Type of residential aged care facility where patient resides
    • • Total number of facility beds
  • 3Inpatient variables
    • • Admitting specialty
    • • Principal inpatient procedure/s
    • • Principal diagnosis
    • • Episode length of stay
    • • Discharging specialty
    • • Separation from hospital description

We used descriptive statistics where appropriate. Proportions were compared using Pearson's χ2 test. For ED length of stay and in hospital mortality, we used multivariate analysis to adjust for the influence of demographic and clinical variables on outcome. For dichotomous outcomes we used a logistic regression model that included variables with P < 0.3 on univariate analysis and backward elimination of interaction terms and non-predictive variables. For continuous outcomes we similarly carried out linear regression analysis, using a square root transformation of variables that were not normally distributed. Data storage was through Microsoft Access (Redmond, WA, USA) and statistical analysis through spss version 17 (spss, Chicago, IL, USA).


Figure 1 summarises case identification. In the 12-month period there were 4680 identifiable presentations from 89 eligible RACF to the six ED. This represents 9.9% of ED presentations in the 65+ years age group and 2.0% of presentations overall. For individual ED, the range of RACF presentations as a proportion of age 65+ presentations was 5.7–15.2%.

Figure 1.

Summary of case identification. ED, emergency department.

Demographic findings

Age distribution (median 85 years) is shown in Figure 2. Relatively more patients were female (63.6%) and this proportion increased with age to 69.2% of patients aged 85+ and 75.1% aged 95+. Most patients (78.5%) spoke English as their primary language, with only 7.5% needing an interpreter. One-sixth (16.6%) had some type of private health insurance.

Figure 2.

Age distribution of residential aged care facilities to emergency department transfers.

ED findings

Almost all patients (93.4%) were transported to ED by ambulance. Most patients (n= 2693, 57.5%) were admitted from ED to an inpatient hospital bed, including 70 transferred from ED to a different hospital. The admission rates from high care (58.5%) and combined care (55.7%) RACF were not significantly different (χ21 3.38, P= 0.07). Another 79 (1.7%) were admitted under inpatient units but spent their entire hospital admission in the ED. The remainder were discharged directly from ED (n= 1203, 25.7%); managed as an ED-only admission or in an ED short-stay unit (n= 623, 13.3%); or died in the ED (n= 63, 1.3%).

Approximately half of all referrals to ED occurred within office hours (n= 2170, 46.4%). No difference in inpatient admission rates for out-of-hours (56.9%) and in-hours (59.3%) referrals was found on univariate analysis (χ21 2.59, P= 0.11). High (47.5%) and combined (45.7%) facilities had similar proportions of in-hours referrals (χ21 1.26, P= 0.26).

The breakdown of patient triage with the Australasian Triage Scale (ATS) was emergency (ATS 1 or 2) 15.7%; urgent (ATS 3) 49.1%; and less urgent (ATS 4 or 5) 35.2%. This is virtually identical to that of all patients aged 65+ (16.7%; 47.9%; 35.4% respectively). Table 1 describes the principal ED diagnosis of patients, excluding the 63 deaths.

Table 1.  Diagnosis at separation from ED
CategorySpecific diagnosisN
  1. AMI, acute myocardial infarction; CAL, chronic airways limitation; CVA, cerebrovascular accident; ED, emergency department; ENT, ear, nose and throat; GI, gastrointestinal; ICH, intracerebral haemorrhage; NOF, neck of femur; SAH, subarachnoid haemorrhage; SDH, subdural haemorrhage; TIA, transient ischaemic attack; UTI, urinary tract infection.

NeurologicalAltered consciousness77
Head injury89
Acute coronary syndrome/chest pain189
Heart failure158
Rhythm disturbance98
GastrointestinalAbdominal pain for investigation83
Bowel obstruction32
GI bleed101
RespiratoryDyspnoea for investigation231
CAL exacerbation75
Urinary retention73
Orthopaedic# NOF197
# upper limb52
# other59
Other surgicalVascular23
PsychiatricBehavioural disturbance including psychosis119
Other medicalRenal failure38
Electrolyte/metabolic disturbance65
Sepsis/fever of unknown source95
Adverse drug reaction22
Minor traumaNo/trivial injury post fall424
Laceration requiring repair80
Soft tissue injury141
Catheter/device problem 108(2.3%)
Not categorised above 78(1.7%)
Planned review in ED 27(0.6%)
Missing data 21(0.5%)
Total 4617(100%)

There was a median (interquartile range (IQR)) length of stay within ED of 5.3 (3.4–7.6) h for discharged patients and 7.5 (5.3–11.4) h for patients not discharged. Comparative figures for all patients aged 65+ are 3.6 (2.0–5.5) h and 6.9 (4.8–10.2) h respectively. On linear regression analysis, several variables (triage urgency, referrals out of office hours and within winter months, and referrals to tertiary teaching hospitals) were associated with increasing ED length of stay for discharged patients. However, despite these variables predicting length of stay, we found no model with an adjusted R2 greater than 20%, suggesting that most of the variance in length of stay for patients discharged is not explained by the factors analysed in our study. When length of stay was divided into 0–4 and >4 h, out of hours referral was not found to be a significant predictor of excess stay (Table 2).

Table 2.  Adjusted odds of ED length of stay >4 h for discharged patients
VariableCategoryOR (95% CI)
  • Reference group. ATS, Australasian Triage Scale; CI, confidence interval; ED, emergency department; OR, odds ratio.

HospitalTertiary referral5.4 (2.9–9.9)
Urban2.5 (1.3–4.6)
SeasonWinter1.4 (1.1–1.9)
TriageEmergency (ATS 1 & 2)2.4 (1.2–4.8)
Urgent (ATS 3)1.8 (1.4–2.3)
Semi-urgent (ATS 4 & 5)1.0

In-hospital findings

The median (IQR) length of stay was 5 (1–10) days for all admitted patients and 6 (3–11) days where ED short-stay admissions were removed from analysis.

There were 8188 procedures or interventions (Table 3) carried out on 3395 admitted patients, with only one-quarter of patients (n= 860) undergoing no coded intervention during their admission. However, more than half of these were non-invasive nonmedical or allied health interventions.

Table 3.  Procedures coded on admitted RACF patients
  1. CT, computed tomography; CVL, central venous line; ECT, electroconvulsive therapy; ENT, ear, nose and throat; GA, general anaesthesia; I&D, incision and drainage; ICC, inter-costal catheter; NIV, non-invasive ventilation; ORIF, open reduction and internal fixation; PEG, percutaneous endoscopic gastrostomy; PPM, permanent pacemaker; PTCA, percutaneous transluminal coronary angioplasty; RACF, residential aged care facility; SPC, suprapubic catheter; TVP, transvenous pacing.

Non-invasiveAllied health and non-medical interventionsDietician, physiotherapy4706
Subtotal 4779
Minimally invasiveImagingCT, angiogram, nuclear1447
OncologicalRadiation therapy15
Orthopaedic# reduction, arthrocentesis39
Other 26
Subtotal 1964
Moderately invasiveAnaesthesiaNeural blockade198
Dental 15
CardiologyTVP, PPM, PTCA69
ENTNasal packing10
GastrointestinalEndoscopy, PEG insertion193
PlasticsWound repair, skin graft84
RespiratoryICC, bronchoscopy17
SurgicalI&D, paracentesis45
VascularCVL insertion55
UrologicalCystoscopy, SPC95
Other 19
Subtotal 831
Major invasiveAnaesthesiaGA331
NeurosurgicalCraniotomy, laminectomy12
SurgicalLaparotomy, cholecystectomy46
Subtotal 614
Total  8188

Three hundred and twenty-two (9.5%) patients died after admission, in addition to the 63 that died within ED before admission, giving an overall case mortality rate of all ED transfers from RACF of 8.2%. Increasing age, male gender, high urgency triage category and transfer from a high care facility are all associated with an increased mortality risk for admitted patients (Table 4). No seasonal, time of day or hospital differences were found.

Table 4.  Adjusted odds of death for admitted RACF patients
VariableCategoryOR (95% CI)
  • Reference group. CI, confidence interval; ED, emergency department; OR, odds ratio; RACF, residential aged care facilities.

Facility typeHigh care1.3(1.0–1.6)
Combined high and low care1.0
TriageEmergency (ATS 1 & 2)8.2(5.8–11.7)
Urgent (ATS 3)1.9(1.4–2.7)
Semi-urgent (ATS 4 & 5)1.0


We have evaluated RACF cases referred to various types of public hospital ED in order to improve the understanding of the casemix and resource requirements of RACF to ED transfer.

From our analysis it appears that while some transfers appear to be in patients with minor illness or injury, most transfers to ED from RACF occur by ambulance in high-acuity patients. Only one in four is discharged directly from ED; for every 12 transfers one patient will die; and even with their advanced age and comorbidities there is a high likelihood that admitted patients will receive some form of invasive intervention. Our results provide important information for the design of acute care systems for this complex and vulnerable population.

The health policy implications of an ageing population ageing are well recognised in Australia and internationally. ED presentations are increasing above population growth, with older people disproportionately represented in this increase.15,16 The likelihood of admission to hospital once a patient has presented to ED also increases with age – in Australia in 2005–2006, patients aged 65 years or over constituted 13% of the population and 47% of hospital bed day occupancy.17 It is estimated that the proportion of the population in this age bracket will double by 2050.18

ED care in Australian hospitals is generally of high quality, but is often unsuited to the needs of chronically ill older people who are more likely to suffer adverse outcome within an ED.3,19,20 Increasing numbers of an ageing and less well population attending ED have led to suggestions of modification to ED-based care for this patient group, through redesign of ED or provision of acute care outside of ED.20,21 The latter particularly has been suggested for older people in RACF due to the chronicity of their condition and the possibilities presented by on-site trained staff in high care facilities. However, before designing and implementing such changes, it is imperative that we understand the casemix and outcomes in this population.

Despite some limitations, notably a lack of interventional and procedural detail for the group discharged from ED without admission, we have provided the most comprehensive multicentre analysis of referrals to ED from RACF in the literature. We have found that referrals tend to occur in patients with significant medical illness or injury, at high risk for requiring an invasive procedure and dying during hospitalisation. These findings do not preclude the exploration of alternatives to ED/hospital care for RACF patients; however, the design of such alternatives would need to incorporate plans for this high burden of illness.

Previous single centre studies, including one from Australia, have attempted to quantify the proportion of transfers from RACF to ED that are ‘avoidable’ or ‘inappropriate’, finding a wide range from 7% to 48%.7,22,23 Rather than use an arbitrary and inherently subjective definition of what is ‘avoidable’, we have substantively detailed the caseload and resource use of current RACF to ED transfer. Any alternative acute care model devised to avoid some or most of these transfers will need to effectively manage this workload. Furthermore, the superiority of any alternative model over current ED care needs to be demonstrated by improvements in one or more of the following parameters: clinical effectiveness, resource efficiency or acceptability to patients. If such cannot be demonstrated, resources may be better spent on improving care within ED.

We have found over half of all interventions carried out in hospitals are non-invasive, hence potentially able to be carried out out of hospital. Although invasive procedures are also common in this population, it should not be extrapolated that no scope exists for alternative acute care models because of the need for procedural interventions. It is likely that at least some of these procedures could be carried out within RACF or on an outpatient basis. Indeed, randomised or quasi-randomised trials have shown that certain subsets of patients within RACF may avoid hospitalisation with enhanced care in the facility.24,25

We did not include exclusively low care facilities in our analysis, but chose to include combined care institutions. Increasingly, combined facilities are burdened by high care patients within low care beds, and in some cases act as quasi-high care facilities. Although it is possible that the same phenomenon may occur in exclusively low care facilities, on balance we believe people in exclusively low care facilities are more likely to be distinct from high care patients, and including such facilities posed a greater threat to the validity of our findings then excluding them. We found no difference in admission rates or referral patterns from high versus combined facilities, although patients from high care facilities were more likely to die during their admission.

Ingarfield et al. compared ED ambulance attendances in patients from RACF with elderly patients living outside facilities.10 Several results from that study conform to our own, such as the in-hospital length of stay (median 6 days) and mortality rate (7.8%) found in RACF patients. In contrast to Ingarfield et al., our study contains clinical information on ED and inpatient casemix, and analyses factors associated with outcomes such as death and length of stay. This additional information will be of importance to clinicians or policy-makers contemplating service re-design for this population.

Several variables were found that predict increased length of stay for patients eventually discharged from ED. Some of these, particularly winter months and tertiary centres, may be considered as indirect markers of ED overcrowding. However, our analysis suggests many unmeasured and possibly unmeasurable variables influenced this outcome, making definite conclusions difficult. Furthermore, at least some ED short-stay admissions may occur in patients that are clinically ready for discharge directly from ED but need specialist transportation, for example, ambulances to facilitate discharge; both admission rates and ED length of stay can be locally influenced by such non-clinical factors. The disparity in ED length of stay between the RACF and broader 65+ populations may be partially explained by this.


Patients transferred from RACF to ED are high-acuity patients with a substantial likelihood of needing hospitalisation, receiving invasive intervention and dying during their admission. Despite this, we should not dismiss the possibility that alternative models of care may lead to improved clinical outcomes, more efficient use of scarce ED resources, or be more acceptable to patients. Randomised controlled trials of alternative care models are needed to fill this evidence gap.


Ms Yueming Li (Senior Statistical Analyst, NSW Health) and Dr Robin Turner (Research Fellow, School of Public Health, University of Sydney) are acknowledged for their advice and assistance with data extraction and statistical analysis.