Journal of the New Zealand Medical Association, 11-September-2009, Vol 122 No 1302
Emergency Department utilisation: a natural experiment
Cecelia Rademeyer, Peter Jones, Stuart Dalziel, Garry Clearwater, Bernard Foley, Mazin Ghafel
Overcrowding in Emergency Departments (EDs) has become an international phenomenon1–5 and has been hotly debated in the medical literature. Overcrowding has been associated with increased morbidity and mortality, raising serious concerns about quality of care.6–9 This problem has also attracted media attention in New Zealand, especially for North Shore Hospital, with newspaper reports of long waits for patients in corridors.10–14
Causes identified internationally include a reduction in the number of EDs and available ED beds, as well as an increase in patient visits.15,16 Attempts to reduce overcrowding have included a number of strategies from central government funding incentives, addressing assessment and treatment times, to increasing ED beds. Despite increase ED bed numbers being suggested as an answer to this problem, little research has focused on the effect of increasing the number of available ED beds within a geographic region on ED utilisation.
In February 2005 a new hospital-based ED was commissioned at Waitakere Hospital (WH) in Henderson, West Auckland. Part of the rationale for opening this new facility was the expectation that it would reduce attendances, and thus overcrowding, in neighbouring EDs.
As the regional population is geographically well defined, the opening of the new facility presented an opportunity to study the impact that a new ED had on patient presentations to neighboring EDs.
Auckland City is located in the top third of the North Island of New Zealand and is bordered by two natural harbours, the Waitemata and the Manukau. These harbours divide the city into northern and southern parts, with only two highways and one rail link connecting them across a narrow isthmus, 800m wide (Figure 1).
Figure 1. Auckland geography: the Auckland isthmus and hospital locations
NSH=North Shore Hospital, ACH=Auckland City Hospital, SSH=Starship Children’s Hospital, WH=Waitemata Hospital, MMH=Middlemore Hospital. Black Arrow=Isthmus.
The area had a population of 1,231,500 in the 2001 Census, and 1,387,780 during the 2006 Census.17 At the northern and southern boundaries of the Greater Auckland area the population density is low and divided from the neighbouring areas (Northland and Waikato respectively) by natural geographic barriers (Figure 2).
Figure 2. Auckland geography: District Health Boards (DHBs)
Source: Sector Accountability & Funding Directorate, Ministry of Health.
Health services within Auckland are provided by three District Health Boards (DHBs): Auckland District Health Board (ADHB) and Waitemata District Health Board (WDHB) which respectively serve the central and northern part of the city, and Counties Manukau District Health Board (CMDHB) which serves the southern part (Table 1).
Table 1. Hospital-based services within the Greater Auckland region
Prior to February 2005, the population north of Auckland City was served by one ED at North Shore Hospital (NSH). In response to projected population growth and overcrowding at NSH ED a new ED was commissioned at Waitakere Hospital (WH). It was postulated prior to the opening of WH ED that this new ED would result in a decrease in presentations at NSH and ADHB hospitals (Auckland City Hospital (ACH) and Starship Children’s Hospital (SSH)).
Ambulance patients from West Auckland would now be taken directly to WH. Self-presenting patients were expected to present to WH rather than NSH or ACH/SSH as this was closer. This study was designed to test that hypothesis.
As Middlemore Hospital (MMH) in the south of the region is geographically separated from the other hospitals (Figure 2) the opening of the new ED was not expected to affect presentations to MMH ED. Furthermore, because of its location at the northern boundary of its catchment area, CMDHB patients usually present to the MMH ED rather than others north of the narrow isthmus. MMH therefore served as a natural control for the study.
Study design—A retrospective analysis of all ED attendances obtained from the electronic databases of the EDs within the Auckland region was undertaken over a 49-month period from February 2003 to February 2007. Clerical staff record patient attendance electronically at the time of first presentation, to be used for funding and audit of hospital performance. The data is believed to be accurate.
The study is an observation of an intervention (at time point N), the opening of the new WH ED. Data obtained included all ED attendances to the four existing EDs for two 12-month periods (Year N-1 and N-2) prior to the opening of WH ED (N), and to all five hospitals for two 12-month periods (N+1 and N+2) after its opening. Data are presented by hospital and by DHB. See below.
The month of February 2005 was excluded from analysis a priori as this was the month the new WH ED opened.
Statistics—The effect of the opening of WH ED on ED presentations was explored using control charts (also known as Shewhart charts or 'process-behaviour charts').18 The control charts were generated using SPSS v14 software (SPPS Inc, Chicago, USA). A control chart is a tool used to study how a process changes over time. It helps distinguish between variation in a process resulting from common causes (i.e. natural/non-significant variation) and variation resulting from special causes (significant variation).
A special cause is anything which leads to an observation beyond a control limit. A control chart presents a graphic display of process stability or instability over time. Data are plotted in time order. A control chart has a central line for the mean, an upper line for the upper control limit and a lower line for the lower control limit. The control limits are commonly set at three standard deviations (SD) from the mean (which corresponds to a false alarm rate of 0.27%). By comparing the plotted data to the mean and control lines conclusions can be drawn about whether the variation in data, and hence the process, is within predictable limits or is unpredictable, in other words affected by special causes of variation.
Special causes can be identified by the following signs:
Control charts can also be used to see if an event occurring at a given time point has a significant impact on the process. For our study data ED attendance (the process) was analysed as monthly data points with control lines determined at three standard deviations from the mean. Eight or more data points in a row on one side of the mean line were considered to be significant.
Population estimates for each DHB (as expected in June of each year) for the years 1996 to 2010 were obtained from Statistics New Zealand19 to see if there were any major population shifts in individual DHB populations that might explain changes in ED presentation numbers to the DHBs.
The population in the region increased a total of 9.4% over the study period (Table 2).
Table 2. Population per District Health Board
Source: Statistics New Zealand19. *Data for 2003, 2004, 2005 and 2007 is estimated. ADHB=Auckland District Health Board, WDHB=Waitemata District Health Board, CMDHB=Counties Manukau District Health Board.
Yearly presentations to each hospital and DHB over the study period are shown in Tables 3 and 4. There is a 74% increase in presentations to WDHB, with only minor increases in the other DHBs. This increase is mainly due to presentations to the new ED at WH.
Table 3. Presentations per study hospital over the study period
ACH=Auckland City Hospital, SSH=Starship Hospital, NSH=North Shore Hospital, WH=Waitakere Hospital, MMH=Middlemore Hospital.
Table 4. Presentations per DHB over study period
ADHB=Auckland District Health Board, WDHB=Waitemata District Health Board, CMDHB=Counties Manukau District Health Board.
The total population of the three DHBs is similar across the region (Table 2). Although WDHB served the largest population, it received only 22% of regional ED presentations in the first two study years (Years N-2 and N-1; (per capita ED attendance rate=0.09). However by the end of the study period ED presentations were more evenly distributed and a closer match to the population distribution, with per capita ED attendance rates of 0.20 at ADHB, 0.15 at WDHB and 0.17 at CMDHB.
Figure 3 is the control chart for ED presentations to NSH over time
NSH=North Shore Hospital
There is a grouping of points below the centre line for the period before June 2004. The increase after this point is explained by the fact that in June 2004 NSH started accepting Orthopaedic patients, coupled with a severe winter. There is no evidence of special cause variation following February 2005. Figure 4 shows the control chart for ED presentations to WDHB over time.
There is evidence of special cause variation after February 2005, with an increase in attendances. The number of presentations to WDHB increased 45% in the year after WH ED opened. Figure 5 shows the change in ED presentations to ADHB over time.
Figure 4. Change in Emergency Department presentations over time to Waitemata District Health Board
WDHB = Waitemata District Health Board (North Shore and Waitakere Hospitals).
Figure 5. Change in Emergency Department presentations over time to Auckland District Health Board
ADHB = Auckland District Health Board (Auckland City and Starship Hospitals)
Two points below the lower control limit in February 2003 reflects a time prior to merger of Obstetric and Cardiothoracic services onto a single hospital site in 2003. There is one point outside the upper control limit in August 2004, reflecting a severe winter. This is a high point on all control charts, reflecting a regional phenomenon. There is no evidence of special cause variation following February 2005. Figure 6 shows the change in ED presentations to CMDHB (MMH) over time (control group).
Figure 6. Change in ED presentations to the control hospital over time
CMDHB = Counties Manukau District Health Board, MMH=Middlemore hospital
The chart displays a clear stable trend of increasing ED presentations over the whole study period.
Recently Han20 reported that ED expansion resulted in increased ED attendance rate and length of stay. It was thought that opening a new hospital ED would help ease the load on neighbouring hospitals in Auckland. However, the opening of the WH ED did not result in reduced patient presentations to other EDs in the region. Instead, the numbers increased in line with population growth (Figures 2 and 4) for both nearby and distant hospitals (Figure 5). In contrast to this there was a marked increase in ED presentations to WDHB, the area in which WH is situated, disproportionate to population growth (45% in the first year and 74% over the first 2 years after the opening of the new ED).
The effect of increasing hospital beds has been known for sometime. In 1961 Roemer et al published a landmark study21 which found that increasing hospital bed availability was responsible for an increase in bed utilisation and length of stay in a community previously thought to be adequately served. Roemer suggested that doctors were the main drivers for the increased use of hospital beds as more beds became available. Our study is the first to report the effect of building a new ED on the patient attendances to neighbouring EDs. The results of this study suggest that patients may directly increase utilisation of a service, independent of the influence of doctors: the increased ED utilisation at WDHB was mainly from lower-acuity self-presentations. (This aspect is subject of a separate paper in publication).
A number of possibilities may explain our study findings. Firstly, the increase in numbers of patients seen at WDHB following the opening of the WH ED may reflect redistribution from the primary health care sector in that area. This behaviour may be motivated by cost: in New Zealand ED care is free while there is a variable part charge for primary care; or the perception that presentation to a hospital ED allows greater access to specialist services.
We sought, but were unable to obtain information concerning after-hours primary care presentations to the after-hours primary care centre near the new facility. It is relevant to note that EDs see only a small proportion of all primary (self-presenting) attendances in a community. The attendance rate in general practice averages 4 visits per capita per annum,22 whereas the average ED attendance rate (in this study) is less than 0.2 visits per capita per annum.
A relatively small redistribution of primary care visits to ED would barely register in all General Practice attendances but could have a significant impact on ED attendances. In this study, the per capita ED attendance rate at WDHB increased from a baseline of 0.09 visits per annum in 2003, to 0.15 visits per annum in 2007: this would barely register within the margin of error for all primary care attendances over the same period.
An alternative explanation is that the need for secondary emergency medical care for the population that lived closest to WH was not being addressed by the available EDs prior to the opening of WH ED. There is some evidence that this may have occurred in that the proportion of total ED presentations that have occurred in the regional DHBs following the opening of the new ED now more accurately reflects the distribution of the population in the region.
It is likely that the increase in total presentations within the WDHB, without a reduction in presentations to the nearby DHBs, reflects a combination of these possibilities. This finding has considerable implications for funding and resource allocation in the future.
As a retrospective observational study, our data may be subject to information bias. However as all EDs in the region use a national electronic database to establish a given individual’s unique hospital number, this is unlikely. The data extracted did not include patient demographics, address, presenting complaint, length of stay or final diagnosis and disposition, which may have informed the debate regarding the appropriateness of given presentations for primary care services.
Ideally, control charts would have been created for more than 2 years before the opening of WH ED in order to identify long-term trends in attendance rate. However in 2001 NSH underwent a major change in their recordkeeping system and it is not possible to accurately compare ED data before and after this change. No change in record keeping occurred in the three DHBs during the study timeframe. Therefore we do not believe the results can be explained by changes in data collection.
Given the large numbers involved in this study, any differences in numbers attending would be likely to reach statistical significance if subjected to traditional statistical tests of significance. We believe that in this situation, use of control charts is more appropriate to detect clinically important differences in the process of patient attendance across the region.
The control chart method determines whether there has been change at a given time-point not explained by natural variation or chance alone. However, it does not determine why that change has occurred. It is possible that the increase in ED presentations to the WDHB area at February 2005 is due to a factor other than the opening of the WH ED. As there were no similar sharp increases in presentations to either the adjacent Auckland, or geographically distinct CMDHBs, we believe that there was no change in regional disease morbidity to account for the increase in presentations.
During the study period there was also no change in provision of primary care facilities in the area. Information about attendances to General Practitioners may help explain where the additional patient visits were created from, however we were unable to access this information.
The opening of WH ED had no effect on the number of presentations to the other hospitals in the region. However it appears to have resulted in a marked increase of presentations to the new facility, thus increasing the total numbers of ED patients seen within the Auckland region.
Conflict of interest: None known.
Author information: Cecilia Rademeyer, Emergency Medicine Registrar, Department of Emergency Medicine, Auckland City Hospital, Auckland; Peter Jones, Director of Emergency Medicine Research, Specialist Emergency Physician, Department of Emergency Medicine, Auckland City Hospital, Auckland; Stuart Dalziel, Specialist Paediatric Emergency Medicine Physician, Children’s Emergency Department, Starship Children’s Hospital, Auckland; Garry Clearwater, Specialist Emergency Physician, Department of Emergency Medicine, North Shore Hospital, Auckland; Bernard Foley, Specialist Emergency Physician, Department of Emergency Medicine, Auckland City Hospital, Auckland; Mazin Ghafel, Public Health Physician/Analyst, Funding and Planning, Auckland District Health Board, Auckland
Correspondence: Peter Jones, Department of Emergency Medicine, Auckland City Hospital, Private Bag 92024, Auckland Mail Centre, Auckland 1142, New Zealand. Fax: +64 (0)9 6309799 (int) 4999; email: firstname.lastname@example.org
issue | Search journal |
Archived issues | Classifieds
| Hotline (free ads)
Subscribe | Contribute | Advertise | Contact Us | Copyright | Other Journals