Understanding variation in ambulance service non-conveyance rates: a mixed methods study

International differences in terminology

Terminology relevant to this topic differs by country. Ambulance service provision (the term used in this report) is also known as Emergency Medical Services or pre-hospital care internationally. In the UK, ambulance services are known as Ambulance NHS Trusts. Non-conveyance (the term used in this report) may be called non-transport in other countries. Telephone advice (the term used in this report) can be known as secondary telephone triage whereby patients are triaged over the telephone by a clinician after their initial triage by a call-handler. Different terms can be used to describe staff working in ambulance services. We use the term ‘call-handler’ for staff taking initial 999 calls, ‘clinicians’ for staff offering telephone advice in the clinical hub of an ambulance service, and ‘crew members’ or ‘paramedics’ for staff making decisions to discharge patients at scene or convey them to a health-care facility.

Variation in definitions used by ambulance services, national quality indicators and this study

‘Hear and Treat’ is a term used in England to describe a key type of non-conveyance, whereby a call-handler triages patients to receive advice from a clinician working within the ambulance service dispatch centre, sometimes called a clinical hub. The clinician may give information or advice to self-manage or make a formal referral to an alternative service (this last option can be called ‘Hear, Treat and Refer’.) Clinicians may also return calls for the dispatch of an ambulance if they decide that this is necessary, resulting in conveyance to an emergency department or an alternative service. The AQI definition in England for ‘Hear and Treat’ is ‘Of calls that receive a telephone or face-to-face response, proportion resolved by telephone advice (Hear and Treat)’. This includes calls ending in telephone advice by call-handlers as well as clinicians. In this study we use the terms ‘calls ending in telephone advice’ or ‘telephone advice only’, which are similar to the AQI definition. We do not use the term ‘Hear and Treat’ because it has a variety of meanings. ‘Telephone advice only’ calls in our study include whatever ambulance services included in their AQIs. They include calls for which an ambulance was not dispatched either by call-handlers or by clinicians in the clinical hub. Our attempts to identify calls ending in telephone advice by clinicians only, and to consider variation between ambulance services for rates of these calls, were not successful (see Chapter 5, Methods).

‘See and Treat’ is a term used in England to describe a key type of non-conveyance. Ambulance services dispatch an ambulance and attending crew members may discharge patients at scene with treatment, information or advice, or formally refer patients to another service. In such cases, patients are not conveyed anywhere by ambulance. The related AQI is ‘Of calls that receive a face-to-face response from the ambulance service, proportion managed without need for transport to Type 1 and Type 2 A&E’. This includes ‘See, Treat and Convey Elsewhere’ calls, which may end in transport by ambulance to a range of services other than an emergency department. In practice, ambulance services calculate this AQI based on calls for which an ambulance is dispatched but the patient is not transported to a hospital with an emergency department. People who are not conveyed because they have died are included in this indicator because they have not been conveyed to an emergency department by an ambulance. In our study we successfully distinguished patients discharged at scene from those transported to alternative services and label these ‘discharged at scene’ rather than ‘See and Treat’. These ‘discharged at scene’ patients include those who are referred to alternative health and social care services but not transported to them.

In our original research proposal we planned to explore the separate group of patients to whom an ambulance is dispatched and who conveyed to a service other than an emergency department. We were interested in these ‘See, Treat and Convey Elsewhere’ responses because of the potential benefits of not conveying patients to emergency departments, and because conveyance to lower acuity services such as minor injury units and urgent care centres posed a potential risk to patient safety. We were able to identify these calls but also learnt enough about the mix of patients within this type of ‘conveyance but not to an emergency department’ to stop at a descriptive analysis (see Chapter 5, Results).

Finally, some paramedics have advanced skills that allow them to administer a wider range of medications, undertake more advanced assessment and diagnosis, and deliver a wider range of treatments than paramedics. They have a number of labels within England and internationally. In this report we call them ‘advanced paramedics’.

Definitions of terms used are summarised in the Glossary.

Non-conveyance rates

Variation between ambulance services for the rate of calls ending in telephone advice was threefold in November 2014 when the data for WP2 were obtained (Figure 3). The level of variation between services was similar over time, although the ranking of services was not consistent over time. There were a number of sudden shifts in rates for individual services, in particular for the Yorkshire Ambulance Service, North West Ambulance Service and East Midlands Ambulance Service. We know through another research study that in April 2015 the North West Ambulance Service stopped following national guidance and started to include hoax calls and calls closed by call-handlers (Janette Turner, University of Sheffield, 2017, personal communication). Although our analysis in WP2 was based on data prior to these sudden shifts, this indicates that individual ambulances services can change the way they calculate AQIs, which may not necessarily reflect their rate of non-conveyance.

FIGURE 3.

Variation between 10 large regional ambulance services in England: AQI percentage of calls ending in telephone advice.

Variation between ambulance services for the rate of patients receiving an ambulance but not being conveyed to an emergency department was twofold in November 2014 when the data for WP2 were obtained (Figure 4). These calls were largely made up of patients who were discharged at scene. The variation between services was reasonably consistent over time, with services largely maintaining the same rank over time. One ambulance service had a much higher rate than the others and one had a much lower rate than the others throughout 2014. There were some sudden shifts in the AQIs for some services. This could indicate a change in the way an AQI was calculated by a particular ambulance service or a change in non-conveyance rates. Sudden shifts are likely to be related to changes in calculations, whereas slow shifts over time are more likely to be related to changes in practice. South East Coast Ambulance and North West ambulance services had sudden shifts in rates, suggesting that rates were calculated differently over this time period.

FIGURE 4.

Variation between 10 large regional ambulance services in England: AQI percentage of calls ending in non-conveyance to emergency department (most of these are discharge at scene).

Variation in total non-conveyance was calculated crudely by adding the two AQIs together, namely the percentage of telephone advice calls and the percentage of non-conveyance to an emergency department. A graph similar to those in Figures 3 and 4 was constructed and this is displayed in Appendix 2. It shows that there was a twofold difference in total non-conveyance rates between services over time, with one service having a consistently higher rate than others over time.

Recontact rates

Variation in recontact rates for the two types of non-conveyance are shown in Appendix 3. There was fivefold variation between ambulance services in recontact rates for calls ending in telephone advice. One service had a consistently high recontact rate and another had a consistently low recontact rate. There was twofold variation between ambulance services for recontact rates when an ambulance was dispatched but patients were not conveyed to an emergency department.

Summary of findings

Ambulance Quality Indicators showed considerable variation between ambulance services in both non-conveyance rates and recontact rates within 24 hours of non-conveyance. Sudden shifts in monthly AQIs for specific services raised concerns about the extent to which AQIs represent the actual non-conveyance rate within an ambulance service, particularly for calls ending in telephone advice.

Implications for the empirical study

1. AQI non-conveyance rates changed over time for individual services so we documented when data collection occurred for different WPs of the study and the AQIs for that time period to help interpret the findings:

   1. April–December 2015 (WP1 interviews in Chapter 4)

   2. November 2014 (WP2 routine data in Chapters 5 and 6)

   3. January–June 2013 (WP3.1 linked data in Chapter 7)

   4. November 2015–September 2016 (WP4.1 observations and interviews reported in Chapter 8)

2. We needed to select a month of ambulance service routine data to analyse for WP2 and WP3.1. At the start of the study in January 2015 we decided to select the latest month available. However, we avoided December 2014 because of small peaks in published AQI non-conveyance rates in December each year. We selected November 2014 instead.

3. One of the 11 services is very small so the study focused on the 10 larger services in England only. Including the small service would introduce uncertainty related to small numbers in the quantitative analyses in WP2 and WP3.1.

4. Sudden changes in monthly rates for individual ambulance services raised concerns about the AQI data, particularly for rates of calls ending in telephone advice.

Design

We undertook a qualitative interview study with three key stakeholder groups: (1) ambulance service managers, (2) paramedics and (3) commissioners. We selected these three groups because ambulance service managers can shape strategic decisions that affect non-conveyance at an organisational level, such as workforce configuration or investment in staff training; paramedics make daily decisions about whether or not to convey patients and can offer views on patient characteristics, workforce practices and organisational factors affecting non-conveyance; and commissioners can encourage ambulance services to undertake non-conveyance by setting targets for non-conveyance rates within contracts and offering additional investment for non-conveyance initiatives. We supplemented these interviews with documentary analysis and sent a request to each ambulance service for information about resources within each organisation.

Data collection and sampling for the interviews

We undertook semistructured interviews with staff and commissioners from all 10 ambulance services. Our intention was to undertake around five interviews per service with staff in the following roles:

• ambulance service managers – a manager who dealt with non-conveyance and a director-level member of staff to address operational and strategic levels of the organisation

• paramedics – an advanced paramedic and an entry-grade paramedic with no less than 1 year’s experience post qualification to include different levels of expertise

• commissioners – the ambulance service lead commissioner

We identified a local collaborator in each service who helped to identify relevant staff for interview. In addition, managers or team leaders sent out recruitment e-mails and the study information to their teams to recruit paramedics. We contacted the relevant CCG for each ambulance service to request an interview with the ambulance service lead commissioner. Each potential interviewee was contacted by e-mail and formally invited to take part in the study. Non-responders to our initial invitation were contacted again within 1 month.

Following receipt of written informed consent, we undertook semistructured interviews mainly by telephone, although a small number of early interviews were undertaken face to face while we tested out the topic guide. One researcher undertook all the interviews. 18 The interviewer was blind to the non-conveyance and recontact rates of the interviewee’s organisation throughout data collection. The same topic guide was used for all participants and was developed based on our research objectives and discussion at our project management group. The topic guide covered perceptions of factors that affected non-conveyance at a local level, and those that were perceived to have an impact on appropriateness of ambulance non-conveyance (see Appendix 4). Interviews were digitally recorded and lasted between 40 and 90 minutes.

Data collection using documents

Prior to starting our research, we identified a number of organisational factors that might explain variation in non-conveyance rates. We were interested in organisational responsibilities and resources such as whether the ambulance service was a provider of the NHS 111 telephone service, and numbers of advanced paramedic and clinical hub staff. While analysing WP1 interviews we identified further organisational issues that we wished to measure for use in our quantitative analysis of variation in WP2 and WP3. We had intended to collect information from services using a pro forma. In practice we identified information from each ambulance service’s annual report where possible (for the period 2014–15),47–56 and where information was missing we sought to identify it from our study contact at each ambulance service. We sought information for 2014 because our quantitative data in WP2 and WP3 were for November 2014 (see Chapter 5). We identified the following characteristics as being of potential interest: population size, geographical size, software used, whether or not the ambulance service was a provider of NHS 111, the number of ‘advanced paramedics’, staff skill-mix, staff turnover, sickness absence rates, income, type of contract with commissioners, system complexity (number of CCGs/NHS acute trusts), and numbers of complaints/serious events.

We had concerns about the consistency of the information collected from documents and our contacts. When the recent National Audit Office (NAO) report was published we chose to use the information within it about individual ambulance services instead,1 with the proviso that the data were for the later period (2015–16) rather than for 2014.

Analysis

We transcribed the digital recordings of interviews verbatim. A researcher checked each transcript for accuracy by listening to the recording and correcting mistakes or filling in gaps where possible. We analysed all interviews using framework analysis. 57 First, we familiarised ourselves with the interviews by reading a sample of transcripts. Second, we developed a thematic framework based on our research questions, the topic guide and reading a sample of transcripts (EK, LBE). The thematic framework consisted of descriptive rather than conceptual themes, including types of non-conveyance, views of rates, calculation of rates, national drivers, local drivers, organisational drivers, collaborative working, workforce issues (skill-mix, culture, working patterns), triage software, patient characteristics, defining and measuring appropriate non-conveyance, commissioning, and respiratory problems. Third, we coded all transcripts to the framework, adding further emergent themes and subthemes throughout the process of coding transcripts. Three researchers undertook the coding (LBE, FF, and NA) using the qualitative software package NVivo (QSR International, Warrington, UK). Lindsey Bishop-Edwards led the process, working closely with Nisar Ahmed and Flavia Saraiva Leao Fernandes to ensure consistency of coding.

One researcher (EK) led the analysis supported by Lindsey Bishop-Edwards. Emma Knowles read extracts of transcripts within a theme within each service. She analysed the overarching theme and documented the variation in perceptions within each ambulance service, discussing findings with Lindsey Bishop-Edwards and Alicia O’Cathain. She repeated this for further themes, considering the connections between themes and refining the thematic framework by collapsing some themes together (e.g. the themes of ‘collaborative working’ and ‘the emergency and urgent care system’ were so interrelated that these were brought together).

The focus of the study was variation between ambulance services. A key difference between our analysis and the analysis for a standard qualitative interview study was the additional process of treating ambulance services as cases within the analysis, which resulted in the need to analyse and present data by ambulance service for use in the quantitative analysis in WP2 and WP3 and in the multiple case study comparison in Chapter 10. 45 A matrix was developed with ambulance services as columns and themes and subthemes as rows. Emma Knowles summarised variation in perceptions about a theme or subtheme on this matrix to allow the team to consider patterns of themes within and across ambulance services. This was undertaken blind to the non-conveyance rates. When this matrix was complete, the AQI non-conveyance rates at the time of the interviews were added to the matrix to consider patterns by non-conveyance rates. We then added data to this matrix from our documentary analysis/the NAO report1 for use in an analysis reported in Chapter 10.

Sample description

We planned to interview around 50 stakeholders. We formally invited 80 individuals to be interviewed, and undertook 50 interviews with those who consented. The main type of stakeholder who did not agree to interview was paramedics (see Appendix 5). On average we interviewed five members of staff working in, or commissioning, each of the ambulance services, ranging between four and seven participants per service. A complete range of stakeholders was obtained in eight services. Of the remaining two services, we were unable to include one of the two managers and one of two paramedics. The digital recording of one interview (a paramedic) was of such poor quality that we were unable to transcribe it and therefore did not include this in our analysis. In total, 49 of the interviews undertaken were analysed.

Interviewees identified a range of factors that they perceived to affect non-conveyance. These are summarised in Table 1. Factors that we could measure in WP2 (see Chapter 5) are simply listed here. Factors that could not be measured, and instances in which there was evidence of variation in perceptions between ambulance services, are detailed in this chapter.

Patient-level factors: patients’ expectations

Some interviewees – mainly paramedics and managers – perceived that a range of patient characteristics influenced decision-making for non-conveyance (see Table 1), with no discernible variation in perceptions across the 10 ambulance services. A key issue, which we were unable to test in WP2 (see Chapter 5), was patients’ and their families’ expectations that an ambulance would be dispatched and patients transported to hospital. Interviewees described trying to persuade patients calling for an ambulance to seek help from other services, or make their own way to an emergency department, and noted the problem of formal complaints arising when patients did not have their wishes met:

[. . .] some people ring an ambulance and they want an ambulance and they want to go to hospital [. . .] so when you’ve assessed them and done everything and said actually, you don’t actually need to go [to an emergency department], you can go to your GP or your walk-in centre or something, some of them are a little bit off, put out a little bit, you know, well ‘I was wanting to go to hospital’. Some people that it doesn’t matter, how hard you try to explain to them that they don’t need to go to hospital [. . .] they insist that they want to go to hospital [. . .] then you’ve got, you’ve got no alternative choice, you know they have to go.

Paramedic, interviewee 41

For patients who had prior experience of non-conveyance, interviewees perceived that patients were receptive to discharge at scene and welcomed the tasks that paramedics undertook on their behalf, such as arranging a GP appointment. One interviewee expressed the concern that the popularity of discharge at scene with patients might lead to supply-induced demand from people who did not need an ambulance:

I think the patients see that they’re getting a lot within the [non-conveyance] encounter. Obviously they get more assessments than they’d get at a GP practice and they get it all in one go whereas a GP practice might bring you back for your 12-lead ECG [electrocardiogram] the next day or oxygen saturations or blood glucose test thing. Those are all sorts of things that they’d send a patient away for, then do at a later date unless they are critically ill. I think some of our ‘See and Treat’ activity going up isn’t due to us being able to deliver higher levels of care [. . .] they were never appropriate for an ambulance service.

Manager, interviewee 37

There was some suggestion that expectations differed by patient characteristics. Patients living in rural areas were described as being more receptive to discharge at scene than those in urban areas, because of the inconvenience of travelling a long distance to an emergency department and arranging their return home from that hospital facility. This was in contrast to patients living in areas of social deprivation whom interviewees perceived as more likely to expect transport to an emergency department:

[. . .] two hospitals, one is about 37 miles in one direction, and the other is about 30-something miles in the other direction, so there’s quite long distances involved. So actually people would prefer to be treated at home. There’s an expectation ‘well I only live down the road from the hospital so you’re going to take me there aren’t you’ there is the opposite here. So, you know, there is a difference in expectations.

Paramedic, interviewee 17

[patients from] socially deprived areas tend to have much more comorbidities and are generally more complex patients. But also, quite often there is an expectation that they just want to go to hospital. It’s important that you discuss with the patient [what their] ideas, concerns and expectations are, and why it would be appropriate for them not necessarily to go to hospital.

Paramedic, interviewee 26

Organisational-level factors

The wider emergency and urgent care system

A consistent message across all ambulance services and stakeholders was the importance of the wider emergency and urgent care system as an enabler of non-conveyance:

[. . .] there’s lots of expectation on ambulance services to single-handedly reduce conveyance to [an emergency department] but, we can, a lot of that is, the success is down to the buy-in from the local health community in terms, if they’ve not got the alternative pathway we can’t use it.

Service C, interviewee 4

Interviewees talked about the importance of the system in relation to connectivity with the ambulance service (how easily the ambulance service can access another health or social-care service), provision (the availability of alternative services or pathways), pressure (high demand or a lack of resource within the system), complexity (numbers of CCGs or consistency between CCGs) and the level of integration with NHS 111.

National-level factors

Summary of findings

There were a large number of factors perceived by stakeholders to influence non-conveyance rates. The factors operated at different levels: patient, call, organisation, the wider emergency and urgent care system, and national. A key finding was the variation in the degree of motivation among senior managers to undertake non-conveyance, with some services actively embracing non-conveyance and others more concerned about the risks than the opportunities. It is possible that this motivation could affect other organisational-level factors, such as how advanced paramedics were used within a service. There was also variation in the way interviewees in different ambulance services talked about their wider emergency and urgent care system. Some systems were perceived to be under pressure, complex or lacking in service provision, all of which acted as barriers to non-conveyance.

Context of other research

Our findings are in line with recent evidence of the complexity for ambulance services of collaborating with so many stakeholders across the geographical areas that they serve. 1 Operational challenges were evident in some services, with interviewees reporting problems with staffing levels. Again, ambulance staff recruitment and retention has been reported as a problem elsewhere. 1 In our study, low staffing levels were perceived to be problematic for undertaking non-conveyance. Some paramedics did not feel supported within their organisation, and this is likely to have an impact on how they practice and the level of decision-making they are prepared to take. Staff satisfaction in the ambulance service, as reported in the NHS staff survey, is poor in comparison to that of other staff in the NHS. 58 Ambulance services also have higher sickness absence rates than other NHS organisations.

Strengths and limitations

This is the first study to take such a comprehensive national view of non-conveyance, drawing on the perceptions of key stakeholders. The inclusion of paramedics of different grades offers further insight into the complexity of decision-making facing ambulance crews that has been noted in other studies. 3,12 This WP had two limitations. First, we analysed the perceptions of a few individuals in relation to each ambulance service to identify organisational characteristics for testing in the quantitative analysis in WP2 and WP3 and for further exploration in the chapter on integration (see Chapter 10). We took care not to be swayed by an individual’s very negative or very positive view of non-conveyance within their organisation. This resulted in us not categorising many services for which mixed or unclear views were apparent. Perceptions are very important; nevertheless, perceptions are not objective measures of organisational practice or priorities. Second, individual paramedics were identified for interview by managers employed by each ambulance service. Therefore, there could have been bias in our sample, with the selection of paramedics made based on the viewpoint that the organisation wished to portray within the study. Although this might have occurred, we reflected throughout data collection and analysis on the amount of ‘account-giving’ within our full data set, whereby individuals wished to convince us of their positions as ‘good commissioners’, ‘good ambulance managers’ or ‘good paramedics’. There is no doubt that some of this occurs within any qualitative interview study. However, overall we felt that interviewees talked to us in an open and informative way, expressing both positive and negative views about non-conveyance within their organisations.

Data request from the ambulance services

We requested 1 month’s worth of routine computer-aided dispatch (CAD) data from each of the 10 large ambulance services in England. The study started in January 2015 and we requested data for November 2014 because it was the last full month for which data were available and November is not an unusual month for non-conveyance historically (see Chapter 3). The data specification form is displayed in Appendix 6.

Ideally, we would have requested these data after undertaking the qualitative interview study in WP1 so that we could request data on factors identified by stakeholders as being important to non-conveyance. However, we knew that obtaining and cleaning this type of routine data is time-consuming so we requested them before we started WP1. When services sent the data, we compared them with AQIs published for that month and worked closely with analysts at each service to reconcile differences between the data and the AQIs. We also requested all CAD data for that month so that we could apply the national guidance on calculating AQIs in a standardised way to the CAD data by applying standard definitions of specific types of non-conveyance.

We intended to link patient-level characteristics available from the 2011 census59 to each call by using the postal sector of the incident and the online census tool to relate this to deprivation and rurality indices from the latest census. Postal sector required special agreements from national information governance structures because they are considered to be identifiable data. As an alternative, we requested that services attach the lower layer super output area (LSOA) of the incident for each call to allow us to link calls with geospatial census data that might explain variation in non-conveyance rates. The Office for National Statistics (ONS) created the current set of LSOAs from 2011 census data59 such that England is entirely covered by a set of non-intersecting geographical areas, all of which have similar population sizes and internally homogeneous populations (assessed on a variety of census measures). LSOAs are the smallest geographical unit for which certain census data are available. They consist of clusters of adjacent areas with a mean population size of around 1600 people (in 2011). We then used the LSOA to link 2011 census data to each call. One service (Service H) could supply only 2001 census-based LSOAs, which we mapped to the most likely 2011 LSOA using population size. 59 The census variables that we linked to the ambulance routine data were informed by early WP1 interviews. These factors were related to the area from which the call was made to the ambulance service rather than to the individual patient.

Data obtained

We obtained data related to our specification from all 10 services. However, one service could not provide data on skill-mix (Service J), one service could not reconcile the differences between AQIs and the data sent (Service C), and one service engaged in a limited way with our team to reconcile differences, which could not therefore be explained (Service H). When we received the data, a team member (TS) checked it against published AQIs. When Tony Stone found disparities he worked with analysts at Business Intelligence/Data Analytics Teams at each ambulance service to resolve differences or explain why they had occurred. The amount of time spent reconciling the data with AQIs was considerable and we paid ambulance services for this. Most services could not fulfil our full specification in terms of providing data on calls sent to clinicians in clinical hubs, free-text on breathing problems and the full CAD data set. Therefore, it was not possible to undertake some of the analyses that we had planned.

Each ambulance service used one of two types of call assessment/triage systems to prioritise calls [Medical Priority Dispatch System (MPDS) and NHS Pathways]; one service used both systems for different localities but the calls associated with each triage system could not be distinguished and the service was categorised as using MPDS in our analyses. These triage systems categorised the reason for the call differently. Services also used different CAD systems, which had different labels for skill-mix and resource/type of ambulance sent. We established a working group of clinicians and experts within our team and a research paramedic from East Midlands Ambulance Service to develop common categories for these variables. We applied the common categories for reason of call to the data from each service. For skill-mix and resource we requested that each service link their specific codes to the common categories.

Some of the variables had limitations, as follows:

• Age – we had to analyse this as a categorical variable because call handlers can sometimes enter only an approximate age for a patient when a caller is unsure of this information.

• Age – telephone advice only is not advised for infants aged 0–2 years in some ambulance services so this age group was identified separately for the analysis.

• Age – ’0’ was used to indicate a baby and to indicate a missing value in some services. We included 0 as a baby.

• Reason for call – this was missing for a considerable proportion of calls, namely calls transferred from NHS 111, calls from health-care professionals, ‘running calls’ reported by ambulance crews on their way to another job, and calls from other emergency services such as police. Calls in which new information comes to light before an ambulance arrives at the scene may result in an upgrade or downgrade of the reason for the call; this was categorised as ‘upgrade’ or ‘downgrade’ rather than a specified reason for call.

• Type of caller – two services had very small percentages of calls from health-care professionals.

• Skill-mix – the skill-mix variable was based on the most senior crew member attending the scene. Analysts from the ambulance services were concerned that data on crew members could be incomplete and out of date and might not reflect who made the decision to non-convey. In addition, the most senior clinician might not be on scene but in the control room giving advice to the on-scene clinician. We had data for on-scene skill-mix only. We developed the common categories of doctor, advanced paramedic or nurse, paramedic, emergency medicine technician (can do basic life support, not registered), co-responding role (student paramedic, emergency care assistant, community first responder), and other (specialist roles, e.g. disasters). Owing to small numbers in some categories, these were collapsed to ‘advanced paramedic or nurse’, ‘paramedic’ and ‘other’ for the analysis.

• Resource sent – different types of vehicles are sent to the scene and this might have an impact on conveyance. We developed the common categories of air ambulance, dual-crewed ambulance (emergency transport vehicle), single-manned vehicle (which can sometimes convey), other conveying vehicle, and non-conveying vehicle (e.g. bike, motorbike, other). We decided not to use this variable in the analysis because it was not raised by interviewees in WP1.

• LSOA – any census variable linked to a call was for the area in which the call was made. The census variables were small area-level characteristics acting as proxies for patient-level characteristics. In addition, the incident may have occurred in a public place, or at the homes of friends or family members rather than a patient’s home.

• Calls sent to a clinician – we wanted to identify calls ending in telephone advice that had been sent through to the clinical hub for management by a clinician so that we could compare the rates of these types of telephone advice calls between services. Six services provided the data with the proviso that they were not accurate. In three services all or most of calls categorised as ending in telephone advice were coded as managed by a clinician: in one no calls were coded in this way, and in the other two around half of calls were coded in this way. We had no confidence in these data so did not use them within the analysis.

Analysis

Presentation of results

The quantitative results from the analysis are presented in this chapter. In addition, findings from the qualitative study in WP1 relating to calculations of rates of non-conveyance are presented here.

Reconciling computer-aided dispatch data with published Ambulance Quality Indictors

In the AQIs, 537,046 calls received a telephone or face-to-face response from one of the 10 large ambulance services in England in November 2014. In our data set, we had 638,254 calls; these included calls from NHS 111, which were not included in the AQIs. We had 93,900 calls from NHS 111 in our data set and 544,354 calls from 999. That is, we had slightly more 999 calls than included in the AQIs. Some ambulance services matched the AQI data closely when responding to our data request but there were large disparities for telephone advice rates for Service H (see Appendix 7).

Applying national guidance for Ambulance Quality Indictor calculations to our data set

During our conversations with analysts in ambulance services we learnt that some services had not followed national guidance for calculating AQIs in terms of inclusion and exclusion of NHS 111 calls and calls from health-care professionals. We applied the guidance to their data to calculate what we termed ‘standardised AQIs’. Applying standardised definitions of AQIs changed the non-conveyance rates for some ambulances services but did not reduce variation between ambulance services. Figure 5 shows the variation in published AQIs for ambulance services A–J (first columns of graphs), variation in AQIs based on the November 2014 data sent to us by services (middle columns of graphs), and variation in AQIs based on the application of national guidance on calculating AQIs applied to the November 2014 data sent to us by services (final columns of graphs). The bold horizontal line on each graph in Figure 5 is the mean of the 10 ambulance services. The similarity of the graphs in each row for each of the four AQIs suggests that variation between ambulance services was not simply related to these differences in interpretation of the national guidance. However, even when services followed guidance it was apparent from our consideration of the data that the range of ways in which calls can end in practice is complex and calculations were open to subjective opinions about the inclusion, exclusion and categorisation of calls (e.g. cancelled calls, standing down an ambulance and definitions of calls ending in telephone advice). Therefore, Figure 5 does not indicate that variation was not explained by differences in calculations between ambulance services.

FIGURE 5.

Comparison of published AQIs by source (see ‘published’ column), AQIs based on the November 2014 data provided to this study by services (see ‘service data’ column), and AQIs calculated using national guidance on the November 2014 data provided by services to this study (see ‘standardised’ column). The horizontal black lines indicate the mean of the 10 ambulance services. (a) The proportion of calls closed by telephone service; (b) the proportion of incidents managed without the need for transport to A&E department; (c) the proportion of patients who recontacted following discharge of care, by telephone within 24 hours; and (d) the proportion of patients who recontacted following treatment and discharge at the scene, within 24 hours.

Deaths

Some people are not conveyed to an emergency department because they are pronounced dead when the ambulance crew arrives or die while the crew is present. These deaths are included in the AQIs and it would be helpful to consider variation with them removed. However, it was not possible to remove all of them consistently from our data set. Deaths are coded differently by the two triage systems, MPDS and NHS Pathways, and can be coded under a variety of headings. Therefore, attempting to remove deaths would have introduced variation between ambulance services into our data set. In WP3.2 (see Chapter 7) deaths could be identified with confidence for a single service: 3% (2099/63,852) of non-conveyed patients in this ambulance service were pronounced dead at scene, indicating the size of the contribution of death to non-conveyance rates.

Description of patient-level factors affecting non-conveyance

Ambulance quality indictor definitions require that NHS 111 and health-care professional calls are included or excluded from certain calculations. We included calls from all sources in our analyses to measure the effect of a range of factors on non-conveyance rates. The description below is based on all calls.

A total of 41% of all calls resulted in non-conveyance to a hospital with an emergency department (Table 3). Most of these were discharged at scene after attendance by an ambulance (29% of all calls). Smaller percentages were seen for calls ending in telephone advice (7%) or the conveying of the patient to an alternative service (4%). Services alternative to an emergency department included a mix of facilities. Names of services rather than types of service were documented in our data set so it was difficult to know what some of these services were. They included walk-in centres/urgent care centres/minor injury units, GP out-of-hours services, different types of hospitals (such as maternity, cancer, psychiatric, hospice/cancer, major trauma with specialist care available), and ‘home’ if an incident occurred elsewhere and a patient was transported to their home. Rates of all non-conveyance to a hospital with an emergency department varied between services, from 28% to 59% (see Appendix 7).

Explaining variation between ambulance services

Explaining variation between services: telephone advice only

There were no skill-mix data available for telephone-advice-only calls so the primary analysis was undertaken on all 10 ambulance services (see Appendix 8). The denominator was all calls. There was considerable variation between ambulance services in terms of their telephone-advice-only rates (between ambulance service variance = 0.0991; Figure 6a). When patient-level variables were tested, this variation between ambulance services remained (variance = 0.1214; Figure 6b), even though telephone-advice-only rates varied by some patient characteristics (see Appendix 8). When ambulance service-level variables were tested, none explained the variation between ambulance services. Therefore, the variation in telephone advice rates between ambulance services could not be explained.

FIGURE 6.

Variation in telephone-advice-only rates between ambulance services explained in regression (based on 10 ambulance services, including calls from NHS 111). (a) Null model; and (b) model with patient-level variables.

Explaining variation between services: discharge at scene

Only nine ambulance services provided skill-mix data in terms of the type of crew attending individual calls. The primary analysis was undertaken on these nine services (see Appendix 9). The denominator was all calls attended by an ambulance. There was considerable variation between ambulance services for their discharge-at-scene rates (between ambulance service variance = 0.1206; Figure 7a). When patient-level variables were tested, this variation between ambulance services remained (variance = 0.1142; Figure 7b) even though discharge-at-scene rates varied by some patient-level variables (see Appendix 9). When ambulance service-level variables were tested, variation in discharge-at-scene rates reduced considerably (variance = 0.0095; Figure 7c). Variation between ambulance services was explained by three ambulance service-level variables after adjustment for patient-level variables:

1. Ambulance services with higher proportions of calls attended by advanced paramedics had higher discharge-at-scene rates [odds ratio 1.05, 95% confidence interval (CI) 1.04 to 1.07], taking into account that patient-level data showed that advanced paramedics had a higher discharge-at-scene rate than paramedics (odds ratio 1.38, 95% CI 1.34 to 1.42).

2. Ambulance services for which interviewees in WP1 described the management as risk averse to non-conveyance had lower discharge-at-scene rates (odds ratio 0.78, 95% CI 0.61 to 0.98).

3. Ambulance services for which interviewees in WP1 described advanced paramedics as an established and valued part of the workforce had higher discharge-at-scene rates (odds ratio 1.82, 95% CI 1.31 to 2.34).

A secondary analysis was undertaken by excluding skill-mix variables from the ambulance routine data and thereby including all 10 ambulance services. This regression failed to explain variation between ambulance services (see Appendix 10), although the odds ratios of variables were very similar to those in the regression that included skill-mix.

FIGURE 7.

Variation in discharge-at-scene rates between ambulances services explained in regression (based on nine ambulance services with skill-mix data, including calls from NHS 111). (a) Null model; (b) model with patient-level variables; and (c) model with patient and ambulance service-level variables. Black dots, odds ratio of < 1; green dots, odds ratio of > 1.

Exploring variation within ambulance services: telephone advice

Telephone-advice-only rates for localities within ambulance services, adjusted for patient-level variables, are displayed in Figure 8. The denominator was all calls. The localities for each ambulance service have the same colour dot. This shows that there is variation in telephone-advice-only rates within ambulance services but that the variation between ambulance services is larger than the variation within ambulance services (Kruskal–Wallis test p < 0.001). A caterpillar plot showed the same pattern (see Appendix 11).

FIGURE 8.

Telephone-advice-only rates for localities within ambulance services, adjusted for patient-level variables (all 10 services, including calls from NHS 111).

Exploring variation within ambulance services: discharge at scene

Discharge at scene rates for localities within ambulance services, adjusted for patient-level variables, are displayed in Figure 9. The denominator was all calls attended by an ambulance. The localities for each ambulance service have the same colour dot. This shows that there is variation in discharge-at-scene rates within ambulance services but that the variation between ambulance services is larger than the variation within ambulance services (Kruskal–Wallis test p < 0.001). A similar graph for all 10 services, not adjusted for skill-mix, showed the same pattern (not shown). A caterpillar plot showed the same pattern (see Appendix 11).

FIGURE 9.

Discharge at scene rates for localities within ambulance services, adjusted for patient-level variables (nine ambulance services only because adjusted for skill-mix, including calls from NHS 111).

Interviewees’ views of measures of non-conveyance

Summary of findings

We could not eliminate variation in calculations of different types of non-conveyance between ambulance services. Only a small amount of variation in non-conveyance rates between services was explained by patient-level variables, most of which were non-modifiable by ambulance services. Variation in telephone-advice-only rates between ambulance services could not be explained. However, a considerable amount of variation in discharge-at-scene rates was explained by three ambulance service-level variables: ambulance services with a higher proportion of calls attended by advanced paramedics, or that reported valuing advanced paramedics for non-conveyance, had higher discharge-at-scene rates; ambulance services reporting risk-averse views of non-conveyance among managers had lower rates of discharge at scene. There was variation in non-conveyance rates within ambulance services but this was smaller than variation between services.

Putting the findings in context

Similarly to our study, ambulance service personnel expressed concerns about inconsistencies in calculations of ‘Hear and Treat’ AQIs in the recent NAO report. 1 A secondary triage service in an Australian ambulance service found that abdominal and back pain were the most common problems dealt with by telephone,21 which aligns with our finding that calls for abdominal pain were more likely to end in telephone advice than other conditions. Population rates of secondary triage have been found to be higher for deprived communities21 but we did not consider population rates in our study. Similarly to our study, in a systematic review and meta-analysis, advanced paramedics have been shown to discharge at scene more than conventional crews,31 although the odds ratio was considerably larger than our study (10.5 vs. 1.4, respectively). We found no other research explaining variation between ambulance services.

Strengths and limitations

This is the first large study attempting to explain variation between different ambulance services. The data set was large and we were able to test a wide range of variables. There were five limitations. First, there were 2 days of industrial dispute in November 2014 when some ambulance staff were on strike. Even given this, November 2014 did not appear to be unusual in any way when compared with previous and future months in AQI statistics. Second, we were unable to construct a data set that eliminated differences in calculations of different types of non-conveyance by ambulance services. Third, data were missing for some factors for some services and we undertook a complete case analysis. Fourth, there may have been differences in interpretation of the term ‘advanced paramedic’ by the different ambulance services. Fifth, some of the ambulance service-level variables tested were derived from the WP1 qualitative interviews. It is possible that knowledge of their ambulance service AQIs and ranking compared with other ambulance services may have led some interviewees to describe their service in a particular way. It is also possible that our coding of qualitative interview data was influenced by our knowledge of the AQIs for each ambulance service. To guard against this, Emma Knowles and Lindsey Bishop-Edwards were blinded to the AQIs until the coding of qualitative findings was complete.

Implications

A large amount of variation between ambulance services in telephone-advice-only rates remained unexplained at the end of this analysis. This variation could be explained by variables that we did not test. For example, in WP1 patient expectations were identified as affecting non-conveyance rates but have not been tested here. However, we would be surprised if this variable explained variation between services because of the range of case-mix variables we tested, some of which are likely to be related to patient expectations. Variation could also be explained by differences in the way telephone-advice-only calls are identified by different ambulance services, an issue we could not test for in this study. A considerable amount of variation between discharge-at-scene rates was explained by factors potentially modifiable by ambulance services: the numbers of advanced paramedics in their service, how these advanced paramedics were used, and whether non-conveyance was seen as an opportunity or a risk by management.

Description of patient-level factors affecting non-conveyance

The recontact rate was 8.8% (3300/37,435) for calls ending in telephone advice and 5.1% (7690/152,106) for patients discharged at scene. This compared with the national AQIs of 8.1% (3737/46,355) and 5.5% (8997/163,205) respectively during that month.

Calls ending in telephone advice were more likely to result in a recontact if they were about older people (recontact rate of > 10% for > 40-year-olds compared with < 5% for < 10-year-olds), from areas with low levels of English not being the first language (12% vs. 6%), from areas with high levels of severe long-term illness (12% vs. 7%), for calls about psychiatric problems and abdominal pain (> 11%), and for low urgency calls (> 11%) (Table 4).

For discharge at scene, there was little variation in recontact rates by characteristics. The largest difference was a higher recontact rate for older people (6% in > 50-year-olds compared with 2% for < 10-year-olds) (see Table 4). There was no evidence of differences by skill-mix.

Explaining variation between ambulance services

Explaining variation between services: telephone advice only

There was considerable variation between ambulance services in recontact rates for calls ending in telephone advice (variance = 0.0790; Figure 10a). This variation was not explained by patient-level variables (variance = 0.0534; Figure 10b). All the variation was explained by ambulance service-level variables after adjustment for patient-level variables (variance < 0.0001; Figure 10c). The regression details are shown in Appendix 12. In summary, the ambulance-service-level variables explaining variation were:

• Recontact rates were higher in ambulance services covering a large number (odds ratio 1.22, 95% CI 1.06,1.40) or a small number of CCGs (odds ratio 1.41, 95% CI 1.19 to 1.67).

• Recontact rates were higher for which the perception of telephone advice provision within an ambulance service reported by our interviewees in WP1 was negative (odds ratio 1.36, 95% CI 1.20 to 1.53).

• Recontact rates were lower in ambulance services for which the cost per call reported in the NAO report1 was high (odds ratio 0.76, 95% CI 0.66 to 0.87).

FIGURE 10.

Variation in recontact rates with ambulance service after telephone advice only between ambulance services explained in regression (based on 10 ambulance services, excludes calls from NHS 111). (a) Null model; (b) model with patient-level variables; and (c) model with patient- and ambulance service-level variables.

Explaining variation between services: discharge at scene

As in Chapter 5, skill-mix data were not available for one ambulance service. The primary analysis was undertaken on the nine ambulance services with skill-mix data. There was considerable variation between ambulance services in recontact rates for calls ending in discharge at scene (variation = 0.0181; Figure 11a). This variation was not explained by patient-level variables (variance = 0.0285; see Figure 11b). All the variation was explained by ambulance service-level variables after adjustment for patient-level variables (variance = 0.0004; Figure 11c). The regression details are shown in Appendix 13. In summary, the ambulance service-level variables explaining variation were:

• Recontact rates were higher in services for which advanced paramedics were reported in the WP1 interviews as being more established (odds ratio 1.16, 95% CI 1.06 to 1.28) and lower in services for which there was reported limited use of advanced paramedics (odds ratio 0.87, 95% CI 0.79 to 0.96). This took into account that, for patient-level data, advanced paramedics had the same recontact rates as paramedics (odds ratio 0.95, 95% CI 0.86 to 1.04).

• Recontact rates were higher in services for which the NAO reported high levels of advanced practitioners within frontline staff (odds ratio 1.38, 95% CI 1.27 to 1.51).

The results were similar when all 10 services were included and skill-mix variables excluded, although some variables were not statistically significant and less variation was explained (see Appendix 14).

FIGURE 11.

Variation in recontact rates for discharge at scene between ambulance services explained in regression (based on 9 ambulance services with skill-mix data, excludes calls from NHS 111). (a) Null model; (b) model with patient-level variables; and (c) model with patient- and ambulance service-level variables.

Ambulance Quality Indicators on recontacts: a problematic measure

During the WP1 interviews (see Chapter 4), some ambulance managers described the two AQIs on recontact rates as a useful source of information because they were indicators of quality and safety, or because they allowed ambulance services to benchmark their performance against that of other ambulance services. For more managers in the sample these indicators lacked credibility because of concerns that actual recontacts were not measured but rather that a proxy was used (recontacts were based on anyone of the same sex calling from the same postcode rather than the same patient calling again about the same issue). Ambulance managers wanted a more reliable approach to measurement using NHS Number to identify genuine recontacts. They also expressed concern about what exactly these measures indicated. The AQIs might include very different issues such as unsafe practices, the fact that the right decision had been made at the time but that some patients’ health had deteriorated over time, or that patient dissatisfaction with the decision not to send an ambulance or transport them to hospital was the cause of the recontact. One manager was concerned that these indicators included frequent callers who called the ambulance service multiple times each week when they did not need an ambulance and that this resulted in difficulty understanding the level of recontacts indicating potential problems.

A useful source of information for commissioners

During the WP1 interviews, commissioners of ambulance services were interested in the monthly data on recontact rates. Both managers and commissioners described having discussions about them, with one manager perceiving that they were ‘held to account’ by commissioners for performance measured by these AQIs. Commissioners described the indicators as useful for a broad overview, allowing them to compare their ambulance service with others, look out for increasing rates, or look for differences by localities within the ambulance service geography (these ambulance services must have produced this small area analysis for the commissioners). Like ambulance managers, some commissioners expressed concerns about the credibility of the data and the importance of the serious incident data to identify inappropriate non-conveyance (see Chapter 7).

Summary of findings

Variation in 24-hour recontact rates for both telephone advice only and discharge at scene were explained by ambulance service-level variables. Recontact rates for telephone advice were higher for ambulance services for which interviewees in WP1 reported negative perceptions of telephone advice, higher for ambulance services with both small and large numbers of CCGs, and lower for services for which the cost per call was high. Recontact rates were higher for ambulance services reporting established advanced paramedics in WP1 interviews, and a high proportion of advanced practitioners in frontline staff. At a patient level, recontact rates for discharge at scene were similar for advanced paramedics and paramedics. Ambulance managers had some reservations about the way the recontact AQIs were calculated and how to interpret them, but some reported using the data with commissioners to consider variation between services and variation over time.

Strengths and limitations

A strength of this analysis is that it brings together recontact data on all ambulance services and identifies factors explaining variation in recontact rates between ambulance services within 24 hours of non-conveyance. A limitation is that we could not reconcile our data set with published AQIs for three services and this suggests variation in the calculation of recontact rates by ambulance services.

Implications

Some findings make sense but others do not. For example, higher rates of recontact for telephone-advice-only calls in ambulance services for which WP1 interviewees expressed negative perceptions of telephone advice makes sense, although there is uncertainty about the direction of this relationship because the perception of interviewees may have been based on knowledge of these recontact rates (see Chapter 5, Discussion). We can find no possible explanation for the finding that recontact rates were higher for ambulance services with both small and large numbers of CCGs. Most confusing of all is the finding that advanced paramedics have the same recontact rates as paramedics for individual patients but that ambulance services with established advanced paramedics have higher recontact rates. Concerns expressed by ambulance service interviewees in WP1 about the recontact data, and the lack of coherence of findings, suggest that caution should be exercised when interpreting the findings of this analysis.

Obtaining and linking data

Data for the 6-month period of January–June 2013 were obtained from one ambulance service. In order to create a clinically meaningful data set, exclusions were applied to the sample: attendances with no electronic Patient Report Form (ePRF), which documents details for calls attended by an ambulance crew; inter-hospital transfers; calls passed to other ambulance services; and duplicate calls for the same incident. An ePRF is never available for calls ending in telephone advice so these calls were included using information in CAD. A narrower definition of calls ending in telephone advice was applied within the PhOEBE study than in AQIs. This definition focused on calls that received input from a clinician: ‘Hear and Treat by Nurse Triage’, ‘Refer and Treat – Nurse Triage referred to other provider’, or ‘Nurse Agreed Plan’. Calls that were dealt with by a call-handler only were not included: ‘Automatically Closed Call’ or ‘Dealt with by NHS Direct’; these accounted for around 5000 calls in that period. The size of the sample we received was 51% of the total number of calls reported in AQIs by that ambulance service during January–June 2013. This was due to the exclusions applied and the fact that AQIs are a call-based data set, whereas our study data set was an individual patient-based data set.

Data linkage was undertaken by NHS Digital, using their data linking algorithm based on NHS number, date of birth, postcode and sex. If NHS number was not available, NHS Digital’s NHS number tracing service looked up NHS numbers using date of birth and patient name. NHS Digital undertook linkage of the calls in the ambulance service data to HES and ONS data to identify whether or not, and, if so, when, patients attended an emergency department, were admitted to hospital, or died. A total of 83% (154,927/187,426) of the patients in the sample were successfully traced; unsuccessful traces were attributable to missing or incomplete patient identifier data from CAD/the ePRF. Unsuccessful traces could not be linked to subsequent health-care information. Subsequent recontacts with the ambulance data were identifiable using a unique HES identification (ID) number generated for each patient. All relevant permissions were obtained for this linkage and for reuse of the data within our study.

Events within a specific time period

There is no consensus on the time period for measuring subsequent events after non-conveyance. Using a short time period may miss events attributable to the decision not to convey, and using a long time period may lead to the inclusion of events that are not attributable to the non-conveyance decision. As described in Chapter 1, researchers have tended to use a range of time periods. In the qualitative interviews in WP1 we sought the views of the ambulance service staff about the most appropriate time period to measure but no clear picture emerged. Interviewees who answered this question said that it depended on the context. We selected 3 days as the primary analysis because events are likely to be attributable to the decision not to convey. We report events up to 1 week to allow comparison of event rates with other studies. Date rather than exact time of hospital admission and death is documented in HES and ONS data. Therefore, 3 days is not 72 hours after the original call but somewhere between 48 and 96 hours. For consistency, we report events within 3 days for ambulance recontact and emergency department attendance (where time of event is recorded).

Recontact with the same ambulance service and Ambulance Quality Indicators

It was possible to identify the recontacts that the same patient had with the same ambulance service. This calculation was superior to that used to calculate AQI recontact rates for which recontact rates are based on whether a call was received from someone of the same sex at the same postcode within 24 hours.

Removal of deaths at scene and end-of-life deaths

Ambulance Quality Indicators and our analysis of CAD data in WP2 include people who are found dead on arrival of the ambulance or who die while the crew is in attendance. We were able to identify these deaths and remove them from the analysis here. In addition, we wanted to identify events that were not expected to occur, so a medical practitioner co-applicant (NS) identified International Statistical Classification of Diseases and Related Health Problems, 10th Revision codes of deaths associated with end-of-life care and we removed these patients prior to analysis.

Perceptions of inappropriateness

In the qualitative interviews in WP1 we sought the views of the ambulance service staff about inappropriate non-conveyance and these views are reported in this chapter.

Analysis

The analysis was descriptive. We undertook the following:

• Compared the characteristics of patients with linked data with patients with unlinked data to consider bias in our sample.

• Calculated the different event rates for telephone advice only and discharge at scene separately.

• Performed a sensitivity analysis. For the sensitivity analysis we included unlinked calls in the denominator, excluded very elderly people who we assumed were very ill and likely to have subsequent events, excluded deaths in very elderly people, and excluded frequent callers. Ambulance services have their own definition of frequent callers but we chose to use our definition because of the complex relationship between calls and recontacts in our data set; we removed anyone who had called the service more than six times in the 6-month period of routine data included here.

Our included calls compared with Ambulance Quality Indicators

The AQIs for the 6-month period for this ambulance service were based on around 16,000 calls ending in telephone advice and 94,000 calls not conveyed to a hospital with an emergency department. Our PhOEBE data set included a smaller number of calls than this (approximately 11,000 and 51,000 calls, respectively) because of the narrower definition of telephone advice used and the inclusion criterion of the need for an ePRF.

Ability to link data

A total of 24% (2521/10634) of calls receiving telephone advice only were linked. Eighty-five per cent (43,108/50,894) of calls discharged at scene were linked (see the flow diagram of included calls in Appendix 15).

There was bias in the linkage process. The quality of data on identifiers for patients attended by an ambulance was better than for patients who received telephone advice only. During this time period, date of birth was not routinely recorded on the CAD system and we failed to link any data for telephone-advice-only calls because this linkage relied on information within CAD. To facilitate linkage, an information analyst at the ambulance service searched for previous and future attendance by an ambulance for all our telephone-advice-only calls to locate the necessary identifiers for linkage. This meant that patients with a history of calling the ambulance service were more likely to be linked for our telephone-advice-only analysis. Therefore, they were more likely to be ill and potentially more likely to have events after their original non-conveyance than those we failed to link.

There was little evidence of differences between patients with linked and unlinked data for those discharged at scene (see Appendix 16). However, there were differences for patients receiving telephone advice only whereby patients with linked data were older than patients with non-linked data (see Appendix 16). Because of the bias within the telephone-advice-only sample, we report event rates for this group with linked calls as a denominator and also all calls (linked and not linked) as a denominator. This offers what are likely to be maximum and minimum event rates for telephone advice only.

Event rates

For telephone advice only, the minimum recontact rate with the ambulance service within 3 days was 7.6%; the minimum emergency department attendance rate within 3 days was 6.2%; the minimum hospital admission rate within 3 days was 2.6%; and the minimum mortality rate within 3 days was 0.06% (6/10,627) (Table 5).

For discharge at scene, the recontact rate with the ambulance service within 3 days was 9.0%; the emergency department attendance rate within 3 days was 12.6%; the hospital admission rate within 3 days was 6.8%; and the mortality rate within 3 days was 0.30% (129/42 796) (see Table 5).

Using an alternative way of expressing the mortality rate, around one person died within 3 days for every 1000 people given telephone advice only and around three people died within 3 days for every 1000 people discharged at scene.

Description of patients having events

A description of patients having different types of events is provided in Appendix 17. For example, a large percentage of patients discharged at scene who subsequently died within 3 days of the original call were aged 81–90 years [42% (54/129)] or > 90 years [12% (16/129)]. This linked data set did not include information on skill-mix, so differences in event rates by crew type could not be calculated.

Sensitivity analysis

The sensitivity analysis showed that the event rates for telephone advice were sensitive to the denominator used and exclusion of frequent callers (see Appendix 18). The event rates for discharge at scene were stable.

Comparison of recontact rates with Ambulance Quality Indicators

The monthly AQIs measure recontact with the ambulance service within 24 hours of the original call. They varied between 3% and 7% per month for this service over this 6-month time period for telephone advice only, averaging around 5%. For incidents in which patients were not conveyed to a hospital with an emergency department, including incidents in which patients were conveyed elsewhere, the AQIs were between 6% and 7%, averaging around 6%. The equivalent statistics in our linked data set were 8% (telephone) and 9% (discharge at scene). The patients included in our data set and the AQIs differed: our calculations were based on 1 day rather than 24 hours, we excluded the small percentage of patients who were dead but categorised as discharged at scene, we excluded patients conveyed elsewhere and we excluded telephone advice calls ended by call-handlers. Nonetheless, the recontact rates we found were of a similar order to those reported in the AQIs.

Timing of recontacts with the ambulance service

The exact timing of recontacts with the ambulance service was available. Timing was plotted by hour after the original call (see Appendix 19). Recontacts for calls ending in telephone advice were usually made within 2 hours of the original call. Recontacts after discharge at scene were usually made within 4 hours of the original call. This suggests that patient expectations for an ambulance, and for transport to an emergency department, rather than changing circumstances, were likely to play a part in some recontacts.

Ambulance staff views of inappropriate non-conveyance

Ambulance staff interviewed in WP1 commented on the utility of AQIs for 24-hour recontact rates with the ambulance service. They also discussed the importance of safety, the challenges of interpreting rates of subsequent events post non-conveyance, alternative approaches to assessing appropriateness, the importance of paramedics learning from their practice, and the need for data on subsequent events:

Summary of findings

Linkage for patients discharged at scene where an ePRF was available was high. A minority of patients had subsequent events within 3 days that might indicate inappropriate non-conveyance. Linkage for patients offered clinical advice by telephone was poor; a minority of patients had subsequent events within 3 days. It is difficult to comment on the acceptability of these rates without comparing them with event rates in other services.

Context of evidence base

The rates of subsequent events reported here can be placed in two contexts. The first of these would be in relation to the rates found by other ambulance services in other studies and countries. Because of the range of ways in which researchers have measured rates of subsequent events, we have summarised these rates from other studies in Appendix 20. Comparison is challenging because researchers used different approaches to identify events (e.g. data linkage, record review, routine data analysis), measured rates over different time periods (e.g. 24 hours, 48 hours, 72 hours, 1 week, 28 days), focused on specific conditions (e.g. falls) or specific groups (e.g. paediatric), included specific types of practitioners (e.g. advanced paramedics only), or excluded certain types of patients (e.g. excluded people who refused to travel). Rates are likely to vary by these factors. For example, patients seen by advanced paramedics had lower emergency department attendance rates and hospital admission rates than patients seen by paramedics. 20 It is also the case that rates are likely to vary by case-mix. Second, we can place the findings in the context of rates for similar services such as emergency department discharge.

Telephone advice

Comparing findings from our study with those from others is not straightforward because our data are about calls ending in telephone advice, whereas published studies focus on secondary triage services, which include calls sent an ambulance as well as ending in telephone advice. 21,34 For example, one secondary triage service of low-acuity calls in Australia referred 68% to an emergency department and 28% of total cases went there by ambulance. 21 The conclusion from a systematic review of secondary triage services – similar to clinical hubs in our study – is that they have been shown to be safe but that further research is needed to determine safety owing to small sample sizes and variation between systems. 34 Concerns have also been expressed about using hospital admission rates and attendance at emergency departments as indicators of lack of appropriateness or safety because some people may need to go to hospital but not need an ambulance. Instead, expert panels are considered a more useful approach by which to determine appropriateness. For example, one study found only 1% of cases in which a panel of experts believed that an ambulance should have been sent. 64 This rate was similar in other studies. 34

An alternative comparison is a survey of patients who were offered telephone advice from ambulance services. 6 Eight per cent contacted the ambulance service once within 7 days, 4% contacted the service twice, and 3% called the ambulance service three or more times. A problem with relying on surveys is that people will not respond if they have died or been admitted to hospital for a serious illness.

Discharge at scene

Focusing on studies that measured events within 3 days or 72 hours, attendance at an emergency department varied, with findings of 8%,65 12% and 15%,29 and 19%66 compared with 13% in our study. Hospital admission was somewhat higher than was found in our study, at 9% and 13% 29 and 10%66 compared with 6% in our study. Death rates were 0.3% of people dying during hospital admission,65 which was similar to our rate of 0.3%. In England and Wales all incidents related to patient death in the ambulance service were analysed. The rate was one safety incident-related death per 168,000 calls received,67 or around 50 per year in England. Most of these were due to delayed response rather than non-conveyance.

A comparison with death rates of patients discharged from emergency departments in the USA shows that the rate of 1.2 per 1000 within 1 week of discharge was lower than our rate of 5.4 per 1000 within 1 week of discharge at scene. 68 The higher rate in our study may be explained by the severity level of illness of patients calling an ambulance compared with those patients attending emergency departments.

Strengths and limitations

A strength of this study is that it was a large study of a wide range of subsequent events post non-conveyance. There were five limitations. First, linkage was poor for telephone-advice-only calls and these findings need to be considered with caution. Second, it was unlikely that we removed all calls from patients dying with end-of-life conditions and some of the deaths and admissions may have been expected. Third, our time period for measuring an event was calculated using date rather than time of event, so is not directly comparable to studies using time of event. Fourth, comparing like-for-like was not possible when placing our event rates in the context of other research because of the wide range of factors that may affect rates. Fifth, if non-conveyance facilitates the capacity of ambulance services to send ambulances to critically ill patients quickly, then the true picture of safety and appropriateness can be gauged only by also considering adverse events attributable to ambulances not responding to some events quickly enough. 67

Implications

Given that ambulance staff are making decisions in situations in which health status can change quickly, that some decisions to discharge at scene are due to the strong preferences of patients and their families, and that the sample is likely to include people in an end-of-life phase, the event rates found here were not large. It is certainly not the case that these non-conveyance decisions simply delay use of emergency departments and hospital beds, because the majority of patients did not have a subsequent event within 3 days of the non-conveyance decision.

Study sample

Telephone advice involving clinician assessment is a two-stage process provided within the ambulance service Emergency Operations Centre. Initial assessment (triage) of 999 calls is conducted by non-clinical call-handlers (variously called call-takers, emergency medical dispatchers and emergency medical advisors) to triage calls for an ambulance response and identify calls potentially suitable for telephone advice. These calls are passed for secondary assessment (secondary triage) by clinicians who are usually paramedics but can also be trained in a range of clinical disciplines including nursing, mental health and midwifery. Callers dealt with by clinicians may be given self-care advice; they may be advised to contact, be referred to or have appointments arranged with other health professionals/services; or an ambulance may be dispatched.

Using monthly AQIs we selected three ambulance services showing variation in the rate of calls ending in telephone advice and in recontact rates. This selection was made based on the average of the 3 months of August–October 2014 to allow sufficient time for the research permissions process. One ambulance service had a low rate of calls closed with telephone advice but one of the highest recontact rates (Service E). One service had a high telephone advice rate but one of the lowest recontact rates (Service H). These two services were selected to represent the extremes of variation for our study. An additional service offering further contrast was selected on the basis of having average rates of both telephone advice and recontact (Service B). At the time of data collection for this component of the study (November 2015–September 2016), the AQI data were different for some of these services: Service E still had a low rate of telephone advice and high recontact rate; Service H had an average telephone advice rate but the recontact rate remained low; and Service B had a high telephone advice rate and a low recontact rate.

Data collection

Data analysis

Observation notes and researcher reflections were recorded using handwritten notes and transcribed into electronic format suitable for analysis after each observation shift. Interviews were digitally recorded and transcribed verbatim. Transcripts were checked for accuracy. The data were analysed using a framework approach57 supported by the use of NVivo qualitative data analysis software. The coding was undertaken by three researchers (LBE coded the observations and NA and ROH coded the interviews). The coding frames developed for both the observation and interview data were reviewed, checked and discussed with a senior researcher (ROH) and the themes were refined and developed through a number of iterations. The data were analysed by Rachel O’Hara. The observation data were analysed first, followed by the interview data and, finally, the data were analysed per service. The emerging findings were discussed at various stages with the study management team and PPI group. Lindsey Bishop-Edwards was blind to the AQI data for services throughout the coding. Rachel O’Hara was aware of the AQIs throughout the analysis.

Presentation of findings

Findings from the non-participant observation and interview data collection have been combined. Differences and similarities across the three participating ambulance services are addressed to explore organisational variation in the provision of telephone advice.

Role of telephone advice: appropriate response or demand management tool?

We explored the perceived role of telephone advice within each of the participating services to understand how this might have an impact on any variations in practice. There was considerable consensus of opinion across participants in the three services regarding the perceived purpose of telephone advice. Key aspects were identified as being telephone triage of patients identified as lower risk/non-emergency; referral of patients to the most appropriate care; signposting patients to the right resource; providing self-care advice; and reducing demand for emergency ambulances, and to a lesser extent, reducing demand on emergency departments:

An ambulance isn’t always the right response for somebody just because they’re unwell [. . .] or taking them to [an emergency department] isn’t always the best thing because obviously that then overloads that, as an emergency service which [the emergency department] really is. And it just causes this massive imbalance that they can’t respond to effectively and it then means that people that genuinely need it aren’t getting the help they need in the right time.

Service E, interviewee 53

I think it’s basically to filter out the calls that don’t warrant an emergency ambulance and to pass them to other pathways.

Service B, interviewee 61

In Services B and H, demand management in relation to the increasing number of emergency calls appeared to be a greater consideration in the provision of telephone advice:

Demand was quickly outweighing supply with regards to ambulances etc. so something had to be done.

Service H, interviewee 65

Clinicians and call-handlers identified an additional aspect of secondary triage as identifying calls that needed an ambulance. Indeed, this was observed on a number of occasions in all services. As identified here by one of the clinicians in Service E, risk management was identified as a key consideration:

I think ‘Hear and Treat’ does work as far as we’re concerned, based on our patients that we do get. As I say, we’re a unit that’s very risk averse. We will send [an ambulance] where there is any query.

Service E, interviewee 70

Some clinicians regarded the organisational rationale for telephone advice as being more aligned with organisational performance and achieving targets for ambulance response times (i.e. demand management). This view was more prominent in Services B and H and is illustrated in the two quotations below. The clinician in Service H appeared to be more accepting of this as a valid aspect of telephone advice that was not at odds with delivering appropriate care for patients:

I think the ambulance [service] have a different view, I think it’s relating to targets.

Service B, interviewee 59

It’s also to minimise the impact on front line motor crews, and to increase their productivity.

Service H, interviewee 64

Clinicians advocated communication of telephone advice response options as an opportunity to educate callers and explain why an ambulance response was not appropriate in this instance, as well as offering information on alterative options to calling 999. Call-handlers also appeared to have a role in this regard, as they often needed to explain to callers why they would be receiving a clinician call-back rather than an ambulance response:

I’m happy when I don’t have to send an ambulance and I can give them health advice on what they should be doing because this, it’s all about education, it’s all about making sure that people know the correct reasons for calling 999, so we educate that way as well.

Service B, interviewee 63

Related to this issue of telephone advice as a demand management tool was the concern voiced by some clinicians and managers that delivering high-quality telephone advice might encourage patients to call 999 for non-emergency health problems in the future:

You would like to think that you are educating them that on this instance you didn’t need to call us, you could have got this service from 111. But because we are very proactive and we will do the best for our patients and we’ll have arranged something such as a district nurse, then some people will call back because we did that so well.

Service E, interviewee 70

Workforce

The triage process

The process of triage – both initial triage by call-handlers and secondary triage by clinicians – is very detailed and can vary according to the level of demand. Coverage of this process here will necessarily focus on key aspects that are relevant rather than providing a detailed account of the process operating in each service. Initial triage of 999 calls is conducted by call-handlers who identify and prioritise calls as either life-threatening conditions that require an immediate emergency ambulance response or conditions that are not life-threatening. More serious non-life threatening conditions are allocated a lower priority ambulance response but for other conditions a non-ambulance response may be more suitable, for example referral to GP, pharmacy, minor injuries unit, urgent care centre or attendance by community services. The general process of call-handler and clinician triage was broadly similar across the three services, although variation was apparent in relation to the specific practices around triaging patients and delivering telephone advice.

Type of work: managing a variety of tasks as well as telephone advice

The effect of very high demand for an ambulance on the triage process

Each of the services have resource escalation procedures that specify the actions required to mitigate the impact of increased demand on patient care. When demand levels are very high the lowest priority calls are given a self-referral response. A standard script is delivered either by the call-handler or by recorded message (to free up the call-handler), explaining that the ambulance service is currently experiencing very high demand and based on the information given they will not be sending an ambulance. Callers are advised to contact other services (e.g. GP, NHS 111, minor injury unit or emergency department). This does not apply to vulnerable patients including the very young, the elderly and elderly patients who have fallen and are on the floor. Depending on the level of escalation, these calls are passed to the clinician team, team-leader or duty manager to check and close the call. The call-handler can exercise their discretion and identify concerns about the response allocated to specific calls for the attention of the clinicians.

Demand levels in Services B and H were very high on a number of occasions during the observations and this had an impact on call-handler and clinician activity. In Service E high demand appeared to have more of an impact on call-handlers than clinicians. On the whole, clinicians in Service E were observed to be less busy than in the other two services. The following quotation from a clinician also points to the view that they are generally sufficiently resourced to meet demand:

As the day progresses, and perceived demand increases, then staffing levels increase. Until we get to sort of 11 o’clock at night and 2 o’clock in the morning, where demand in theory should start to wane, then we are left with, the manning should be, again two people at each site. Some days it can go to just one person, and in those instances yes demand can outstrip the ability to do the ‘Hear and Treat’. But generally if we’re fully staffed, there is two staff both sites 24/7 with an incremental number of staff coming in across the busy period.

Service E, interviewee 70

In Service H the activity of clinicians was less affected by high demand. In Service B clinicians identified very high demand as an ongoing problem that affected the delivery of telephone advice. Several calls were observed being allocated a ‘no send’ script after call-handler triage. When concerns were raised about some of the ‘no send’ outcomes, they were passed for clinician assessment. For the clinicians, the impact of this ongoing high demand on their ability to deliver telephone advice was a source of frustration, whereby a significant amount of their time was spent doing ‘welfare calls’ for delayed ambulances, rather than telephone advice:

I think it is sheer volume of workload, because we’re having to do welfare calls as opposed to ‘Hear and Treat’ [. . .] that now seems to be taking over at least 50% of our role. We allocate at least two, three or four people every day, every shift to purely be doing nothing but welfare calls, which means those three or four clinicians aren’t hearing and treating, they aren’t, they aren’t doing the job properly, they’re just literally ringing up saying sorry.

Service B, interviewee 61

Although it might be expected that this would reduce the scope for the delivery of telephone advice that results in closing calls without dispatching an ambulance, a number of instances were observed in which an ambulance response was cancelled as the caller or patient decided to use their own transport to get to a hospital.

In all three services, clinicians were expected to call back patients within 1 hour to conduct their assessment. This could be time-consuming, involving a number of call attempts, and when contact with the caller was not possible within this time frame, the call would be upgraded for an ambulance response. In Service B, in which clinicians were busy dealing with welfare calls, this could limit their ability to conduct timely call-backs. One of the clinicians suggested that additional staff could help to address this problem.

Patient factors

Availability of telephone advice response options in wider emergency and urgent care system

Variation was observed in relation to telephone advice response options in each of the three services. All services offered self-care advice and recommended self-referral to other services including daytime GP or an out-of-hours service, NHS 111, walk-in centres, urgent care centres, minor injury units or emergency departments as appropriate. This advice could include details of the location of services. A common problem concerned the variability of response options available within each ambulance service, whereby interviewees described some areas as having quite a lot of resources but others as having less. There was also inconsistency in referral criteria and how services were accessed. Where the referral process took a long time, it could result in the call being upgraded to a full emergency response, particularly when the patient had deteriorated.

Organisational factors

Summary of findings

A summary of variation between services and the possible direction of effect of factors on rates of telephone advice and recontact rates are displayed in Table 8. It is noticeable that Service E appears to be very different from the other two services, with fewer factors that might increase rates of telephone advice and more factors that might keep rates low. This is consistent with it having consistently low rates in the AQIs. There are differences between Services B and H, with Service B having more factors that might keep rates low. The change in AQIs for these two services over the time of the study makes it difficult to relate these findings to the AQIs. The only thing we can say with any confidence is that we would expect higher rates in these services than in Service E given our observations and interviews, and this is apparent in the AQIs.

We identified few factors that might affect recontact rates but we did identify two factors and they aligned with the different rates in that the more risk-averse service (Service E) had high recontact rates and the service with more response options that patients were likely to be satisfied with had low recontact rates (Service H). The effect of whether staff engaged solely or largely with telephone advice rather than other tasks was unclear to us because this would depend on how calls related to other tasks were coded by each service.

Context of other research

The findings highlight a range of areas of variation in the delivery of telephone advice across the three ambulance services. Although the triage process appeared to be broadly similar, there were some specific differences in practice in terms of the system triage tools used, the staff involved, the type of work they did, and response options offered to patients. This is similar to variation in the organisation and delivery of NHS 111 across five different sites. 70

One aspect of variation concerned the degree of alignment of this type of non-conveyance with either demand management or risk management. The former was more a feature in Services B and H, with the latter being very prominent in Service E. This has the potential to influence both the amount of telephone advice given and recontacts in 24 hours. In Service E, staff were more likely to view their role as ensuring that anyone who might need an ambulance got one, whereas staff in Services B and H focused on only offering an ambulance to people who needed one so that ambulances were available for critical cases.

It was expected and observed that call-handlers have a role in the identification of potential calls for telephone advice. However, the findings also showed that non-clinical call-handlers deliver telephone advice and do not simply pass calls to clinicians for secondary triage. In all three services, telephone advice responses were delivered at the initial triage stage. The circumstances when this is permitted and who actually closes the call varied across the services. Similarly, non-clinical staff in NHS 111 also undertake call assessments and have different levels of autonomy according to the organisations they work in. 70

The most notable variation in relation to skill-mix concerned the availability of mental health nurses in Service H. Patients experiencing a mental health crisis were identified by staff as particularly challenging to deal with. Therefore, the experience and professional links that these specialist nurses brought appeared to offer the potential for better care and outcomes that did not involve an ambulance response or referral to an emergency department. The three services also had different sets of alternative response options for delivering telephone advice, with Service H having a greater variety of alternatives to an emergency ambulance. A systematic review of secondary triage indicated that its effectiveness is limited by the range of options available to clinicians and suggests that the availability of alternative options may also reduce the level of recontacts following telephone advice only. 34 Access to a range of alternative options would therefore appear to be an important factor in the delivery of telephone advice and minimising recontacts. To this end, the NHS England document Transforming Urgent and Emergency Care Services in England. Clinical Models for Ambulance Services, designed to help Urgent and Emergency Care Networks (UECNs) commission ambulance services, states that ‘clinicians working in the 999 system – through “Hear and Treat” or “See and Treat” models – should have unrestricted referral rights to all other services in the UECN, including social care services.’73 However, it appears that there is still some way to go to achieve this given the current inconsistency in access to services.

Ambulance service staff talked about caller expectation as a challenge in delivering telephone advice, as the ability to communicate confidently is important to convey why an alternative to an ambulance may be better. A 2013/14 survey of 999 callers in England who had been classified as receiving telephone advice (‘Hear and Treat’) found that the majority of respondents were positive about their experiences of initial call-handler triage and their interaction with the second person they spoke to (presumably clinicians conducting further assessment). 6

Some ambulance service staff expressed reservations that delivering high-quality telephone advice might promote 999 as the first place to call for similar non-emergency health problems in the future. This concern was also stated in a paper detailing a similar secondary telephone triage service for ambulance service patients in Australia21 in which the authors commented that recruiting a new ‘customer base’ for the ambulance service through the provision of telephone advice needs to be considered against its impact on managing existing demand. A number of staff across the three services referred to a need for public education to reduce 999 calls related to minor health problems. Some clinicians advocated communicating educational information to callers when conveying telephone advice.

Strengths and limitations

A key strength of this study was the use of non-participant observation to explore the processes involved in delivering telephone advice. The observation of staff in their ‘natural’ work environment combined with semistructured interviews has enabled us to gain an understanding of organisational variation in the provision of telephone advice within three ambulance services. Although there have been a number of studies using observation to explore the actions of health-care professionals in similar settings,69–71 we are not aware of any other study examining telephone advice in response to 999 calls and specifically focusing on variation. The participation of three ambulance services allowed the comparison of practices across three organisations with recorded variation in the amount of telephone advice and recontacts. The same researcher conducted all data collection with support from members of the study team. Data were presented to the team at various stages for feedback.

In terms of limitations, this was a relatively small-scale study comprising five observation shifts and a small number of interviews at each site. Although this provided a considerable number of qualitative data, it does to some extent represent a ‘snapshot’ of what was happening on the particular occasions that the services were observed. We attempted to capture potential variation across times of the day and days of the week, but there may have been variations over time that we did not capture. For the interview data, participants were either self-selected or had been invited by their manager. Although the interviewees provided a range of views and insights, we acknowledge that these do not necessarily reflect all staff views. Although we have used AQI data to identify participating ambulance services on the basis of recorded variation in the amount of telephone advice and recontacts, we cannot verify that these data have been recorded consistently by services (see Chapter 5) and therefore offer a cautious approach to relating variation in the AQI data with variation in processes across the three services. Finally, it was not possible to identify how NHS 111 affected telephone advice in response to 999 calls and whether this was another source of variation depending on how effective NHS 111 services are at diverting low-acuity calls away from ambulance services.

Implications

The aim of this WP was to explore organisational variation in the provision of telephone advice within clinical hubs. The findings provide insights on the complexity of the process and key areas of variation across the three participating ambulance services. We have considered the potential impact of these variations in practice on rates of telephone advice and recontacts as measured by AQIs. However, having confidence in the accuracy of rates of telephone advice is needed to establish the extent to which the various factors are related to these outcomes in order to understand what model of telephone advice works best for different outcomes.

Variation was identified in relation to the staff involved in telephone advice and highlights the need to consider the specific workforce requirements (staff numbers and skill-mix) for telephone advice to meet the needs of the local population. For example, although staff in all three services identified patients experiencing a mental health crisis as particularly challenging, only one service had mental health clinicians to manage these calls at the time of our research. The issue of variation in response options also seems worthy of attention at the commissioning level; this has already been identified in other research as a key factor influencing the effectiveness of telephone advice and reducing recontacts.

Finally, it is worth considering the role of telephone advice in relation to the wider emergency and urgent care system. Advice to attend an emergency department using their own transport, or transport there by taxi, are potential outcomes from telephone advice. Findings from WP3.2 (see Chapter 7) indicate that hospital admission within 1 day of receiving telephone advice occurred for approximately 2% of calls (see Table 5). Thus, although telephone advice reduces demand for ambulances, some patients may still be admitted that day and it is possible that an ambulance response with discharge at scene may have prevented this.

Not a reason for calling that raises concerns

During qualitative interviews in WP1 and WP4.1 ambulance staff identified health conditions or reasons for calling for an ambulance that resulted in unnecessary conveyance to emergency departments. Respiratory problems were not mentioned as a condition about which staff had particular concerns. In WP1 interviewees frequently raised concerns about the unnecessary conveyance of people with mental health problems.

Initial call: taking the patient’s word

During observation and interviews within three ambulance services offering telephone advice (WP4.1), some call-handlers making initial triage decisions and their managers discussed the challenge of assessing breathing problems during the initial call. They described having to make a triage decision based on patients’ views of their breathing difficulties rather than their own perceptions of patients’ breathing based on listening to patients. They found the short length of time available to make a triage decision problematic and were frustrated at having to send an ambulance to patients who might not need one:

[. . .] they may be having breathing problems, but they are completing sentences, they’re not fighting desperately for every breath. But then they may still receive an ambulance anyway. It’s hard to know until you get to the end of the assessment [. . .] these things like the breathing problem, where I mentioned, somebody says ‘yeah it’s getting worse in the last hour’, we can’t, if they said that is the case, even if they’re breathing really, really well on the phone, we cannot go against what they’ve told [us].

Call-handler, interviewee 53 from WP4.1

Non-conveyance for breathing problems versus other reasons for calling an ambulance

When the reason for the call was available in the CAD data in WP2, 11% (47,737/434,494) of calls from the 10 services were for breathing problems. Breathing problems had lower levels of non-conveyance than all other reasons for calling combined (χ² = 2190, df = 3, p < 0.001): 3.5% compared with 8.0% for calls ending in telephone advice and 27% compared with 31% for patients discharged at scene. It was the sixth most common reason (5%) for telephone advice after being sick or unconscious (26%), having an injury (17%), having abdominal pain (10%), having had a fall (9%) and having cardiovascular problems (5%). It was the fifth most common reason (10%) for discharge at scene after reported falls (21%), an injury (15%), being sick or unconscious (14%) and having cardiovascular problems (11%).

Types of respiratory conditions not conveyed

During WP1 qualitative interviews, ambulance staff mainly discussed discharge at scene of patients with exacerbations of COPD. Asthma was sometimes mentioned, with staff describing how nebulisers could usually be used to sort out problems without the need for conveyance to an emergency department.

Initiatives to facilitate non-conveyance: pathways and individual care packages

Staff in some services in the WP1 qualitative interviews described specific services that could increase rates of discharge at scene or conveyance to alternative services for people with chronic respiratory conditions. In particular, ambulance crews could convey or refer directly to respiratory clinics or specialist services, or they could access patients’ individual care packages. The latter option enabled access to clinical information for a specific patient to allow staff to judge what was clinically normal for the patient, and also allowed staff the opportunity to discuss care with a community matron or specialist nurse to support their decision-making.

Staff often described these referral and information-sharing initiatives as specific to some localities rather than something that was available across the whole ambulance service region. Some paramedics also described informal pathways in their locality, built up through historical relationships with local services rather than through formal agreements. Their descriptions of helpful initiatives that facilitated non-conveyance for people with chronic respiratory problems included having proactive staff in services in the wider emergency and urgent care system and having services that were easy to access, responsive and able to instil confidence in crew members that patients would receive timely and quality care if not conveyed to an emergency department. If these characteristics were not present then interviewees described how paramedics would cease to use available initiatives for referral or conveyance to alternative services and simply convey patients to an emergency department.

Variation between ambulance services

Safety and appropriateness of discharge at scene

In WP3 the linked data set for a single ambulance service showed that 12% (5316/45,310) of calls were made for breathing problems, similar to the 11% found in WP2 for all 10 ambulance services.

Three days after discharge at scene, there was some evidence that more patients had events after non-conveyance than patients for all other conditions:

• 12% (607/5083) of patients with breathing problems recontacted the ambulance service compared with 9% for all other conditions (p < 0.001);

• 13% (678/5083) of patients with breathing problems attended an emergency department compared with 13% for all other conditions (p = 0.137);

• 9% (437/5083) of patients with breathing problems were admitted to hospital compared with 6% for all other conditions (p < 0.001);

• 0.77% (39/5083) of patients with breathing problems died compared with 0.30% for all other conditions (p < 0.001).

Caution is required when interpreting the p-values because small differences have small p-values when numbers are so large.

Patients who died within 3 days were all > 40 years of age. A total of 54% (21/39) of deaths were in patients aged > 80 years.

Summary of findings in the context of other research

Breathing problems were a common reason for calling an ambulance service (11% of calls). This was similar in the USA, where 12% of emergency medical services encounters have been categorised as occurring in response to respiratory distress. 25 Patients were less likely to be non-conveyed and more likely to have subsequent events within 3 days of non-conveyance than patients calling with all other conditions. Factors explaining variation between ambulance services were similar to those found for all conditions reported in Chapters 5 and 6, although the complexity of the emergency and urgent care system in terms of serving large numbers of CCGs resulted in lower discharge-at-scene rates for breathing problems. Internationally, a small number of research studies have been undertaken to describe the epidemiology and outcomes of ambulance service users with respiratory distress or breathing problems. 25–28 As stated in the introduction to this report, these studies focused on patients who were transported to emergency departments rather than those who were not conveyed. A study of the impact of advanced paramedics on admission rates of people calling 999 with breathing difficulties identified a hospital admission rate of 15% (18/119) within 72 hours for those attended by an advanced paramedic. This was higher than the rate found in our study, in which the admission rate was 9% for patients discharged at scene.

Strengths and limitations

We could find no research focusing specifically on non-conveyance of people calling an ambulance service with breathing problems, so this study is a unique contribution to the evidence base. During the qualitative WPs of our study, staff had little to say about this reason for calling; thus, the data are sparse and do not offer the ‘thick’ description associated with high-quality qualitative research. Routine data held in CAD systems did not have sufficient detail to allow an analysis of non-conveyance for respiratory problems or specific types of respiratory problems, such as COPD. Clinical data collected by ambulance crew attending a scene or collected within triage software used in clinical hubs in which telephone advice is offered, would be needed to identify patients with respiratory problems who are not conveyed to an emergency department. The quantitative data reported here for ‘breathing problems’ comprise calls for a range of health conditions with breathing problems, including cardiac-related conditions.

Implications

Although it is interesting to consider the specific reasons for calling an ambulance service, and although non-conveyance rates differ by reason for the call, there was nothing particularly distinct about the non-conveyance of patients calling with breathing problems. Indeed, ambulance staff expressed concerns about other reasons for calling, particularly people calling with psychiatric problems, rather than concerns about the conveyance of people with COPD or asthma. These concerns were not about people with COPD who also had anxiety and depression75 but rather about users of mental health services who were suffering crises.

Our respiratory physician co-author was surprised by these findings, expecting to see much higher levels of non-conveyance for breathing problems than identified here. He viewed these findings from the perspective of a physician working with people with COPD in the community when, in practice, the category studied here included a range of conditions. Nonetheless, he expected to see higher levels of non-conveyance given national guidelines around home care and alternative care options for people with COPD and asthma. His personal experience of developing local community initiatives for managing chronic respiratory conditions was that the ambulance service was not part of their multiagency working and that there was opportunity to expand the use of care at home through working closely with the ambulance service. It would be worthwhile undertaking additional research on identifying and evaluating ways of increasing appropriate non-conveyance for chronic respiratory conditions given that this patient group is likely to make up a significant proportion of callers to the ambulance service.

Is there an optimal rate for non-conveyance?

The recent NAO report suggests that there is a need to identify an optimal rate for non-conveyance. 1 Our study has not identified an optimal rate; this would require a consideration of the safety and cost-effectiveness of different rates of non-conveyance based on a stable case-mix and we have not done this. However, the AQIs show that it is possible to achieve a rate of 10% of calls ending in telephone advice and 35% of calls ending in discharge at scene or conveyance to an alternative service. Some services may be achieving this by including types of calls that policy-makers may not define as non-conveyance. Some services have put considerable effort into increasing their non-conveyance rates and have achieved these, or higher, rates. Policy-makers cannot necessarily expect the rate to continue to increase in the future. Increases in rates will depend on the following:

• How non-conveyance in current performance indicators is counted – calls ending in telephone advice in current published indicators include calls referred to the clinical hub, calls terminated by call-handlers and, for some ambulance services, a mix of other calls such as hoax calls. Non-conveyance could be defined based only on calls sent to a clinical hub that end in telephone advice, which would reduce current rates in AQIs. If all ambulance services adhered to guidance by including only real calls for an ambulance that are ended by either call-handlers or clinicians, then this would reduce current rates in AQIs for some ambulance services. If people who have died prior to the ambulance arriving or while the crew are on scene are removed from AQIs then current rates would reduce.

• Changes in the urgency levels of calls to the ambulance service – demand for the ambulance service has increased over recent years. NHS 111’s direct referrals to the ambulance service have been identified as contributing to this increase in an evaluation of NHS 111 pilots. 80 This relatively new service in the emergency and urgent care system is likely to have changed the urgency levels of cases attended by an ambulance and, therefore, to have increased the potential for discharge at scene. Other changes in case-mix in the future may affect the potential for ambulance services to increase non-conveyance: increases in low-acuity calls will increase the opportunity for telephone advice and discharge at scene.

• The actions ambulance services are allowed to take in response to low-urgency calls – if ambulance services are allowed to retriage referrals from NHS 111, or if ambulance service call-handlers are allowed to refer low-urgency callers to NHS 111,8 then the potential for non-conveyance by ambulance service clinicians will decrease. These actions may be a more appropriate response to some calls than clinical input by an ambulance service either by telephone or face to face.

• The investment services make in non-conveyance processes – investment can include numbers of staff, types of staff, the education of staff, the training offered to staff, how staff are valued within their workplace, and the types of work that staff are asked to undertake. This investment may be helped or hindered by commissioners’ views of the importance of different type of non-conveyance.

• The changing characteristics of the wider emergency and urgent care system – if other services in the emergency and urgent care system, such as emergency departments, acute hospital beds, social services or daytime general practice, are under pressure, then demand for ambulance services may increase, or the ability of ambulance service clinicians to discharge at the scene in the confidence that the patient will have contact with another appropriate service may be reduced. Conversely, pressure at emergency departments with long handover times may increase discharge-at-scene rates.

Should NHS 111 clinical advice and 999 telephone advice in clinical hubs merge?

One can argue that NHS 111 calls referred to a clinician within that service and 999 low-urgency calls sent to a clinical hub for telephone advice could be the same types of calls, which are simply being dealt with by two different services. Having two separate pathways to this telephone advice is essential because having both NHS 111 and 999 allows patients to self-triage their problem as urgent or emergency, which, in the main, they do very well. However, some patients understandably do not self-triage appropriately and, therefore, the relationship between NHS 111 clinical telephone advice and ambulance service clinical telephone advice needs careful consideration, especially given that ambulance services provide NHS 111 in large areas of England. In 2016/17 three ambulance services provided NHS 111 to all their population and a further six provided it to part of their populations. In some ambulance services, there is co-location of NHS 111 and clinical hub clinicians but this does not mean that there is integrated working. There was some evidence within our WP1 interviews that some services were moving towards aspects of integrated working, but there is an opportunity to consider integrating staff and practices in ways that might increase efficiency. Furthermore, other aspects of the relationship between ambulance services and NHS 111 could be reviewed in terms of the ability of the two services to refer to the other. There is the option that ambulance service call-handlers triage directly to NHS 111. This was evaluated when NHS Direct was in operation. 8 It had some problems in terms of a proportion of calls being sent back to the ambulance service, which was less than ideal, as some patients could have felt like they were being bounced around the emergency and urgent care system. 76 However, additional assessments of this practice, with evaluation, would be useful.

Is non-conveyance cost-effective?

We have not undertaken an economic analysis of non-conveyance. A recent systematic review of secondary telephone triage identified that this was an important thing to do. 21 The recent NAO report estimated that ambulance trusts potentially saved themselves £74M and emergency departments £63M in 2015/16 by increasing non-conveyance rates compared with rates recorded in 2011/12. 1 The report points out that this could be an underestimate because it did not include hospital admissions in this calculation, and non-conveyance to emergency departments is associated with lower emergency admission rates. 3 However, the authors also point out that they did not include the additional costs of community services that ambulance services may refer to as part of the non-conveyance decision. 1 Nor did they consider that it may take more time to treat a patient at home than convey them to an emergency department in some circumstances (low-urgency calls took 88 minutes to deal with in 2015/16 compared with 75 minutes in 2011/12). 1 It is also the case that a proportion of patients subsequently get an ambulance or attend an emergency department after the original decision to non-convey (see Chapter 7). Finally, the high quality of the advice provided to patients during ambulance service non-conveyance may result in supply-induced demand, encouraging patients to use the service in preference to alternatives. 13 This is related to the issue of recursivity in urgent care provision whereby how services act encourages future patient demand. 81 In summary, the economics of non-conveyance are complicated but should be addressed in future studies.

Is supply-induced demand likely?

The possibility that supply-induced demand is contributing to increases in demand for ambulance services should be seriously considered. It was a concern raised by interviewees in our study and has been raised as a concern about secondary triage services internationally. 21

How can risks be balanced?

There are a number of risks related to non-conveyance. First, there is a risk that an ambulance service might not be able to meet demand unless non-conveyance is an option. Second, there is a risk that ambulances may not be able to make timely responses to life-threatening conditions because they are responding to low-acuity patients. Third, there is a risk that a patient may die or experience a permanent adverse event as a result of a decision not to convey. Fourth, there is a risk that ambulance crews may be disciplined or lose their jobs if something goes wrong after a decision not to convey. This can lead to ‘disproportionate risk aversion’, as identified in a study of decision-making at transition points in the ambulance service. 12 Fifth, there is a risk that an ambulance service may face a loss of reputation if it receives large numbers of patient complaints about decisions not to convey. There is no right balance of risks; the balance is open to debate and discussion between policy-makers, commissioners, national patient groups such as Healthwatch, and senior managers and frontline staff in ambulance services is necessary.

Formalising definitions of non-conveyance

When policy-makers and researchers assess non-conveyance rates for ambulance services they may wrongly interpret these rates, especially for calls ending in telephone advice. It is easy to interpret published AQIs as rates of patients offered advice over the telephone by clinicians when in practice this includes an array of reasons for the call ending without an ambulance being dispatched. The evidence suggests that policy-makers might benefit from considering what they wish to see measured and putting in place plans to produce AQIs that can measure this. There remains concern about the validity of the data used to calculate AQIs, especially for calls ending in telephone advice. Real variation between services, and especially variation in calls receiving clinical advice that end in telephone advice, could easily be larger or smaller than rates reported in AQIs. A first step to better understand the important aspect of clinical telephone advice by ambulance services is to investigate the development of new indicators that separate out what might be called ‘true hear and treat’ from calls dealt with by call-handlers or calls that should simply never be classified as ending in telephone advice, such as hoax calls. Improving the accuracy of the AQIs, and monitoring adherence to the guidance for calculating AQIs, would help ambulance services to consider real differences between themselves and other services.

Reducing variation so that all ambulance services offer the optimum rate of non-conveyance

It is possible that the ambulance service with the highest rate of non-conveyance at the time of the research did not have the optimal rate of non-conveyance. Establishing what the optimal rate is requires modelling of the cost-effectiveness, and safety and appropriateness, of different rates of non-conveyance. Some approaches to non-conveyance may not be cost-effective, may result in increased mortality or increased severity of illness, or may simply delay conveyance. This study did not set out to establish the ‘right’ rate of non-conveyance and cannot shed light on this issue. A further research study is required to address this, which repeats the linked data analysis undertaken here for one service only for all ambulance services in England and considers the relationship between mortality and subsequent service use for different rates of non-conveyance. A future study would also be necessary to identify the cost-effectiveness of different approaches to non-conveyance.

How to reduce variation

The variation in overall rates of non-conveyance mainly concerned the small number of ambulance services with extreme rates, namely one service with a relatively high rate of discharge at scene and some with relatively low rates of discharge at scene. There was no indication that the high rate was problematic (although safety was not measured for this service and a limited measure of appropriateness was considered). There was evidence of opportunities to increase the discharge-at-scene rate for ambulance services with lower rates. Some of this unwarranted variation lay within the control of the ambulance service and was determined by their workforce policies and motivation to undertake discharge at scene. National policies cannot be the cause of variation because some ambulance services achieved high non-conveyance rates with these policies in operation. However, a change in policies – in particular response-time targets – may help services with low rates to prioritise investment in non-conveyance. A key issue is the considerable amount of variation in ambulance processes in terms of CAD systems, triage software for dispatch, triage software by telephone or face-to-face clinical input, paramedic nomenclature, and how staff with different skills are used. Much of this variation is not necessary and standardisation of these processes could contribute to reductions in the variation in outcomes such as non-conveyance.