a1111111111 a1111111111 Background Some patients calling ambulance services (known as Emergency Medical Services interna- tionally) are not transported to hospital. In England, national ambulance quality indicators show considerable variation in non-transport rates between the ten large regional ambu- OPENACCESS lance services. The aim of this study was to explain variation between ambulance services Citation: O'Cathain A, Jacques R, Stone T, Turner J in two types of non-transport: discharge at scene and telephone advice. (2018) Why do ambulance services have different non-transport rates? A national cross sectional Methods study. PLoS ONE 13(9): e0204508. https://doi.org/ 10.1371/journal.pone.0204508 Mixed model logistic regressions using one month of data (November 2014) from the Com- Editor: Belinda J. Gabbe, Monash University, puter Aided Despatch systems of the ten large regional ambulance services in England. AUSTRALIA Received: April 12, 2018 Results 41% (251 677/615 815) of patients calling ambulance services were not transported to hos- Accepted: September 10, 2018 pital. Most were discharged at scene after attendance by an ambulance (29% n = 182 479) Published: September 21, 2018 and a small percentage were given telephone advice (7% n = 40 679). Discharge at scene Copyright:© 2018 O'Cathain et al. This is an open rates varied by patient-level factors e.g. they were higher for elderly patients, where the rea- access article distributed under the terms of the son for calling was falls, and for patients attended by paramedics with extended skills. Creative Commons Attribution License, which permits unrestricted use, distribution, and These patient-level factors did not explain variation between ambulance services. After reproduction in any medium, provided the original adjustment for patient-level factors, the following ambulance service level factors explained author and source are credited. variation in discharge at scene rates: proportion of patients attended by paramedics with Data Availability Statement: The authors of this extended skills (odds ratio 1.05 (95% CI 1.04, 1.07)), the perception of ambulance service study did not collect the data themselves. The data staff that paramedics with extended skills were established and valued within the workforce are owned by the ten National Health Service (NHS) Ambulance Trusts who agreed to participate (odds ratio 1.84 (1.45, 2.33), and the perception of ambulance service staff that senior man- in the research. The contracts the authors signed agement viewed non-transport as risky (odds ratio 0.78 (0.63, 0.98)). Variation in telephone with these ambulance services stated that the data advice rates could not be explained. provided was solely for the purpose of the project. We do not have permission to share the data. Other researchers can access the data used in this Conclusions article in the same manner as the authors, by Variation in discharge at scene rates was explained by differences in workforce configura- emailing the Research Leads in each ambulance tion and managerial motivation, factors that are largely modifiable by ambulance services. service to request use of the data. The contact details are: email@example.com; tanya. PLOS ONE | https://doi.org/10.1371/journal.pone.0204508 September 21, 2018 1 / 20 Why do ambulance services have different non-transport rates? A national cross sectional study firstname.lastname@example.org; email@example.com. Introduction uk; firstname.lastname@example.org; rwh-tr. Not all patients calling ambulance services, or Emergency Medical Services as they are known email@example.com; Research.Development@nwas. as in some countries, are transported to a hospital. Non-transport can occur because patients nhs.uk; firstname.lastname@example.org; jasmine. email@example.com; julia.williams@secamb. refuse to go to hospital or because ambulance clinicians make decisions not to take patients to nhs.uk; firstname.lastname@example.org. a hospital. In some countries refusal to travel is the main reason for non-transport (sometimes called non-conveyance), but in other countries ambulance services have policies and guide- Funding: This article presents independent research funded by the National Institute for Health lines to allow staff to make decisions about whether to transport patients to hospital. Non- Research (NIHR). The views and opinions transport rates vary by country and research study.[2, 3] For example, a systematic review of expressed by authors in this publication are those falls in older people identified that these varied between 11% and 56%; a recent systematic of the authors and do not necessarily reflect those review that included studies from North America, Europe, Australia, Asia and Africa identified of the NHS, the NIHR, NETSCC, the HS&DR larger variation in non-transport rates for the general patient population of between 4% and programme or the Department of Health. The grant number was 13/54/75. The funders had no role in 94%. study design, data collection and analysis, decision There are different types of non-transport to hospital. In the United Kingdom (UK) there to publish, or preparation of the manuscript. are currently three main types in use: telephone advice to self-manage or contact another ser- Competing interests: The authors have declared vice, given by clinicians based in the ambulance service; discharge at scene after a face-to-face that no competing interests exist. contact with an ambulance crew where the crew offer treatment and advice to patients; or transport by ambulance to a service other than a hospital with an emergency department, such as a walk-in or urgent care centre. These types of non-transport to a hospital are also in use internationally. For example, telephone advice was offered to one in ten calls to the ambulance service in a region in Australia. Ambulance services make an important contribution to health care provision, and must be considered when attempting to understand the quality and safety of health care. Non-trans- port to hospital has the potential to improve the quality of care for a large number of patients each year, although evidence of these benefits is lacking currently. Avoiding a trip to hospital could potentially deliver benefits for patients, emergency departments and ambulance services. For patients, treatment at the scene without transport to a hospital could be the most appropri- ate clinical response to patients' needs, thereby reducing the inconvenience of attending an emergency department far from their home. For emergency departments, ambulance services dealing with patients safely and appropriately without transport to hospital could reduce demand for emergency departments and therefore demand for emergency hospital beds, allowing emergency departments to reduce waiting times for patients who need the level of clinical care they provide. For ambulance services, concerns have been expressed within the media about ambulances needing to queue outside overcrowded emergency departments in England. Non-transport of patients has the potential to remove the delays caused by this queu- ing, allowing ambulances to respond more quickly to other calls. Variation in non-transport rates between ambulance services As well as variation in non-transport rates between countries, there is variation between ambu- lance services within countries. In England, Emergency Medical Services are provided by 11 ambulance services, of which 10 regional services cover 99% of the population of 55 million people. The national policy is to promote non-conveyance (as it is called in England) as a way of offering care close to home. Ambulance services are fulfilling this remit by treating a large proportion of patients at the scene and not transporting them to hospital. Rates of non- transport are monitored as indicators of the quality of ambulance services and are published monthly for each ambulance service. At the end of 2016, the rate of calls ending in tele- phone advice varied between 5% and 17%, the rate of calls sent an ambulance but not trans- ported to a hospital varied between 23% and 51%, and overall non-transport rates varied between 40% and 68% for the 10 large regional services. PLOS ONE | https://doi.org/10.1371/journal.pone.0204508 September 21, 2018 2 / 20 Why do ambulance services have different non-transport rates? A national cross sectional study Variation in practice within health care can raise concerns about quality. It is important to identify the causes of variation in health care and encourage actions to deal with it. Some variation may be warranted due to patient need or preference, and some may be unwarranted. [10, 11] Unwarranted variation can be related to differences in evidence-based practice, prefer- ences held by service providers, or supply of resource. For example, variation in non-transport rates between ambulance services caused by differences in the supply of paramedics with extended skills would be unwarranted. It is also the case that some variation may be modifiable by ambulance services, e.g. workforce configuration, and some variation may not, e.g. the types of problems patients call an ambulance for. Determinants of non-transport A recent systematic review identified a range of patient characteristics associated with non- transport including age, gender, ethnic group, geography, reason for call and vital signs. There was little consistency in the direction or size of effect of patient-level determinants between studies included in this review; for example, higher non-transport rates were associ- ated with males in some studies and females in others. There was consistent evidence of higher rates in children and elderly people, and for some reasons for calling the service such as falls and diabetic hypoglycaemia. Factors influencing the decision-making process around trans- port were related to the patient, professional, healthcare system and availability of decision support tools. Patient influences included the physical health and desire for transport of patients, and healthcare system influences included access to general practitioners and other healthcare services to allow discharge at scene. Ambulance personnel with extra training or extended skills had higher non-transport rates. This latter finding was supported by another systematic review and meta-analysis identifying that paramedics with extended skills had higher non-transport rates than conventional paramedics, although concerns were expressed about whether all potential confounders were adjusted for within analyses of individual stud- ies. The evidence base does not always distinguish between different types of non-transport to hospital and tends to focus largely on discharge at scene. Nor does it address factors affecting variation in non-transport rates between ambulance services and therefore the influence of ambulance service characteristics. Variation in non-transport rates between ambulance ser- vices may be explained by differences in the types of people calling each ambulance service. This would not cause concern because it reflects differences in case-mix between ambulance services. However, variation explained by differences in practices between ambulance services would be a cause for concern because patients with the same health need in different parts of the country would be receiving different health care. A national study of Variation in Ambu- lance Non-conveyance (the VAN study) focused on understanding why there was variation in non-transport rates between ambulance services in England. This mixed methods study included a qualitative interview study of staff in ambulance services to identify factors affecting non-transport and a quantitative study using routine data from ambulance services to identify factors explaining variation between ambulance services. The quantitative component of this study is reported here, aiming to explain variation in different types of non-transport between the ten large regional ambulance services in England. Methods Setting and context In England in the UK Emergency Medical Services are provided by 11 ambulance services within the National Health Service (NHS) to the population of 55 million people. Ten of these PLOS ONE | https://doi.org/10.1371/journal.pone.0204508 September 21, 2018 3 / 20 Why do ambulance services have different non-transport rates? A national cross sectional study ambulance services deal with over 99% of emergency ambulance calls from the population of England. Most calls are from patients calling 999 to request an ambulance±approximately four in five calls±with the remaining calls being passed directly through from patients calling a tele- phone helpline for urgent care called `NHS 111', or through direct referral from a health pro- fessional requesting an emergency ambulance. Calls are taken by non-clinical staff who use software to identify the priority of each call. The ten ambulance services use one of two differ- ent priority dispatch systems: Medical Priority Dispatch System or NHS Pathways. A small percentage of calls that are categorised as low priority are passed to clinicians within each ambulance service for secondary telephone triage. These clinicians, physically based in each ambulance service, use a decision support software to determine whether the patient needs an ambulance or can be offered telephone advice only. Telephone advice includes self-care advice or referral to a service such as primary care; an ambulance is not dispatched. For calls not sent for secondary telephone triage, an ambulance is dispatched and patients may be attended by paramedics, paramedics with extended skills, or emergency care technicians with basic emer- gency training. The ambulance crew assesses and treats patients at scene and can discharge people at scene, transport them to a hospital-based emergency department, or transport them to another health facility such as a walk-in centre. Decisions are made by ambulance crews in conjunction with patients and their families. Local protocols exist for making decisions relat- ing to transport of patients with some health conditions. Sometimes non-transport occurs due to patient refusal to travel. Wider study The research reported here was part of a sequential mixed methods study of a qualitative com- ponent followed by a quantitative component. The quantitative component, a statistical analysis of one month of routine data from the ten ambulance services, is the focus on this paper. However, because the qualitative findings were used within the quantitative analysis reported here, a brief description of the qualitative research is given. Further details of this quali- tative component are reported elsewhere. A qualitative interview study was undertaken with five staff from each of the ten large ambulance services in England (totalling 49 interviews). The aim was to identify factors perceived to affect different types of non-transport specifically within each interviewee's ambulance service. Three types of staff were purposively selected to represent different perspectives within each ambulance service. Two managers were selected to offer a strategic view of non-transport within their ambulance service e.g. workforce configura- tion and training. These included Operational and Medical Directors and clinicians managing teams providing non-transport. Two paramedics were selected to offer the perspectives of staff making decisions about non-transport for individual patients. The healthcare commissioner for each ambulance service was selected to offer an external view of how the ambulance service engaged with non-transport. In England, healthcare commissioners hold budgets to buy health services for their geographically-based population. Each ambulance service has a lead healthcare commissioner who negotiates contracts about how the ambulance service will provide health care, including non-transport. Ambulance service-level factors were derived from this qualita- tive research and then tested in the quantitative component of the study (see section on `ambu- lance service-level factors' for details). The National Research Ethics Service Committee North West±Greater Manchester West (REC reference 14/NW/1388)±approved this study. Data on non-transport rates Routine data was requested from the Computer Aided Despatch systems of each of the 10 large ambulance services in England. Data on all emergency calls that received a telephone or PLOS ONE | https://doi.org/10.1371/journal.pone.0204508 September 21, 2018 4 / 20 Why do ambulance services have different non-transport rates? A national cross sectional study face-to-face response from the ambulance service was requested for the single month of November 2014. The requests for data were made in May 2015. The patient transport service, providing pre-planned non-emergency transport, was not included in the requested data set. This single month of data was expected to consist of around 540,000 calls to the ambulance service (based on published ambulance quality indicators for England in November 2014). Selection of a longer time period would have resulted in a very large dataset, with statistical tests identifying very small differences as statistically significant. Selection of a smaller time period such as a week would have raised concerns about the potential for some ambulance ser- vices to experience unusual events within that week as a cause of variation between ambulance services. Call outcomes were telephone advice only, discharge at scene, transport to hospital with an emergency department, and transport to an alternative service. All datasets were fully anonymised and deidentified by ambulance services prior to sending to the research team. Patient-level factors Potential patient-level factors were identified from the literature and the qualitative interviews undertaken as part of the wider study. Routine data was not available for some of the factors identified, in particular patient desire for transport to hospital. Routine data on factors was identified from three sources (see Table 1). First, the ambulance service Computer Aided Des- patch (CAD) systems. Second, data from the 2011 census for England. Variables identified from the literature and qualitative research as affecting non-transport were identified within the publically available census data. These variables were linked to the CAD data using the lower super output area (LSOA) of each call which was provided by ambulance services with the CAD data. LSOAs are the smallest geography for which some census data are available, with a mean population size of around 1600. The LSOA based census factors described the area from which the call was made to an ambulance service rather than characteristics of the individual patient. Third, each ambulance service holds routine data about the skill-mix of crews attending each incident. We requested that the highest skill-mix of any crew attending each incident be linked to the CAD data. Labels used to describe crew members differed by ambulance service so the research team standardised labels (see Table 1 for details). Ambulance service-level factors Data on ambulance-service level factors were identified from three sources (see Table 1). First, three variables were identified from each ambulance service: from the CAD system, routine data held on skill-mix of crews attending incidents, and from the information team at each ser- vice (see Table 1). Second, eight variables were identified from the qualitative interview study in the wider mixed methods study in the following way. Interviews were transcribed verbatim. Transcripts were read to identify factors perceived to affect non-transport within that ambu- lance service. Each transcript was coded using this set of factors. For each factor, all relevant transcript excerpts from interviewees within an individual ambulance service was read and a summary statement about the factor for that ambulance service was produced. This was repeated for each factor and each ambulance service. Each factor was categorised as one of three levels for each ambulance service±positive/high, no views/mixed views, or negative/low. The middle category was used where there was conflicting information from different inter- viewees within an ambulance service about a specific factor, or where there was not enough information to make a judgement. This process is called `quantitizing'  and was under- taken by a single researcher who was blind to the non-transport rates of each ambulance ser- vice. A matrix was produced with ambulance services as columns and factors as rows, with cells categorised as 1,2 or 3 . This matrix was passed to the statistician for inclusion in the PLOS ONE | https://doi.org/10.1371/journal.pone.0204508 September 21, 2018 5 / 20 Why do ambulance services have different non-transport rates? A national cross sectional study Table 1. Description of factors tested in the regression. Level of Factor Groups Description, including justification for groups Source of data % Missing values factor Patient Patient age 0±2, 3±10, 11±20, 21±30, 31± Age was grouped because age is sometimes given CAD 7.4% 40, 41±50, 51±60, 61±70, 71±80, by a caller who guesses the patient age. 0±2 group 81±90, > 90 was used because some ambulances services reported different non-transport policies for children under 2 years old. Patient sex Female, Male Only two categories available CAD 6.2% Time of call Out of Hours, In Hours Time of call was dichotomised into `In CAD 0% hours = 8am-6pm weekdays' and `Out of hours = all other times'. It was grouped because qualitative interviews identified perceptions that the availability of services that facilitated non- transport was better during normal working hours for health services. Source of call 999, 111 In England patients call 999 directly or are passed CAD 0% to 999 after calling the urgent care telephone service NHS 111 Type of caller Patient All calls from a patient, family, friend or CAD Variable missing for one Health practitioner bystander were labelled as `patient'. Community ambulance service nurses and general practitioners can call for an (11.7%) emergency ambulance on behalf of patients and were labelled `health practitioner' Reason for call Falls, Abdominal Pain, There are many codes for reason for call and they CAD 29.4% Breathing difficulties, differ by the two triage software systems used by Cardiovascular, Fitting, Injury, ambulance services. A small working group of Psychiatric, Sick or clinicians, experts within the research team, and a Unconscious, Other research paramedic met to develop common categories from the two triage software systems and identified specific codes where a large proportion were not-transported; all other reasons were classed as `other' Assessment of Red 1 & 2 (emergency), Green Codes such as Red 1 and 2 were grouped together CAD 1.1% urgency 1 & 2 (urgent), Green 3 & 4 due to small numbers (low acuity) Indices of Multiple Quintiles Q5 (Least Deprived), The IMD is the official measure of deprivation of Census 3.7% Deprivation (IMD) Q4, Q3, Q2, small areas in England, ranking every small area Q1 (Most Deprived) from 1 (most deprived) to over 30,000 (least deprived). It is based on seven aspects of deprivation including income and employment. Quintiles were used because this is a common approach to using IMD in regressions. The variable describes the area from which the call was made. Urban-rural status Urban, Rural The Rural Urban Classification is an official Census 3.7% statistic used in the census to distinguish rural and urban areas. There are 4 urban and 6 rural categories. The urban/rural dichotomy was used. Rural areas are outside settlements with more than 10,000 resident population. The variable describes the area from which the call was made. % population with Quintiles The percentage of the population in the small area Census 3.7% no central heating from which the call was made that reported having no central heating in the census. Used to represent the quality of housing of a patient. % population living Quintiles The percentage of the population in the small area Census 3.7% alone from which the call was made that reported living alone in the census. Used to represent informal support unavailable for patients. (Continued ) PLOS ONE | https://doi.org/10.1371/journal.pone.0204508 September 21, 2018 6 / 20 Why do ambulance services have different non-transport rates? A national cross sectional study Table 1. (Continued ) Level of Factor Groups Description, including justification for groups Source of data % Missing values factor % population with Quintiles The percentage of the population in the small area Census 3.7% English not as their from which the call was made that reported not first language having English as their first language in the census. Used to represent ethnic groups where communication might affect decision to transport. % population with Quintiles The percentage of the population in the small area Census 3.7% severe long term from which the call was made that reported illness having a severe long term illness. Used to represent health status of patients. Skill-mix Paramedic, Paramedic Each ambulance service uses different labels and Ambulance Variable missing for one extended skills, Other codes for the skill-mix of crew attending the routine data ambulance service and scene. A small working group of clinicians, some missing values for experts within the research team, and a research all other services paramedic met to develop common categories for (14.4%) skill-mix. Each ambulance service was requested to link the labels and codes they used to these common categories, based on the highest skill- mix of attending ambulance crew. This data was not available in CAD but held in another routine dataset by ambulance services and was linked to the CAD data by the ambulance service. Paramedics with extended skills, or advanced paramedics as they are called in England, are defined by the national College of Paramedics as experienced autonomous paramedics with masters degrees in a subject relevant to their practice. `Other' mainly included emergency medical technicians and a small number of doctors and nurses. Ambulance Workforce % patients attended by The patient-level variable on skill-mix was used to Ambulance Missing for one service service configuration paramedics with extended skills create an ambulance service-level variable of the routine data (14.4%) percentage of calls attended by paramedics with extended skills to represent the size of the workforce made up of paramedics with extended skills within each ambulance service Complexity of Medium, high, low In the qualitative interviews interviewees CAD 0% emergency and described how ambulance crews having to move urgent care system between areas run by different healthcare commissioners (these are called clinical commissioning groups) reduced the ability to discharge at scene because each area had different services with different referral pathways which ambulance crew needed to know about in order to discharge at scene. Also, although a lead healthcare commissioner worked with the ambulance service to devise a contract for providing non-transport, sometimes individual commissioners from these clinical commissioning groups set up their own contracts with the ambulance service. The CAD system identified the number of clinical commissioning groups covered by each ambulance service to represent the complexity of the external system that an ambulance service had to deal with. Type of triage AMPDS One ambulance service used two types of software Ambulance 0% software NHS Pathways in different geographical regions and was coded as Information using the software triaging the majority of callers Team (Continued ) PLOS ONE | https://doi.org/10.1371/journal.pone.0204508 September 21, 2018 7 / 20 Why do ambulance services have different non-transport rates? A national cross sectional study Table 1. (Continued ) Level of Factor Groups Description, including justification for groups Source of data % Missing values factor Stability of the No changes, Significant Staff perceptions of changes occurring to senior Qualitative 0% organisation changes management or the effects of external assessments study of service quality Organisational No view or mixed views, Staff perceptions of motivation of senior Qualitative O% motivation for non- viewed as opportunity, risk management to undertake non-transport study transport aversion How extended No view or mixed views, in Staff perceptions of whether paramedics with Qualitative 0% paramedics are used limited capacity, established extended skills were used in the ambulance study and valued service Fear of retribution No evidence, Low levels of fear, Staff perceptions of level of fear of retribution Qualitative 0% Evidence of fear amongst paramedics if non-transport resulted in study adverse events Provision of services Inconsistent views, lacking in Staff perceptions of availability of services in the Qualitative 0% in the wider system provision, good provision wider emergency and urgent care system that study facilitated non-transport Connectivity with Inconsistent views, lacking Staff perceptions of how connected an ambulance Qualitative 0% wider system connectivity, good connectivity service was to other services within the wider study emergency and urgent care system Commissioners Worked with some localities Staff perceptions of the quality of the relationship Qualitative 0% only, poor, good between the ambulance service and their health study care commissioners Telephone advice Limited use, negative views, Staff perceptions of provision of this type of non- Qualitative 0% enthusiastic senior transport within their ambulance service. study management Applicable to telephone advice analysis only Cost per call Medium, low, high Cost per call was calculated by the National Audit National Audit 0% Office by dividing an ambulance service's urgent Office and emergency care income by the number of calls presented to its switchboard. It was calculated to represent cost-effectiveness. It was tested in the regression because it was available and varied by ambulance service rather than there being a clear rationale for its potential influence on non-transport. 10 ambulance services grouped into three groups of high, medium and low Cost per face-to-face Medium, low, high See explanation for `cost per call'. 10 ambulance National Audit 0% attendance services grouped into three groups of high, Office medium and low Staff absence rate Medium, low 10 ambulance services grouped into three groups National Audit 0% of high, medium and low. No service had a higher Office rate so only two categories were used % frontline staff with Medium, high The percentage of frontline staff with extended National Audit 0% extended skills skills is similar to the workforce configuration Office variable above. However, it also includes staff offering telephone advice and does not measure the percentage of incidents attended by different skill-mix. 10 ambulance services grouped into three groups of high, medium and low Income per head of Medium, low, high 10 ambulance services grouped into three groups National Audit 0% population of high, medium and low Office https://doi.org/10.1371/journal.pone.0204508.t001 regression (see Table 1). Third, advantage was taken of a published national audit of ambu- lance services in England, where tables of ambulance characteristics were presented.  Five variables were identified from this National Audit Office report where there was variation between ambulance services in factors potentially important to non-transport (see Table 1). Variables were categorised as high, medium and low by a researcher. PLOS ONE | https://doi.org/10.1371/journal.pone.0204508 September 21, 2018 8 / 20 Why do ambulance services have different non-transport rates? A national cross sectional study Analysis Ambulance services linked their CAD data and their routine data on skill-mix. They included LSOA in the dataset and sent the data to the research team. The research team made efforts to reduce bias in these datasets by examining the data from each service in detail and interacting with information staff in each ambulance service to understand how CAD variables were coded and calculated. The research team used LSOA to link census data to the CAD data. Although the census variables were at small-area level rather than patient level, they were treated as patient level in the regression. The ambulance service-level variables (from the quali- tative interviews, National Audit Office and routine ambulance data) were added to this data- set to allow a multi-level analysis to be undertaken on patient and ambulance-level variables. Specific hypotheses were not tested due to the variation in direction of effect of factors identi- fied in a recent systematic review. Mixed effects logistic regression models with ambulance service as a random intercept were fitted in the statistical software R using the lme4 library. Regressions were undertaken sep- arately for calls ending in discharge at scene and calls ending in telephone advice. No analysis was undertaken on calls ending in transport to facilities other than hospitals with emergency departments because these included such a mix of acuities: low-acuity calls taken to minor injury units, end-of-life calls taken to hospices, and high acuity calls taken to specialist tertiary care e.g. patients with suspected stroke to a hyper-acute stroke unit. The denominator for the discharge at scene analysis was all calls sent an ambulance. The denominator for the telephone advice analysis was all calls. There were two stages to the analysis. The first stage was to investigate which patient-level variables were statistically associated with non-transport. This was undertaken for discharge at scene because this accounted for the majority of non-transport. All patient-level variables listed in Table 1 were tested in the regression with the exception of one variable. Type of caller (patient or health practitioner) was removed from the analysis because there was no data avail- able on this variable for one ambulance service and no healthcare professional calls recorded for another. The skill-mix variable was also missing for one of the ambulance services but, as this was identified in the evidence base as an important factor in non-transport, the analysis was conducted on the 9 ambulance services that provided skill-mix (the primary analysis) and then the final models were refitted on data from all 10 ambulance services with the skill-mix variable excluded (sensitivity analysis). Using a likelihood ratio test, all single patient-level vari- ables that were significantly related to the outcome at p<0.1 were entered into the multivari- able model building analysis. A backward elimination strategy with variables removed at P>0.05 was used to identify the subset of variables that independently predicted the outcome. As a final check, all variables excluded in the first stage of model selection were added to the model to see if they became important in the presence of others. First order interactions between predictor variables were investigated but, due to the large sample size, small effects were statistically significant and made the models difficult to interpret. Because of this all inter- actions were removed from the models. The patient-level variables identified at this stage were used to adjust for case-mix when testing ambulance service-level variables in both regressions (for discharge at scene and for telephone advice). The second stage of the analysis was to consider ambulance service-level variables. All ambulance service-level variables listed in Table 1 were tested in the regression. Each ambu- lance service-level variable was added to the model with patient-level variables. All ambulance service-level variables from either the model with case-mix included or excluded with p<0.1 were entered into this stage of the analysis. All of these variables were added to the model and backward elimination was used to removed ambulance service-level variables with P>0.05. PLOS ONE | https://doi.org/10.1371/journal.pone.0204508 September 21, 2018 9 / 20 Why do ambulance services have different non-transport rates? A national cross sectional study Variability between ambulance services was assessed visually using caterpillar plots. The odds ratios of the random intercepts (ambulance service) were plotted for (i) the null model or intercept only model (i.e. the model with no patient or ambulance service-level variables); (ii) the model with only patient-level variables; and (iii) the model with both levels of variables. By comparing these plots it was possible to see how variability between random intercepts (ambu- lance services) changed as different levels of variables were added. The approach used to investigate variation in discharge at scene rates was repeated for tele- phone advice only rates. For this analysis, urgency level, skill-mix and source of call were excluded because they were not relevant to these types of calls or data on these variables were not available: most telephone advice only calls are categorised as low acuity, there was little or no skill-mix data available, and national policy at November 2014 was that calls from a particu- lar source (NHS 111) should not be re-triaged by clinicians in ambulance services so only calls sourced from 999 were included in the dataset. Results Description of types of non-transport The dataset of 615 815 calls was larger than the one reported in national ambulance quality indicators for non-transport (538,865) in November 2014 because it included calls from NHS 111 which are excluded from the denominator of the national ambulance quality indicators. 41% (251 677/615 815) of patients calling ambulance services were not transported to hospital. Most non-transported patients were discharged at scene after attendance by an ambulance (29% n = 182 479 of all calls). Discharge at scene rates varied between 21% and 46% by ambu- lance service. A small percentage were given telephone advice only (7% n = 40 679 of all calls) or transported to an alternative service (4% n = 28 519 of all calls). Telephone advice only rates varied between 3% and 11% by ambulance service. Denominators and missing data The reason for calling an ambulance was missing for a large proportion of calls (Table 1). Con- sideration was given to not including this variable in the analysis but it is a strong and consis- tent predictor of non-transport so it was included. For the discharge at scene analysis, the denominator was all patients with face to face contact with an ambulance crew including dis- charged at scene, transported to an emergency department and transported to an alternative service. A complete case analysis was undertaken on 343 875 patients of the 546 916 patients in the nine ambulance services with skill-mix data. For the sensitivity analysis on all ten ambu- lance services the complete case analysis was based on 370 656 of the 575 136 calls. For tele- phone advice, the denominator was all calls. A complete case analysis was undertaken on 400 630 of the 615 815 calls. Determinants of discharge at scene rates Only nine ambulance services were included in the `discharge at scene' analysis because a key variable `skill-mix' was not available for one ambulance service (see Table 1). There was con- siderable variation in discharge at scene rates between the nine ambulance services which pro- vided data on skill-mix (see Fig 1A). When patient-level variables were tested, rates varied by age group (lower for children aged below two and higher for elderly patients), were slightly higher for men, lower in hours (8am-6pm weekdays), lower for any reason for calling that was not about falls, higher for calls classified as low urgency, higher for calls sourced from NHS 111, higher for calls attended by paramedics with extended skills, and higher for calls made PLOS ONE | https://doi.org/10.1371/journal.pone.0204508 September 21, 2018 10 / 20 Why do ambulance services have different non-transport rates? A national cross sectional study Fig 1. Variation in discharge at scene rates between ambulances services (based on 9 ambulance services with skill-mix data). https://doi.org/10.1371/journal.pone.0204508.g001 PLOS ONE | https://doi.org/10.1371/journal.pone.0204508 September 21, 2018 11 / 20 Why do ambulance services have different non-transport rates? A national cross sectional study from areas of social deprivation (see Table 2). Although different patient-level variables were associated with variation in discharge at scene rates, they did not explain the variation between ambulance services (see Fig 1B). Patient-level variables in the final model were age group, sex, time of call, source of call. reason for call, assessment of urgency, % population with severe long term illness, indices of multiple deprivation and skill-mix (see Table 2). When ambulance ser- vice-level variables were tested, variation in discharge at scene rates reduced considerably (see Fig 1C). Variation between ambulance services was explained by three ambulance service-level vari- ables after adjustment for patient-level variables: discharge at scene rates were higher for ambu- lance services with higher proportions of calls attended by paramedics with extended skills, and ambulance services where staff described paramedics with extended skills as an established and valued part of the workforce; rates were lower for ambulance services where staff described the management as risk averse to non-transport (see Table 2 for variables included in final model). A sensitivity analysis was undertaken by excluding skill-mix variables and thereby including all 10 ambulance services. This regression failed to explain variation between ambulance services, although the odds ratios of both patient-level and ambulance service-level variables were very similar to those in the regression that included skill-mix (final model shown in Table 2). Determinants of telephone advice rates There was no skill-mix data available for telephone advice only calls so the primary analysis was undertaken on all 10 ambulance services. There was considerable variation between ambulance services for their telephone advice rates (see Fig 2A). When patient-level variables were tested, rates were lower for older age groups, slightly higher for men, lower for calls made in-hours and higher where the reason for the call was abdominal pain (see Table 3 for patient- level variables in final model). Variation between ambulance services was not explained by patient-level variables (see Fig 2B). No ambulance service-level variables were statistically sig- nificant. Therefore variation in telephone advice rates between ambulance services could not be explained by the variables tested here. Discussion Summary of findings 41% (251 677/615 815) of patients calling ambulance services were not transported to hospital. Most non-transported patients were discharged at scene after attendance by an ambulance (29% n = 182 479) and a small percentage were given telephone advice only (7% n = 40 679). Discharge at scene rates were higher for older patients, men, patients calling out of hours, where the reason for calling was falls, calls categorised as non-emergency, patients calling from areas of social depri- vation, and patients attended by paramedics with extended skills. These patient-level factors did not explain variation between ambulance services. After adjustment for patient-level factors, three ambulance service-level factors explained variation in discharge at scene rates. Rates were higher for services with a higher proportion of patients attended by paramedics with extended skills, and where the perception of ambulance service staff was that paramedics with advanced skills were established and valued within the workforce. Rates were lower where the perception of ambulance service staff was that senior management viewed non-transport as risky. Variation in telephone advice rates between ambulance services could not be explained by the variables tested here. Context of other research The patient-level factors explaining variation in non-transport rates in this study were similar to those found in a recent systematic review of non-transport: age, gender, and reason for call. PLOS ONE | https://doi.org/10.1371/journal.pone.0204508 September 21, 2018 12 / 20 Why do ambulance services have different non-transport rates? A national cross sectional study Table 2. Factors explaining variation in discharge at scene rates. Variable 9 ambulance services (with skill-mix data) 10 ambulance services (no skill-mix data) Patient-level No of calls in No of calls resulting in Unadjusted Adjusted P-Value Unadjusted Adjusted P-Value denominator discharge at scene Odds Ratio+ Odds Ratio Odds Ratio+ Odds Ratio (95% CI) (95% CI) (95% CI) (95% CI) Age group 0±2 13,116 2,982 1 1 - 1 1 - 3±10 9,204 2,523 1.30 (1.22, 1.35 (1.27, <0.001 1.29 (1.22, 1.35 (1.27, <0.001 1.37) 1.44) 1.37) 1.43) 11±20 24,133 7,764 1.60 (1.53, 1.66 (1.58, <0.001 1.60 (1.53, 1.66 (1.58, <0.001 1.68) 1.75) 1.67) 1.74) 21±30 36,413 12,453 1.82 (1.75, 1.91 (1.82, <0.001 1.81 (1.73, 1.89 (1.81, <0.001 1.90) 2.00) 1.88) 1.98) 31±40 29,271 9,345 1.66 (1.59, 1.76 (1.68, <0.001 1.65 (1.58, 1.75 (1.67, <0.001 1.73) 1.84) 1.72) 1.83) 41±50 31,789 9,722 1.53 (1.46, 1.63 (1.55, <0.001 1.52 (1.46, 1.62 (1.55, <0.001 1.59) 1.71) 1.58) 1.70) 51±60 30,517 9,214 1.49 (1.43, 1.60 (1.52, <0.001 1.47 (1.41, 1.58 (1.51, <0.001 1.56) 1.68) 1.54) 1.66) 61±70 35,076 10,632 1.46 (1.40, 1.52 (1.45, <0.001 1.44 (1.38, 1.50 (1.44, <0.001 1.53) 1.60) 1.50) 1.58) 71±80 50,162 16,039 1.57 (1.51, 1.54 (1.47, <0.001 1.53 (1.47, 1.52 (1.45, <0.001 1.63) 1.62) 1.60) 1.59) 81±90 61,972 22,920 1.91 (1.84, 1.75 (1.67, <0.001 1.88 (1.81, 1.73 (1.66, <0.001 1.99) 1.83) 1.95) 1.81) > 90 22,222 9,248 2.29 (2.19, 1.99 (1.89, <0.001 2.25 (2.15, 1.97 (1.87, <0.001 2.39) 2.09) 2.35) 2.07) Sex Female 163,599 52,053 1 1 - 1 1 - Male 180,276 60,789 1.08 (1.06, 1.02 (1.00, 0.034 1.07 (1.06, 1.01 (1.00, 0.064 1.10) 1.03) 1.09) 1.03) Time of call Out of Hours 224,933 75,996 1 1 - 1 1 - In Hours 118,942 36,846 0.87 (0.85, 0.87 (0.85, <0.001 0.87 (0.85, 0.87 (0.86, <0.001 0.88) 0.88) 0.88) 0.88) Reason for call Falls 58,411 25,828 1 1 - 1 1 - Abdominal Pain 8,321 2,046 0.36 (0.35, 0.34 (0.32, <0.001 0.36 (0.35, 0.34 (0.32, <0.001 0.38) 0.36) 0.38) 0.36) Breathing difficulties 43,341 11,729 0.49 (0.48, 0.62 (0.60, <0.001 0.49 (0.48, 0.62 (0.60, <0.001 0.50) 0.63) 0.51) 0.63) Cardiovascular 58,189 13,040 0.37 (0.36, 0.46 (0.45, <0.001 0.37 (0.36, 0.46 (0.45, <0.001 0.38) 0.48) 0.38) 0.47) Fitting 16,625 4,766 0.53 (0.51, 0.64 (0.62, <0.001 0.54 (0.52, 0.65 (0.62, <0.001 0.55) 0.67) 0.56) 0.67) Injury 44,616 14,099 0.53 (0.52, 0.55 (0.54, <0.001 0.54 (0.53, 0.56 (0.55, <0.001 0.55) 0.57) 0.55) 0.58) Psychiatric 9,227 3,293 0.74 (0.70, 0.74 (0.71, <0.001 0.74 (0.71, 0.74 (0.70, <0.001 0.77) 0.78) 0.77) 0.77) Sick or Unconscious 52,478 17,063 0.63 (0.61, 0.71 (0.69, <0.001 0.63 (0.61, 0.70 (0.68, <0.001 0.64) 0.72) 0.64) 0.72) Other 52,667 20,978 0.70 (0.69, 0.77 (0.75, <0.001 0.67 (0.66, 0.74 (0.73, <0.001 0.72) 0.80) 0.69) 0.76) Urgency Level (Continued ) PLOS ONE | https://doi.org/10.1371/journal.pone.0204508 September 21, 2018 13 / 20 Why do ambulance services have different non-transport rates? A national cross sectional study Table 2. (Continued ) Variable 9 ambulance services (with skill-mix data) 10 ambulance services (no skill-mix data) Patient-level No of calls in No of calls resulting in Unadjusted Adjusted P-Value Unadjusted Adjusted P-Value denominator discharge at scene Odds Ratio+ Odds Ratio Odds Ratio+ Odds Ratio (95% CI) (95% CI) (95% CI) (95% CI) Red 1 & 2 (emergency) 167,581 45,759 1 1 - 1 1 - Green 1 & 2 (urgent) 139,292 50,417 1.45 (1.43, 1.25 (1.23, <0.001 1.44 (1.42, 1.24 (1.22, <0.001 1.48) 1.27) 1.47) 1.26) Green 3 & 4 (low acuity) 37,002 16,666 2.25 (2.20, 1.90 (1.86, <0.001 2.21(2.16, 1.87 (1.82, <0.001 2.30) 1.95) 2.26) 1.91) Source of call 999 336,915 109,995 1 1 - 1 1 - NHS 111 6,960 2,847 1.33 (1.27, 1.46 (1.38, <0.001 1.35 (1.29, 1.43 (1.37, <0.001 1.40) 1.53) 1.41) 1.50) Skill-mix Paramedic 278,804 88,851 1 1 - - - - Paramedic with extended skills 23,189 10,111 1.36 (1.33, 1.38 (1.34, <0.001 - - - 1.40) 1.42) Other 41,882 13,880 1.00 (0.98, 0.96 (0.94, 0.002 - - - 1.03) 0.99) % severe long term illness Q1 (Lowest) 68,123 25,529 1 1 - 1 1 - Q2 70,079 22,553 0.96 (0.94, 0.95 (0.93, <0.001 0.96 (0.94, 0.95 (0.93, <0.001 0.98) 0.97) 0.99) 0.98) Q3 70,302 23,443 1.01 (0.94, 0.99 (0.96, 0.301 1.00 (0.98, 0.98 (0.96, 0.162 0.98) 1.01) 1.02) 1.01) Q4 70,225 23,334 1.00 (0.98, 0.97 (0.95, 0.039 1.00 (0.98, 0.97 (0.95, 0.048 1.03) 1.00) 1.03) 1.00) Q5 (Highest) 65,146 20,983 1.05 (1.02, 1.01 (0.98, 0.368 1.04 (1.03, 1.00 (0.98, 0.798 1.08) 1.04) 1.07) 1.03) Indices of Multiple Deprivation Q5 (Least Deprived) 44,140 15,378 1 1 - 1 1 - Q4 47,526 18,953 1.00 (0.98, 1.02 (0.99, 0.265 1.01 (0.99, 1.02 (0.99, 0.111 1.03) 1.04) 1.04) 1.05) Q3 46,859 22,807 1.03 (1.00, 1.04 (1.02, 0.002 1.03 (1.01, 1.05 (1.03, <0.001 1.05) 1.07) 1.06) 1.08) Q2 46,891 25,994 1.01 (0.99, 1.05 (1.02, 0.001 1.02 (1.00, 1.05 (1.03, <0.001 1.04) 1.08) 1.04) 1.08) Q1 (Most Deprived) 44,163 29,710 0.99 (0.97, 1.05 (1.02, 0.002 1.00 (0.98, 1.05 (1.02, <0.001 1.02) 1.08) 1.03) 1.08) Ambulance service-level % calls attended by paramedics 343,875 112,842 1.04 (1.02, 1.05 (1.04, <0.001 - - - with extended skills 1.08) 1.07) Organisational Motivation No view or mixed views 260,597 85,540 1 1 - 1 1 - Opportunity 36,625 16,925 1.74 (1.36, 0.80 (0.58, 0.176 1.88 (1.14, 1.38 (0.64, 0.407 2.23) 1.11) 3.09) 2.97) Risk Aversion 46,653 10,377 0.58 (0.47, 0.78 (0.61, 0.036 0.63 (0.42, 0.70 (0.39, 0.255 0.72) 0.98) 0.93) 1.29) How extended paramedics are used No view or mixed views 138,210 40,360 1 1 - 1 1 - In limited capacity 153,896 49,926 1.16 (0.96, 1.06 (0.91, 0.460 1.29 (1.11, 1.19 (0.75, 0.460 1.41) 1.24) 1.49) 1.87) (Continued ) PLOS ONE | https://doi.org/10.1371/journal.pone.0204508 September 21, 2018 14 / 20 Why do ambulance services have different non-transport rates? A national cross sectional study Table 2. (Continued ) Variable 9 ambulance services (with skill-mix data) 10 ambulance services (no skill-mix data) Patient-level No of calls in No of calls resulting in Unadjusted Adjusted P-Value Unadjusted Adjusted P-Value denominator discharge at scene Odds Ratio+ Odds Ratio Odds Ratio+ Odds Ratio (95% CI) (95% CI) (95% CI) (95% CI) Established and valued 51,769 22,556 1.71 (1.43, 1.82 (1.31, <0.001 1.95 (1.57, 1.52 (0.81, 0.191 2.05) 2.34) 2.42) 2.84) + these are odds ratios from a univariable logistic random effects model with ambulance service as a random intercept rather than a true unadjusted value (i.e. from a logistic regression model with no random effect) https://doi.org/10.1371/journal.pone.0204508.t002  The higher rate of non-transport found in this study for paramedics with extended skills has been found in a recent systematic review, although the odds ratio of 1.4 was considerably lower than the pooled odds ratio of 10.5 found in the review. However, a number of poten- tial confounders were adjusted for in the study reported here, unlike in individual studies included in the systematic review. A large number of variables were tested but they did not explain variation in telephone advice rates. Telephone advice offered to patients calling for an emergency ambulance is under researched. A recent systematic review identified such a small number of studies that the con- clusion was that it was difficult to generalise from the evidence base. Strengths and limitations This is one of the first studies attempting to explain variation in non-transport rates between different ambulance services. The dataset was large and a wide range of variables was tested. There were six limitations. First, although there is national guidance on the types of calls to cat- egorise as non-transported, ambulance services have different CAD systems and different staff produced the data sets amalgamated for this study. It is possible that some variation between ambulance services was explained by these differences rather than actual differences in rates. The research team attempted to check for differences and standardise the data but it was not possible to remove all differences. Second, data was missing for some variables for some ambu- lance services and a complete case analysis was undertaken. This reduced the size of the dataset and limited the analysis to cases were characteristics were available. This is likely to have intro- duced bias but the size and direction of bias is difficult to determine given the limited under- standing of the types of cases more likely to have missing data. Third, there is little consistency of language when describing skill-mix within ambulances services. We addressed this by send- ing each ambulance service a set of generic skill-mix labels to link with their skill-mix codes but there was still room for different interpretation of the skill-mix types by the different ambulance services. Fourth, some of the ambulance-level variables tested were derived from qualitative interviews with ambulance staff. Only five interviews were undertaken within each ambulance service so data saturation was not necessarily achieved, leaving concerns that only a partial view of non-transport within each ambulance service was obtained. It is also possible that knowledge of their ambulance service performance for non-transport rates relative to other ambulance services may have led some interviewees to describe their ambulance in a way that matched their rate. It is also possible that coding of qualitative interview data was influenced by the research team's knowledge of the rates for each ambulance service even though researchers were blinded to this until the coding of qualitative analysis was complete. It is not possible to estimate the magnitude of any bias introduced by this but the regression is likely to overestimate the amount of variation explained by these factors. Fifth, some factors PLOS ONE | https://doi.org/10.1371/journal.pone.0204508 September 21, 2018 15 / 20 Why do ambulance services have different non-transport rates? A national cross sectional study Fig 2. Variation in telephone advice rates between ambulances services (based on 10 ambulance services). https://doi.org/10.1371/journal.pone.0204508.g002 have been identified elsewhere as influencing paramedic decision-making in non-transport and were not tested here due to lack of availability of data e.g. pressures of demand for ambu- lance services, and the level of training paramedics receive. Finally, the data was from November 2014 and is being published in 2018 due to unexpected delays between requesting the data and obtaining and checking datasets from each ambulance service. There were no sig- nificant changes to national policy or guidelines for non-transport in England between 2014 and the publication of this article. The only major change to the ambulance service in England related to response times rather than non-transport, giving services longer to make a decision about the level of resource to send to a call. This is unlikely to have affected the relevance of the findings of this study. Nonetheless, it is important to acknowledge that this analysis is based on 2014 data and that it is necessary to continue to measure and assess the variability of non- transport rates between different ambulance services. There may also be a limitation in terms of using only one month if use of ambulances services differs by time of year. Implications Variation in health care is difficult to interpret unless the gold standard level is known. There is no consensus on the optimal non-transport rate for an ambulance service. It is easy to draw PLOS ONE | https://doi.org/10.1371/journal.pone.0204508 September 21, 2018 16 / 20 Why do ambulance services have different non-transport rates? A national cross sectional study Table 3. Factors explaining variation in telephone advice rates. Calls Telephone Unadjusted Adjusted P-Value N Advice Odds Ratio+ Odds Ratio N (95% CI) (95% CI) Patient-level Age group 0±2 15,463 1,448 1 1 - 3±10 11,134 1,228 1.20 (1.12, 1.29) 1.22 (1.12, 1.32) <0.001 11±20 28,778 2,629 0.98 (0.92, 1.03) 0.81 (0.75, 0.86) <0.001 21±30 43,886 4,678 1.15 (1.09, 1.22) 0.93 (0.87, 0.99) 0.019 31±40 34,790 3,461 1.07 (1.01, 1.13) 0.91 (0.85, 0.97) 0.004 41±50 37,357 3,006 0.85 (0.80, 0.90) 0.78 (0.73, 0.84) <0.001 51±60 35,657 2,646 0.78 (0.74, 0.82) 0.77 (0.72, 0.82) <0.001 61±70 40,604 2,686 0.69 (0.65, 0.73) 0.70 (0.65, 0.75) <0.001 71±80 57,545 3,307 0.59 (0.56, 0.62) 0.70 (0.56, 0.64) <0.001 81±90 70,687 3,814 0.55 (0.52, 0.58) 0.61 (0.53, 0.60) <0.001 >90 24,729 1,071 0.43 (0.40, 0.46) 0.45 (0.41, 0.49) <0.001 Sex Female 190,346 14,205 1 1 - Male 210,284 15,769 1.00 (0.98, 1.03) 1.02 (1.00, 1.05) 0.100 Time of call Out of Hours 263,370 21,290 1 1 - In Hours 137,260 8,684 0.76 (0.74, 0.78) 0.81 (0.79, 0.83) <0.001 Reason for call Falls 62,881 2,595 1 1 - Abdominal Pain 11,950 3,098 9.10 (8.61, 9.61) 7.36 (6.94, 7.80) <0.001 Breathing difficulties 45,972 1,578 0.83 (0.78, 0.88) 0.69 (0.64, 0.73) <0.001 Cardiovascular 61,421 1,614 0.62 (0.59, 0.66) 0.55 (0.52, 0.59) <0.001 Fitting 17,560 469 0.64 (0.58, 0.70) 0.48 (0.43, 0.53) <0.001 Injury 54,291 4,961 2.32 (2.22, 2.43) 1.89 (1.80, 1.99) <0.001 Psychiatric 10,600 1,005 2.46 (2.29, 2.64) 1.86 (1.72, 2.02) <0.001 Sick or Unconscious 61,557 7,806 3.37 (3.22, 3.52) 2.93 (2.79, 3.07) <0.001 Other 74,398 6,848 2.69 (2.57, 2.82) 2.18 (2.07, 2.29) <0.001 % Severe Long Term illness Q1 (Lowest) 79,009 6,803 1 1 - Q2 80,039 6,289 0.94 (0.91, 0.97) 0.97 (0.94, 1.01) 0.176 Q3 80,673 6,029 0.92 (0.89, 0.96) 0.96 (0.92, 1.00) 0.034 Q4 81,359 5,839 0.93 (0.90, 0.96) 0.96 (0.92, 1.01) 0.093 Q5 (Highest) 79,550 5,014 0.85 (0.81, 0.89) 0.89 (0.85, 0.93) <0.001 Indices of Multiple Deprivation Q5 (Least Deprived) 51,250 3,637 1 1 - Q4 63,422 4,568 1.03 (0.99, 1.08) 1.02 (0.97, 1.07) 0.493 Q3 76,211 5,536 1.05 (1.01, 1.10) 1.01 (0.96, 1.06) 0.692 Q2 93,931 7,163 1.11 (1.06, 1.15) 1.01 (0.96, 1.06) 0.714 Q1 (Most Deprived) 115,816 9,070 1.24 (1.19, 1.29) 1.11 (1.06, 1.17) <0.001 https://doi.org/10.1371/journal.pone.0204508.t003 the conclusion that the services in England with the highest rates are best and that the other services must change their practices to increase rates. Establishing the optimal rate would require further modelling of the cost-effectiveness of non-transport options, and the safety PLOS ONE | https://doi.org/10.1371/journal.pone.0204508 September 21, 2018 17 / 20 Why do ambulance services have different non-transport rates? A national cross sectional study and appropriateness of different rates of non-transport. Some approaches to non-transport may not be cost-effective, may result in mortality or increased severity of illness, may increase health service costs through additional subsequent service contacts, or may simply delay con- veyance to hospital. Evidence suggests that concerns around safety and appropriateness may be unfounded for both discharge at scene  and telephone advice, although it is still unknown whether adverse event rates are dependent on rates of non-transport. In a con- text where national policy promotes care closer to home, some patients call the ambulance service when they could have contacted a GP, patients can gain the reassurance they seek without being transported to hospital, and where there is little evidence to suggest that non-transport is unsafe,[2, 17] this study identified ways in which ambulance services with lower rates of discharge at scene could increase those rates. Some of the unwarranted variation between ambulance services lay within the control of the ambulance service, determined by their workforce policies and motivation to undertake discharge at scene. It is important to reduce variation in practices between ambulance services so that patients within the same country receive a similar service. There is a national policy drive currently in operation in England to improve ambulance services. Increasing non-transport rates so that care is offered close to the patient's home is a key part of this national initiative. The focus of this policy includes standardising practices between ambulance services and skilling the work- force to increase non-transport rates safely. The findings of this study have been presented to the leaders of the national ambulance improvement initiative to help contribute to planned improvements. There is also national guidance under development for emergency and acute medical care in over 16s with a recommendation for more non-transport. The findings are based on one month's data in 2014 for the 10 large ambulance services in England which is likely to be generalizable to other time periods because publicly recorded non-transport rates did not vary by month during the year of 2014. The findings are likely to be transferable to England in future years because although non-transport rates increased over time, and definitions of ambulance quality indicators changed over time, the ranking of ambulance services remained stable over time. However, it is important to continue to mea- sure the variability of non-transport rates between ambulance services and determine the rea- sons continued variability. The results presented here could be used as a baseline for measuring change in variability. The transferability of findings to ambulance services outside England is dependent on how these services are provided within different countries, whether there is a statutory obligation to transport patients to hospital, the level of non-transport in operation, and the extent to which non-transport rates are based on ambulance crew decision- making or patient refusal to travel. Conclusions The factors explaining variation in non-transport rates between ambulance services differ by type of non-transport. Variation in discharge at scene rates between ambulance services in England was found to be largely unwarranted. It was explained by differences in workforce configuration and managerial motivation, factors that are largely modifiable by ambulance ser- vices. Variation in telephone advice rates between ambulance services could not be explained by variables tested in this study. Acknowledgments We are grateful to members of the Business Intelligence/Data Analytics Teams within each ambulance service for providing us with the data. We are grateful to wider team members: Emma Knowles, Lindsey Bishop-Edwards, Joanne Coster, Cathryn James, Maggie Marsh, PLOS ONE | https://doi.org/10.1371/journal.pone.0204508 September 21, 2018 18 / 20 Why do ambulance services have different non-transport rates? A national cross sectional study Rachel O'Hara, Aloysius Niroshan Siriwardena, Julia Williams. We are grateful to members of our public and patient involvement group. Author Contributions Conceptualization: Alicia O'Cathain. Data curation: Tony Stone. Formal analysis: Richard Jacques. Funding acquisition: Alicia O'Cathain, Richard Jacques, Janette Turner. Investigation: Alicia O'Cathain, Richard Jacques, Tony Stone. Methodology: Alicia O'Cathain, Richard Jacques. Project administration: Alicia O'Cathain. Resources: Janette Turner. Software: Richard Jacques, Tony Stone. Supervision: Alicia O'Cathain. Validation: Tony Stone. Writing ± original draft: Alicia O'Cathain. Writing ± review & editing: Alicia O'Cathain, Richard Jacques, Tony Stone, Janette Turner. References 1. Alrazeeni DM, Sheikh SA, Mobrad A, Ghamdi MA, Abdulqader N, Gadgab MA, et al. Epidemiology of non-transported emergency medical services calls in Saudi Arabia. Saudi Med J 2016; 37(5):575±8. https://doi.org/10.15537/smj.2016.5.13872 PMID: 27146623 2. Ebben RH, Vloet LC, Speijers RF, Tonjes NW, Loef J, Pelgrim T, et al. A patient-safety and professional perspective on non-conveyance in ambulance care: a systematic review. Scandinavian Journal of Trauma, Resuscitation and Emergency. 2017; 25(1). 3. Mikolaizak AS, Simpson PM, Tiedemann A, Lord SR, Close JC. Systematic review of non-transporta- tion rates and outcomes for older people who have fallen after ambulance service call-out. Australasian journal on ageing. 2013; 32(3):147±57. https://doi.org/10.1111/ajag.12023 PMID: 24028454 4. Eastwood K, Morgans A, Smith K, Hodgkinson A, Becker G, Stoelwinder J. A novel approach for man- aging the growing demand for ambulance services by low-acuity patients. Australian Health Review. 2016; 40(4):378±84. https://doi.org/10.1071/AH15134 PMID: 26568037 5. Bray BD, Steventon A. What have we learnt after 15 years of research into the `weekend effect'? BMJ Qual Saf 2017; 26:607±10. https://doi.org/10.1136/bmjqs-2016-005793 PMID: 27903756 6. O'Cathain A, Knowles E, Maheswaran R, Pearson T, Turner J, Hirst E, et al. A system-wide approach to explaining variation in potentially avoidable emergency admissions: national ecological study. BMJ quality & safety. 2013:bmjqs-2013-002003. 7. NHS-England. Transforming urgent and emergency care services in England. 2013. 8. Kay I. Ambulance Quality Indicators 2011±2017 [Available from: https://www.england.nhs.uk/statistics/ statistical-work-areas/ambulance-quality-indicators/. 9. Appleby J, Raleigh V, Frosini F, Bevan G, Gao H, Lyscom T. Variations in health care. The good, the bad and the inexplicable London: The King's Fund. 2011. 10. Wennberg JE. Unwarranted variations in healthcare delivery: implications for academic medical cen- tres. British Medical Journal. 2002; 325(7370):961. PMID: 12399352 11. Wennberg JE. Time to tackle unwarranted variations in practice. British Medical Journal. 2011; 342: d1513. https://doi.org/10.1136/bmj.d1513 PMID: 21415111 12. Tohira H, Williams TA, Jacobs I, Bremner A, Finn J. The impact of new prehospital practitioners on ambulance transportation to the emergency department: a systematic review and meta-analysis. Emerg Med J. 2014; 31(e1):e88±e94. https://doi.org/10.1136/emermed-2013-202976 PMID: 24243486 PLOS ONE | https://doi.org/10.1371/journal.pone.0204508 September 21, 2018 19 / 20 Why do ambulance services have different non-transport rates? A national cross sectional study 13. O'Cathain A, Knowles E, Bishop-Edwards L, Coster J, Lawson R, Marsh M, et al. Understanding varia- tion in ambulance service non-conveyance rates: a mixed methods study. Health Services and Delivery Research (HS&DR). 2018; 6(19). 14. Sandelowski M, Voils CI, Knafl G. On Quantitizing. J Mix Methods Res. 2009; 3(3):208±22. https://doi. org/10.1177/1558689809334210 PMID: 19865603 15. NHS Ambulance Service. London: National Audit Office; 2017. 16. Bates D, Ma È chler M, Bolker B, Walker S. Fitting linear mixed-effects models using lme4. arXiv preprint arXiv:14065823. 2014. 17. Eastwood K, Morgans A, Smith K, Stoelwinder J. Secondary triage in prehospital emergency ambu- lance services: a systematic review. Emergency Medicine Journal. 2015; 32(6):486±92. https://doi.org/ 10.1136/emermed-2013-203120 PMID: 24788598 18. O'Hara R, Johnson M, Hirst E, Weyman A, Shaw D, Mortimer P, et al. A qualitative study of decision- making and safety in ambulance service transitions. Health Services and Delivery Research. 2014; 2 (56). 19. Turner J, Snooks H, Youren A, Dixon S, Fall D, Gaze S, et al. The costs and benefits of managing some low priority 999 ambulance calls by NHS Direct nurse advisers. Report for the National Institute for Health Research Service Delivery and Organisation programme. 2006. 20. Fisher JD, Freeman K, Clarke A, Spurgeon P, Smyth M, Perkins GD, et al. Patient safety in ambulance services: a scoping review. Health Services and Delivery Research. 2015; 3(21). 21. Booker MJ, Simmonds RL, Purdy S. Patients who call emergency ambulances for primary care prob- lems: a qualitative study of the decision-making process. Emerg Med J. 2014; 31:448±52. https://doi. org/10.1136/emermed-2012-202124 PMID: 23535018 22. Togher FJ, O'Cathain A, Phung V-H, Turner J, Siriwardena AN. Reassurance as a key outcome valued by emergency ambulance service users: a qualitative interview study. Health Expectations 2015; 18:2951±61. https://doi.org/10.1111/hex.12279 PMID: 25303062 23. NICE. Emergency and acute medical care in over 16s.2018 Available from: https://www.nice.org.uk/ guidance/service-deliveryÐorganisation-and-staffing/acute-and-critical-care#panel-pathways. PLOS ONE | https://doi.org/10.1371/journal.pone.0204508 September 21, 2018 20 / 20
PLoS ONE – Public Library of Science (PLoS) Journal
Published: Sep 21, 2018
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.
Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.
All the latest content is available, no embargo periods.
“Hi guys, I cannot tell you how much I love this resource. Incredible. I really believe you've hit the nail on the head with this site in regards to solving the research-purchase issue.”Daniel C.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud
“I must say, @deepdyve is a fabulous solution to the independent researcher's problem of #access to #information.”@deepthiw
“My last article couldn't be possible without the platform @deepdyve that makes journal papers cheaper.”@JoseServera