Expanding emergency department capacity: a multisite study

Additional keywords: ambulance, data linkage, outcomes, service delivery.

Approximately 23% of the 7.1 million emergency department (ED) presentations made to Australian public hospitals in 2007–08 were via ambulance.¹ Compared with other Australian states and territories, Queensland reported the largest average annual increase in ED presentations (7.7% p.a.) from 2006–07 to 2010–11.² The utilisation rates of ambulance services in Queensland have also increased considerably at an average annual rate of 5.4% between 1999–00 and 2009–10 for urgent dispatches.³ Escalating growth in demand from emergency patient services has placed increasing strain on both ambulance and hospital resources.^3,4 Negative outcomes, such as ambulance diversion, access block (an ED length of stay (LOS) of >8 h for patients requiring hospital admission⁵) and increased risk of hospital mortality, as a result of ED and hospital crowding have been reported in Australia and overseas.^5–7 Meeting healthcare targets within this environment can be difficult, but is becoming increasingly mandated, monitored and reported upon.

Improvements in or expansions of healthcare-related services are required in order to meet the healthcare needs of the community in a safe and sustainable fashion.^8–11 In Australia, the Federal Government (through the National Partnership Agreement) has committed to provide funding exceeding A$3 billion for new subacute beds, to meet ED and elective surgery targets and for capital and recurrent projects to improve access for patients accessing public hospital services.¹⁰ Following a staged annual increase commencing in 2012, it is expected that by 2015 90% of ED presentations should be admitted, transferred or discharged within 4 h, thereby meeting National Emergency Access Targets (NEAT).¹⁰ Although descriptions of opening an additional ED or expanding the size and number of beds in an existing ED are noted within the literature,^12–16 little research exists evaluating the impact these measures can make to patient, ambulance and health service delivery outcomes.

Expanding ED capacity interacts with overall service provision and patient outcomes. As such, the aim of the present study was to identify the characteristics and predictors of hospital admission and describe outcomes for patients who arrived via ambulance to three Australian public EDs before and after the opening of 41 additional ED beds within a health service.

The present study was a retrospective comparative cohort study undertaken in three regional public hospitals located in south-east Queensland, Australia. These three hospitals, along with three private hospitals, served a total population of approximately 800 000.¹⁷ All three EDs were teaching facilities and treated both adult and paediatric patients. Hospital A had 45 ED beds and 473 hospital beds; Hospital B had 36 ED beds and 290 hospital beds; Hospital C had 41 ED beds and 200 hospital beds. Located approximately 15 km apart, Hospitals A and C shared operational capability; a further 50 km away, Hospital B was slightly more isolated, located midway between two main groups of larger hospitals. The ED at Hospital C opened on 3 September 2007. Seventeen permanent Queensland Ambulance Service stations were located in the direct catchment area, plus a rotary wing retrieval service.

All patient presentations arriving to three EDs over a 24-month period (3 September 2006–2 September 2008) were included in the linkage of data sources from ambulance, ED and hospital admission. Figure 1 shows the sample inclusion process. Some patient presentations were excluded from the dataset during and following the data cleaning and data linking process due to mode of arrival incorrectly coded (was not via ambulance), no name, no gender and incomplete date and time data. A power calculation showed that the sample size was more than adequate to detect a difference in the primary outcome (hospital admission) and the chance of a Type II error was negligible. Our sample size could provide 99% power, based on an α level of 0.05, model (Nagelkerke) R² of 0.384 and 21 predictors for the primary outcome (hospital admission).

Fig. 1.  Sample inclusion flow diagram (data from three hospitals, 24 months).

The specific data used (sourced from the Queensland Ambulance Service (QAS) Emergency Department Information System (EDIS) and Hospital Based Corporate Information System (HBCIS)) are given in Table 1. The variables chosen were based on previous research and related literature.^18,19 The Australasian Triage Scale (ATS) is an indicator of urgency, where a number (on a scale of 1–5) corresponds to the time frame in which patients should be seen by a doctor.²⁰ Patients allocated an ATS Category 1 should be seen immediately, Category 2 within 10 min, Category 3 within 30 min, Category 4 within 60 min and Category 5 within 120 min.

Table 1.  Data collected from each health information source
DRG, diagnostic-related group; ED, emergency department; ICD 10, International Statistical Classification of Disease and Related Health Problems (10th revision)

We used Health Data Integration (HDI; Australian eHealth Research Centre, Herston, Qld, Australia), an automated deterministic linking approach developed by the Commonwealth Scientific and Industrial Research Organisation (CSIRO), to link data from the three separate health information systems databases (QAS, EDIS and HBCIS). The HDI linking strategy has been tested previously for accuracy with high sensitivity, specificity and positive predictive values.²¹

Descriptive statistics were used to describe the profile of all patients presenting to three EDs via ambulance. These statistics included median and interquartile range for age, time to see doctor, ED LOS and hospital LOS, and frequency and percentages for categorical variables (i.e. age group, gender, ATS category, day presented, season, admission, diagnostic-related group, offload time >30 min, seen within ATS time frame, ED LOS >4 h, ED LOS >8 h for admitted patients and in-hospital mortality). Inferential statistics were used to identify differences between groups before (3 September 2006–2 September 2007) and after (3 September 2007–2 September 2008) opening of the ED. The non-parametric Mann–Whitney U-test was used for continuous data; Chi-squared tests were used for categorical variables. Univariate followed by multivariate logistic regression models were built (using the enter and forward stepwise method, respectively) to assess the individual variable contribution followed by adjusted contribution for the main outcome, hospital admission. Multivariate procedures were used to adjust for the confounding effect because there were at least three predictor variables.²² Predictors entered into the regression model included age, gender, ATS category, time of presentation (in 8 h blocks), day of week (as weekday or weekend), International Statistical Classification of Disease and Related Health Problems – 10th revision (ICD 10) diagnosis code, season, ambulance offload delay >30 min, ED LOS >4 h and hospital. The ICD 10 are internationally recognised diagnostic codes that account for diagnoses, descriptions of symptoms and cause of death.²³ In Australia, patients who present to the ED are assigned a modified ICD 10 code by the treating medical officer or nurse and are entered prospectively into the ED database.²⁴ Crude and adjusted odds ratios (OR) are provided for the logistic regression models with results presented as OR and 95% confidence intervals (CI). Two-sided P < 0.05 was considered significant. Reference groups were based on previous research, cell size or the most logical comparison. Data were analysed using SPSS version 18.0 (SPSS, Chicago, IL, USA).

Ethics approval was obtained from the Health Service Districts Human Research Ethics Committee, the Queensland Ambulance Service and Queensland Health’s Research Ethics and Governance Unit to use health information.

In total, 286 037 patients presented to the three EDs in the study period. Of these presentations, 79 196 analysable patient presentations were via ambulance, with the overall number of ambulance arrivals increasing by over 2000 from one year to the next. The characteristics of patients presenting via ambulance differed between the periods before and after ED expansion in terms of the demographic and clinical characteristics of age, gender, ATS category and diagnosis (Table 2). Although a 2-year increase in median age is statistically significant, it is unclear whether this finding is clinically significant. However, a 1.5% increase in Category 1 and 2 presentations, although small, may have clinical significance. Table 3 presents the ED characteristics of ambulance-arriving presentations. Although proportions appear relatively unchanged from a clinical perspective, some characteristics differed statistically between the periods before and after ED expansion (weekday or weekend, season), whereas others did not (shift, day of week).

Table 2.  Clinical characteristic of patients arriving at the emergency department (ED) via ambulance before and after ED expansion
IQR, interquartile range; ATS, Australasian triage scale; ICD 10, International Statistical Classification of Disease and Related Health Problems (10th revision)

Table 3.  Emergency department characteristics of patients arriving at the emergency department (ED) via ambulance before and after ED expansion

The proportion of ambulance-arriving patient presentations that required admission within each potential predictor entered into the univariate regression model is presented in Table 4. Table 4 also displays the crude OR, 95% CI and P-value of each predictor. All univariate predictors were entered into the multivariate logistic regression analysis model. When stratified according to the period before and after ED expansion, similar numbers (15 and 14, respectively) and types of independent predictors, indicating higher odds of hospital admission, were identified. Predictors with OR >2 across both time frames were age ≥65 years, ATS Category 1, 2 or 3, ICD diagnoses relating to circulatory and respiratory diseases and an ED LOS >4 h.

Table 4.  Independent predictors of hospital admission for ambulance patients stratified according to pre- and post-emergency department expansion
OR, odds ratio; CI, confidence interval; ATS, Australasian triage scale; ED, emergency department; ICD 10, International Statistical Classification of Disease and Related Health Problems (10th revision); LOS, length of stay

Outcomes for patients arriving at the ED via ambulance are presented in Table 5. All outcomes differed significantly between the periods before and after ED expansion. The only outcomes that improved related to in-hospital mortality, which decreased by 1.5% based on the patient’s last index of admission. Outcomes that did not improve included the proportion of patients not offloaded within 30 min, admitted and access blocked; these increased by 4%, 4% and 11%, respectively. The proportion of patients seen within the recommended ATS time frame also did not improve, decreasing from 44% to 39%. Median time to see a doctor and ED LOS for both admitted and non-admitted patients did not improve, increasing by 4, 65 and 21 min, respectively. These differences were statistically significant. Due to the large sample sizes included in the study, the in-hospital LOS differed significantly; however, this was not necessarily clinically meaningful because the median LOS remained constant at 2 days. Compared with National public hospital data,²⁵ a favourable outcome for ambulance-arriving patients included ED LOS for those who did not require hospital admission; less favourable outcomes included time to see a doctor, ED LOS for patients requiring hospital admission, being seen within ATS recommended times, admission requirement and in-hospital mortality.

Table 5.  Outcomes for patients arriving at the emergency department (ED) via ambulance before and after ED expansion
Data are presented as the number of patients in each group, with percentages in parentheses, or as the median (interquartile range), as appropriate. Time to see doctor, based on 75 129; ED length of stay (LOS) (total): based on 79 191 patient presentations; ED LOS (not admitted), based on 46 605 patient presentations; ED LOS (admitted), based on 31 928 patient presentations; seen within Australasian Triage Scale (ATS), based on 75 129 ED presentations. AIHW, Australian Institute of Health and Welfare; IQR, interquartile range; N/A, not available

Within Queensland, more people present to the ED, arrive via by ambulance, have longer waiting times, require hospital admission, have a shorter hospital LOS and have higher mortality, higher admission costs and more potentially preventable hospitalisations compared with most other Australian states and territories.²⁶ The ability to link patient level records across and within systems enabled us to obtain a more informed understanding of the patient journey that encompasses the ambulance, ED and hospital admission episodes of care.

The proportion of ambulance-arriving patient presentations made to the EDs in the present study was higher than that of the national average (33% vs 23%).²⁵ With such large sample sizes and with distributions that were highly skewed, although the differences in demographic and clinical characteristics of patients presenting in the two time periods (before and after ED service expansion) reached statistical significance, clinically these small differences would make little impact on service provision. The one difference that could have influenced outcome was the 1.5% increase in ATS Category 1 and 2 patients arriving after the ED bed numbers had expanded.

Patients who present to the ED via ambulance have a relatively high admission rate. Compared with all presentations made to Australian public hospital EDs,²⁵ patients arriving to the EDs in the present study, via ambulance, were a group with higher admission requirements (27%²⁵ vs 40%). Our admission rate was similar to that reported in a Canadian study²⁷ (38% admission rate of ambulance arrivals), but lower than that seen in a study undertaken in Singapore (59% admission rate for ambulance arrivals).²⁸ The Canadian study further revealed that people arriving to the ED via ambulance had admission odds 3.15-fold those of people arriving via other means.²⁷ These figures indicate that further identification of specific predictors of admission used to facilitate targeted service delivery and patient flow may be useful.

Although the odds of admission changed slightly for some predictors in our study, the actual predictors themselves remained relatively unchanged between the periods before and after ED expansion. Prominent predictors of hospital admission for patients arriving to the ED via ambulance in our study are easily identifiable and include age ≥65 years, ATS Category 1, 2 or 3, ICD diagnoses relating to circulatory and respiratory diseases and an ED LOS >4 h. With the exception of ED LOS, all the predictors identified in our study can be recognised very early on in a patient’s ED episode of care.

Our findings are relatively congruent with other studies identifying the prevalence and predictors of hospital admission; however, other studies were not limited to ambulance-only arrivals and some focused on the predictors of specific demographic or illness groups. With this in mind, our discussion here focuses on the main predictors of hospital admission. Regarding age, several studies have identified an admission rate of approximately 65% for people aged ≥65 years.^28,29 Within this demographic, higher odds of admission were related to increasing patient age, higher heart rate, lower blood pressure, lower triage score and several chief complaints, such as pneumonia and stroke.²⁹ Knowing that older people (particularly those arriving via ambulance) comprise a high proportion of and likelihood for admission, it is possible to use this and other predictive information, such as that offered by LaMantia et al.,²⁹ to order an in-patient bed at the point of triage, thus avoiding delays in admission.

Admission requirement for people arriving by ambulance and triaged ATS Category 1, 2 or 3 was relatively high (~84%, 66% and 40%, respectively). Although these proportions were reasonably reflective of all ED presentations made to Australian public EDs during this time,²⁶ they are higher than those noted elsewhere.²⁷ Identifying higher odds of admission for people triaged as more emergent (i.e. ATS Category 1, 2 or 3) compared with the less emergent ATS Category 5 ambulance patient presentations was not surprising and, rather than ‘over-triaging’ (a practice where nurses allocate a triage category of higher acuity than required),³⁰ the finding is most likely reflective of genuine urgency and illness.

Circulatory (80% admission rate) and respiratory (55% admission rate) diseases were indicative of hospital admission for ambulance-arriving presentations made to the three study site EDs with odds six- and twofold higher than for people not admitted with these conditions. A more recent Canadian study involving all adult ED presentations also identified (to a lesser extent) high admission rates (50%) for circulatory conditions compared with other illnesses.²⁷ An Australian study undertaken in five Melbourne EDs focusing on patients admitted with a specific respiratory disease (chronic obstructive pulmonary disease) revealed this as a group with a higher admission rate (77%).³¹ Although these two studies were not confined to ambulance-arriving patients, they do indicate that people with circulatory and respiratory conditions are highly likely to be admitted. Furthermore, these patients may benefit from early intervention and expedited admission processes, such as those offered by medical admission units.³²

People arriving by ambulance spending more than 4 h in the ED were admitted in higher proportions (60% vs 40%) and had odds of admission sixfold higher than those spending <4 h in the ED. This finding may be explained by an inadvertent pressure on medical staff to admit, rather than discharge. Although the present study was conducted before the introduction of NEAT targets,¹⁰ it provides evidence to indicate a need for practice and process changes. With the imperative to improve patient flow from a whole-of-hospital approach,^33–35 this finding may be useful to justify early planning and management purposes. It is suggested that planners use additional clinical information (e.g. timely reporting on diagnostic investigations)³⁶ and evidence-based admission prediction tools³⁷ to enhance predictability and avoid inappropriate admissions.

Our study identified both statistically and clinically significant differences for patients arriving to the ED via ambulance following ED expansion. Even though ED capacity expanded from 81 to 122 beds within a health service area, difficulty meeting several ED performance targets continued. These ongoing challenges included offloading patients within 30 min, seeing patients within the recommended ATS time frame and ED LOS >8 h.

Most patients were able to be offloaded from ambulance to ED within 30 min. Although some (11% before and 15% after ED expansion) were delayed by >30 min, the proportions are lower than those noted in recent research undertaken in New South Wales, where 17% of patients experienced a delay of >30 min.³⁸ This finding indicates that further research understanding specific elements of this process is required. Although some of the Queensland processes result in better outcomes than in other states, Queensland outcomes are still not perfect and alternative service models may be useful in targeting the initial 30 min for patients arriving via ambulance.

The issue of ambulance offload delay has been of concern to patients, hospitals and the Queensland State Government. A recent Queensland report into this issue presented 15 recommendations for implementation in EDs across the state.³⁹ Examples of these recommendations include: having an off-stretcher key performance indicator within the health service; the development of performance-based hospital access incentive-funding framework; aborting hospital authority to ‘bypass’ or divert ambulances to another hospital; rest responsibility of sharing ambulance distribution with the ambulance service; that major public hospitals introduce senior clinical nurses to care for waiting patients; and the establishment of a high-level Emergency Services Management Committee to provide policy advice to the Minister on issues affecting consumer access to (and delivery of) public hospital emergency services.³⁹ The extent to which and outcomes from the implementation of recommended changes are, to date, unknown. In Canada, a dedicated advanced practice nurse, tasked with receiving care of and providing early management for ambulance-arriving patients, has been a model in place (funded by the ambulance service) for several years.⁴⁰ A formal evaluation would benefit other health services interested in implementing a similar model.

The results of the present study indicate that expansion of ED bed capacity did not lead to improvements in compliance with ATS categories for time to be seen by a doctor for ambulance-arriving patients. Furthermore, in this study setting, greater proportions of patients exceeded ATS category cut-offs for the time to be seen by a doctor, both before and after ED expansion, compared with the national average for all patient presentations (56% before and 61% after ED expansion vs 32% nationally).²⁵ Other service delivery models to address this stage of the patient journey are required. Nurse practitioners and/or doctors to assist in triage⁴¹ or shortly thereafter⁴² may be worth trialling.

Accessibility (which includes waiting times for ED care) is one of the health system performance indicators of the National Health Performance Framework.²⁵ Waiting time (from arrival to treatment) is one of three distinct phases of a patient’s journey through the ED.¹⁹ The other two are time to triage (from arrival) and LOS, both for non-admitted and admitted patients.¹⁹ Regarding LOS, our findings indicate that the EDs performed well on the portion of the indicator for non-admitted ED LOS, which was 72 min (before) and 51 min (after ED expansion) shorter than the national average of 281 min.²⁵ This may be explained, in part, by the use of efficient models of care for minor injuries.⁴³ However, on the portion of the indicator for ED LOS for admitted patients, ED LOS was far higher (by 147 and 212 min before and after ED expansion, respectively) than the average of 244 min.²⁵ This long ED LOS for admitted patients is reflected in the high levels of access block also identified (35% and 47% before and after ED expansion, respectively) and the 4% increase seen in the admission rate for ambulance-arriving patients. Additional in-patient beds did not accompany the ED expansion in the study hospitals, and this may explain, to some extent, the reason why further improvements in these outcomes were not identified.

For those patients requiring hospital admission, hospital LOS remained unchanged (at 2 days), which was considerably less than the 3.1 days (for all acute care admissions) and 6.3 days (excluding same day separations) noted in a national report.²⁵ In-hospital mortality rates found in our study were higher than the national average of 1.2%.²⁵ This finding may be reflective of the high acuity and admission likelihood of our study group (patients arriving via ambulance only). However, in-hospital mortality rates did decrease between the before and after ED expansion periods. This may be explained by the following: (1) patients were discharged earlier due to access block pressures and they may have died elsewhere; (2) there may have been better care, but this was unable to be captured in the parameters we collected and presented herein; or (3) after ED expansion, the proportion of patients who presented in ATS Categories 1 and 2 increased by 1.5%. Although it may be argued that improvements beyond those identified here may have been expected, it may be that the high access block rates reflect a crowded system that extends beyond the ED. As has been noted by Han et al.,¹⁶ if the volume and complexity of presentations increase and in-patient beds are not available to enable transfer out of the ED, then this limits the ED’s capacity to improve performance.¹⁶ A whole-of-hospital approach that involves shared responsibility and communication between the ED and wards to better meet ED targets and reduce overcrowding and mortality rates has been reportedly successful in Western Australia⁴⁴ and warrants consideration by other Australian states.

There are several limitations to the present study. First, because this was a retrospective analysis of prospectively collected data, there may have been inaccuracies within the data provided. However, we did implement data-cleaning measures in order to account for and manage inaccuracies. In addition, with the intention of understanding service delivery from an area-wide perspective, our analysis reflects pooled data from three EDs. Therefore, improvements to individual sites are not evident. Second, results indicating statistical significance may not necessarily relate to clinically meaningful significance due to the extensive volume of data used.⁴⁵ However, not all outcomes were significant, indicating that sample size was not the only factor determining significance. Third, our study was limited to the impact of opening additional ED beds only and comprised patients arriving to the ED via ambulance only. No accompanying hospital beds were opened at the same time, and we did not compare those arriving to the ED via ambulance with those arriving via other means. In addition, factors outside the scope of the present study, such as changes to working practices within the pre-hospital and ED settings and changes to health policy and funding arrangements, may have impacted on our results. Thus, the interpretation of our findings should consider these facts. Fourth, the present study did not include the effect on the three private hospitals within the local area or the other seven public hospitals within the broader geographic region. As such, there may have been a network effect that extended beyond the sites included in this study. Fifth, although we found ED LOS >4 h to be correlated with admission likelihood, it is difficult to ascertain whether this was caused by the wait or contributed to the cause for admission. Given that the ED LOS was shorter than the national average for non-admitted patients and longer than the national average for admitted patients, this indicates that the long ED LOS is likely caused by difficulties moving the patients from the ED to the hospital. However, this (i.e. ED LOS >4 h) and other predictors used were chosen based on related literature and clinical expertise. Finally, research examining structures, processes and other outcomes not described within our study (e.g. costs, time to triage, re-presentation, readmission rates and other crowding measures noted by Hwang et al.⁴⁶ and Liu et al.⁴⁷) can be useful in providing further insight into the care quality and outcomes of this change in service and to this patient group. Further research accounting for these limitations is warranted.

The aim of the present study was to identify characteristics and predictors of hospital admission and describe the outcomes for patients who arrived via ambulance to three EDs following the opening of 41 additional ED beds within the area. The results demonstrate that for this group of patients, after the additional emergency capacity was operational, improvements to in-hospital mortality outcome could be realised. However, improvements to other outcomes, such as time to be offloaded from an ambulance, time to see a doctor, ED LOS and hospital LOS, were not realised and further research related to novel models of care are required. Easily identifiable predictors for hospital admission were noted in both pre- and post-ED expansion cohorts and included older age, assignment of higher (more emergent) triage categories and certain diseases pertaining to circulatory, respiratory and mental health groups. Understanding these predictors may be useful for implementing new strategies and managing certain patient groups who arrive to the ED via ambulance. Focused care pathways for these patients likely to be admitted, but spending lengthy times in the ED, may be required. Although conducted before NEAT, our findings (higher admission rate in patients with prolonged ED LOS and lower mortality rate) provide important information regarding the use of time-related targets.

The authors declare they have no competing interests.