Explaining variation in emergency admissions: a mixed methods study of emergency and urgent care systems

Aim

To identify system-wide factors explaining variation in potentially avoidable emergency admissions in different emergency and urgent care systems.

Objectives

1. Calculate the ‘age- and sex-adjusted potentially avoidable emergency admission rate’ for each emergency and urgent care system in England.

2. Explain variation in the ‘age- and sex-adjusted potentially avoidable emergency admission rate’ in different systems using routine data on population, health and system characteristics.

3. Undertake in-depth research in systems with high and low ‘age- and sex-adjusted potentially avoidable emergency admission rate’ to identify further factors which may be influencing variation.

4. Identify modifiable factors which affect potentially avoidable emergency admissions to help policy-makers, commissioners and service providers implement changes to reduce avoidable admissions.

Geographically based systems

Numbers of emergency admissions for the set of 14 conditions for each emergency and urgent care system were calculated using Hospital Episode Statistics (HES) for the 3 financial years April 2008 to March 2011. Admissions from all sources were included. The most common admission route was EDs, with an average of 69% of avoidable admissions through this route, although this varied by system between 44% and 92%.

The condition code for the first finished consultant episode was used. PCT mid-2009 resident populations were then used as the denominator to calculate the rate of potentially avoidable emergency admissions per 100,000 population. The directly age- and sex-standardised admission rates per 100,000 per year were calculated for each PCT for the 3-year period using seven age groups (0–4, 5–14, 15–44, 45–64, 65–74, 75–84, 85+ years), standardised to the whole population for England in 2009. A 3-year period was selected to ensure that the effect of annual variability in emergency admission rates at a system level was minimised. It is important to understand that this rate is an indicator of avoidable admissions. It is based on conditions rich in avoidability; that is, not all admissions from these conditions were avoidable in practice. For shorthand, we call this the SAAR even though it was a directly standardised rate.

Acute trust-based systems

Catchment populations for acute trusts were needed to calculate admission rates for the 14 conditions. There are different ways of calculating catchment populations for hospitals, with debates about which approach is best. 31,32 We used estimates of acute trust catchment populations for emergency admissions in 2009 calculated by Public Health Observatories in England. 33 These catchment areas were defined as the number of people in each sex and age group who live in the catchment of the acute trust. They were calculated using HES data between April 2006 and March 2009 to count the number of patients in each age and sex group admitted from small areas called Middle Super Output Areas. These areas have a minimum population of 5000, with an overall mean of 7200. Mid-year population estimates for 2009 were supplied by the Office for National Statistics (ONS). Within each 5-year age and sex group, the proportion of patients who went to each acute trust as a proportion of patients who used any acute trust was calculated. For each small area, this proportion was multiplied by the resident population in that age and sex group to give the small area catchment population for each acute trust. Then the small area catchment populations for each acute trust were summed to give the total catchment population for each acute trust.

We calculated the SAARs per 100,000 per year for each acute trust for the 3-year period April 2008 to March 2011 using seven age groups (0–4, 5–14, 15–44, 45–64, 65–74, 75–84, 85+ years) standardised to the whole population for England in 2009. A 3-year period was selected to ensure that the effect of annual variability in emergency admission rates was minimised.

Some specialist acute trusts offer care to specific age, sex or condition groups only. We wanted to compare similar types of acute trusts and focused on general acute trusts because they account for the majority of emergency admissions. 1 We included any acute trusts where the estimated population in each age and sex group was > 1000. This was an arbitrary cut-off point which successfully excluded children’s hospitals, women’s hospitals and condition-specific hospitals. It also excluded some general acute trusts located near children’s hospitals because they did not admit children.

Geographically based systems

There were 152 geographically based systems (based on PCT populations). The reliability of the HES data was checked by looking for consistency between numbers of emergency admissions per year within each system. The only large differences between years for any system were two systems with zero emergency admissions for 2010–11. We excluded them from the analysis, leaving 150 systems.

The median SAAR was 2258 [interquartile range (IQR) 1808–2662], with a 3.4-fold variation between systems ranging from 1268 to 4359, and a 1.9-fold variation between the 10th and 90th percentiles. Geographical variation in the SAAR was apparent, with highest rates clustering in the north-west of England, the north-east of England and east London (Figure 2).

FIGURE 2.

Variation in SAAR in 150 geographically based systems in England.

Acute trust-based systems

There were 131 acute trusts after exclusion of specialist hospitals. Two acute trusts had missing data for admissions in 2010/11 and were removed, leaving 129 acute trusts. The median SAAR was 1939 (IQR 1676–2331) per 100,000 population per year, with threefold variation between acute trusts ranging from 1194 to 3601, and a 1.8-fold variation between the 10th and 90th percentiles (Figure 3).

FIGURE 3.

Variation in SAAR in 129 acute trust-based systems in England.

Standardised avoidable admissions rate and all emergency admissions

There was a strong association between the SAAR (2008–11) and the directly standardised rate for all emergency admissions for geographically based systems (2009–10) (Figure 4). Systems with high SAARs tended to have high rates of emergency admissions (Pearson’s r = 0.88, p < 0.001).

FIGURE 4.

Relationship between SAAR and directly standardised rate of all emergency admissions rate by geographically based systems. DSR, directly standardised rate.

It is also worth noting that five of the conditions in our SAAR appeared in the top 10 diagnostic groups contributing to an increase in emergency admissions over recent years. 1

Standardised avoidable admissions rate and ambulatory emergency admissions

Ambulatory or primary care sensitive conditions are conditions for which hospital admission could be prevented by interventions in primary care. There are various definitions34 but the King’s Fund list is the most frequently used in England. 22 There is some overlap between these 19 ACSCs and our 14 conditions (indicated by * below):

• influenza and pneumonia

• other vaccine-preventable conditions

• asthma

• congestive heart failure

• diabetes complications*

• COPD*

• angina*

• iron-deficiency anaemia

• hypertension

• nutritional deficiencies

• dehydration and gastroenteritis

• pyelonephritis

• perforated/bleeding ulcer

• cellulitis*

• pelvic inflammatory disease

• ear, nose and throat infections

• dental conditions

• convulsions and epilepsy*

• gangrene.

Our SAAR is based on injuries as well as illness and includes mental as well as physical health problems. We could not locate ACSC rates for our systems but we did locate the rate of admissions with emergency ambulatory care conditions (EACCs) per population by PCT. 35 This is a directly age-, sex- and deprivation-standardised admission rate based on 49 emergency conditions with the potential to be managed on an ambulatory basis. The 49 conditions were developed by the National Institute for Improvement and Innovation. Figure 5 shows that there is correlation between our SAAR and the EACC (Pearson’s r = 0.34, p < 0.001) but that it is weaker than that between our SAAR and all emergency admissions.

FIGURE 5.

Relationship between SAAR and admissions with EACCs by geographically based systems.

Identification of factors

Based on previous research on variation in emergency admission rates, and the focus here on the emergency and urgent care system, it was necessary to locate factors relating to the population, geography, health and range of emergency and urgent health services for the 150 geographically based systems. We searched national databases of routinely available data for relevant data on factors reported at PCT level. Where factors were available by financial year, we selected 2009–10, which was the mid-point year of the SAAR, or the calendar year 2009, or summed quarterly data for 2009. The source of data for factors is reported in Appendix 1.

Analysis

We undertook general linear modelling in SPSS version 20 (IBM Corporation, Armonk, NY, USA) weighted for the size of the system population to account for larger uncertainty of estimates for smaller systems. The dependent variable was the SAAR for 150 systems. The independent variables were tested in a hierarchical multiple regression in two blocks, using forward stepwise regression within each block. Variables were included if the p-value for the t-test was < 0.05. The blocks were determined by the extent to which service providers could affect a factor. Block 1 was population-related factors. We used the residuals from block 1 as the dependent variable in block 2 and then tested the effect of service-related factors.

After we had completed the analysis and moved on to phase 2 of the study, we felt that this regression did not capture enough of the conceptual framework of the study. Although our conceptual framework led us to include factors which other researchers had not – that is those from services other than primary care and hospitals – we had ignored the fact that factors might operate in sequence. That is, services outside the hospital would affect numbers who turned up at the hospital. For example, difficulty getting a GP appointment has been shown to explain higher rates of attendance at EDs. 26,27 We undertook a three-block analysis to reflect this:

• block 1: population-related factors: population, geography, health

• block 2: out of hospital service-related factors – GP access, GP quality, ambulance service performance, GP OOH service

• block 3: in hospital service-related factors: ED attendance, conversion rate at ED, short length of stay in hospital.

Univariate linear regression

We display descriptions of the variables tested in the multiple regression (Table 4). It is interesting to note that two factors explained a very large amount of variation in the SAAR. Both were population factors related to deprivation: income deprivation and employment deprivation, explaining 60% and 72% of variation respectively. These are likely to be correlated. We explore correlation between factors in the next section.

Correlations between factors

Some of the factors were correlated with others. It is important to describe these correlations to facilitate interpretation of the final multiple regression. Employment deprivation was positively correlated with income deprivation and the prevalence of some of the 14 conditions (Table 5). Urban/rural status was positively correlated with a number of factors including the proportion of elderly people in a system and some access to general practice factors.

Factors affecting standardised avoidable admissions rate in a multiple linear regression

We then undertook multiple linear regression to identify the combination of factors which predicted the SAAR (Table 6). We undertook the block 1 analysis – which tested population-related factors – and identified two factors which predicted the SAAR: employment deprivation and urban/rural status. The more deprived populations and more urban populations had higher SAARs. These two factors together predicted 75% of the variation. In the block 2 analysis – which tested service-related factors – those related to EDs, hospitals, emergency ambulance services and primary care explained further variation (r² = 85% in a non-hierarchical analysis). Systems with higher SAARs had higher attendance rates at EDs, higher rates of conversion of ED attendances to admissions, higher proportions of very short-stay patients, higher rates of ambulance calls transported to hospital and better perceived access to general practice (see Table 6). This last factor was in the opposite direction from the univariate linear regression and counterintuitive in the context of the conceptual framework of the study. We discuss this in Chapter 10.

Regression metrics are reported in Appendix 2, showing that the model was a good fit. There was no evidence of multicollinearity in block 2: variance inflation factors varied between 1.1 and 1.4. Some of the factors in the multiple regression were correlated with each other even though there was no multicollinearity: employment deprivation was negatively correlated with the proportion of ambulance callers not transported to hospital; urban/rural status was correlated with perceptions of access to general practice and the proportion of ambulance callers not transported to hospital (see Table 5). Because of the large amount of correlation between factors, it is likely that a number of regressions would fit the data equally well, so there is some uncertainty about the model presented here.

Test of a stronger system-based theoretical model

The first block was the same as block 1 in the primary analysis reported above. In block 2 – out-of-hospital factors – ambulance performance was the only factor adding to the regression. That is, factors about availability of, access to and quality of primary care did not explain further variation at this stage. In block 3 – in-hospital factors – the factors explaining further variation were those in the two-block analysis reported earlier. That is, this approach made little difference to our findings.

Characteristics of acute trusts

We searched databases for routinely available data on the characteristics of acute trusts in England. Ideally we wanted to include characteristics of the acute trust catchment populations because these were important when explaining variation in the geographically based SAAR (see Chapter 4). The best variable we found was the deprivation level of the population in the geographical area in which the acute trust was located. We used the postcode of the ED of each acute trust and identified the percentage of households in poverty in the area around the postcode using the ONS data on Middle Super Output Areas. Where there were two or more EDs based at different geographical sites, we took the mean of these percentages. We found routine data on 11 acute trust characteristics including demand, supply, accessibility and management variables.

Analysis

We undertook linear regression in IBM SPSS Statistics version 20 weighted for acute trust catchment population to account for larger uncertainty of estimates for smaller populations. The dependent variable was the SAAR for 129 acute trusts. The independent variables were tested in a hierarchical multiple regression in two blocks, using forward stepwise regression within each block. Variables were included if the p-value for the t-test was < 0.05. These blocks were determined by the extent to which the factors tested were modifiable by acute trust managers. In block 1 we tested population-related factors: percentage of household poverty in the area in which the hospital was located and demand for the ED in terms of attendance rate. We then used the residuals from block 1 as the dependent variable in block 2 and tested the additional effect of acute trust characteristics.

Univariate linear regression

We display descriptions of the factors that were tested in the multiple regression in Table 8.

Correlation between factors

Use of the ED and poverty levels around the acute trust were correlated (Pearson’s r > 0.4). There was a negative correlation between use of the ED and the conversion rate (Pearson’s r > 0.4): as use of the ED increased, the percentage of attendances converted to admissions decreased.

Multiple regression

In block 1, only demand for the ED was statistically significant because of the correlation between this factor and poverty levels around the acute trust (Table 9). The residuals from block 1 were used as the dependent variable to then test the added effect of acute trust characteristics. Five factors explained some variation in block 2: supply determined by the number of acute beds per 1000 population, management or coding of admissions determined by percentage of admissions with length of stay < 1 day, management within the ED determined by the conversion rate from ED attendance to emergency admission, the management of the hospital determined by its foundation trust status, and access to planned care determined by percentage of people waiting less than 1 month for a hospital outpatient appointment. Regression metrics are reported in Appendix 2, showing that the model was a good fit. The variance inflation factor for block 2 was approximately 1 for each factor so there was no indication of collinearity. The final multiple regression explained 53% of the variation in the SAAR when all the factors in Table 9 were entered in a single block. Because of the correlation between some factors, it is likely that a number of regressions would fit the data equally well so there is some uncertainty about the model presented here.

Numbers of cases

There is a need to balance selecting a large enough number of case studies to allow patterns to be determined across cases and a small enough number to make full use of the strength of the case study approach to explore issues in depth. Yin points out that ‘A fatal flaw in doing case studies is to conceive of statistical generalisation as the method of generalising the results of the case study’ (p. 32). 39 Rather, analytic generalisation is important, where a theory is generated within one case study and replication – or not – of the theory is considered in the other case studies. We selected six case studies to offer a balance of pattern identification and depth of understanding.

Selection of case studies

The regression undertaken in phase 1 identified 27 candidate systems for in-depth case study, that is with SAARs which could not be fully explained by the factors in the regression reported in Table 5. We explained in Chapter 4 (section Link with phase 2) how we selected the case studies, whereby we purposively selected a mix of high, medium and low SAARs and over- and underprediction by the phase 1 regression. Early in the study the health-care commissioners in two selected cases did not want to participate in the study or did not respond to our invitation. At the time we felt that commissioners were an essential stakeholder because they helped us to define the services providing care to the population of the system. Therefore, we replaced each system with another system within the same geographical cluster. When this occurred again later in the study we kept the original systems because we realised that we could make use of the commissioner’s website to obtain the information we needed about service use by their population.

Data collection

Case studies can draw on six key sources of evidence: documents, records, interviews, direct observation, participant observation and physical artefacts. 39 Because of the complexity and size of each system, observation was difficult. Therefore, within each case we used three types of data: documents, routine data and interviews. The interviews were central to the case studies, with support from documents and routine data.

Operationalising the definition of a case

Each case was an emergency and urgent care system defined by the main services used by a population served by the health-care commissioning bodies known as PCTs at the time of the data collection (2011–13). Services included in a case could be within or outside the geographical boundaries of the geographically based systems in the phase 1 regression. We relied on local commissioners and their websites to identify key services to include in a case.

Context in which cases were operating

We were interested in 2008–11, for which the SAAR was calculated. We interviewed stakeholders in 2012 and 2013. As stated earlier, we asked interviewees to clarify when initiatives they discussed were in operation so we could relate these to the findings in the phase 1 regression.

The interviews were undertaken at a time of major NHS reorganisation. PCTs were in the process of being disbanded to make way for Clinical Commissioning Groups (CCGs). Patient representation structures were also changing, with Healthwatch introduced in April 2013, taking over the function of LINks. This made it difficult to identify some stakeholders for interview because they were in the process of leaving their job or had yet to be appointed.

The subject of emergency admissions and their reduction was high on the political and commissioning agenda. Tariffs were in operation nationally which reduced payments to service providers for emergency admissions occurring over the rate of previous years.

Analysis

Description of cases and interviewees

Overarching themes

As well as helping to identify differences between the six systems, the case studies identified general issues about emergency and urgent care systems and avoidable admissions. The cases were selected because they were more likely to identify issues we had not measured in the phase 1 regression. These cases could also identify issues that might be common to all systems. We report the overarching themes below before describing how these issues occurred within each case. We selected quotes which illustrated the issues we found, labelling them by type of stakeholder but not by case in order to protect the anonymity of our interviewees.

Why HU1 was selected: a really high standardised avoidable admissions rate

This case was part of a geographical cluster of two systems with very high SAARs which were underpredicted by our phase 1 regression (H = high, U = underpredicted). It is helpful to have HU1 in the case studies because:

• It is a system with a very high SAAR. At 3493 it was ranked at 147 out of the 150 systems. The other system in this geographical cluster also had one of the highest SAARs.

• It had an underpredicted SAAR in our regression. The predicted SAAR in our regression was 3172, ranked at 144 out of 150, so it was still one of the highest predicted SAARs. The other system in the cluster shared the underprediction and both had two of the worst ranked predicted SAARs.

Description of factors in the phase 1 regression

Its SAAR was ranked at 147 out of 150. It was ranked high for most factors in the regression in phase 1 (Table 13). It had low rankings for two factors: it was ‘other urban’ and had a low conversion rate.

Condition-specific standardised avoidable admissions rates: a system-wide problem

HU1 ranked 147 out of 150 overall for the SAAR and generally ranked highly for all conditions with the exception of ‘pyrexial child’ (Table 14). That is, the high avoidable admission rate was likely to be system-wide rather than related to some parts of the system specific to dealing with some conditions. There was some possibility that the system might operate differently for children and adults but the numbers of ‘pyrexial child’ admissions were small compared with other conditions and small numbers may account for the different ranking of this system. There was no obvious relationship between the SAAR and interviewees’ views of pathways for these conditions (see Table 14). In the documentary analysis, COPD and mental health were identified as priorities within the system, with initiatives aiming to improve COPD care in primary care and the early identification of mental health symptoms.

System configuration: a complicated system

The population of the system is approximately 300,000 people. The services used by the system population are shown in Figure 6, based on information provided during the interviews and the documentary analysis. The system is unusual in our six case studies because the two main acute trusts used by the population are located outside the geographical boundaries of the system. The system also has two local authorities and two GP OOH services. There are two walk-in centres provided by the community trust and one minor injury unit provided by an acute trust, two of which offer radiography.

FIGURE 6.

Emergency and urgent care system configuration for HU1. LA, local authority; MIU, minor injuries unit; WIC, walk-in centre.

Acute trusts: both high standardised avoidable admissions rates, particularly acute trust B

Emergency admissions from this population went to two acute trusts with EDs, both of which were outside the geographical boundaries of the system. In the documentary analysis we identified that one of these acute trusts opened a new hospital building with a new ED at the early part of the SAAR period. Interviewees mentioned this too. One acute trust (A) had a lower SAAR than the other (B), but both had the highest SAARs in our sample of 129 acute trusts (Table 15). The SAARs were similar to the geographically based SAAR of 3493. Acute trust B – with the highest SAAR – was used by the other system in this geographical cluster of two systems with high unexplained SAARs.

There were differences between the two acute trusts for the factors included in the phase 1 regression (see Table 15) in that acute trust A’s factors tended to be ranked lower than acute trust B’s. Nonetheless, both acute trusts had higher SAARs than predicted by the regression. Note in particular the conversion rate from ED to admission, which was one of the highest in the country. This did not tally with the conversion rate for the geographically based system in Table 13 which was of middling rank and calls into question the accuracy of the conversion rate for the geographically based systems given interviewees’ perceptions of a high rate. The low ranking for percentage of households in poverty identifies a problem with this factor in that it is based on the small geographical area in which the ED is located rather than the catchment population of the acute trust.

Population factors

Geographical factors: poorly located services affect demand

Interviewees described how the geographical location of services affected their inappropriate use. First, the area had two walk-in centres and one minor injury unit for a population of 300,000. However, interviewees noted that the distribution of services across the geography incentivised use of the EDs because people in some communities needed to drive past an ED to get to their nearest walk-in centre or tackle a geographical barrier of a bridge. Interviewees also discussed how the acute trusts were distant from many communities so if people travelled to them then they tended to stay or be admitted, especially from deprived communities with low car ownership. In phase 1 we did include a factor about the urban/rural status of a system but this factor did not address the location of services.

Incentives to keep admissions high

Ambulance service

In the phase 1 regression the ambulance service providing care to this system had the lowest non-conveyance rate in the country. The interviewees from the ambulance service were interested in non-conveyance and discussed the positive relationships they had with some health services. However, they were concerned about the lack of consistent provision of services, which limited their ability to keep people in their own homes or take them somewhere other than hospital. The ambulance service annual reports described missed response targets and plans to enable crews to make more use of community-based facilities such as district nursing and walk-in centres.

The emergency department of acute trust B struggled to deal with high demand

Out-of-hours provision was poor

Interviewees identified that admission avoidance services (delivered in the acute and community sectors), which were available 5 days a week, worked well but needed to be available 7 days a week and in the evening as well as daytime. For example, the ambulance service described how they could not contact a district nurse to come out at 02.00 to unblock a catheter when they could do this at 14.00 on weekdays. This was a common complaint made by a wide range of stakeholders about community, social and mental health services. This was despite both local authorities describing intermediate care services delivered 7 days a week. The community trust interviewees were aware of gaps in community provision and were willing to provide OOH services but identified commissioner unwillingness to fund this. Some interviewees expressed concerns about the lack of availability of GP OOH services locally and how this led to patients attending the EDs. The CCG wanted to place a GP in the ED OOH as a solution but had struggled to find GPs to do this.

Lack of access to intermediate beds

Interviewees discussed the need for community beds for patients who did not need a hospital bed but were admitted for social reasons. There are 30 intermediate care beds available in this area but these were seen as inaccessible because they took patients with only some health problems. In addition, it was highlighted that the geographical location of these beds served a distinct part of the population, and these beds were underutilised, with an interviewee describing them as ‘40% full’. Some interviewees described how these beds were seen as early discharge beds rather than admission avoidance, fitting in with the sense that admission avoidance was not a priority in the area.

There is limited capacity and the acute trusts tend to use the limited capacity for step down, because they would prefer to step down than avoid an admission, because of the step down they have generated an in-patient spell which comes with a finance package associated with it. Whereas a step up they avoid the patient being admitted in the first place.

You’ve got to be fitter than Linford Christie to get a bed in Intermediate Care.

It was a joke, because basically you could only be admitted if you were very well.

In the documentary analysis we identified that acute trust A had opened, and closed, an intermediate care facility with approximately 40 beds during the SAAR period. No mention was made of this during the interviews.

Difficulties accessing primary care

Fragmentation but determination to integrate

Interviewees described new initiatives, meetings, an urgent care board and a general willingness of service providers to work together. Progress was described, for example recently seeing social services and mental health support available onsite in the EDs. However, this progress was recent, sometimes focused on early discharge rather than admission avoidance, and described by some interviews as ‘add-on’ rather than integrated:

[. . .] the District Nursing team and the Rapid Response Team do a good job. [. . .] I think lots of the teams are working well. I think where the difficulty sits for them is they’re working well in isolation, and they need to start to look at it a little bit more holistically, and that’s kind of the area that we are now embracing really.

An enthusiastic social care sector but a lack of connection with health

Two local authorities served this system; staff were interviewed from both of these organisations. Both interviewees enthusiastically described their involvement in admission avoidance and highlighted some positive relationships between themselves and health care colleagues. In one local authority there was a feeling that social services were being more proactive in admission avoidance than their health colleagues:

I’m really struggling to think of a positive example that helps prevent unnecessary admissions that the council doesn’t lead . . . it’s ploughing through the constant reasons for not doing things because in health there’s always a reason why you shouldn’t do it.

Interviewees from social care and community services acknowledged that the relationship between these two sectors could be improved. A community interviewee perceived these barriers as driven by a lack of colocated working, and working with performance indicators which put health and social care at odds with each other.

Explanatory analysis

The interviews were highly supportive of the results of the phase 1 regression and appeared to explain some of the findings. None of the interviewees was surprised to hear that the system had a high avoidable admission rate. They raised issues about the deprived unhealthy population, the high attendance at the ED and the high conversion rate at the ED. The conversion rate was seen to be caused by consultant shortages and slow diagnostics. Another issue was coding, in that one acute trust needed to fund their new hospital, so it had a perceived incentive to code people as admissions if they were using alternatives to staying in the ED. The SAAR might be higher than predicted because of:

1. poor availability of services OOH

2. a needy sick population

3. acute trust coding of people in ‘holding areas’

4. geographical layout of services

5. lack of senior review in ED

6. lack of integration

7. a hospital-centric system with little access to community beds

8. a focus on early discharge, not admission avoidance.

Why this case was selected

It is helpful to have LO2 in the case studies because:

• It is a system with a very low SAAR. At 1572 it was ranked at ninth lowest out of the 150 systems.

• It had an overpredicted SAAR in our regression. The predicted SAAR in our regression was 1950, ranked at 43 out of 150, so it was still low.

• It is one of two systems in a geographical cluster and also one of six systems following a similar pattern.

Description of factors in the phase 1 regression

Its SAAR was ranked as ninth out of 150. Its ranking was low for many factors in the phase 1 regression but it had a high ranking for the conversion rate from ED attendance to admission, employment deprivation, short length of stay in hospital and access to GP (Table 16).

Condition-specific standardised avoidable admissions rates

LO2 ranked ninth out of 150 overall for the SAAR and generally ranked low for all conditions with the exception of blocked catheters, angina and falls (Table 17). That is, the low avoidable admission rate was likely to be system-wide rather than related to some parts of the system specific to dealing with some conditions. On the whole, positive interviewee perceptions of specific conditions correlated with the quantitative data: chest pain, cellulitis and UTI were ranked lower in this system than in other systems and management of these conditions was viewed positively by interviewees in various parts of the system. An exception was mental health, which was ranked low in comparison with other systems, but there were negative interviewee perceptions regarding the mental health pathway. In addition, a local public health report during the early part of the SAAR period highlighted the need to develop NHS mental health services in partnership with the voluntary sector and other providers to ensure an integrated service was provided in the community.

System configuration

The population is approximately 750,000 people across a very large geographical area. The services used by the system population are shown in Figure 7. The system has two acute trusts in different parts of the area, and a large number of community units indifferent localities. The ambulance service provides the GP OOH service.

FIGURE 7.

Emergency and urgent care system configuration for LO2. LA, local authority; MIU, minor injuries unit; WIC, walk-in centre.

A tale of two acute trusts

Emergency admissions from this population go to two acute trusts with EDs. One acute trust (B) has a considerably lower SAAR than the other trust (A). The population within which each trust is based is very different, with acute trust A located in a much poorer area than B. Acute trust B has one of the lowest SAARs in the country and neighbours the other system in this geographical cluster. Two-thirds of the population of the geographically based system go to this hospital, so it has a considerable effect on the SAAR. The conversion rate is much higher than expected given the SAAR for both of the acute trust-based systems.

Population: two very different areas within the system

In this geographical area there was a sense of two systems in operation, one focused around each acute trust. The catchment population around each trust was described as very different. One side of the area (west) was described by interviewees as having a population which needed or used emergency care more than others: elderly people, deprived communities and immigrant communities. The second area (east) was less well described by interviewees but was seen as having a stable community. These two very different populations were apparent in the acute trust data in Table 18, where the poverty levels around each hospital were very different, as were all the statistics for each trust.

Geography

A strong community service focus

Some interviewees noted a lack of acute beds in their area, although they were in the locality served by an acute trust with a relatively high rate of acute beds per 1000 population. Some interviewees said this was made up for by having lots of community beds available to them and others by the use of community matrons and home care packages:

We have good access to spot purchase beds, and we have access to social care beds, planning beds, rehabilitation beds, intermediate care, domiciliary home care. We have access to therapy, voluntary, mental health.

Proactive approach to admission avoidance

There was a contrast between this case and HU1 in that interviewees described a large, proactive and accessible set of services focused on admission avoidance. There was a rapid assessment team in acute trust A which worked within the ED proactively identifying people who could be discharged. This team was seen by four interviewees as having the single most positive impact on avoiding admissions. Community matrons and district nurses were highly valued by others in the system. District nurses were also highlighted as being ‘excellent’ during the OOH period. There were a dozen community matrons in one locality who were known by and respected by the other interviewees, who viewed them as key to avoiding admissions. Overall, there was a sense that they were proactive about admission avoidance:

as a Trust, admission avoidance has always been, in the 8 years I have been doing this team, it’s just been at the forefront all the time and I think that’s because we are a small, it’s a small acute trust, it’s very small, we have got few community beds so I think there has always been that keep-at-home ethos embedded in the Trust really.

Interviewees also described good community nurse-led services focused on specific conditions: falls service, COPD service, continence and palliative care at home.

Good-quality primary care

The view that LO2 provided good-quality GP care was described by interviewees across the system (acute trust B, ambulance service, GPs and Healthwatch). Interviewees alluded to the geography, population demographic and workforce stability within the area, which contributed to this. These factors were also highlighted in MU4, the only other system that stood out because interviewees discussed high-quality GP care.

The average quality of a GP in [area] is actually very good [. . .] the link-up between the patients and the GPs is still quite 18th century; you know where the patients they think about the GPs as their GPs, my GP, you know. It’s not, let’s say I have got certain old elderly population, they will not see anybody else besides me . . . there’s a good continuity of care.

One connection that appeared well established was that between GPs and paramedics. In the regression analysis, LO2 had high rates of ambulance non-conveyance and this relationship between GPs and the ambulance service was described as facilitating non-conveyance:

you know most GPs if you make a reasonable case to them [are happy with a referral]. And sometimes we don’t even if we’re phoning a GP surgery in hours quite often we don’t even speak to the doctor. Quite more often than not the receptionist will make an appointment for us on our say so . . . I think probably most GPs are used to getting phone calls from the ambulance service and it just becomes part of the way things are done.

A strong emphasis on admission avoidance, out-of-hours service provision and integration

Ambulance service could do more

In an ambulance service annual report the ambulance service regarded itself as having made a significant contribution to reducing hospital admission within this system, and indeed had this as a priority. Good performance in non-conveyance was evident in the quantitative data. Surprisingly, interviewees wanted the ambulance service to do more to keep people at home and away from EDs. Community services wanted a better relationship with the ambulance service and wanted ECPs to offer more widespread services. However, this seemed to be about wanting more, that is expansion, because there were examples given of the ambulance service making direct referrals to community services. There was a falls pathway to allow ambulance services to refer directly to community services, but the community felt more could be done around falls. There was a sense of connection between services and the ambulance service but frustration expressed that that more could be done.

Explanatory analysis

The effect of population and geographical make-up on the SAAR is very large (based on phase 1 regression). This area had low employment deprivation and large distances to travel to a hospital, which were likely to produce low avoidable admissions. This area had a high rate of non-conveyance of ambulance calls and this may fit with the large range of community-based services that the ambulance service could make use of both in and out of hours. The regression may have overpredicted the SAAR because of:

• integration between services in this system

• the emphasis on community services

• proactive approach to admission avoidance

• availability of services OOH.

Why this case was selected

It is helpful to have MO3 in the case studies because:

• It is a system with a medium SAAR. At 2481 it was ranked at 100 out of the 150 systems.

• It had an overpredicted SAAR in our regression. The predicted SAAR in our regression was 2710, ranked at 118 out of 150, making it high.

• It is one of two systems in a geographical cluster.

Description of factors in the phase 1 regression

Its SAAR was ranked as 100 out of 150. It had this ranking for most factors in the regression in phase 1 (Table 19). It had a low ranking for attendance at EDs.

Condition-specific standardised avoidable admissions rates

MO3 ranked 100th out of 150 overall for the SAAR and generally ranked in this medium to high area for most conditions with the exception of mental health, blocked catheters and pyrexial child, where it had relatively low admission rates (Table 20). That is, the medium to high avoidable admission rate was likely to be system-wide rather than related to some parts of the system specific to dealing with some conditions. There was some correlation between the rank of the condition and interviewee perceptions; for example blocked catheter, falls and hypoglycaemia, which ranked relatively low in this system (that is low avoidable admissions) and were highlighted by interviewees as having effective pathways. Documentary analysis also highlighted that the ambulance service had improved its links with the falls team and identified a reduction in elderly fallers in the middle part of the SAAR period. There were mixed perceptions of mental health, so it was surprising that this was the second lowest-ranked condition within this system.

System configuration

The population is approximately 500,000 people across a large geographical area. The services used by the system population are shown in Figure 8. A key difference between this system and the others so far (HU1 and LO2) is the number of UCCs. During 2008 the emergency and urgent care system closed a number of EDs and opened GP-led UCCs around the geographical area. It also introduced telephone access to urgent advice around the same time to direct people to appropriate levels of care and transport them to the UCCs.

FIGURE 8.

Emergency and urgent care system configuration for MO3. LA, local authority.

Acute trust

Emergency admissions from this population go to two hospitals with EDs, both of which are managed by a single acute trust. The data are available nationally only at trust level. The community and acute trust merged two years before our interviews, after the time period of the SAAR. The trust had a high ranking of number 102 of the 129 acute trusts included in our phase 1 regression and its SAAR was considerably higher than predicted in our regression (Table 21). This is the opposite of the situation for our geographically based system analysis. Few of the factors we included in the regression aligned with this high-ranked SAAR.

Population

The population was described by interviewees as diverse in terms of health, socioeconomic status and age. One subgroup described was a stoic rural population suffering from COPD caused by mining. This group did not use health services until their condition was really serious. A number of interviewees referred to the elderly population and the high number of nursing homes in the area, which they felt impacted on demand for emergency services.

Interviewees felt that demand for their services was extremely high, especially for the EDs, and that nothing they did seemed to impact on this. This was a very surprising view given the low rate of demand for the ED shown in Table 21 compared with other acute trusts in England. This may be due to time differences between our data in Table 21 (2009–10) and our interviews (2012–13), or it may be about capacity versus demand in the ED.

I think we put a lot of effort into stopping people coming into hospital and for all our effort it doesn’t seem to me that it bears much fruit.

Geography

The area is largely rural with one city and a number of smaller towns. The difficulties in accessing care in rural areas identified in rural LO2 were also highlighted for this system. However, interviewees felt that services were in the right place and in particular that the UCCs were based in the right localities. One of the EDs was in the town centre and therefore considered to be easily accessible.

Rapid assessment teams in the hospitals avoid admissions

Interviewees within and outside both hospitals spoke highly of the rapid assessment teams in each hospital. One team had been available for 4–5 years, and one for only 2 years, so only one was available in our SAAR timeframe. These teams were described as having senior doctor and nurse cover, fast access to diagnostics and some OOH availability in that they worked up until 20.00 5 days per week. These services were described as high-volume, seeing 140 patients in a busy week. Interviewees felt that there was a need to extend these services more into weekends and have the diagnostics support available OOH. Interviewees viewed these services as successful at avoiding admissions, describing them as accessible and responsive and turning people round the same day:

rapid response people are really good and what they do is they come and they make an assessment, decide that that person can go home with x, y and z and then they just organise the x, y and z. They sort it all out, they book the transport, they’re brilliant, they’re absolutely brilliant.

A crowded emergency department

Interviewees across the system described the EDs as busy, with many highlighting one of hospitals as being particularly busy:

they [the hospitals] are overwhelmed and often closing their doors saying they’ve got no beds left . . . and A&E, always look busy and bursting at the seams, in [named hospital] in particular.

Interviewees described how they had put a lot of effort in recently to sort out problems with patients backed up in their ED by setting up systems to streamline minor cases to nurse practitioners and the UCC which was nearby. The really busy hospital had also recently implemented a senior review model for all patients to deal with the fact that junior doctors were making decisions to admit and were requesting lots of tests to support their decisions. Even though the statistics show that attendance at the ED was relatively low compared with other parts of England (see Table 21), interviewees working in the ED described it as normal to have ambulances queuing up outside. These new schemes were established after our SAAR time period and it may be that the need to find solutions in recent years indicates problems occurring during our SAAR time frame.

Primary care

The UCCs are GP led, with some accessible 24 hours a day, so this system appeared to have high levels of access to GPs. This was not apparent in the quantitative data in Table 19, where patient-perceived access was average. GPs appeared to have good connections with other providers across the system. Community nurses were aligned to GP practices (rather than being centralised), which contributed to relationship building, GPs were active in providing weekly education for community nurses, district nurses described ‘open-door’ access to GPs, and paramedics described telephone contact with GPs regarding patients.

A range of community services

Integration is partial and informal

The main formal integration described was the merger between the acute trust and the community trust, which occurred 2 years before our interviews (after our SAAR time frame). This had offered an opportunity to manage the ED and UCCs together, with an emphasis on taking the pressure off the ED. It also raised the issue for hospital-based staff that they needed to know what was in the community so they could refer to these services. There was also some colocation of services in the EDs which facilitated integrative working. However, much of the integration described was more informal than that, relying on relationships between individuals and services. Indeed some of the interviewees felt that there were many opportunities to talk to other services, which they appreciated, but that actions were needed too. Documentary analysis highlighted much discussion regarding integration. Acute trust annual reports, during the SAAR period and beyond, iterated a desire to maximise integrated pathways and care. Such integration was seen as desirable in improving access to unscheduled care but was also seen as a factor in reducing avoidable admissions.

Explanatory analysis

Possible explanations for the SAAR being lower than predicted:

• use of UCCs as a central part of the emergency and urgent care system; however, the low use of the EDs – which may be due to the high use of these centres – is already in the regression

• rapid response team

• the integrated health and social care teams in the community

• the difference between the acute trust and the geographically based system.

Why this case was selected

It is helpful to have MU4 in the case studies because:

• It is a system with a medium SAAR. At 1905 it was ranked at 48 out of the 150 systems.

• It had an underpredicted SAAR in our regression. The predicted SAAR in our regression was 1690, ranked at 22 out of 150, making it a low SAAR.

• It is one of three systems in a geographical cluster.

Description of factors in the phase 1 regression

Its SAAR was ranked at 48 out of 150. The population factors were also ranked low but the ED factors were not (Table 22). This system had very high demand for EDs and a very low conversion rate to admission. It has the highest perceived access to GPs in 48 hours. At the beginning of the SAAR period (2008–9), documentary analysis identified that the PCT set out a 5-year framework aimed at reducing avoidable emergency admissions, and reported in the same year that reductions in hospital admissions had been achieved by the commissioning of new services, better social care and GPs working with community hospital matrons.

Condition-specific standardised avoidable admissions rates

MU4 ranked 48th out of 150 overall for the SAAR and generally ranked in this medium to low area for most conditions with the exception of DVT and abdominal pain, where it had relatively high admission rates (Table 23). That is, the medium to low avoidable admission rate was likely to be system-wide rather than related to some parts of the system specific to dealing with some conditions. There was extensive discussion of condition specific pathways in this system. On the whole, pathways for the lower-ranked conditions (i.e. COPD, hypoglycaemia and UTI) were viewed positively by interviewees.

Although our interviewees did not mention it, the documentary analysis identified investment (just prior to the SAAR period) in mental health services to reduce admissions. During the same period the ambulance service also reported the introduction of 24/7 access to advice from mental health specialists. This fitted with the relatively low SAAR for mental health but did not tally with ambulance service interviewee perceptions of difficulties in accessing mental health services.

System configuration

The population is approximately 500,000 people across a large geographical area. The system has two acute trusts (Figure 9). There are a large number of minor injury units and community bed units within this system. The ambulance service provides the GP OOH service.

FIGURE 9.

Emergency and urgent care system configuration for MU4. LA, local authority; MIU, minor injuries unit; WIC, walk-in centre.

Acute trusts

Emergency admissions from this population go to two acute trusts with EDs within the system boundary. A further two acute trust in other areas were mentioned during the interviews, which a minority of the population was described as attending. The two local acute trusts are based in two towns of equal size, accounting for only 20% of the population of the system. That is, a large part of the population lives at a distance from the acute trusts.

The two acute trusts had a similar SAAR, which was similar to the geographically based SAAR. However, the two hospitals acted quite differently in the regression. The regression underpredicted acute trust A’s SAAR, similar to the geographically based SAAR, but overpredicted acute trust B’s SAAR.

This acute trust-level data did appear to relate to the geographically based system data (Table 24). In the geographically based system regression the use of EDs was very high (see Table 22) but there was no sign of this for the acute trusts (Table 24). This may indicate incorrect data about EDs in the geographically based system regression, as suspected for some of our other case studies, or it may be related to the large number of minor injury units (which are classed as type 3 EDs in routine data), which may have been included in some routine data sources and not others. This lack of consistency was also apparent for the conversion rate in the ED. The conversion rate was very high for each acute trust.

Population

Geography

This is a very rural area and its rurality was seen as causing difficulties similar to those raised in the other rural cases of LO2 and MO3. Interviewees felt that the population living outside the two major towns had some distance to travel to an acute trust. However, the distribution of minor injury units throughout the system led one interviewee to describe how the population was not further than a 30-minute drive from a unit or an ED. Interviewees discussed the need to balance putting care close to people and not so close that it generated demand. Similarly to HU1, the large distance to the hospitals was seen as causing increases in avoidable admissions because once people got to the ED they could not get home so they were admitted. That is, they raised a reason working in the opposite direction from another reason for distance to ED affecting admission rates: people living close to EDs being more likely to attend and perhaps then more likely to be admitted. The large distances created a need to find ways of providing care over a large geography such as using ECPs in the ambulance and GP OOH services.

Good chronic obstructive pulmonary disease care

A number of interviewees discussed the range of initiatives that prevented and avoided COPD admissions. They offered this information without prompting and also in response to us pointing out a low SAAR for COPD. They described their outreach hospital team, a good respiratory network, a good pathway and the use of telehealth. This did not seem as if interviewees were trying to find a reason for a low SAAR, because when we mentioned the low UTI admission rate they could think of no reason for it. The documentary analysis showed emphasis on avoiding emergency admissions for long-term conditions.

Emergency departments

Some issues that arose in other cases also arose here around lack of space within the EDs and the need for more consultants. Some services were available within the EDs such as psychiatric liaison and professions allied to medicine offering assessments. Interviewees were keen on having GPs in the EDs to discharge attendances and had tried a scheme that they felt had created demand because it was not integrated within the ED. Now they had a new approach and felt positive about it. For all the similarities with other cases, there was no sense of crisis here; in fact one interviewee described how ambulance queues were less common here so ambulances were more likely to take patients to these EDs. There was also a sense that, while interviewees acknowledged that there had been issues within EDs, they were very much moving forward, identifying problems and implementing solutions, including extending the ED in acute trust A. Nothing interviewees said explained the high demand for EDs in our geographically based system quantitative data in Table 22. One of the EDs (acute trust B) was described as having had problems with consultant presence at the time of our SAAR, although nothing in Table 24 supported this as a problem affecting avoidable admissions.

Confusion about assessment versus admission

Interviewees described confusion about the role of the medical assessment unit in acute trust A because it admitted people to its assessment beds. Scenarios were described whereby social care packages for patients attending the unit were cancelled by family or social care because there was an assumption that the patient was taken to the unit to be ‘admitted’ rather than ‘assessed’. This arose only within MU4.

The medical assessment unit haven’t really got a clear identity, because we call it a medical assessment unit but really it’s a ward and it’s got beds. The staff use a ward routine and I think their mentality to a certain extent isn’t an assessment unit, it’s an admissions unit. So some people call it an assessment unit, some people call it admissions unit so I think in some ways it needs some work, you know, because it’s only, we’ve only got 20, 22 beds, but we know that we get between either 40 and 60 admissions a day.

Out of hours is problematic

There was a general feeling among a range of service providers in this system that there were problems with OOH provision: ‘some of the services we access during the day aren’t available at night either’. Some services such as the minor injury units were described as providing a 7 days a week service but interviewees felt that too many services stopped running at 17.00 and later provision of support services was needed. Indeed, of the eight minor injury units, one was operational overnight, with none offering X-ray facilities beyond 17.00. While overnight opening of another unit had been trialled, it was acknowledged by others in the system that this was not financially viable in a rural community. This system had an urgent referral telephone service open to any health-care professional to facilitate the referral of patients to community hospitals, nursing/residential care and minor injury units. It was operational 7 days a week but was not open overnight (which interviewees wanted):

it [telephone referral service] suddenly stops at half past seven and you’re left on your own again and you have to try and sometimes make do and mend it until the next day . . . I’d love it if they kept [the referral service] open overnight . . . I think would make our job easier in my mind . . . having somebody to refer to on the path when you’ve finished your shift would certainly help in my mind and I would certainly feel more confident about leaving people at home. The referral service should be open 24 hours.

A well-connected ambulance service focusing on non-conveyance

A number of interviewees described working closely with the ambulance service. The service was described as treating people at home and had a directory of services to allow them to refer and transport to services other than the ED. Some of this was attributed to the rural nature of the system:

In the rural area you tend to spend a little bit more time seeing if the GP could manage that . . . which probably costs more resources, [but is] ultimately better for the patients. But they tend to happen in the more rural areas where it costs more to transport.

Indeed the phase 1 regression in Table 22 shows that this service had high non-conveyance rates compared with its counterparts. The documentary analysis supports this. Ambulance service annual reports throughout the SAAR period highlighted the achievements made in non-conveyance.

Primary care

A lot of community services

Social care: a hindrance to admission avoidance?

While there were some positive comments regarding social care, these were offered by the interviewee from the local authority. On the whole, negative views regarding social care were offered from the acute setting. As in HU1, there were references to the difference in the speed of decision-making within social care from health, and different organisational cultures. Interviewees in both acute settings described scenarios whereby they felt that social care potentially hindered admission avoidance:

my concern would be once you start to get social services and that sort of people involved it can delay discharge sometimes rather than necessarily speed things up . . . so I’m not convinced that having social workers in the [emergency] department necessarily would add great benefit to our patients . . . I think they tend to work on a slightly different timescale too, to the way we do. You know, they haven’t worked in an environment where they have had to make those rapid decisions as much as we do.

More integration needed

There were mixed views about how much integration occurred in this system. Here, it seemed patchy with plans to improve it.

I think we’ve got lots and lots of work going on but sometimes it’s not very joined up.

We’re just creating an urgent and emergency care strategy in [system] the idea of which is to pull all the organisations together so we do have a shared vision and purpose for the whole community.

Coding of admissions

Some interviewees described how attendances were coded as admissions once they left the ED and moved to other parts of the hospital, such as the clinical decision unit:

the system doesn’t allow us to code as an ambulatory care patient or whatever. They get admitted. And then that causes us a problem particularly if we bring them back the next day because then they are classed as a readmission.

Explanatory analysis

This case was very similar to LO2 in terms of the population and geography yet MU4 had a much higher SAAR (MU4 = 1900 versus LO2 = 1600). Possible reasons for a higher SAAR than predicted are:

• a lack of proactiveness of the rapid response team seen in LO2

• the lack of integration described in LO2

• the perception of poor OOH provision, particularly GP OOH service.

Why this case was selected

It is helpful to have LO5 in the case studies because:

• It is a system with a low SAAR. At 1776 it was ranked at 33 out of the 150 systems.

• It had an overpredicted SAAR in our regression. The predicted SAAR in our regression was 2300, ranked at 82 out of 150, so it was still medium but considerably higher than the actual SAAR.

• It is one of two systems in a geographical cluster.

Description of factors in the phase 1 regression

Its SAAR was ranked as 33 out of 150. It had very high ranking for most factors, which was not in line with the low SAAR (Table 25).

Condition-specific standardised avoidable admissions rates

LO5 ranked 33rd out of 150 overall for the SAAR and generally ranked in this low area for most conditions with the exception of blocked catheters and pyrexial child, where it had relatively high admission rates (Table 26). That is, the low avoidable admission rate was likely to be system-wide rather than related to some parts of the system only. Pathways in acute trust A for the 14 SAAR conditions were generally viewed positively within the interviews. COPD and falls were low-ranking conditions where perceptions were positive across the emergency and urgent care system. Documentary analysis indicated that the ambulance service had been working with GPs regarding the management of falls during the SAAR period, and highlighted successes in referring more falls patients to GPs following this.

A number of interviewees discussed the range of initiatives that prevented and avoided COPD admissions. They offered this information without prompting and also in response to us pointing out a low SAAR for COPD. They described a valued outreach hospital team providing hospital at home, good specialist community nurses and a good respiratory network which linked acute and community care. These initiatives were in place during the time period of the SAAR.

System configuration

The system has a population of 350,000 within a metropolitan area. The services used by the system population are shown in Figure 10. One acute trust is within the geographical boundaries of the system and the other is in another system. The community trust was contracted to deliver community services in this system towards the later part of the SAAR period.

FIGURE 10.

Emergency and urgent care system configuration for LO5. LA, local authority; WIC, walk-in centre.

Acute trusts: another tale of two trusts

The population uses two acute trusts mainly but has access to a range of trusts because it is situated in a large city with many hospitals. The catchment populations of the two main trusts are 300,000 (acute trust A) and 450,000 (acute trust B). These acute trusts are used by populations other than our system population but B is located within the system boundaries and interviewees reported that two-thirds of the system population used it. Both acute trusts had low SAARs, but one had a very low SAAR (acute trust A). Both SAARs were overpredicted by the phase 1 regression. They showed a similar pattern for factors to that of the geographically based analysis, with higher ED attendance and conversion rates than expected given their SAARs. Acute trust B was based in an area with higher poverty levels and had fewer acute beds than acute trust A (Table 27).

During the interviews, acute trust A was described in much more positive terms than acute trust B for admission avoidance. It was described as proactive, engaging with community services and offering senior decision-making in the ED. This was supported by the acute trust A annual reports, which used phrases such as ‘integrated working’, ‘working together’ and ‘productive relationships with other organisations’. Acute trust B adopted services established in acute trust A after a number of years and not in the time frame of our SAAR. This stark difference described by interviewees was not fully apparent in Table 27, where we might expect to see much lower conversion rates in acute trust A than B. However, the difference was extreme for the percentage of short-stay admissions, with acute trust B having the highest percentage in the country.

Population

Primary care

Poor services: out-of-hours provision and social services

Many of the interviewees described how services which were good for admission avoidance did not operate at weekends or overnight. These included overnight cover at walk-in centres and the community rapid response teams. Community beds were theoretically available 24/7 but described as difficult to access OOH. There was a vast difference in the provision of OOH described by interviewees for the two acute trusts. Acute trust A (with a very low SAAR) operated 7-day access to admission avoidance schemes and urgent radiology. In contrast, acute trust B did not describe 7-day working practices and vocalised more negative views about the OOH period:

I think not having OT, physio, social worker input particularly out of hours is a big problem . . . the last referral for OTPT is 4 o’clock which is not adequate, it’s not good enough.

Interviewees from acute trusts A and B, and the community trust, offered negative views about the local social care sector. The service was described as being under pressure and criticisms focused on the ‘long-winded’ process to get care, and the lack of staff available at weekends to discharge patients from the ED:

Well there’s a lot of bad feeling with the social services because social services don’t provide a good service. Whatever the reason that is we don’t feel they provide a good service so in general you have to speak to a social worker out of hours for example you ring them up and you have to leave a message with somebody at the civic centre and then you might get a call back in 2 hours and it may be that you’ve got a patient that is being discharged medically fit but they haven’t . . . their house is in a state or the door’s been broken into because the neighbours broke the door down because they were worried about them. You can’t do anything about it, you can’t get anybody to help you at that time . . . It’s not responsive I mean whether it’s because of how many staff they’ve got I don’t know.

Social care themselves acknowledged that, historically, integration between themselves and health had not been strong but the impetus towards making these connections was under way.

Evidence of integration for one acute trust only

Explanatory analysis

This case was unusual in that it was described not only as affluent – and this was evident from its low employment deprivation level in the phase 1 regression – but also as having minority ethnic populations making use of private health care. Reasons why the system had a lower SAAR than predicted are:

• One acute trust had proactive rapid assessment teams.

• Use of private care was described as high.

Why this case was selected

This case was geographically close to a set of systems with high SAARs but had a medium SAAR of 2421, ranked at 93 out of the 150 systems. The predicted SAAR was 2748, ranked at 123 out of 150. It offered some useful contrasts with some of our other six cases. It was covered by the same ambulance service as HU1 but had a different SAAR. It had a similar SAAR to MO3, offering the opportunity to compare and contrast two systems that appeared to perform the same.

Description of factors in the phase 1 regression

Its SAAR was ranked as 93 out of 150 but it had a low conversion rate from ED attendance to admission (Table 28).

Condition-specific standardised avoidable admissions rates

MO6 ranked 93 out of 150 overall for the SAAR but had very low ranks for some conditions (DVT, cellulitis, blocked catheters, UTI, falls) and high ranks for others (angina, mental health, abdominal pain, pyrexial child). The low-ranking conditions were issues dealt with by community services (Table 29). Interviewees had positive perceptions of community pathways.

System configuration

The system has a population of 265,000, most of which is served by a single acute trust. The services used by the system population are shown in Figure 11. It is a very simple system with one acute trust. During the SAAR period this system had an NHS walk-in centre, which has since closed.

FIGURE 11.

Emergency and urgent care system configuration for MO6. LA, local authority.

Acute trusts: low conversion rate

The acute trust SAAR was ranked 94th out of 129 trusts, with a very similar SAAR to the geographically based system (Table 30). The SAAR was underpredicted in the phase 1 regression in contrast to the overprediction for the geographically based system. The conversion rate from ED attendance to admission was very low for the acute trust, that is similar to the geographically based system.

Population

Interviewees described a deprived population with high prevalence of health problems and expectations of quick access to urgent care, as well as stoic older people. The high deprivation levels could be seen in the quantitative data presented in Table 28. One would expect a much higher SAAR given these deprivation levels.

Coding of admissions

An issue that stood out in this system compared with the other six was the way in which attendances at the ED could be placed temporarily, including overnight, in units where they did not count as admissions. There were a number of these units, including the clinical decision unit. The most long-standing of these units was for PCT residents only and designed specifically for admission avoidance. This offered an explanation for the low conversion rate from ED to admission.

And they specifically set up [the unit] for admission avoidance. Now they are admitted in the traditional sense, they are on a ward, but [. . .] not on the mainstream hospital, not in the investigation part, they’re there purely for the purpose of getting them back where they came from.

Proactive management of avoidable admissions

Interviewees consistently described services, units and teams in the hospital and community that GPs and others could contact to avoid an admission. Some of the services established, which interviewees believed reduced avoidable admissions, were GP access to medical and surgical advice, a senior medical physician proactive within the ED, a GP admissions unit to which GPs could refer directly, a hospital-based unit established for residents of the area only and a rapid response team to which GPs referred. The community rapid response team, made up of nurses, OTs, physiotherapists and social care, with access to equipment, worked in a highly responsive mode to offer short-term care to keep people in their own home and was perceived as avoiding ‘an awful lot of admissions’. It had been established in some form for many years, and dealt with frail elderly, cellulitis and UTIs – conditions which had low SAARs in this system (see Table 29). Some of these services had been set up in the previous 2 years and would not have affected the SAAR, but one unit was a long-established unit with beds, which was highlighted as an asset in a public health annual report and highly valued by interviewees within the system:

one of our biggest assets is [unit], a unit which was set up specifically to admit people for a few hours and then return them to their own homes. So admission in the very loosest sense of the word because they’re there to be seen by social workers and by physiotherapists and OTs and then returned to their own homes. Generally we’ll send them from the ED there so if they’re waiting for transport, waiting for relatives, they’ve got outstanding things that need doing, [. . .] which avoids about 14 admissions a day.

Interviewees described not access to community beds but rather a focus on helping people to return to their own homes from the ED. The number of initiatives described was large and indeed one interviewee commented that this issue needed to be addressed on a number of fronts: ‘it’s multi-faceted [. . .] I don’t think there’s one thing because [unit 1] works, [unit 2] works, some of the internal pathways we’ve got work.’

A busy emergency department

The ED was described as busy by both its staff and the ambulance service interviewees. Interviewees in all six of our systems described busy EDs and increasing demand but the language used to describe this department was more emotive, suggesting an extreme situation. The quote below is from an interviewee from the ambulance service who was likely to have experienced a number of other EDs:

it’s really busy . . . always at full capacity in terms of the major areas/bays . . . I think that the department is not big enough [for the number of patients attending] . . . there’s quite a lot of stress within the staff. I don’t know how they’ve stayed there for so long in comparison to other A&E departments, with the work load or with the way it’s managed and managed to stay sane [. . .] it is a good department, it’s just basically at breaking point! . . . [it’s been overloaded] for a long time.

There was some quantitative evidence of high demand for this ED because it had one of the highest attendance rates for an acute trust catchment area (see Table 30). It did not have a particularly high attendance rate for its geographically based system (see Table 28). There was some explanation for this within the interviews in that some of the demand was seen to be exacerbated by the location of the acute trust (on the geographical periphery of the system) and the good transport links around it, making it a popular destination for ambulance crews. More than one interviewee described that one-fifth of ED attendances came from populations residing in neighbouring systems, impacting on the ability to discharge patients from the ED. The low conversion rate seems all the more surprising given this context.

Good out-of-hours provision but wanting more

The ED in this system was unique among our six case studies because it had access to a hospital-based 24/7 admission avoidance service which was described as operating consistently over the 24-hour period. The ED also utilised GPs in the department 7 days a week, which was viewed positively by interviewees. Even so, ED interviewees wanted more access to OOH provision for community services.

A financially challenged acute trust

The positively viewed units and initiatives described earlier were based in the acute trust. Yet we had initially had problems with access to interviewees in this system because the acute trust was receiving adverse media attention. The interviewee below describes it as ‘a failing trust’, yet there did not seem to be evidence of this in our routine data or interviews.

You’ve hit us at a very good time you see, because as you know [we have] been a failing trust until recently. We’ve got Monitor with us all the time. Because we’ve failed financially and we’ve failed all our A&E targets last year, so we’re on red, both for our clinical targets and for our finances. And we’ve got a turnaround team at the moment trying to put us back on track. We know the only way to turn around the trust is to reduce hospital admissions.

Explanatory analysis

One would expect a system with this level of deprivation, demand for its ED and a ‘failing acute trust’ to have a much higher SAAR. Reasons why it might have an overpredicted SAAR were:

• having many proactive admission avoidance initiatives, some of which were available OOH

• having a rapid response team accessible by GPs and the ED

• coding of admissions that was the opposite of that in case study HU1 because ‘holding areas’ did not code people as admissions.

Level of integration between services in the system

We found no variable to measure the level of integration between services in different systems.

Amount of proactive admission avoidance initiatives

We found no variable to measure the amount of proactive admission avoidance initiatives in systems.

New data on attendance and conversion rates at emergency departments for geographically based systems

We used the HES A&E 2009–10 data for type 1 and 2 EDs only, to recalculate attendance rates and conversion rates for EDs for geographically based systems. We repeated the phase 1 regression with these two new variables and found that the regression results were similar but not identical to phase 1: length of stay < 1 day, the conversion rate and the ambulance non-transport rate still explained variation in the SAAR in the multiple regression. However, attendance rate at the ED and seeing a GP within 48 hours were no longer significant predictors within the multiple regression. The residuals from this new regression for our six case studies were still large enough to be selected as cases (< 1.3 or > 1.3), and still in the same direction, with the exception of HU1. HU1 had a very small residual because its new conversion rate was extremely high. This was supported by perceptions of interviewees from that case study who reported that ED attendances were coded as admissions and problems within the ED with consultant availability.

Availability of services out of hours

We would expect systems with poor OOH availability to have higher rates of avoidable admissions OOH. Ideally we wanted to test the ratio of admissions in hours (Monday to Friday, 09.00 to 17.00) to those OOH, but time of admission is not available in HES. Instead we created a proxy variable of the percentage of admissions occurring at the weekend. If there is a uniform distribution of admissions by day of week, we would expect 29% of avoidable admissions to occur at the weekend, with this percentage higher in systems with poor OOH support services. The mean was 24% for our geographically based systems, ranging from 20% to 26%. That is, the admission rate was lower at the weekend than week days, which does not support the conceptual framework that access to services in the wider system may avoid admission. There was also little variation between systems in the percentage of weekend avoidable admissions. This variable explained a small amount of variation in the SAAR in the univariate linear regression (Pearson’s r² = 4%, p = 0.019) but did not add to the multiple regression once deprivation and geography were included.

Amount of resource within systems

We used routine data we had obtained for another study about the amount of surplus or deficit a PCT had in 2006–7. 41 We used this as a proxy for financial status in 2008–11. This factor explained a large amount of variation in the SAAR in the univariate linear regression (r² = 14%, p = 0.001) but did not add to the multiple regression once deprivation and geography were included.

Opportunity to test further emergency department-related factors for acute trust-based systems

One of our authors (SG) is an expert in chest pain research and worked with a medical student, Tom Bidmead, to explore variation in admissions for chest pain between acute trusts using the HES data from our study and a survey of how EDs deal with chest pain. We added two general issues to the questionnaire which we felt were relevant to our study: the amount of senior cover in EDs and the perceptions of the amount of support services available to EDs for dealing with social admissions. The response rate to the survey was 105 of the 142 trusts (74%), with data available for 94 of the 129 acute trusts in our analysis in Chapter 5.

Variation in admission rates

We found a large amount of variation in potentially avoidable admission rates between geographically based systems and acute trust-based systems. The order of variation found is not unusual for emergency admissions overall11,13,18 and for ambulatory care-sensitive conditions in England22 and in the USA. 24

Deprivation affects admission rates

Employment deprivation explained a very large amount of variation in our study and this is not unusual. Deprivation has been identified as an important explanatory factor of variation in emergency admissions within multiple regressions exploring variation between PCTs,12 general practices10,12,13,18–20 and hospitals. 24 For example, age and deprivation explained 64% of variation between practices,13 and age, sex, ethnicity and deprivation explained 64% of preventable hospital admission variation between practices. 20 Researchers have used different indicators of deprivation such as receipt of social security benefits,21 the Index of Multiple Deprivation12 and the Carstairs score. 11 The type of indicator may not be important given that the two we used (employment and income deprivation) were highly correlated and both explained large amounts of variation. Not all studies have found a relationship between deprivation and emergency admissions. 15

Interviewees in our case studies proposed that deprived communities had higher levels of morbidity, poor access to transport and high expectations of immediate access to emergency care. The reasons proposed for deprivation explaining so much variation in other studies were that people in deprived areas may be more ill, may present later for treatment or may not have social support at home. 18 Recent evidence shows a strong association between deprivation and multimorbidity. 42 Researchers also suggest that people from deprived communities may have difficulty negotiating access to services when they are ill. 43 In the context of our study, this may result in underuse of alternatives to EDs and community-based services which could help people to avoid a hospital admission. A study of variation in health-care utilisation in the USA related high use to low social capital. 44

Supply-induced demand

We found that the number of acute beds per 1000 population was associated with emergency admission rates. This is supported by other research. The supply of acute beds has been shown to be related to admissions overall45 and for terminally ill people. 46 Numbers of beds in these studies were considerably higher than in our study, varying between 2.1 and 4.3 per 1000 population compared with our variation of 1 to 2.9 beds per 1000 population. This relationship between supply and demand has not been found consistently. 24 Our finding may be explained by supply-induced demand. There was some support for this in our interviews, because interviewees described initiatives put in place in times of high demand and full beds to avoid admission. Supply-induced demand has been put forward as affecting evaluations of interventions to reduce emergency admissions. 47 In a discussion of variation in health care more widely, Wennberg discusses variations in supply-sensitive care and how clinicians find it hard to believe that the capacity of their health-care system affects their practice. 48

What does the large percentage of short length of stay patients mean?

The percentage of all emergency admissions staying less than a day explained variation in our multiple regression. In the UK, these short-stay admissions have been shown to be responsible for some of the recent increase in emergency admission rates. 1 It is unclear what this factor represents in our study. We surmised from our case studies that variation in this factor might be associated with different hospitals coding attendances differently. ED attendances that are moved to another part of the hospital to avoid breaching the 4-hour wait target in EDs may be coded as attendances or admissions depending on the hospital. Standardisation of coding might help to understand how best to avoid admissions appropriately by allowing researchers, commissioners and service providers to focus on actual admissions.

Conversion rate in emergency departments

The conversion rate from ED attendance to admission explained variation in potentially avoidable admissions in our study. Based on our case study findings, this factor may have been representing the availability of ED consultants or the use of admission avoidance schemes such as senior review or multidisciplinary teams offering rapid assessment within EDs. There are studies showing considerable variation in admission rates between doctors in EDs which was not related to illness severity or associated with higher mortality rates. 49 The conversion rate has been increasing in England over recent years. 50

Busy emergency departments

Our interviewees described relentless demand for emergency services and very busy EDs. These departments can be busy because lots of patients arrive or because it takes a long time to admit or discharge patients, resulting in crowding. On the issue of use of this service, a number of research studies have reported on why patients choose to attend EDs rather than alternatives. 51,52 Qualitative interviews with ED attenders have identified lack of access to general practice, anxiety about the presenting problem, perceptions of the availability of treatments in the ED and a lack of alternative pathways as important predictors of attendance rates. 52 On the issue of crowding, our interviewees identified that junior doctors are more likely to use diagnostic tests which patients have to wait for. A US group of researchers studied change in occupancy in an ED over recent years and identified that intensity of practice, whereby patients had more tests and treatments than in previous years, accounted partly for increased crowding. 53 This intensity of treatment has been related to increased admission rates from EDs over recent years. 54

Four-hour target

Our interviewees who were based in EDs described the effect of the 4-hour target on their practice. They were keen not to breach the 4-hour waiting time and moved patients out of the ED to facilitate this. In some hospitals this resulted in ED attendances being coded as emergency admissions, while this did not occur in other hospitals. Concerns have been expressed about the potential adverse effects of this target but research evidence does not support the potential for it to increase short-stay emergency admissions; in fact it decreased short-stay admissions. 55

Improving access to primary care?

We found that, in our univariate regression, perceived access to general practice explained variation in potentially avoidable admission rates, with poor access associated with high SAARs. This fitted with our conceptual framework. Additionally, our interviewees perceived this to be the case, identifying that patients would attend the ED for immediate care rather than their general practice and thus increase their likelihood of admission. Our univariate analysis concurred with other studies which showed that perceived poor access to primary care was associated with high admission rates. 10,12,15 Aspects included patients reporting being able to see a particular GP and having access to a GP within 2 days, which are relevant to avoidable emergency admissions because people with acute exacerbations may not be able to access urgent appointments at their general practice or may perceive this to be a barrier. However, in our multiple regression the relationship reversed, with good perceived access associated with high SAARs. We may have found a reverse relationship with access to general practice because of the other variables included in our regression. We included ED attendance rates, which in other studies were predicted by perceived access to general practice. 26,27 Therefore, the ED attendance rate may have been a measure of access to general practice in our regression. It is also worth noting that the relationship between poor perceived access to general practice and high use of EDs has not always been found. 56 This is also true for its relationship with emergency admissions. 20

Effective interventions for reducing emergency admissions

In our case studies, managers and clinicians had introduced schemes or initiatives to avoid admissions and our interviewees spoke highly of some of them. These included multidisciplinary rapid assessment teams available to the acute trust or ED to facilitate discharge from EDs mainly and medical admission units sometimes; condition-specific pathways which allowed a co-ordinated approach to managing patients; senior review in EDs by medical or emergency consultants; GPs in EDs; easy access to community beds for hospital and primary care staff; education or individual GP for nursing homes; and rapid response teams or community matrons in the community stopping patients getting to the hospital in the first place. Our PPI coapplicant viewed these as examples of good practice to promote to the rest of the NHS. However, before recommending them, it is important to explore the research evidence base about their effectiveness. We could reference positive and null findings from a number of studies evaluating some of the schemes in use within our case studies. However, this should be done in the context of a systematic review which pays attention to use of robust evaluative designs and systematic searching. Purdy et al. 57 published a series of systematic reviews of interventions to reduce unplanned admissions and concluded that the majority of interventions tested did not reduce admissions. They also highlighted insufficient evidence on some types of initiatives. This concurred with a non-systematic review on a wide range of approaches aimed at reducing emergency admissions in people with long-term conditions and frail elderly. 58 It is interesting to note that few of the schemes/initiatives being used in practice within our six case studies have been found to be effective, or evaluated at all with randomised controlled trials in this comprehensive systematic review. 57 Pathways had an evidence base but the conclusion was that there was no convincing evidence that they worked. 57 There was a lack of studies of ED interventions, including GPs in EDs. 58

More recent research has been published about some of the schemes discussed by our interviewees. For example, the effectiveness of a team of allied health professionals offering co-ordinated care to older people in EDs (similar to our rapid assessment teams) found a very small reduction in emergency admissions albeit using a non-randomised design. 59 The use of geriatric assessment in an ED (a type of senior review initiative) identified a lower conversion rate to admission, again using a non-randomised design. 60 However, it is important to evaluate robustly some of the types of schemes discussed in our study, using randomised controlled designs if possible.

Is integration the way forward?

It is important to remember what we have been unable to include in our regression analysis. There were no routinely available data on availability and quality of community/intermediate/social care services, or integration between services, so we could not assess the effect of these potentially important factors. Integration of different types of services (health and social care, acute and community care) was described differently in our six case studies. Our interviewees felt that integrated practice could facilitate admission avoidance. The evidence base is less positive than our interviewees. Bardsley et al. 61 reflected on the 30 integrated and community-based interventions aimed at reducing emergency admissions that the Nuffield Trust had evaluated over time. These included the integrated care pilots which used structured care for at-risk older people. The results of their work were overwhelmingly negative in terms of reducing emergency admissions, including the integrated care pilots. They concluded that structural change does not necessarily deliver integration in practice. This aligns with other research on structurally integrated health and social care for mental health which showed that it was not sufficient for integrated care in practice and did not necessarily lead to better outcomes. 62 Evans et al. 63 undertook a review of the literature on integration in health care and discussed the belief that integration can offer better quality of care, efficient care and higher patient satisfaction. Their conclusion was that we do not necessarily know what integration is (even though many definitions and models exist) or how to achieve it. The literature does not support enforced structural integration but points more to integration in terms of shared vision and trust. 62 Evans concurs with this, recommending a move from a focus on structure to one on culture. Indeed our study highlighted the complexity of integration and the importance of communication and knowledge to build up trust between services.

There was some indication in our study that integration was easier between an acute trust and the main PCT population they served. There is some support for this in the literature where emergency bed use for older people was found to be lower in PCTs with a high proportion of their population going to a single acute trust. 64

Is 7-day working the way forward?

Our interviewees wanted more OOH provision of services that help to avoid admissions such as mental and social services. Two recent national reviews of emergency care in England have supported consideration of 7-day working. 50,65 There would be a substantial cost to the NHS of providing this and cost-effectiveness studies would be required.

Will educating the general public about how to use urgent care help?

Our interviewees wanted to see more education of the general public about which services to use under different circumstances when they had perceived urgent health needs. Our previous research showed that recent users of emergency and urgent care had a lack of awareness of alternatives to GPs and EDs and were confused about when to use newer services. 4 Research focusing on people with long-term conditions suggests that they understand which services are available and choose those that suit them most. 66 They see newer services such as walk-in centres as not offering the technological input of the ED or the established relationship of general practice. The permeability or ease of access of a service affects their choice, with EDs viewed as more permeable than general practice. 66 For people with long-term conditions, simply educating them on choices available would not work. However, it might be useful to explore whether any approaches to educating the general public about how to use emergency and urgent care services have been effective. It would also be useful to review the literature on why people use emergency and urgent care services and how this might affect avoidable emergency admissions, as well as to interview patients and their families about factors affecting avoidable emergency admissions.

Non-conveyance by ambulance services

We found that the large variation in non-conveyance rates by ambulance services explained some variation in avoidable admissions. This is a unique and important finding. Non-conveyance rates have increased considerably in recent years in England and there is support for it nationally as a way of managing demand for emergency care. 65 A review recommended that research be undertaken into how best to deliver appropriate and safe non-conveyance;67 some of this research is ongoing. 68

Easy and safe option

A prominent theme throughout our interviews was hospital admission or ED attendance as the easy or safe option. This is supported by a study of non-conveyance in the ambulance service for epilepsy, where ambulance crew were concerned about litigation and therefore transported people to hospitals. 69 Additionally, hospital admission has been seen as the only option for ED management of people with mental health problems in rural areas in Australia. 70

Reflecting on incentives

It is worth considering how incentives in the emergency and urgent care system appeared to affect the behaviour of staff within it. We heard in our case studies about the perceived incentive to admit to make money for an acute trust in need of financial support; the incentive to save time by admitting patients because of the workload involved in finding alternatives; the incentive to meet targets such as 8-minute response for ambulance services and 4-hour waiting at EDs leading to admission or coding as admission; and the incentive to avoid winter crises in bed use through schemes to avoid admission. Making alternatives to hospital admission the easy and safe options could act as an incentive for staff to avoid some admissions.

Resources

We could not use resource allocation data in our regression because they include emergency admission rates. The data we located about PCT surplus and deficit were limited in terms of their age and focus, so we could not measure the effect of variation in resources available in different systems. Lack of resources was raised in our case studies, however, with discussions about how it affects the input of social services and the loss of some disciplines from multidisciplinary teams providing proactive admission avoidance schemes.

Strengths

The strengths of the study were that it focused on admissions for conditions rich in avoidable admissions rather than all admissions; it moved beyond previous research which has focused on either primary care or hospital factors explaining emergency admissions; and the design used a combination of quantitative and qualitative research to understand the factors affecting avoidable emergency admissions.

Limitations

Project Advisory Group

We would like to thank members of the Project Advisory Group who offered support, encouragement and evaluations of the content and progress of the study: Dr Sarah Purdy (Chair, GP, researcher at the University of Bristol), John Soady (Sheffield Clinical Commissioning Group/Public Health), Beryl Darlison (PPI).

Patient and public involvement

We would like to thank members of the SECF for providing a patient and public perspective on the design, findings and interpretation of the study over and above the input from Enid Hirst as management team member and Beryl Darlison as Project Advisory Group member.

Researcher support

Colin O’Keefe helped to code the qualitative data within NVivo.

Robin Chatters contributed to the documentary analysis.

Tim Pearson calculated the SAAR and other variables using data from the HES.

Tom Bidmead undertook a survey of management of chest pain in EDs in England and shared his data with us for phase 3 of our study.

Administrative support

Susan Proctor offered administrative support to the project.

Marc Chattle and Richard Campbell managed data for phases 1 and 3.

Participation in the study

We would like to thank all the interviewees who spent time out of their extremely busy work days to share their views of their emergency and urgent care systems.