Developing feasible person-centred care alternatives to emergency department responses for adults with epilepsy: a discrete choice analysis mixed-methods study

Work Package 1

1. To systematically identify alternative CPs being considered by the NHS for epilepsy and describe the extent to which service users have been involved in their design.

2. To understand the decision-making processes of PWE and their SOs for seeking or not seeking ED care, and their concerns and expectations regarding alternative CPs.

3. To identify the attributes of postseizure emergency care that PWE and their SOs consider important and identify if this differs by the context in which the seizure occurs.

Work Package 2

1. To determine patient and SOs’ preferences for postseizure emergency care using DCEs and identify subgroup differences.

2. To estimate utility of different CP configurations to service users and subgroup differences.

Work Package 3

1. To hold expert workshops at which stakeholders, informed of the DCE evidence, would identify the optimal alternative CP(s) for seizures that could be recommended for implementation and evaluation.

NHS context and the need to manage demand on services

Like other health systems around the world, the NHS has been operating within a context of rising demand, slow funding growth and increasing operating costs. In 2015–6, this culminated in an aggregate funding deficit of £1.85 billion for NHS providers and commissioners. 2,3 The NHS Long-Term Plan4 and the Five Year Forward View5 challenged the NHS to make substantial savings, while, at the same time, working with service users to improve care experience and outcomes and reduce health inequalities.

Avoidable attendances and epilepsy identified as an area where opportunities exist to reduce demand

One way proposed to generate savings and manage demand is by finding innovative ways to reduce clinically unnecessary use of expensive emergency health services. 6 Visits to type 1 EDs that are clinically unnecessary are termed ‘avoidable attendances’. 7 These refer to those where the patient does not require the full facilities of a type 1 ED to manage their health problem. A type 1 ED (hereafter referred to as ED) is one that offers a consultant-led 24-hour service, with full resuscitation facilities and designated accommodation for the reception of accident and emergency patients. 8 In the UK, the terms ED and accident and emergency (A&E) department are often used interchangeably, the latter being particularly common in lay parlance.

Different estimates regarding the proportion of attendances at EDs that are ‘avoidable’ are available. Using their definition (now adopted by NHS Digital9) which is based on actual care received, O’Keeffe et al. 7 conservatively estimated ~15% of adult attendances are ‘avoidable’. Based on 2018–9 attendance data, this equates to ~2.2M first ED attendances. 10

Epilepsy has been identified as one condition that accounts for some of these ‘avoidable’ visits11 and where health inequalities are implicated. 12

Epilepsy and its epidemiology

Epilepsy is the recurring tendency to have unprovoked seizures. With a prevalence of ~1%,13 this chronic relapsing condition is the UK second most common serious neurological disorder. It affects people across the age spectrum. Antiseizure medication is the mainstay of treatment. People with epilepsy, particularly the ≤ 48% who continue to experience seizures,14 are at increased risk of injury, psychological and social disadvantage and premature death. 15,16

Use of emergency hospital services for epilepsy and societal impact

In the UK, up to 20% of PWE visit an ED each year. 17–19 In England, there are around 100,000 such visits to ED each year. 20 In 2015–6, these cost the NHS ~£70 million. 21,22

One reason costs are high is because half of the PWE visiting EDs are admitted to the hospital;19,20,23–25 85% of admissions for epilepsy occur on such an unplanned basis. 26 Epilepsy and convulsions are the second leading cause of unplanned hospital admissions amongst chronic ambulatory care-sensitive conditions. 25

Re-admissions further drive costs up;27,28 ≤ 60% of PWE reattend ED within 12 months. 29 This rate of return is higher than seen for other chronic relapsing conditions, such as asthma and diabetes. 30,31 Amongst chronic ambulatory care-sensitive conditions, epilepsy is the second most common cause of unplanned hospital admissions (17%). 25

Use of emergency hospital services for epilepsy is often clinically unnecessary

Seeking emergency care for a suspected seizure, including those caused by epilepsy can be clinically appropriate, even life-saving. Reasons include a first seizure and status epilepticus. 32–34 Most PWE attending ED for a seizure do not attend for such reasons.

Our group leads the UK-wide National Audit of Seizure Management in Hospitals (NASH) initiative. 35,36 Data from > 8000 seizure-related ED attendances from 85% of UK acute hospitals were examined for the first and second rounds of NASH. Most PWE attending ED were found to have already diagnosed, rather than new epilepsy, and to be presenting with a nonemergency state not requiring the full facilities of ED.

Data23,37 indicate the leading presentations for epilepsy are a person with established epilepsy who has (1) experienced an uncomplicated seizure that is in line with their usual presentation, (2) experienced a seizure in public who cannot be ‘left at scene’ (e.g. because they are alone, in a post-ictal state, been incontinent or had a minor injury) or (3) experienced a seizure that has self-terminated, but which is different to the person’s ‘normal’ seizure(s).

While potentially dramatic and frightening, seizure presentations such as those above do not require the person to be cared for within an ED. 38–40 They will typically only require rest and reassurance and will return to baseline health without medical intervention. Respiration following a seizure usually rapidly returns to normal after convulsion cessation. Oxygen support postseizure will rarely be required,41 with supplemental oxygen only recommended if a person’s oxygen saturation level is < 94%. 40 Patients will often be left confused and aching (this is the post-ictal state), but this is not a cause for concern and usually resolves relatively rapidly leaving the patient alert.

In addition to being typically unnecessary from a clinical perspective, an ED visit after the aforementioned seizures risks iatrogenic harms arising from unnecessary investigations and interventions. 42,43 They can also be disruptive to the lives of patients, their families and hospital service functioning. The third and latest round of NASH shows that while most (~93%) ED visits for seizures last < 4 hours, a high proportion (~50%) do result in some sort of hospital admission (albeit for < 24 hours). 44

Unmet need in those with epilepsy visiting the emergency department

While the acute episodes leading PWE to visit ED do not typically require ED facilities, many of the presentations are expressions of a need for an improvement in the person’s ambulatory care.

The National Audit of Seizure Management in Hospitals20 found most (~65%) people visiting ED were not under the care of a healthcare professional specialising in epilepsy but were often requiring their support. Many patients were, for instance, on suboptimal drug treatment and thus at increased risk of experiencing avoidable seizures, side effects, psychosocial dysfunction and avoidable death. 45

Some PWE visiting ED also appear in need of enhanced self-management support, such as that provided by epilepsy nurse specialists (ENSs). This could mean the PWE and their families can learn more about their condition and how to manage seizures by themselves. 29,46–48 Unlike for other chronic conditions, there is no routine course PWE can go on to learn about epilepsy once diagnosed. Consequently, knowledge and self-management confidence can be low, particularly in those PWE who are older and those with low levels of formal education. 49,50

Momentum to reduce emergency department visits for epilepsy by introducing alternative care pathway

The policy agenda4,51,52 and the publicity given to the NASH findings21,53 created a vigour to reduce visits for seizures and enhance patient outcomes. Systematic examinations of the literature54 show there is no evidence on how to do this. An idea which has been gaining traction is for the ambulance service to assume a greater role. 55

The reason for this is that most (~90%) people who visit EDs for seizures have been transported there by an emergency ambulance. 24,35,36 Suspected seizures are the seventh most common call-out. 56 At present, ambulances convey nearly every person they attend to who has experienced a suspected seizure to ED. 56–58 This is despite most not demonstrating a clinical need for it and ambulance care guidelines not recommending it.

Dickson et al. 56 reviewed the ambulance records for suspected seizure incidents presenting to one regional English ambulance service and found medical emergencies were uncommon. The seizure had self-terminated before the emergency vehicle arrived in > 90% of cases and airways were clear. Breathing was found to be normal in > 96% of cases, and in only 8% of cases, emergency drugs were administered (by the crew or a SO beforehand). Despite this, ambulance crews still advised transport to ED in 89% of cases.

The Joint Royal Colleges Ambulance Liaison Committee (JRCALC) guidelines40 describe ambulance clinicians’ permitted scope of practice. They acknowledge the complexity of the management of seizures. Nonetheless, they state that the ‘overall risk of adverse events is low and most patients do not require the facilities of a hospital emergency department’ (p. 197). For those patients who are no longer seizing, JRCALC stipulate only that those who are pregnant, those who have experienced a first seizure and those treated with benzodiazepines should always be conveyed to ED (unless they have a care plan that states otherwise).

Conveyance decisions and factors behind high conveyance rate for epilepsy

A sizeable research literature exists on how ambulance clinicians decide whether to convey a person to ED. Two recent systematic reviews59,60 – which, between them, considered evidence up until April 2018 – have drawn this literature together. They show the complex nature of conveyance decisions, with multiple factors, beyond patient need, potentially affecting them. Oosterwold et al.’s60 conceptual framework (reproduced in Report Supplementary Material 1) summarises the macro, meso and micro factors.

Oosterwold et al.’s framework does not assert the importance or interplay between the different factors. It does, though, highlight the quality of evidence on each and opportunities for intervention. One factor associated with high-quality evidence, and which helps explain the attention being given to alternative CPs, is paramedics often lack access to alternatives to ED. This is highlighted by the following quotes from paramedics:61

We struggle for alternatives to ED. While we might be directed towards primary care, in practice they are lacking.

There was a big investment in a new urgent care centre locally but they won’t take people who’ve had a seizure. I’ve had patients … in a postictal state who need maybe half an hour until they come round … but [the centres] have this crazy idea that if somebody’s had a seizure then they need to have a CT scan … Sometimes there is no alternative but to take them to ED.

Other key factors that have been stated as contributing to high rates of conveyance and compound problems are (1) the limited external support ambulance crews receive with their conveyance decisions, (2) time pressures and (3) the limited access crews have to inform about the medical history of the person they are managing.

In terms of medical history, crews will typically not have access to the person’s full medical record. Instead, they will rely on information provided by informants on-scene and those who called for the ambulance.

Crews do increasingly have some access to NHS Summary Care Records. 62 However, in their standard form, these provide limited information pertinent to seizure care (i.e. only medications and allergies). A minority of PWE will implement procedures to overcome this limitation to information sharing, often at their own expense (e.g. by carrying a document, wearing medical alert jewellery and phone apps). However, crews still report finding information on the person’s history to be a challenge. 61 There is a National Institute for Health and Care Excellence (NICE) quality standard which states persons with epilepsy should have an agreed, written emergency care plan. 63 It only applies to adults with a history of prolonged or repeated seizures.

Regarding time pressures, some paramedics describe how under current funding arrangements they can feel able to only spend a limited amount of time ‘on scene’ to assess patients and permit them to recover. As such, transporting a person to ED can represent the most time-efficient management option. Dickson et al. 56 found that within one regional ambulance service, the mean time clinicians spent on scene for people presenting with seizures was 27 minutes. The following quotes61 show how time can sometimes impact upon paramedics’ conveyance decisions:

As a practitioner, if I don’t hit my times I’ll be pulled in by my manager, so it does pile the pressure on, forcing crews to think, ‘just put them on the truck and take them to hospital’.

If someone has a seizure outside of the home, we wouldn’t take them home … It’s not necessarily the right option for that patient … But by taking them home, which is further away, we will be tied up for longer.

There is a vision therefore of what could help: ambulance service access to some form of alternative CP whereby those seeking help for seizures judged not to require ED could be cared for within less costly, alternative environments. 64 The exact nature of the CP is not entirely clear, and most efforts are in the early stages. However, what is apparent is different English regions and services are considering different formats, and development is occurring in an unco-ordinated way.

Via positions that members of our team hold (e.g. professional bodies, scientific advisory committees and patient support organisations), we were aware of a range of packages envisioned. They included the introduction of policies for ambulance crews to leave patients at scene with the offer of a telephone call from a general practitioner (GP), a pharmacist practitioner or a nurse within 24, 48, 72 or indeed 120 hours. Other regions were considering allowing ambulance crews to transport patients home after seizures rather than taking them to ED. Others were looking to work with urgent treatment centres (UTCs), so those with seizures can be cared for within them.

‘Urgent Treatment Centre’ is the label that, following the Urgent and Emergency Care Review, has been given to most English walk-in centres, minor injury units and urgent care centres in an effort to reduce confusion. 65 Further details on them is provided in Report Supplementary Material 2.

Alternative care pathways and the ambulance service in general

The Keough Urgent and Emergency Care Review66–68 envisioned that demand on acute services could be reduced by expediting the transformation of the ambulance service. Rather than ambulances transporting most people to hospital, it envisions ambulances becoming mobile treatment centres, treating more patients at scene and making greater use of alternative CPs.

Alternative CPs are not new. Ambulance clinicians have not been obliged to transport everyone they see to ED since ~1997. Snooks et al. 69 developed some of the first CPs in 1999 and piloted them with the London Ambulance Service. These were so-called ‘see and treat’ protocols. In brief, crews were provided with protocols for different presentations. They would select the applicable protocol and work through a series of questions to establish if the presentation satisfied certain criteria. If they did, they were guided towards alternatives to ED. This ranged from them contacting the patient’s primary care team, advising the patient to contact NHS Direct, asking the patient to arrange an appointment with their GP to offering advice.

‘See and treat’ CPs are now common,70–77 and other alternative response options have also become available, such as ‘hear and treat’ (i.e. where advice is given to patients over the telephone and they are signposted to other care providers). 78,79

Recently, there have also been attempts to widen the aims of ‘see and treat’ CPs so they not only avoid unnecessary ED conveyances but also more explicitly seek to improve patient outcomes. 80–83 An illustration is the CP for hypoglycaemia developed by diabetes specialist nurses. 80 Patients with acute hypoglycaemia for whom an ambulance has been called are assessed and treated on scene by paramedics, rather than being transported to ED. Patients are then referred directly by ambulance staff to the specialist nurses, who subsequently contact the patient by telephone.

What is known about the use and risks and benefits of alternative care pathways

As is often the case with behaviour change interventions,84,85 the design and implementation of CPs have been ‘theory-lite’. The evidence on the utility of alternative CPs is nevertheless generally positive, albeit not definitive. 86–88

In their review of potential revisions to the urgent and emergency care system, the Nuffield Trust, for instance, identified greater ambulance/paramedic triage in the community as having the most positive evidence. 55 Some CP proponents highlight headline data suggesting their introduction helped ambulance services manage their own increasing demand. It is estimated that ambulance trusts saved themselves ~£74M and EDs ~£63M in 2015–6 by decreasing ED conveyance compared to 2011–2. 89 In 2018–9, England’s ambulance services managed > 8M ‘999’ calls; 6% of these received telephone advice (‘hear and treat’), 31% were treated at scene (‘see and treat’), 63% were conveyed to ED and the remaining 6% were transported to a lower-level facility (‘see and convey elsewhere’). 90

Further supporting the use of alternative CPs is the significant expert support that exists for paramedics’ ability to manage patients without conveyance. 91 There is also evidence that paramedics are willing to use CPs92 and that they can do this safely. 93 The Prehospital Outcomes for Evidence-Based Evaluation programme project looked at all types of presentations to the ambulance service94 and found that within 3 days of non-conveyance, 83% of non-conveyed cases experienced no subsequent health event (9% of patients recontacted the ambulance service, 12.6% attended ED, 6.3% were admitted and 0.3% died; patients could experience more than one event).

Finally, some studies have also indicated that CPs can be associated with increased patient satisfaction,86,87 with reports that it can be frustrating to be conveyed to ED when not needed. 95–102 For example, an initial evaluation of the aforementioned hypoglycaemia CP found patients benefitted: 88% of patients reported greater understanding of their condition and 73% felt better able to self-manage future episodes. 71 In some cases, the uptake of CPs has though been lower than anticipated69 and, to date, epilepsy is a presentation that has received little attention. 103

Potential of an alternative care pathway for epilepsy

There are several reasons for the momentum to introduce an alternative CP for epilepsy.

Firstly, the rate of transportation for seizures and subsequent hospital admissions is high and has remained so, despite wider NHS initiatives. 25 It is probable therefore that this presentation will continue to be a challenge for the NHS.

Secondly, paramedics have indicated that they are keen (if they receive additional training and the necessary support) to take on more responsibility for managing seizures. 104 This is important as acceptability of an intervention to those expected to deliver it is key to intervention feasibility. 105

Thirdly, changes to how ambulance services are funded may mean the use of CPs becomes financially incentivised and paramedics might feel able to spend more time ‘on scene’. 106

Finally, and most importantly, an alternative CP has the potential to work and improve patient outcomes. Qualitative research61,107,108 provides the beginnings of a theoretical basis for the use of an alternative CP in epilepsy. On the basis of this work, possible mechanisms by which the intervention could reasonably make a difference are that it may (1) increase awareness and likelihood that paramedics will consider non-conveyance and referral pathways as an option in appropriate cases, (2) increase paramedics’ clinical knowledge of how to make appropriate non-conveyance decisions, (3) increase paramedics’ knowledge of alternative care providers that are acceptable to service users and (4) increase paramedics’ confidence about making a non-conveyance decision and reducing anxiety about risk.

By paramedics operating as part of an integrated healthcare team, some of the CPs importantly have the potential to reduce health inequalities and improve the support PWE receive. This is likely to come by the CP introducing a mechanism by which PWE ‘in need’ are brought to the attention of specialist epilepsy services.

Despite indications that specialist input could help those attending ED,109–111 and NICE recommending it,112 most PWE (~80%) attending ED do not currently receive it. Most (~60%) are also not referred on either. 20 PWE who are older, those not known to the service and those live in the most deprived areas are the least likely to be referred. 113 The only communication regarding the event that typically occurs is a letter sent by the ED to the persons GP stating that the episode occurred. 44

Ensuring that PWE in need of specialist care receive it is a long-standing challenge in the UK. Potential reasons why persons are not always identified and supported are (1) primary care is not remunerated (as it is for other conditions) to identify PWE with uncontrolled seizures, where there is diagnostic uncertainty or treatment failure, (2) there are fewer neurologists per head than in other developed nations114 and (3) there are fewer nurse specialists than recommended. 115 An indication of the challenge is only ~50% of PWE in the UK are currently seizure-free. 14 Trial evidence indicates this should be ~70%. 116 Importantly, it is those in the most socially deprived areas who fare worst. Ashworth et al. 117 found the difference in the proportion of PWE living in the least and most deprived quintiles who were seizure-free was 12%.

Current configuration of care for those presenting with established epilepsy

While there is some variation between English regions, the information so far presented allows one to provide a broad indication as to the configuration of care an adult who presents to the ambulance service for established epilepsy is likely to receive. The ambulance crew managing them will typically not have access to relevant information about the person’s medical history and most (~70%) would ultimately be conveyed to ED. The time being cared for in ED would be ~3–4 hours. The person’s GP would typically be notified of the event by letter, but the person will not be seen by or referred on to an epilepsy specialist (such as an epilepsy nurse or neurologist).

Acceptability of alternative care pathways to target population is unknown

While a CP for epilepsy has much potential, it is unclear whether any of the CPs being considered could deliver on it. This is because their acceptability to patients and their SOs is unknown since they are not being involved in CP design. Acceptability of an intervention to its intended recipients and providers is a fundamental criterion an intervention needs to satisfy if it is to be positioned to achieve its intended outcome. 105

Care packages for epilepsy are seemingly being developed in a largely ‘top-down’ fashion. This is despite evidence that decisions about whether or not to access healthcare services can be informed by how a patient or SO perceives their situation118–120 and that patients/SOs and health professionals’ can sometimes have different views about what constitutes an ‘emergency’ situation. 121–124 It is possible therefore that the nature and content of the envisioned CPs will not align with what patients and their SOs would consider to meet their needs.

To our knowledge, only one epilepsy CP has been reported on to date. Piloted for 11 months by the Yorkshire Ambulance Service, it allowed paramedics, when they felt it appropriate, to not convey adults with known epilepsy to ED but instead directly refer them to an ENS (who would telephone the patient within 5 days). 125

An evaluation of this CP underscored the opportunities that existed to improve the management of those from the target population. Telephone consultations, for example, between the nurse specialist and patients lasted for only a median of 10 minutes. Despite this, an independent review of the nurses’ notes revealed positive action/outcomes occurred in most cases. Examples included (1) contacting the patient’s pharmacist or GP to optimise medication, (2) arranging an appointment in the neurology clinic, (3) referring the patient to a psychotherapy service and (4) booking the patient an appointment in the ENS clinic.

The problem was that over the 11-month evaluation period, < 10% of eligible PWE attended to by paramedics were put onto the CP. The reasons for this low uptake rate were not explored. An explanation is the CP’s configuration was not acceptable to PWE. To date, it has been assumed all the target population do not want to be conveyed to ED and will readily accept any alternative to ED that is offered. The actual evidence on the preferences of PWE is unclear.

We certainly know not all people who experience a seizure will attend ED. Around 48% of PWE in the UK experienced a seizure in the past year, but less than half attended ED. 17–19 This includes people who do not seek emergency care in the first instance and others declining a visit ED even after an ambulance has attended. 95,126 Figure 1 provides a representation of this.

FIGURE 1.

Indication of experience of seizures and use of ambulance service and attendance at hospital EDs by PWE in England. Notes: Reproduced from McKinlay et al. 1 with permission. Segments of figure are not to scale. a, Based on ~1% of UK population having epilepsy13; b, Moran et al. 14 indicate 48% of PWE will have had a seizure in prior 12 months; c, When a call is received by ambulance service described as relating to a ‘convulsion’, ‘fit’, ‘seizure’, the call handler endeavours to ask standardised questions to gauge, such things as severity and potential aetiology. For those using Advanced Medical Priority Dispatch System, according to Protocol 12, one question is ‘Is s/he an epileptic?’. Most callers should be able to answer question as most (~70%) seizure calls made by relative, friend or carer. 56 Audit data from two regional ambulance services (North West Ambulance Service; Yorkshire Ambulance Service) indicate that in 2018 of those attended to ~70% of those who indicate a history of epilepsy are conveyed to ED, and 30% are not; d, Hart and Shorvon17 found that ~20% of PWE had attended an ED in the prior 12 months.

However, this does not mean that it can be assumed all PWE do not want to be conveyed to ED when it is clinically safe. The following points suggest a need to be cautious.

Firstly, the characteristics of PWE who visit ED might mean they are less ready to accept not being transported to ED during the intense and worrying moments of a seizure and aftermath. Compared to the wider epilepsy population, they have lower epilepsy knowledge, report more clinical anxiety, report greater perceived epilepsy stigma and are more likely to live in a socially deprived area. 29,127–131 Up to 20% also have an intellectual disability. 35,36

In the few studies that have interviewed PWE about their visits to ED it has been found some PWE express a need for immediate access to urgent care when they have a seizure. This includes them explaining how they and their SOs (to whom care decisions are often delegated when the patient is unconscious or lacks capacity) are often fearful of seizures, including the possibility of death and brain damage. They also say they are unsure about how to manage them. 47,48 This, they say, leads them to call for an ambulance when they are about to have, or have had, a seizure. For many, this is an established pattern of behaviour. One study found the median years diagnosed for PWE visiting ED was 11. Some telling quotes from interviews48 with PWE and SOs include:

Cancer, you’re awake. I know you can die, but you’re awake. I’d prefer something like that … Having epilepsy, you’re going into a fit. You don’t know if you’re going to wake up or die. That’s why I call [999]!

Person with epilepsy

[I was] just worried because I don’t know anything about epilepsy … I only know the bad things … I know you can die … I am so worried I decided just to ring an ambulance … better safe than sorry.

Significant other to person with epilepsy

Another important piece of evidence comes from work with paramedics. 61 It highlights that although some PWE do not want to be conveyed to ED, others may.

when they talk to us about what their preference is it’s mixed … many been taken to hospital before. They know that that’s the pattern and they expect it … they feel nervous if you start talking about them staying at home … Some feel very vulnerable and want to go through the motions of seeing the paramedic, the doctor, the nurse …

Each of the CPs being considered around the country requires financial expenditure. This is difficult to justify. This is because it is unclear whether the outlay would translate into a service that PWE and SOs would use and benefit from.

How to make the development of alternative pathway’s more rational and cost-efficient

While it is beyond the scope of the current project, our ultimate aim is to work with the ambulance service to implement the most promising CP for PWE, rapidly evaluate it and use the evidence to transform service organisation nationally. However, as patients and SOs from the target population have yet to be consulted, it is not known what sort of alternative CPs have the potential to be most effective and should be prioritised for evaluation/implementation.

What is needed is information about the target population’s preferences for seizure care, including what elements are most important to them, what things do they not want and to what extent would they be willing to give up one or more elements for the presence of others. A clearer understanding of their decision-making processes for seeking or not seeking ED care is also needed, as is information on any concerns and expectations they may have regarding alternative CPs. It would also be important to understand whether this differs according to seizure presentation. Together this information would help establish to what extent the target population is open to the possibility of alternative management and could be used to codesign a CP. It could be designed to encapsulate the features of most importance to patients and their SOs.

Patient preference information is defined by the US Food and Drug Administration as ‘qualitative or quantitative assessments of the relative desirability or acceptability to patients of specified alternatives or choices among outcomes or other attributes that differ among alternative health interventions’ (p. 6). 132 It reflects what treatment characteristics matter to patients, how much these matter and how patients make ‘trade-offs’ between them.

From an economist’s perspective, there are two approaches to the measurement of preferences: revealed and stated. The former is derived from observed market activities. Stated preferences in contrast are derived by asking persons for their views, such as via surveys. 133 An obvious advantage of the latter approach is that it can permit one to explore people’s preferences for products which remain hypothetical. In the current project, the focus has been on such stated preferences obtained by using DCEs.

Discrete choice experiments

Discrete choice experiments are one of the most popular and robust quantitative ways to obtain stated preferences. 134,135

The DCE method is an attribute-based survey methodology. It is underpinned by random utility theory and relies on the assumptions of economic rationality and utility maximisation. 136 These state that any ‘good’, including healthcare packages, can be described by their constituent characteristics – known as ‘attributes’ – and that the extent to which an individual prefers a good will depends on the levels these attributes take. 137 In the case of a healthcare package, attributes could include both process attributes (e.g. where the ambulance takes you; the healthcare professional responsible for your care) and outcome attributes (e.g. likelihood of follow-up).

Following the above-stated theories, if a person is presented with alternative ‘goods’ described by a set of attributes, it is assumed the person will make ‘trade-offs’ between attributes to select the good that yields his/her highest individual benefit, known as utility. Within a DCE, a person’s stated preferences are therefore elicited by presenting them with a scenario (e.g. having X health condition) and asking them to choose which of two (or more) care packages they prefer. Each care package is described according to the same attributes (e.g. where the ambulance takes you) and in terms of levels that vary (e.g. ED, UTC). Changing the levels (using an experimental design) produces a set of alternative care package options (choice sets), that respondents are asked to choose between.

By observing how participants change their responses across choice tasks, DCEs allow one to infer the extent to which different care attributes and levels are driving preference, the direction of their effect and the extent to which persons will accept certain ‘trade-offs’ between attributes.

To help show what a DCE looks like, we present an example in Appendix 1. It has been annotated to highlight key DCE terms.

Discrete choice experiments have increasingly been used to determine healthcare preferences, including for epilepsy. 138–140 They have been recognised as a suitable method for eliciting evidence of patient preferences to inform regulatory benefit–risk assessments of medicines. 141,142 Chief amongst the reasons for this is that they offer a robust methodology to estimate which attributes are important in decision-making.

There is also good evidence that when a DCE is designed well, the preferences expressed within them can show good congruence with real-world behaviour. 143–145 Quaife et al. 146 recently completed a meta-analysis on this topic. With a pooled sensitivity estimate of 88%, they found DCEs to be particularly good at predicting who would ‘opt-in’ to use a product or service that a respondent does not currently use. Specificity was less impressive, indicating DCEs were less good at predicting who will ‘opt out’ and not choose not to use a product or service. This did though appear to be influenced by how many choices the person had to make in the DCE.

Some key design considerations for discrete choice experiments

Good practice guidelines for designing, conducting and reporting DCEs are available. 133,147,148 One of the first and critical stages is identifying the attributes that will be used to describe the options within the DCE that people are asked to choose between and for which the DCE will provide quantitative preference weights for. The levels that the attributes take will also need determining and should be theoretically feasible.

Coast et al. 149 provide guidance for the selection of the attributes to ensure random utility theory, the psychological basis for DCEs, is not violated. Firstly, it is important that the attributes include all those that might be important for an individual in coming to a decision. Ignoring an important attribute risks an underspecified model. Selecting and defining the attributes necessitate an adequate understanding of the target population’s perspective. 150 If evidence is not already available, qualitative research can be completed with the target population to model their likely decision calculus.

Secondly, the attributes should not be too close to the latent construct that the DCE is investigating. For example, where utility is the underlying latent construct, an attribute within the DCE that is labelled as utility or expresses overall satisfaction with the good should not be included.

Thirdly, attributes should not be intrinsic to a person’s personality and should be capable of being manipulated experimentally. For example, attitude towards taking medicines is a characteristic of the person, rather than the good.

Using discrete choice experiment evidence to inform service design

As highlighted by Ryan and Farrar,151 DCEs can provide crucial information for service and policy designers and those allocating resources. Evidence on service user preference by itself is, however, potentially insufficient to help choose with confidence which CP configuration amongst the various options is best positioned to achieve its goals and should be prioritised for implementation/evaluation. This is because other feasibility factors are important.

Table 1 outlines Michie et al.’s105 so-called affordability, practicability, effectiveness, acceptability, side-effects and equity (APEASE) criteria for determining the potential of different interventions. It highlights key factors that can, to differing extents, be important in determining promise and has been used by a range of bodies to help select interventions. 152

The factors include affordability, practicability, effectiveness, acceptability, side effects and equity. To expand further, is the intended intervention likely to be acceptable to all those involved – including those expected to deliver it and their employing organisations? The latter will, for instance, need to be willing to assume any potential risks and consequences that may occur from the intervention if something goes wrong.

Is the envisioned intervention practicable? Can the service wanted by the intended users be implemented in the NHS? Is there the required capacity and information sharing resources? Thirdly, how effective and cost-effective will the service be in achieving the desired objectives? Fourthly, can it be afforded within budget? It arguably does not matter how acceptable or preferred an intervention is if it cannot be afforded. On a related point, if responsibility for commissioning an intervention is shared by different groups, will it be possible to get them all to agree to fund it? Fifthly, how far does one anticipate that the intervention might lead to unintended adverse outcomes? And finally, equity. How far is the intervention likely to increase or decrease known differences between the more advantaged and disadvantaged in our society?

Given the other potentially important factors raised by the APEASE model, a DCE could be seen as providing robust evidence on service user acceptability. This could then be used as the basis for discussions by informed stakeholders who would be asked to evaluate potential CP options against all the APEASE criterion. They could then advise which configuration is judged to strike the optimal balance between patient acceptability and NHS feasibility.

Design

A short piloted online survey ran from 1st April to 30th June 2019. It used a set of responsive open and closed questions. As the language for alternative CPs is still evolving and not fixed, the survey asked respondents whether the NHS organisation they were representing (or the local services they worked with) had made or were planning any changes to any policies, schemes or pathways that relate to the management of those with epilepsy (or indeed seizures generally), with a view to minimising ED visits (see Appendix 2).

If a change was reported to have occurred by the respondent or was planned, they were asked about its nature, to what extent service users had been involved, and how, if at all, they anticipated the change would address health inequalities. Respondents could upload a copy of any associated paperwork and inform us of what their experience with the change had been.

The time frame specified within the questionnaire for changes that had been made was the past 5 years. The time frame for future changes was the next 12 months. Five years was chosen as this corresponded with the timing of dissemination to local sites of findings from the first NASH.

Services not reporting or planning changes were asked why and shown three known care innovations and asked if they comprised usual practice within their organisation. Respondents were asked, where necessary, to consult with managers of their service to ensure responses incorporated both current and planned redesign.

All respondents were asked about the extent to which reducing unplanned hospitalisations for chronic ambulatory care sensitive conditions (generally and then by condition) was a priority for their organisation.

Survey respondents were able to indicate if they wanted to be informed of opportunities to take part in subsequent WPs.

Ethical approval was not required as the survey was deemed a service evaluation. 155

Recruitment

Personalised invitations to participate, followed by two reminders, were sent to the clinical leads/directors of England’s regional ambulance trusts (n = 10) (and to the Isle of Wight NHS Trust who provide its ambulance service),156 regional neuroscience centres (n = 25)157 and neurology centres (n = 16). 158,159 Invites were also sent to a random sample of 25% (n = 35) of England’s ‘type 1’ EDs160 (stratified by geographical area and attendance size). 161

Respondents could delegate completion of the survey to someone else within their service.

Analysis

Characteristics of the responding organisations were examined using descriptive statistics and representativeness judged against publicly available information.

To determine the different types of CPs being considered and the proportion of responding institutions considering each, the first 50% of submitted surveys was selected and imported into NVivo 10 (Lumivero, Burlington, USA). The responses given within these were read by a qualitative researcher (AM) to identify recurrent types of CP. This was used to develop an initial thematic coding framework capturing the main types. The framework was then applied to the full data corpus to identify and code all responses. Responses that reflected CPs insufficiently identified by the original coding framework were used to modify the framework. Frequent discussions between the researcher and a multidisciplinary research team helped ensure interpretations and conceptualisations of the data were credible, valid and shared.

Sample

Seventy-two (82.8%) services responded: 36 (85.7%) neurology and neuroscience centres, 26 (74.3%) EDs and 10 (90.9%) ambulance services. The survey took respondents a median of 8 minutes to complete [interquartile range (IQR): 5–16.75].

Responding services were spread across England (Figure 3), and their characteristics were similar to non-participating sites (see Report Supplementary Material 3). Responding individuals were doctors (72.2%), paramedics (13.9%) and nurses (13.9%). Analyses relating to service changes are based on responses from the 68 (94.4%) services whose responses included no missing data.

FIGURE 3.

Geographical location of (a) regional ambulance services, (b) neuroscience centres and neurology services and (c) EDs by their survey participation status.

Priority

The pooled response indicated reducing unplanned hospitalisations associated with chronic ACSs was a ‘high priority’ (median 4; IQR = 3–5). In ranking eight different ambulatory care sensitive conditions, ‘epilepsy and convulsions’ was given a pooled median rank of 3 (IQR = 2–4). Neurology/neuroscience centres gave it a median rank of 2 (IQR = 1–3), ambulances services 3.5 (IQR = 2.75–5) and EDs 4 (IQR = 3–5).

Service changes

Forty-one (60.3%) services said they or service(s) they worked with had made and/or were planning a service change(s) (see Appendix 3). Thirty-four (50.0%) reported a change(s) had occurred. Neurology/neuroscience sites (n = 30, 93.8%) were most likely to report changes, and EDs (n = 4, 15.4%) were least likely.

Types of change

Eleven change types were identified. Table 2 describes them. Most addressed the care of those with established, rather than new epilepsy. They fell into three categories according to which part of the patient’s care journey they focused on:

Before emergency help is sought

Seventeen (25%) services reported such changes. These related to care planning, attempting to prevent a person’s condition from requiring emergency help. The most widely made change was the introduction of Rapid Access clinics, providing specialist epilepsy support to patients between routine appointments as needs arose. Less common changes included implementing education for PWE on seizure first aid and – to enable more proactive and risk-stratified care – access by specialist services between appointments to data on a person’s seizure control and medication.

When person is being cared for by the emergency services

Nineteen (27.9%) services reported these changes. A key change reported was, as anticipated, the introduction by ambulance services of alternative CPs to support non-conveyance to ED and alternative care arrangements for PWE with uncomplicated seizure presentations. Protocols, with flow charts, were available for the CPs with a view to supporting crews’ decision-making. For those not conveyed to ED, protocols recommended a range of differing actions from leaving patients at home, ‘on scene’ or in some instances conveying to an UTC.

A few of the protocols also included mechanisms by which patients could be referred on to other services. In some instances, this was in the form of ambulance crews having access to a directory of local services and their contact details. In one instance, an e-referral system allowed crews to electronically notify GPs of the attendance and its details. In two areas, patients could be referred to the epilepsy service, with an ENS contacting the patient within 1–5 days by telephone. Eligibility criteria differed. In one area, it was open to anyone with established epilepsy. In another, it was only for patients already under the epilepsy service and in certain geographical locations.

The other common service change under this category was the introduction or expansion of an acute neurology service. Here, an ENS or neurologist was available to review PWE once they were within the ED, either face-to-face or virtually. The aim being to facilitate discharge and identify support needs.

Follow-up care

Twenty-three (33.8%) services reported these changes, with most expanding neurology services for those with established epilepsy. These sought to reduce waiting times for ED referrals (aim 1–4 weeks). Some services offered telephonic clinics, others face-to-face appointments but within primary care locations to increase accessibility. To further minimise referral times, one neurology service allowed EDs to directly book patients into their service’s appointment slots. Two other services had implemented processes to automatically notify them of seizure-related ED attendances.

Usual practice

Nine (33.3%) of the 27 services that had not made, and did not plan, any changes, reported usual practice comprised at least one of the three presented service innovations (see Appendix 3). Five (18.5%) said epilepsy services were automatically informed of patients attending ED. Three (11.1%) used protocols to divert people presenting with an uncomplicated seizure away from ED. Two (7.4%) reported medical records for PWE were accessible to ambulance staff.

Service user involvement

Of the 34 services that had implemented a change, only 7 (21.2%) had consulted service users.

Study design

Qualitative interviews were completed with PWE, with or without the presence of one of their SOs. Methods were guided by the COREQ checklist164 and Coast et al. 149 checklist for DCE formative work.

Feelings of stigma, the sensitive nature of the topic, lack of driving licences and anxiety about travelling due to seizures meant interviews, rather than focus groups were considered the preferable data collection mode.

The study received ethics approval by the King’s College London Psychiatry, Nursing and Midwifery Ethics Committee (LRS-18/19-10353).

Participants were reimbursed with a £20 shopping voucher for their time.

Recruitment

To be eligible, people needed to (1) be aged ≥ 18 years, (2) have a diagnosis of epilepsy for ≥ 1 year, (3) be prescribed an antiseizure medication, (4) have had emergency service contact in the previous 12 months for epilepsy (be it a visit to ED and/or ambulance contact) (living in the South-East area of England), (5) be able to provide informed consent and (6) participate in a qualitative interview in English. They could have any epilepsy syndromes or seizure type. No medical records were accessed to confirm eligibility. All were based on self-report.

Persons with severe current psychiatric disorders (e.g. acute psychosis) or life-threatening medical illness were excluded, as were those who resided within a care or nursing home or had no fixed abode. We excluded persons living in care or nursing homes since, in these circumstances, it is typically not the patient or their SOs who are responsible for care decisions when a seizure occurs.

Purposive recruitment methods were used with the intention of capturing a comprehensive range of perspectives, and recruitment was to continue until there was no new variation in observations (data saturation). Purposive sampling is a type of nonprobability sampling. Here, a researcher selects participants from their sampling frame because they have characteristics that the researcher wants present with a view to enabling a research question to be answered. It is not about creating a representative sample with the intention of making generalisations. We sought to ensure variation in age and sex of participants, as well as for those who had had recent seizures at home and in a public place to be adequately represented. Based on the previous DCE design work,139,149 we anticipated obtaining interview data from ~20 to 30 PWE.

To identify SO participants, participating PWE were asked if they wanted to choose a SO/informal carer to take part with and be interviewed with. Despite greater social isolation, up to 90% of PWE still identify a spouse, parent or friend who can act as an informal carer for them. 165 Joint interviews and eliciting the concerns of both parties to the same event are regarded as particularly appropriate for studying complex health decisions. 95,126

The first recruitment process involved AMc contacting by mail PWE who took part in the SMILE (self-management education for adults with epilepsy) trial in the South-East of England and asking them if they were interested in future projects. 166 SMILE was a phase III trial comparing a two-day education course to treatment as usual alone. Inclusion criteria were ≥ 2 seizures in the prior 12 months, having established epilepsy, being prescribed antiepileptic medication and an ability to independently complete questionnaires in English. At 12-month follow-up, no differences between trial arms on the primary (quality of life) or secondary outcome measures were found. 167 The number of people expressing an interest in future projects was ultimately lower than anticipated, with AMc receiving permission to send invites to 47 PWE. Of these, only 15 responded and were screened; eight were found to be eligible for the current study and were ultimately interviewed.

The second recruitment process was completed with the assistance of Epilepsy Action, a user-led charity in the UK. They assisted with recruitment by advertising the study on social media platforms, at in-person community events in South-East England and in their newsletters. Interested PWE were asked to contact AMc to register their interest. Of the 37 people who made contact, 21 were eligible, of whom 4 subsequently dropped out. This process provided a further 17 participants for the current study.

Recruitment concluded at 25 interviews, as no new themes were identified.

Procedure

One-off interviews were held with the participants in locations (home, public place or a university office) and at times of their choosing. Interviews were conducted by AMc – a female postdoctoral researcher with a PhD in mixed-methods psychology. She has experience of conducting interviews with PWE. No non-participants were present during the interviews.

An interview guide was used which was developed by AMc, AN, LR and MM (two research psychologists, one neurologist, and one medical sociologist, respectively). Interview guides were reviewed by two PWE and one ambulance staff member, who provided insight based on their experience and then further refined by pilot interviews with two further PWE. The topic guide is shown in Appendix 4.

All participants provided written informed consent, and interviews were audio-recorded. No field notes were taken. No prior relationships existed between the research team and participants. Participants were aware that the researcher was not affiliated with their usual care provider.

Interviews began by asking participants for contextual information about their epilepsy and recent seizures and episodes of ED care. The interviewer then sought evaluative information, asking respondents to describe positive and negative perceptions concerning emergency care and any alternative methods of care they have experienced. The researcher explored the decision-making processes that surrounded patients and SOs’ seeking or not seeking ED care following recent seizures and the reasons for their choices.

Strategic data were then obtained from respondents by describing some aspects of possible alternative CPs and asking them for their views. They were asked what they thought may be positive aspects and their concerns if they had been offered them following a stated seizure of theirs. As WP1b occurred at the same time as WP1a, the CPs discussed with the interviewees included ones already known about by the team (see Appendix 4).

Finally, at the end of the interview, AMc worked with each participant with epilepsy to complete a ranking exercise to prioritise the importance of care attributes. The findings from it are reported in Chapter 4.

Data analysis

Data collection and analyses proceeded in an iterative manner. Audio recordings were transcribed verbatim. Transcripts formed the data set. These were anonymised to protect participant confidentiality. They were not returned to participants for comment or correction.

To analyse the data, a framework approach168 was employed as it is suited to policy-oriented research examining patient experiences. 169 Transcripts were read with themes and codes identified. Data were then managed using NVivo 12 (Lumivero, Burlington, USA).

Following preliminary analysis, two matrices were developed by AMc and MM to summarise individual cases, with seven coded categories relating to decision-making (self-care, decisions by self/ambulance/public, informing others); care preferences (experiences of using emergency services, satisfaction, suggestions); and three categories regarding views and experiences relating to potential CPs. Initially, both AMc and MM summarised cases to examine consensus, with AMc completing coding. A constant comparative approach was used to interrogate the data and identify factors influencing individual participant decisions and preferences.

Where necessary, quotes with potentially identifiable information were edited to preserve anonymity and ensure clarity of meaning.

Participants

Twenty-five people (15 female, 10 male), along with 5 of their SOs, completed one-off interviews between April and September 2019. Interview duration ranged from 28 to 168 minutes (mean: 69).

Report Supplementary Material 4 describes the sample’s characteristics in detail. In brief, the mean age of participants with epilepsy was 37 years (range: 19–73). Participants described their ethnicity in the following ways: White British (n = 22); white others (n = 2); Black Caribbean (n = 1). The sample was well educated, with n = 18 (72%) having postsecondary-level qualifications (beyond General Certificate of Secondary Education level) (this compares to ~44% in the UK170), and all had seen an epilepsy specialist in the prior 12 months (compared to an expected 48% based on findings from the National Audit of Seizure Management in Hospitals44). Their mean years diagnosed with epilepsy was 21 (range: 4–50); 17 (68%) were at the time being prescribed two or more antiseizure medications. Fifteen participants said they had a comorbid health condition (physical health condition: 11; mental health condition: 7), with 3 reporting dissociative seizures.

Twenty of the participants said they had a friend or family member who helped them with their epilepsy (parent: 11, child: 2, partner: 9).

Qualitative findings from interviews relating to decision-making and preference

Alternative care pathways

When discussing alternative CPs, participants highlighted how they felt the appropriateness of the different configuration options would vary on an event-by-event basis:

There shouldn’t be a set rule of that, yeah, you must go to A&E or yes, you must go to the Urgent Care Centre, it all depends on after doing the observations they do in the ambulance, where they think you should go and also how, depending on how the individual is feeling.

p. 19, Male, age 30s

Ranking exercise methods

Decisions and reflections of discrete choice experiment design following ranking exercise

Knowledge exchange event methods

Knowledge exchange event results

Decisions and reflections of discrete choice experiment design panel following knowledge exchange exercise

The panel welcomed the KE participants’ feedback. They accepted most of their suggestions. The changes served to create V1.2 of the attributes in Table 5. The details are as follows:

1. Access to care plan

The panel agreed the DCE would require an introduction that provided a detailed explanation of the term ‘care plan’ to participants and what they can contain. They also stated that the care plan attribute would need to explicitly refer to access to medical records since this is an aspiration for current practice and for many this is what ‘care plan’ in an emergency comprises.

1. Location of care

The panel, with input from clinicians within the research team, endorsed all the changes recommended by the KE participants. The exception to this was the comment from one epilepsy specialist that suggested the level ‘Stay where you are’ should not be used for the atypical seizure presentation. The reasons for the panel declining that last suggestion were as follows.

Firstly, the parameters of the atypical seizure presentation to be used within the DCE were restrictive. The presentation as described does not meet the requirements for ED care on clinical grounds (i.e. it was an epileptic seizure in a person with established epilepsy from which they had recovered from, but which was different in some way to their usual experience, and no injuries were incurred).

Secondly, the DCE was only asking for a person’s care preferences in a hypothetical situation. It was not making recommendations as to the suitability of the approach.

Finally, one reason the representative expressed a need for all atypical seizure presentations to visit ED was based on the view that not going to ED would deprive the person from input from an epilepsy specialist. This does not though reflect clinical reality. While NICE guidelines112 suggest this should happen in some form, most people with established epilepsy who visit ED are not seen or referred to a specialist. 44 Moreover, within the DCE, it was still possible for a person to stay where they were or be conveyed to an UTC and still receive specialist follow-up.

1. Time

The panel accepted the recommendations of the representatives. They noted that the experimental design would need to allow for conditions between ‘Time’ and ‘Location of care’ to ensure the choices remained plausible.

1. Involvement of specialist services

On reflection, the panel said this attribute should be refined to focus on specialist advice during the care episode, thus minimising overlap with the other attribute ‘Follow-up’.

1. Follow-up

The panel agreed that the attribute should not indicate that health professional with specialist training in epilepsy would be known to the patient.

When presented to the KE participants, the attribute ‘Follow-up’ had four draft levels (i.e. not referred; within 5 days by telephone; 2–4 weeks; and 4 weeks +). The panel on reflection stated that the level for this attribute of ‘Not referred’ be removed given the earlier noted NICE recommendation of follow-up in all cases. The panel also suggested collapsing the remaining levels to help reduce the size of the experiment to a more manageable number of choice-sets.

1. Contact with your GP

The language used for this attribute was amended in line with the participants' suggestion.

Pilot interviews

Results

Pilotees

Twelve PPI members were enrolled and provided feedback on the DCE survey. Saturation in terms of feedback was occurring after around six pilot interviews. However, to verify the refinements made to the DCE based on their feedback were sufficient, an additional six PPI members completed the DCE.

Feedback received and changes made to the discrete choice experiment survey

The survey was well received by the PPI members and their understanding of the majority of the DCE and wider non-DCE was as intended. Respondents did nonetheless request some changes which the panel made. The changes served to create the finalised attribute set which is labelled V1.4 in Table 5. The set is also presented in Figure 4; this time along with the images used to represent the attributes levels.

FIGURE 4.

Finalised attributes and levels used in all DCEs (V1.4) along with images.

The changes made based on PPI feedback were as follows. Firstly, greater clarity of meaning for the two attributes describing current and future involvement of a health professional with specialist training in epilepsy was requested. PPI members instructed that ‘Involvement of specialist services’ be changed to ‘Epilepsy specialist today’, while ‘Future contact with an epilepsy specialist’ should be changed to ‘Additional contact with epilepsy specialist’.

There was variation in how the PPI members interpreted the two ‘yes’ levels for the attribute ‘Involvement of specialist services today’. Consequently, the panel decided to simplify this to a single ‘Yes’ level.

Patient and public involvement members also challenged the levels used for the attribute ‘Additional contact with epilepsy specialist’. They had commonly experienced ‘No’ follow-up following instances of urgent care and so requested that this level be included to help ensure face validity. The panel made this change. Finally, following the recommendation of PPI members, Contact with GP was simplified to GP told.

In addition to the above changes requested by PPI members, the DCE panel also decided it was necessary to amend the survey in view of the coronavirus disease 2019 (COVID-19) which had started to impact the UK while piloting was occurring. Government requests to ‘socially distance’ were starting, and changes in the way people were using emergency services were becoming evident. 177 Therefore, respondents were asked to make their choices based on their preferences for care outside of the pandemic, termed ‘normal times’ in the survey. Feedback was specifically received on this amendment to wording from pilots. The animation was also updated, and the close of the survey was updated to include a signpost to information on COVID-19 and epilepsy.

Design

The DCE survey was cross-sectional. A person with epilepsy could take part with or without one of their SOs and vice versa. People with epilepsy and SOs completed the survey independently, and the responses of any patients and SOs known to each other were not linked.

Eligibility criteria

To participate, patients needed to be aged ≥ 18 years, have had contact with the ambulance in the prior 12 months for epilepsy and have already been diagnosed with epilepsy at the time of this contact. We recruited SOs of such people who were aged ≥ 16 years. We excluded persons with a terminal condition, a severe psychiatric disorder and those who lived in a care facility.

The full inclusion and exclusion criteria are described in Table 7. These are provided according to the route by which a person was recruited. The key difference between the eligibility for the two recruitment routes is where in England the person needed to live.

Sample size

Definitive sample size calculations for DCEs require the finalised design of the DCE to be known. 180 Therefore, to permit decisions regarding recruitment to be made in advance of the formative work being completed, we estimated the likely design of the DCE and used Orme’s181 commonly used182 formula. Orme’s formula is (n*t*a)/c must be ≥ 500, where n is the number of respondents, t is the number of choice tasks, a is the number of options per task and c is the number of analysis cells (for main effects, the largest number of levels for any one attribute).

We estimated that for each seizure scenario, the DCE would involve the person being asked to choose between two CP options (a = 2), that they would complete around nine choice tasks (t = 9), and that three would be the maximum number of levels for any one attribute (c = 3). Using these parameters, the required number of participants providing complete data for each of the seizure scenarios was calculated as n = 84. As the intention was for data to be obtained for three seizure scenarios, but participants would be randomised to complete a combination of just two, a minimum sample of 126 participants would be required. To allow for variation in the actual number of choice tasks and levels ultimately used, we increased this to 150. It was then further inflated to 174 to accommodate for the ~15% of participants that (based on prior epilepsy DCEs) were anticipated to have insufficient data for inclusion in the analyses. 139,140,183

As one of the study’s planned subgroup analyses was a comparison of the care preferences of PWE (and SOs) who had and had not been conveyed to ED in the prior 12 months (see Analysis), it was necessary to recruit 174 PWE (and SOs) who had visited ED in the prior 12 months and 174 PWE (and SOs) who had not. This meant the target sample size for the DCE was n = 348.

Deviations from original recruitment protocol

The two recruitment routes used

Route 1 – via ambulance service

North West Ambulance Service NHS Trust is the second largest of England’s 10 regional NHS Ambulance Trusts. It acted as a Participant Identification Centre for the survey. Figure 5 shows the process by which they identified and invited participants. The specifics are as follows.

FIGURE 5.

Illustration of identification and recruitment process for participants via ambulance service (with indicative figures based on prior years). AMPDS, Advanced Medical Priority Dispatch System; dx, diagnosis; ep, epilepsy; psych., psychiatric diagnosis; 12B01E, fitting with effective breathing (not verified), epileptic or previous diagnosis of fitting, < 35 years; 12C02E, fitting during pregnancy – epileptic or previous diagnosis of fitting; 12C03E, diabetic patient fitting – epileptic or previous diagnosis of fitting; 12C05E, history of stroke or brain tumour – epileptic or previous diagnosis of fitting; 12C06E, overdose/poisoning (ingestion) – epileptic or previous diagnosis of fitting; 12D01E, not breathing (after key questioning) – epileptic or previous diagnosis of fitting; 12D02E, fitting agonal/ineffective breathing – fitting history; 12D04E, fitting effective breathing (not-verified) – epileptic or previous diagnosis of fitting > 35.

North West Ambulance Service NHS Trust first completed an electronic search of the emergency calls it had received in the prior 12 months to identify incidents coded under Advanced Medical Priority Dispatch System (AMPDS) Protocol 12 and assigned an ‘e’ suffix. Protocol 12 is used to record and manage calls which are described as relating to a convulsion. The ‘e’ suffix indicates the incident was related to a person who was reported by the caller to have epilepsy or a previous diagnosis of fitting. Further details on NWAS’ characteristics, the AMPDS and which of the Protocol 12 subcoded incidents were eligible are provided in Report Supplementary Material 5.

North West Ambulance Service NHS Trust restricted its search to mean they only identified persons with a Protocol 12 incident who were aged ≥ 18 years and living within ~30 miles of Liverpool city centre. The latter was necessary for two reasons. Firstly, to make the original plan of a researcher travelling to some participants practicable. Secondly, the Service Support Costs provided by the North-West Coast Comprehensive Research Network (NW CRN) to reimburse NWAS for the time spent identifying participants was only for persons resident within the footprint of the NW CRN. Report Supplementary Material 6 provides a map of the catchment area. Approximately 4.1M people reside within 30 miles of Liverpool city centre.

Based on prior health DCEs184 and work with the target population,22,46 we anticipated 30–60% of the PWE invited by NWAS might participate. We used the lower estimate for uptake of 30% and the target sample sizes for PWE who had (i.e. n = 174) and had not experienced conveyance to ED following ambulance attendance (i.e. n = 174) to determine how many individuals NWAS needed to identify and invite. Namely, n = 580 people who had not been conveyed to ED following attendance and n = 580 people who been conveyed to ED following attendance. As conveyance and non-conveyance to ED do not occur at the same rate, the time periods that NWAS would need to search within to identify sufficient persons from these two groups was permitted to differ.

Having completed the electronic search, two NWAS research paramedics (DW and KH, see Acknowledgements) then reviewed the ambulance report [called the Patient Report Form (PRF)] for each of the identified incidents by hand. This was to exclude persons ineligible. Those who remained had their status checked against the NHS ‘Spine’ service to further exclude any persons who had died and/or who had ‘opted out’ from contact for research.

People who continued to be eligible were posted an invitation pack by NWAS. This included a covering letter – signed by NWAS’ chief consultant paramedic – informing them about the study. A repeat invite was sent ~3 weeks later.

To identify SO participants, invited PWE were asked to identify a SO to take part with should they wish this to happen. They were asked to pass on recruitment documents to them contained within their invitation pack.

People with epilepsy and SOs interested in participating were asked to visit the survey page. A short, simple web address was provided. They could also contact the research team by phone, e-mail or by using a FREEPOST return slip included in their invitation pack. Those wanting to access the survey and take part were asked to provide – via the survey page – eConsent. 185

Each person completing the survey had the option of providing their personal details to the research team to receive a £20 shopping voucher.

Approval for this recruitment route was provided by the Health Research Authority and the National Research Ethics Service Committee, West Midlands, Solihull (19/WM/0012).

Procedure

Data quality checks and curation

A member of the research team (PD) screened out any persons submitting a survey response that included any answer/s indicating they were ineligible (e.g. < 18 years old, did not live in England, insufficient contact with the ambulance service). Persons who did not submit sufficient DCE responses (defined as completion of at least one choice task from one of the DCEs) were also excluded.

For each participant, an area-based measure of social deprivation was calculated to indicate the relative social deprivation of the area in which they lived. It was calculated by using the postcode the person reported via the survey and linking this with the Index of Multiple Deprivation (IMD) 2019 score188 for their lower super output area (LSOA). Each person was classified according to the deprivation quintile they lived in (1 being the most deprived). For participants for whom an IMD was missing (because they did not supply a sufficiently complete or valid postcode), multiple imputation was performed as per Goodman and Gatward’s approach. 189

Describing sample

Reasons for exclusion, the extent of missing data and the outcomes of randomisation were calculated. These will be reported in Chapter 6, as will the characteristics of the samples recruited by the different recruitment routes and as a whole.

As data were provided by participants from two separate recruitment routes, a decision was needed about what data management approach would provide data from people most representative of the target population. Therefore, the characteristics of the participants recruited from the different routes needed to be compared to the parent population. High-quality data on the characteristics of the target population (i.e. segment 3 of Figure 1) are not available. What is available is data from the latest round of the NASH-344 on the characteristics of persons with established epilepsy prescribed antiseizure medication who are seen at ED for a seizure (i.e. segment iv of Figure 1). We compared the samples from the two different routes, and when combined, to the NASH-3 data. As will be seen when the results are presented in Chapter 6, it was judged appropriate to merge the data from the two recruitment routes to maximise representativeness.

Preference data

Experimental design

The six attributes identified in the formative work, with associated levels, resulted in 288 possible CP configurations. Twenty-four of these were paired into 12 questions, in which respondents had to choose between two possible CPs, using an efficient design generated by Ngene experimental design software (Ngene 1.2.1, Choice Metrics) using the D-error measure for an efficient design for the MNL model. This reduced the number of pairwise choices required for robust estimation while incorporating conditions on location and time, that emerged during the formative work. The Ngene syntax used is in Report Supplementary Material 8.

To ensure the CPs used in the design were plausible ‘Location of care’ and ‘Time’ were specified as individual attributes (V1.3 in Table 5), with conditional levels, rather than being combined as previously decided (V1.2 in Table 5). All three DCEs (‘Atypical seizure’, ‘Home typical seizure’ and ‘Public typical’) used the same design (12 pairwise choices that a participant would be asked to make) to describe CP options in the given seizure scenario. Figure 6 provides an example of what one of the pair-wise choice questions in the DCE looked like.

FIGURE 6.

Example of a pairwise choice question used within the DCE survey (for a participant who had epilepsy themselves). Notes: The DCE for a seizure scenario contained 12 such choice questions. For these, the attributes in the shaded left-hand column stayed constant, but the levels (i.e. the descriptions for option A and option B) varied.

Ranking of care package configurations according to preference

Using the coefficients (β) from the respective regression models, the expected utility for each of the 288 configurations was calculated for each of the three different seizure scenarios by sample (PWE or SO), as described above.

Utility can be understood as the total value a CP configuration offers a person, estimated by summing the gains or losses associated with the attribute levels the configuration has. The expected utility model was parameterised using the preference weights elicited in the DCE and attribute level selection to represent all 288 possible configurations. The β coefficients for each attribute level (derived from the DCE regression analysis) were summed to estimate the total utility for each configuration:

where β = value of the attribute coefficients represented in the individual CP.

We also investigated the contribution of individual attribute levels, by measuring the change in total utility observed by a change in a single level, while all other attributes remained equal. This is expressed as a percentage.

The 288 configurations were ranked according to expected utility. A rank of 1 indicates the configuration most favoured for that scenario, while a rank of 288 indicates the least favoured. Ranks were constructed separately for each seizure scenario and by both PWE and SOs. Thus, six different sets of ranks were generated. The rank of each CP configuration across all six models was compared visually, using a heatmap generated in Microsoft Excel.

To determine how current ambulance care was meeting participants preferences, we selected from the 288 CP configurations the one deemed to best approximate current care and its ranking for the different seizure scenarios is reported. Based on evidence presented in the earlier chapters, the configuration chosen to represent current care was, according to the six attributes and levels, comprised: (1) ‘The paramedic has access to medical records or a care plan’: no, (2) ‘What happens next’: A&E, (3) ‘Time’: 3 hours, (4) ‘Epilepsy specialists today’: no, (5) ‘GP told’: yes and (6) ‘Additional contact with an epilepsy specialist’: no. It is represented in Report Supplementary Material 10.

We anticipated service providers would be keen to understand the possibility of implementing a single CP configuration for seizures when non-conveyance was deemed clinically safe (rather than needing to offer slightly different care package configurations depending on the specifics of the seizure scenario and the person the crew was primarily liaising with ‘on scene’). To support such discussions, we determined the top-ranked care package configuration for the different seizure contexts and estimated its uptake within the other seizure contexts, by estimating: P=exp(utility CPi)/Σexp(CPj).

Identification and invitation

Route 1 – via North West Ambulance Service

Prior to this route being paused by the NWAS due to COVID-19, NWAS had screened n = 2634 incidents and identified n = 981 individuals for invite. Of these, n = 456 (46.5%) people were identified because of an incident that resulted in conveyance to an ED, and n = 525 (53.5%) individuals were identified due to an incident that ended with non-conveyance.

The leading reason for an incident not resulting in a person being identified for invite was that the individual had already been identified based on an earlier incident (25.7%) (Figure 7). Other key reasons included that the person resided within a care facility (19.8%); and that the person was not noted on the PRF completed by the ambulance crew at the time as having diagnosed epilepsy (18.4%). In 14.7% of cases the PRF for the incident could not be retrieved or the writing was illegible.

FIGURE 7.

Flow chart of the screening process and identification of individuals for invitation via ambulance service. n, number; NFA, no fixed abode.

Following the NIHR’s project ‘Restart’,192 the ambulance recruitment route was reactivated and (after a further ‘status’ check) invitation packs were dispatched on 31 July 2020. A repeat was sent approximately 3 weeks later.

The time between the incidents that had led these persons to contact NWAS and them being sent an invitation ranged from 10 to 21 months. The wide range reflects the different periods of time needed to identify enough persons for invite. To identify sufficient people following a conveyed incident for invite, NWAS needed to screen 5 months of incidents (i.e. June 2019 to October 2019). In contrast, to identify a sufficient number of non-conveyed individuals, NWAS needed to screen 11.5 months of incidents (i.e. October 2018 to October 2019).

Having been sent an invitation pack, n = 18 persons informed the research team they were unable to complete the survey online. A further n = 4 persons informed the team they were ineligible.

Participants

Recruitment

Submissions to survey

Recruitment via the ambulance service route ran for 5 months and via the public advertisement route for 5.5 months. Both closed to new recruits on 25 November 2020. During these time periods, there were a total of n = 1687 submissions via the two recruitment routes (n = 156, 9.2%, via the ambulance service route; n = 1532, 90.8%, via public advertisements). Report Supplementary Material 12 shows recruitment by time.

Usable submissions to survey

Figure 8 provides a participant flow chart. It shows that of the submissions received, n = 772 (45.7%) did not involve the person answering a single question after having consented to participate. Hence, there were n = 916 (54.3%) submissions that could be considered for analysis (n = 662 PWE; n = 254 SOs).

FIGURE 8.

Responses to survey by participant type, recruitment route and extent of completeness of response to DCE. a, NB some respondents fulfil more than one exclusion criteria.

Of the n = 916 submissions, n = 205 (22.4%) (n = 148 by PWE; n = 57 by SOs) needed to be excluded as they were submitted by persons whose answers indicated they were ineligible. A further n = 117 (12.8%) submissions (n = 87 by PWE; n = 30 by SOs) were excluded as they did not involve the person answering a single DCE question.

From the two-recruitment routes, there were thus useable submissions to the DCE survey from n = 594 participants (n = 427 PWE; n = 167 SOs). Of these, most (81.1%; n = 358 PWE, n = 124 SOs) were recruited via the public advertisement route and 18.9% (n = 69 PWE, n = 43 SOs) via the ambulance service route.

Response rate

The estimated response rate for recruitment via the ambulance service route was 11.7% (112/954). The denominator here is the number of invites dispatched by the ambulance service (n = 981) minus the noted n = 4 invites sent to persons who said they were ineligible and the n = 23 submissions from persons who were ineligible. The calculation assumes the remaining invites were received by persons who were eligible, that they were sent to the correct address and treats responses by PWE and SOs as independent.

It was not possible to calculate a response rate for the public advertisement route since the denominator is unknown.

Characteristics of participants by recruitment route, sample representativeness and justification for merging

The headline characteristics of the samples recruited by the two different routes and their representativeness relative to the target population (as indicated by NASH-3) are shown in Table 9. The full characteristics of the samples are shown in Appendix 7.

TABLE 9 - Headline characteristics of samples recruited by the two routes and their representativeness individually and when combined

N, number; NASH-3, National Audit of Seizure Management in Hospitals, audit round 3.

Analysis data set	Route 1, via ambulance service	Route 2, via public adverts	Combined	NASH-3
	N = 112	N = 482	N = 594	N = 1676
Participant type, n (%)
PWE	69 (61.6)	358 (74.3)	427 (71.9)	1676 (100)
SO	43 (38.4)	124 (25.7)	167 (28.1)
Age of PWE, median (range)
Reported by PWE	36 (26–51)	37 (27–49)	37 (27–49)
Reported by SO participants	32.5 (26–46)	28 (23–39)	29 (24–41)
Combined	34 (26–49)	35 (26–48)	35 (26–48)	43 (29–58)
Missing	1	45	46
Sex of PWE, female, n (%)
Reported by PWE	37 (53.6)	280 (78.2)	317 (74.2)
Reported by SO participants	20 (47.6)	30 (37.0)	50 (40.7)
Combined	57 (51.4)	310 (70.6)	367 (66.7)	783 (46.8)
Missing	1	43	44
Intellectual disability in PWE, yes, n (%)
Reported by PWE	7 (11.1)	27 (11.4)	34 (11.4)
Reported by SO participants	18 (42.9)	19 (23.5)	37 (30.1)
Combined	25 (23.8)	46 (14.5)	71 (16.8)	297 (17.8)
Missing	7	165	172
PWE Index of Multiple Deprivation, n (%)
Reported by PWE
Quintile 1	24 (38.1)	60 (25.5)	84 (28.2)	517 (30.9)
Quintile 2	12 (19.1)	56 (23.8)	68 (22.8)	395 (23.6)
Quintile 3	11 (17.5)	40 (17.0)	51 (17.1)	304 (18.1)
Quintile 4	13 (20.6)	44 (18.7)	57 (19.1)	292 (17.4)
Quintile 5	3 (4.8)	35 (14.9)	38 (12.8)	168 (10.0)
Missing	6	124	129
PWE seen epilepsy specialist in prior 12 months, yes, n (%)
Reported by PWE	48 (76.2)	205 (87.2)	253 (84.9)
Reported by SO participants	36 (85.7)	76 (93.8)	112 (91.1)
Combined	84 (80.0)	281 (88.9)	365 (86.7)	815 (48.6)
Missing	7	166	173

The age of the PWE taking part in the survey, or being reported on by significant others, was similar across the two recruitment routes. Via ambulance the service, the median was 34 years (IQR = 26–49). Via the public advertisement route, it was 35 years (IQR = 26–49). However, the sex of PWE taking part in the survey, or being reported on, did differ by recruitment route.

Specifically, persons taking part via the ambulance service route were more likely to be male (48.6% of participants) than those participating via the public advertisement route (29.4% of participants). Significant other participants were also more likely to take part by the ambulance service route than by the public advertisement route. Of the participants who took part via the ambulance route, 38.4% were SOs, compared to 25.7% of the participants taking part via the public advertisement route.

Table 9 shows that combining the n = 594 participants from the two recruitment routes created a sample which was more like the target population. It had a more similar proportion of PWE in it (or that were being represented) that had an intellectual impairment (16.8%). It also had broadly the same amount of people in the different quintiles of deprivation represented within in, including from the most deprived one (28.2%). Merging the samples also meant the proportion of male PWE taking part or being represented (33.3%) better aligned with the target population data.

Figure 9 shows the geographical distribution of the participants in the merged sample. It shows a spread of participants from across England’s regions, with persons from the North West being overrepresented due to the two recruitment routes both having operated there.

FIGURE 9.

Distribution of participants in the final analysis data set by English region that they reside in. Notes: Map separates England into local government region. The figures in red relate to individuals recruited via the ambulance service route (n = 112), whereas figures in black indicate individuals recruited via the public advert route (n = 482). Postcodes for 173 individuals (n = 166 from public advert route; n = 7 from the ambulance route) were partial and did not permit confidence in determining local government region.

Even after merging the samples, the median age of the participants (35 years, IQR = 26–48) remained 8 years younger than the average for the target population. Another important difference was the much higher proportion of participants in the survey samples that had purportedly seen a health professional specialising in epilepsy in the prior 12 months. Potential reasons for this difference are proposed later in the chapter.

Further characteristics of the merged sample

The full characteristics of the n = 594 participants are shown in Appendix 7.

The median years that PWE taking part (or being represented by a SO) had been diagnosed with epilepsy was 12 (IQR = 4–26) and most (51.5%) reported having experienced 10 or more seizures (of any type) in the prior 12 months.

Most (64.8%) experienced multiple seizure types, with 79.6% selecting the tonic–clonic seizure description as best representing the type that typically led them or the person they knew to have contact with the ambulance service. A significant minority (16.8%) of PWE were noted as having an intellectual impairment.

Most (74.4%) were found to have had one to three contacts with the ambulance service in the prior 12 months for epilepsy. There was evidence of non-conveyance amongst the sample, with 36.8% of participants reporting fewer ED visits than ambulance contacts.

The main ways the SO participants knew someone with epilepsy was that they were their parent (59.4%), followed by being their spouse/partner (24.4%). The demographics, epilepsy profile and service use reported by PWE who took part in the survey and that reported by SOs on behalf of a PWE differed in some respects.

Specifically, SO participants tended to be representing PWE who were younger, more likely to be male, more likely to have an intellectual disability and more likely to be prescribed a rescue medication. The people being represented were also more likely to be said to be in the company of someone who they knew when ambulance contact occurred, tended to have had more contact with the ambulance service and when this occurred, for this to result in conveyance to ED (Table 10).

TABLE 10 - Comparison of the characteristics and epilepsy profile of the PWE taking part in or being represented in the survey

n, number; p/12 months, past 12 months.

	Reported by PWE themselves (n = 427)	Reported by significant others (n = 167)
Age of PWE, median (range)	37 (27–49)	29 (24–41)
Missing	2	44
Sex of PWE, female (%)	74.2	40.7
Missing	0	44
PWE has an intellectual disability, yes (%)	11.4	30.1
Missing	128	44
Rescue medication prescribed to PWE, yes (%)	26.5	39.8
Missing	129	44
Ambulance contacts in p/12 months, %
1–3	77.2	67.5
4–6	13.8	17.1
7–9	4.4	6.5
10 or more	4.7	8.9
Missing	129	44
ED contacts in p/12 months, %
0	15.3	8.9
1–3	64.5	66.7
4–6	13.6	12.2
7–9	3.7	4.1
10 or more	3.0	8.1
Missing	129	44
When ambulance comes, how often is PWE with someone who could help, %
Never	7.4	1.7
Rarely	16.8	9.9
About half the time	20.5	14.9
Most of the time	35.4	32.2
Always	19.9	41.3
Missing	130	46

Randomisation and completion of allocated discrete choice experiments

Scrutiny of the randomisation outcome data showed that the randomisation function worked well. Each combination of seizure scenarios and presentation order was equally represented in the data set with useable data (see Report Supplementary Material 13).

Three hundred of the PWE (70.3%) and 123 of the SOs (73.7%) answered all of their 24 DCE questions (12 × two seizure scenarios). The remainder of the participants answered between 1 and 23 of the 24 that they were asked to do.

Of a potential 1188 DCEs (i.e. 594 participants allocated 2 each), 88% were completed. DCE completion rate ranged from 85% for the ‘Home typical seizure’ scenario when completed by PWE, to 91% for the DCE with the ‘Atypical seizure scenario’ when completed by SOs. Seventy-six per cent of participants started their second DCE. Participants were most likely to continue to a second DCE if they completed the ‘Home typical seizure’ scenario first. They were least likely to continue if the ‘Home typical seizure’ scenario came second (see Report Supplementary Material 13).

Familiarity with seizure scenarios

The majority of participants reported that they (or the person that they knew in the case of SOs) had ‘often’ or ‘sometimes’ been in each of the DCE seizure scenarios (see Appendix 8).

The extent of familiarity with the individual scenarios was similar for PWE; 69–73% said they had been in each of them. In contrast, SO participants reported more familiarity with the ‘Atypical seizure’ scenario (82%) than the ‘Public typical seizure’ (66%) or ‘Home typical seizure’ (67%) scenarios.

Experience of and feelings towards COVID-19

During the 5.5-month period of recruitment, the median daily COVID-19 infection rate was 3497 (IQR = 343–33,470) (see Report Supplementary Material 14). 193

Less than 7% of the participants believed they themselves had at some point contacted COVID-19 (see Report Supplementary Material 15). Both PWE and SOs did, however, ‘agree’ that COVID-19 had reduced their willingness to use ED after a seizure or for the person they knew to do so. For SOs, COVID-19 also meant they were less willing for the PWE that they knew to use UTCs or to attend a hospital outpatient appointment after a seizure. COVID-19 did not appear to impact on the preferences of either PWE or SOs for seeing a GP.

Discrete choice experiment responses

As described in Chapter 6, the survey responses from 594 participants were eligible for analysis; 72% of which were from PWE.

Each participant was asked to complete DCEs on two different seizure scenarios. The participants were therefore altogether asked to complete a total of 1188 DCEs. Of these, 1047 (88%) were completed.

Model specification: testing for non-linear effects

Table 11 shows the results of testing the linearity of the attribute ‘Time’. Model A uses a continuous specification for time, in which the results would be interpreted as linear. However, Model B suggests that when time is categorised (using effects coding to allow for estimation of all four levels), the size/direction of the coefficients moving between levels is not as expected (not linear). The most pronounced example is for SOs responding to the ‘Home typical seizure’ scenario; this indicates strongest preferences of 2 hours, rather than 1 hour as would be anticipated.

Visual inspection of plots of the effects coded coefficients illustrates this deviation from the linear assumption by seizure contexts (see Appendix 9). The base-case model was therefore respecified as

U_i = utility derived by an individual

β₀ = constant term

β_i = estimated coefficient for each attribute (variable)

ε = error term (assumed to vary by seizure scenario)

While it was no longer appropriate to calculate willingness to wait (marginal rate of substitution of time), the impact on total utility of each level change was calculated as the percentage change in utility score for a categorical level change, all other attributes remaining equal, within each model.

Statistical significance of attributes and direction of their influence on preference

The results of the regression models for the preferences of PWE and SOs in the different seizure scenarios are presented in Tables 12 and 13, respectively. The β-coefficient illustrates the magnitude of preference of the attribute level. Where an attribute level does not reach statistical significance, we cannot state with certainty that that characteristic would influence a person’s preference for a CP.

Ranking of care package configurations by preference

Highest ranking care package configurations

Figure 10 shows the top-ranked CP configuration (i.e. rank 1) for each of the six different seizure contexts. There was similarity across the seizure contexts in terms of which attribute levels featured in the top-ranked configurations. In fact, three of the contexts had the same top-ranked configuration. Thus, only four different CP configurations appear in the figure despite there being six contexts. The configurations are labelled A–D.

FIGURE 10.

Top-ranked care package configuration for the different seizure contexts. The labels for the top-ranked configurations have been given solely for the purpose of the present report. They were not labelled as such within the DCE. ^a p < 0.05 based on 95% CIs.

The three contexts that had the same top-ranked CP configuration – namely, A – were when PWE considered an ‘Atypical seizure’ and when SOs considered a ‘Home typical seizure’ and a ‘Public typical seizure’. The probability of preferring the alternative (highest-ranked) CP over the current practice configuration ranged from 0.75 to 0.94.

For all contexts, the highest-ranked configuration included the paramedic having access to medical records or a care plan, an epilepsy specialist being available to advise the attending paramedic, the patients GP being notified of the incident and an appointment with a specialist within 2–3 weeks being booked for the patient. The only attributes that varied within the four top-ranked configurations were ‘What happens next’ and ‘Time’, which is explained by the differences in statistical significance of these attributes between across scenarios.

With respect to ‘What happens next’, none of the top-ranked configurations for the six scenarios involved the person being conveyed to ED. For five of them, the top configuration involved the patient staying ‘where they were’. For the remaining one – namely, for an ‘Atypical seizure’ being considered by a SO – the top configuration involved the patient being conveyed to a UTC. For this scenario, SOs did not have a significant preference for any location, but the location with the greatest positive coefficient was UTC.

The preferred ‘Time’ level varied between the contexts. PWE preferred 2 hours for an ‘Atypical seizure’, as did SOs for a ‘Home typical seizure’ and a ‘Public typical seizure’. For a ‘Home typical seizure’, PWE preferred 1 hour, as did SOs for an ‘Atypical seizure’. For the remaining context – namely, a ‘Public typical seizure’ – PWE preferred 3 hours.

How care package configuration approximating current care positioned within rankings

The top-ranked configurations differed in important ways from how current care is typically configured. The only attribute level within all the top-ranked configurations consistent with much of current practice was that participants wanted the patient’s GP to be notified of the incident. It is unsurprising therefore that for all six seizure contexts, the CP configuration representing current care was amongst those least favoured.

For PWE, the highest rank the current care configuration achieved across the six contexts was 230/288. For SOs, the highest rank it achieved was 220. The only CP configurations that ranked lower than the current care configuration was those which did not involve the patient’s GP being notified of the incident (see Report Supplementary Material 16).

Figure 11a and b show a heatmap of the rankings by seizure scenario and participant type and where current care appears (horizontal line Figure 11b). This data visualisation shows how preference varies across the 288 CPs and indicates even minor changes (i.e. single attribute level changes) would be estimated to result in a CP configuration that was typically more favoured by participants than current care. It also confirms how not all attribute level changes would be equal in the extent to which they result in a CP that aligns with participant preference.

FIGURE 11.

Heatmap of the ranking of 288 CP configurations across the six seizure contexts according to extent to which they align with service user preference. (a) Care pathway configurations ranked 1–144; (b) CP configurations ranked 145–288. Green = highest rank/most preferred; red = lowest rank/least preferred; vertical black line at rank 271 represents current care configuration.

Subgroup analysis

Design

The design adopted was relatively novel. It was informed by experience within the investigator team of completing KE endeavours and findings from the literature. Wilkins and Cooper’s175 definition of KE noted in Chapter 2 informed our approach. Our study design asked stakeholders to consider and debate the findings from the DCEs on service user preference, for them to share their views on the feasibility of users preferred attribute levels and to then identify together what they considered to be the optimal CP configuration.

Because of the following assumptions, a group, rather than an individual, approach to KE was considered preferable: a group of people should be less likely to arrive at a wrong decision than a single individual is; a group is more likely to lend some authority to the decision produced and decisions are improved when opinions are openly discussed, potentially challenged and views need to be justified. 194

Eligibility criteria

For the workshops, we sought a group of informed individuals/‘experts’ deemed to have a high level of knowledge and clinical experience of epilepsy, neuroscience and urgent and emergency care policy and practice. Persons from the following groups were considered the key stakeholder groups in this regard: ambulance clinicians, epilepsy specialists (neurologists, ENS and neuropsychiatrists), emergency medicine consultants, GPs, commissioners and managers.

To take part, persons needed to be aged ≥ 18 years, live in the UK and be able to provide informed consent and participate independently in English in a workshop.

At each workshop, we sought to have persons from each of the different groups represented. Evidence suggests heterogeneity in a decision-making group can lead to a better performance than homogeneity in terms of considering all relevant aspects of the topic. 195 We also sought to have representatives from a range of different English regions present at each workshop. This was important since ambulance regions can differ in their non-conveyance rates. 92 Moreover, infrastructure relevant to CP has historically differed between regions. 196

In addition to the participant groups noted above, at each workshop, we also sought to have at least one service user representative and for them to be an active participant. Their representation was primarily with the intention of helping the other participants, where needed, interpret the reasons for the preferences that service users expressed within the survey. It was not about seeking further evidence on the preferences of service users.

Deviations from original protocol

Recruitment

Clinical representatives were recruited from organisations participating in the WP1a survey of service providers. Commissioning and managerial representation was secured by asking the National Ambulance Commissioners Network and the Association of Ambulance Chief Executives National Ambulance Strategy and Transformation group (NASAT) to circulate an advert to appropriate members. NASAT comprises strategy and transformation directors with senior clinicians from each ambulance service being represented. Service user representation was secured by inviting members from the project’s PPI group.

Fifty people were ultimately selected and sent invitation packs that included a participant information sheet. A sampling matrix was used to facilitate broadly equal recruitment of persons from the different target groups. We intentionally sought to over-recruit by ~30% with the anticipation that some would not attend workshops they were scheduled to attend. 201

Persons receiving the invites could pass on the invitation to someone else if they themselves were not able, willing or suitable to take part. Repeat invites were sent if no response was received from them within 2 weeks. Those willing to participate were asked to inform the research team who sent them a link to an e-consent form.

To promote transparency, records of the workshops were kept, and they were audio-recorded and subsequently transcribed verbatim.

The KE workshops received Health Research Authority approval and approval opinion from the NRES Committee, West Midlands, Solihull (19/WM/0012).

Service user representatives were offered a shopping voucher of £20 as an acknowledgement of their time.

Procedure

Overview of structure and facilitation

Figure 12 illustrates the three-part structure of the workshops. Workshops were facilitated by BM, a university-based qualitative research assistant with a specialist interest in health services research but no specialist knowledge of the ambulance service. BM had not had any involvement in any other WP.

FIGURE 12.

Structure of WP3 KE workshops.

Analysis

The CP configuration(s) specified as representing the optimum by the different workshop groups were recorded and are presented in a matrix, alongside the DCE evidence.

The qualitative data from parts 1, 2 and 3 were thematically analysed using an approach informed by the work of Braun and Clarke. 202 It enabled scrutiny of the data to capture delegates’ views and justifications for their preferences. It was conducted deductively (with the identification of pre-existing themes underpinned by previous research) and inductively (with the identification of themes grounded in the data to identify patterns and themes related to the study objectives).

Beth Morris led the qualitative analysis and was supported by AN. Familiarity with the data was developed by them independently listening to the audio-recordings and reading the transcripts. BM generated codes through open coding and categorised these thematically. AN reviewed the codes and their application and suggested alternative interpretations until consensus was reached about the interpretation that best fitted the data.

Quotations are presented within the body of the results section to illustrate themes. There has been minor editing to preserve anonymity and ensure clarity of meaning.

When presenting quotes from neuroscience representatives, those from ENSs are presented separately to those of neuroscience doctors (i.e. neurologists and neuropsychiatrists). This is to preserve potentially different views. The merging of comments from neurology and neuropsychiatry was required to help maintain anonymity.

When describing participants’ views on the feasibility of different attribute levels a brief reminder is given each time as to the preferences of service users (as determined by the DCE).

Findings from part 3 of the workshop were used to refine the list of possible attribute levels that a CP for this target population would be recommended to assume. This would reduce the number of CP configurations for consideration from the possible to 288 (presented in Chapter 7) to only those which stakeholders considered optimal and feasible. Descriptive statistics were used to describe the expected utility of these remaining configurations.

Participants

From the 50 representatives invited, 35 confirmed a willingness to take part and were booked to attend a workshop. Twenty-seven of these ultimately attended a workshop: 14 females and 13 males. They included 10 ambulance clinicians (from 7 different regional services), 8 neuroscience representatives [from 5 different ambulance regions (3 neurologists, 1 neuropsychiatrist and 4 ENSs)], 5 commissioners (from 4 different ambulance regions) and 4 PPI representatives (from 4 different ambulance regions).

Two of the workshops lasted 2 hours, while one lasted 2.5 hours. One had 8 participants, one 10 and the other 9.

Themes

Analysis of the transcripts provided insights into the extent to which the DCE evidence aligned with the representatives’ clinical, professional or lived experience and the perceived feasibility of the preferred attribute levels. These shall be expanded upon in the following sections (with additional illustrative quotes for the themes being provided in Report Supplementary Material 19).

Feasibility of attribute levels and number of care package configurations required for seizure scenarios

Participants believed the different attribute levels preferred by participants would broadly speaking be feasible to implement in the NHS in the next 5–10 years. Moreover, they believed that a single CP for all the seizure scenarios was preferable and justifiable. They were of the view that there was sufficient commonality in service user preferences across the seizure contexts for this to occur. They considered a single CP would also be preferable from an administrative and commissioning perspective:

There are some challenges for ambulance staff in terms of quality versus performance … but er, I think for a lot a lot of ambulance services it’s [preferred CP configurations] probably not that, that far of a stretch actually.

Paramedic, Male 1

As shown in Figure 13, the three workshop groups were of the same view as to which configuration represented the optimum. It comprised ambulance clinicians having access to medical records, the person typically staying where they were, the time taken being < 6 hours (be it 1, 2 or 3 hours), for crews to be able to be advised by a specialist on the day, for the GP to be notified and for the incident to result in an appointment being made for the patient to subsequently have a follow-up appointment with an epilepsy specialist (be it within 1 or 2–3 weeks).

FIGURE 13.

Attribute levels specified by KE groups as representing optimal balance between NHS feasibility and service user preference. For columns presenting ‘Summary of evidence from DCE’: a green cell indicates an attribute level the respondents significantly preferred for the CP to have in that scenario; a red cell means an attribute level that respondents significantly preferred to not have in the CP for the scenario; white cells indicate those that did not reach statistical significance.

While the groups considered it to be NHS-feasible, they did note some instances when some of its preferred attribute levels might not be achievable and so some flexibility in permitted attribute levels would be required. These instances mainly centred around periods of ‘winter pressure’ and when a specialist who knew the PWE being attended to was not available to advise paramedics. Their qualifying statements are noted in Figure 13.

The specific feedback that the participants had for the feasibility of the different attribute levels is as follows. Their comments informed why they believed a CP could, in due course, be offered that was largely in line with service users’ preference.

Attribute 1: the ambulance clinician has access to medical records or a care plan

Reminder of DCE findings: Service user representatives wanted ambulance clinicians to have access to their medical records or a care plan across the seizure scenarios.

The consensus was this was achievable in the next 5–10 years. Indeed, some stated that it could happen sooner. Their justification being that in some regions, mechanisms were in place for the sharing of more rudimentary medical records (e.g. NHS Summary Care Records) with crews:

we’ve got access to Summary Care Records, and we’ve got access to a system called ‘Co-ordinate my Care’ … a purpose-built platform for er sharing care plans. [So implementing medical record or care plan access is] Very doable. You could do it tomorrow.

Paramedic, Female 1

Participants believed the key benefit of ambulance clinicians having access to such information was it could help them identify persons suitable for consideration of non-conveyance and increase their confidence:

[It could] give them that bit of reassurance … [paramedics] don’t work in an ED department where there’s somebody on hand to ask the question and get that second opinion … for me, it doesn’t have to be that physical person, it can be that well documented care plan that will give them the confidence to make that decision.

Paramedic, Male 3

In the future, the information was seen as also holding the potential to positively influence the vehicle/crew type that was dispatched:

… if this person’s preferred place of care is at home and … they have a history of epilepsy, you may well think actually we need to send a response vehicle with a senior clinician on it to this patient rather than a double crewed ambulance erm, to make sure that we’re providing exactly the right response …

Paramedic, Male 6

All participants were united in stating the information from the medical record that crews were given access to needed to be pertinent and concise:

… in the heat of the moment to kind of trawl … years of clinic letters or hand-written medical notes is you know, only half useful … the development of a very specific document … a care plan is where er, significant gains can be had.

Neuroscience doctor, Male 1

In terms of what was pertinent, participants said it should cover ‘the baseline for that patient as an absolute minimum’ and have ‘some representation of that patient’s wishes’ (Paramedic, Male 6).

Ambulance clinicians can differ widely in terms of their training and experience. 61,108 For this reason, participants emphasised the information needed to be in a format and language accessible to all. Participants also reflected on some limitations of what is already being made available to crews for other presentations. They said lessons should be learnt to ensure the information provided is consistent and easy to use:

We cover over 30 CCG [clinical commissioning group] footprints, so we’ve got er over 30 variations in documentation that we may see on scene.

Paramedic, Male 6

Attribute 2: what happens next – are patients conveyed and, if so, where to?

Reminder of DCE findings: Service user representatives expressed a tendency to prefer being cared for ‘where they were’ and avoid ED. For typical seizures at home, PWE and SOs had a statistically significant preference to remain where they were. PWE also significantly preferred avoiding ED and UTCs for seizures occurring at home and to avoid ED for typical seizures in public.

Workshop participants were mostly in agreement that it was feasible to follow service users’ preferences to stay where they were following a typical seizure and that this is becoming more common practice:

… do we think this is feasible … patients with diagnosed epilepsy with a typical seizure presentation 100%. We would engage with this … I think that’s absolutely been the best option for a while [stay at home] and paramedics are gaining confidence in that in their current practice.

Paramedic, Female 2

Participants were more circumspect in their support for the preference of PWE (1) to stay where they were following typical seizures in public and (2) to not be conveyed to ED following an atypical seizure. It was noted how caring for patients where they are after an atypical presentation, rather than conveying them to ED, would represent a major change in practice for many clinicians:

… with atypical seizure presentations, most of us have quite low thresholds to take patients to ED.

Paramedic, Female 2

Much discussion was also had about the range of possible presentations that can be captured by the term ‘atypical’. They stated the CP preferred by services users might be suitable for some, but not all of them. They acknowledged the parameters of the atypical seizure scenario used in the DCE and agreed that for this specific variation, the patient staying where they were should be feasible.

The difficulty with referring to ‘atypical’ is it’s a wide term. If a patient’s seizure lasted 30 seconds longer than normal that may be construed as atypical, but they are probably still safe to stay at home …

Paramedic, Male 4

Because of the potential elevated risk of atypical seizure presentations, participants highlighted how consideration might need to be given as to which grades of ambulance clinician would be permitted to use the CP for them. They highlighted this was already done for some CPs used for other presentations. They also noted that the issue would become most pertinent during periods of high demand when ambulance services are sometimes supported by voluntary staff and private services.

… there aren’t just paramedics throughout ambulance trusts, there are non-registrants that go out to patients on their own as well …

Paramedic, Female 3

Attribute 3: time taken to be assessed, monitored and treated

Reminder of DCE findings: Service users, across all seizure scenarios, expressed a statistically significant preference to avoid being assessed, monitored and treated by an emergency healthcare professional for more than 6 hours. SOs also had a significant preference, following a typical seizure at home, for the time to be 2 hours.

Workshop participants believed a CP that reflected patient preferences for time would be feasible:

… absolutely achievable and probably for the most part er, that is something that we achieve with … cases already.

Paramedic, Male 4

They also highlighted how this could theoretically be achieved if the seizure occurred in public:

where would you allow them to recover safely … One thought was in the back of an ambulance …

ENS, Female 3

Nonetheless, participants did highlight how feasibility might reduce during periods of high demand and that there might be operational challenges if crews stay with persons for long recovery times (rather than being available to respond to other incidents):

… there will always be tensions between you know, call volumes … some days it would be possible to, to maintain that kind of stance [2-hour timeframe] but on other days … It just may not always be possible …

Commissioner, Female 2

Attribute 4: paramedic access to specialist advice on the day of seizure presentation

Reminder of DCE findings: Across all seizure scenarios, service user participants expressed a statistically significant preference for their emergency healthcare professional to have access to a healthcare professional with specialist training in neurology for advice. The attribute did not specify that the professional would be someone they had previously received care from.

While workshop participants believed this to be feasible within the next 5–10 years, they were not in consensus as to who should be providing the advice. Some epilepsy service representatives believed that for the specialist – be it an epilepsy nurse or neurologist – to offer meaningful support they needed to be personally familiar with the patient. With existing capacity, they stated this would not be likely to happen:

… if it’s not someone who knows them or who really understands their epilepsy and their background then actually the value of that specialist advice I think is massively diminished.

Neuroscience doctor, Female 1

They were of the view that access to medical records that include high-quality emergency seizure care plans would be more helpful than specialist advice:

… if they’ve got a seizure care plan, if they know their treatment plans and it’s all written out, actually they [crews] won’t need this … they don’t need the specialist advice.

Neuroscience doctor, Male 1

However, it was acknowledged that not all PWE being attended to by the ambulance service are ‘known’ to a specialist. They might not have been seen by them for a substantial period, or indeed ever, in some instances. Moreover, as reflected by a service user representative at the workshop, even some of those that are ‘known’ to specialist services do not currently have emergency seizure care plans:

I don’t have a care plan, and I do wonder how many other people with epilepsy don’t really have a care plan.

PPI, Female 2

In contrast, ambulance clinicians were keen to emphasise any sort of advice from an epilepsy specialist would be welcomed – be it from a healthcare professional who knew the patient or not. They emphasised how they work in a largely isolated way and so would value access to any sort of additional expertise to decisions:

… when you’re working in the ambulance setting … you can’t just kind of pop out to like the, the cubicle and catch your colleague …

Paramedic, Female 1

They felt access could be particularly helpful when crews were managing atypical seizure presentations and for newly qualified paramedics. Whoever it is that provides the specialist advice, it was agreed that any scaled system of support would need to be robust for ambulance clinicians to have confidence utilising it:

… there’s got to be a reaction, there’s got to be someone picking up that phone [when you call for advice] … otherwise you get a couple of er hangings on or no pick up and very quickly that can … lead to lack of confidence … crews … saying … you know don’t bother … it’s never working and we just end up conveying anyway …

Commissioner, Male 1

In terms of how access would occur, ambulance participants explained the infrastructure is already in place in many areas as they use it to access advice from different specialities for other presentations.

Attribute 5: general practitioner informed of seizure presentation via report from ambulance clinician

Reminder of DCE findings: Service user representatives for all seizure scenarios expressed a significant preference for the patient’s GP to receive a written report from the ambulance service.

There was a consensus at the workshops that this would be feasible in the next 5–10 years. In fact, participants pointed out that in those regions where crew members complete patient report records electronically, it was already happening:

When we discharge someone on the scene, the GP is automatically emailed … as long as we can trace the patient on the [system] a … record from us of our attendance. So yeah absolutely, easily achievable …

Paramedic, Male 3

… we’re a little bit behind the curve on erm, electronic er report forms, but I, I don’t think that would hold us back in the, the timescales that we’re talking about.

Paramedic, Female 1

Attribute 6: additional contact with an epilepsy specialist arranged by ambulance service

Reminder of DCE findings: Service user representatives had a statistically significant preference for the emergency healthcare professional treating them on the day to arrange for them to have a follow-up appointment with an epilepsy specialist. They did not have a significant preference for whether the appointment occurred within one or within 2–3 weeks.

Overall, workshop participants believed this attribute level was feasible to achieve in the next 5–10 years.

Neuroscience representatives highlighted how other parts of the urgent and emergency care system – particularly ED staff – can already arrange follow-up clinic appointments for the patients they see. Extending this to ambulance crews was considered viable.

Ambulance clinicians said they already arrange follow-up appointments for other presentations with different specialities. They stated they would prefer for the referral process to be via an online service. If telephone calls were required to arrange it, they wanted this be delegated to another part of the service, such as to those within the ambulance service’s central hub, NHS 111 or primary care:

Yeah, it’s absolutely doable and we do it for other areas and er other clinical conditions.

Paramedic, Male 3

what we don’t want to do is leave our crew on scene … chasing telephone numbers and referrals and doing lots in the back of an ambulance.

Paramedic, Male 6

Neuroscience representatives wanted to note that the feasibility of the patient having an appointment with a specialist could vary between regions and potential inequities in access existed. It was highlighted how some services were already struggling to meet wait time standards for the referrals they were receiving from more traditional referral routes. This issue was not, however, deemed to be an insurmountable barrier. Participants believed that with expansion in capacity and/or a change in how existing resources were utilised, that providing the desired follow-up would be feasible:

Where there isn’t a larger group of epilepsy nurses or epilepsy specialists, then they would have to look at investing and training and setting up services. But I think that would be possible …

ENS, Female 3

… we’ve looked at the way we run our services and made quite a radical change … we’re not booking routine follow-up appointments … they can contact us. … that’s allowing more capacity … that’s now our mission – that we get back to calls within the day because they’re, they’re patients or healthcare professionals that really need to speak to us.

ENS, Female 2

Survey of service providers

The cross-sectional survey – to which over 72 services responded – confirmed the impressions that instigated the ‘Collaborate’ project. Specifically, nearly all responding ambulance services were considering, or were working towards implementing an alternative CP for crews to use for epilepsy. Moreover, there was a lack of consistency and equity in the nature of the CPs being considered.

The CP configurations being considered differed in terms of where the non-conveyed person would be taken. In some instances, they would remain at home, in some ‘on scene’ or in others taken to an UTC. The CPs also differed in the extent to which they would help potentially stimulate improvements in the person’s subsequent ambulatory care and potentially address known health inequalities. Only a few included a mechanism by which the patients would be referred for a review of their epilepsy, and then they differed in terms of eligibility, timing and who conducted the review.

A final important finding was that, despite being a statutory obligation, almost none of the responding services had consulted service users on the proposed or intended changes. Thus, even having surveyed providers, it remained unclear which CP configuration was best positioned to deliver benefits since the extent to which they were acceptable to service users was unknown. This underscored the need for the subsequent WP elements.

In helping clarify the characteristics of the CP configuration being considered, the survey provided the evidence needed to help us to ensure that the CP options presented to PWE and SOs within the DCEs were plausible options. This ensure were therefore modelling preference for realistic CP configurations.

Qualitative interviews with service users and ranking exercise

In-depth interviews were completed with 30 service users. They provided important insights into some of the challenges PWE continue to experience, how disruptive seizures can be and how a lack of information transfer and co-ordination among services poses challenges for this population.

The interviews also highlighted some concerns that service user would have with some potential CP configurations. These would be important to address, even if it were simply addressing possible misunderstandings on their behalf (e.g. that the opening times of UTCs should be sufficient to capture most seizure incidents in people with established epilepsy).

One of the additional purposes of the interviews was to help identify attributes of care that were important to service users and which would need to be used to describe the CP options presented within the DCE. To do this, the interviewer worked with participants to complete a ranking exercise, towards the end of the interview. The approach helped us manage the tension that can exist between the complex and nuanced findings produced by qualitative research and the reductive nature of attribute development. 149 The exercise was important as service users identified attributes of potential importance that the research team and literature had not initially considered.

What emerged from the interviews and ranking exercise was an approximation of the factors contributing to the ‘decision calculus’ that service users use when considering postseizure care options. At this stage, while their relative weight of importance was not known, it was becoming apparent that service users would likely want to know where the CP would take them (or not, in the case of non-conveyance); whether the ambulance crew would have access to their medical records or care plan; how long it would take for them to be assessed, monitored and treated during the emergency episode; the extent to which the emergency care provider could request advice from an epilepsy specialist (e.g. epilepsy nurse, consultant); to what extent the CP would mean their usual care providers would be informed of the incident; and, finally, whether any follow-up from a epilepsy specialist would be instigated.

Discrete choice experiment survey

Aided by the evidence from the survey and qualitative interviews, a DCE survey was developed and completed by almost 600 service user participants. It elicited care preferences for three hypothetical seizure scenarios (‘Home typical seizure’, ‘Public typical seizure’ and ‘Atypical seizure’). The findings were illuminating.

For all seizure scenarios, both PWE and SOs (i.e. close family and friends) were found to want a CP configuration that differed markedly from what they are currently offered. Specifically, in all instances, they preferred: for the paramedic to have access to their medical records or a care plan; for a health professional with specialist training in neurology (be it an epilepsy nurse, neurologist and not someone necessarily known the patient) to be available to advise paramedics; for their GP to receive a report from the ambulance service; and for the incident to result in an appointment being arranged for them with an epilepsy specialist in the future. Whether an appointment occurred within 1 week or within 2–3 weeks was not found to be a significant determinant of preference. In terms of ‘What happens next’, there was a pattern of preferring to avoid conveyance to ED and to remain where they were, or in one instance to go to a UTC. With regard to ‘Time’ (i.e. how long the CP involved the person being assessed, monitored and treated), everyone wanted to avoid the longest duration of 6 hours in all seizure scenarios.

That service users want something clearly different to what is currently offered was most starkly illustrated when we ranked 288 different CP configurations according to their expected utility to service users. The configuration representing current care was amongst the least favoured in all scenarios.

Perhaps more positively, the DCE indicate the improvement in CP utility that would be expected to result from specific attribute changes for each seizure scenario. If only limited changes are considered possible, these could help service providers identify which changes to prioritise. The attribute levels with the most consistent and largest impact on preference across the scenarios were (1) providing the crews with access to medical records or a care plan and (2) having a specialist who is able to advise the ambulance crew during the episode. Table 16 shows the expected utility for each individual configuration.

There were instances of variation between seizure scenarios and participant types in terms of exact preference. These are important to note and will be discussed shortly. However, what was most apparent was the sizeable similarity across and between them in terms of what care attributes were important. This raised the possibility that a single CP configuration might be possible to deploy for all target seizure scenarios. This was something explored with stakeholders in the WP3 workshops.

Interestingly, the slight differences that did exist between seizure scenarios and participant types in terms of the direction and strength of effect different attributes appeared to hold face validity. For instance, in terms of the different seizure scenarios, participants ascribed most value to paramedics having access to a patient’s medical records or a care plan when a seizure occurred in public. Our PPI group helped the research team interpret this. They said it was likely because, in public, a PWE is more likely to be unaccompanied by a SO who could share medical history information. Also, it is in these instances that the call for an ambulance would often have been made by a concerned member of the public and the presence of the medical record would provide important context for the incident. Indeed, it could help avoid misinterpretation of the episode. There is the risk that unusual behaviour associated with some seizures can be misinterpreted (e.g. as stemming from intoxication, mental illness or representing disorderly conduct). 203–205

Participants also expressed a stronger preference for an epilepsy specialist to be able to advise the paramedics on the day and for their GP to notified of the event when the seizure presentation was atypical. Our PPI group again helped the research team interpret this. They were of the view that this made intuitive sense. They stated that while still potentially distressing, typical seizures are the manifestations PWE and SOs would have most knowledge of. Consequently, they are the ones they might feel most comfortable in managing. Atypical seizures, however, were the manifestations they would be most concerned about and thus would likely most value specialists guiding paramedics on. The PPI members said the need for attention from an epilepsy specialist after an atypical seizure was a key message frequently relayed to them by their own usual care providers and by seizure first aid resources.

Responsibility for care decisions after a seizure is often delegated to SOs and so it was important to capture their preferences. We found few differences existed between PWE and SOs. The degree of similarity in their preferences lends further support to the possible deployment of a single CP configuration for the target population.

The similarity for the most part in the views of PWE and SOs is perhaps remarkable. SOs were after all instructed to express via the DCE their preferences for the care of the PWE they knew (even if it was different to what they expect the PWE they knew would say). Also, SOs were (as we had anticipated) often ‘representing’ a different part of the target population.

One of the areas of divergence that did exist between PWE and SOs in their preferences related to the attribute ‘Time’ for a ‘Home typical’ seizure. Like PWE, SOs wanted the ‘Time’ it took to be assessed monitored and treated to be < 6 hours. Uniquely though, SOs also expressed a preference for the time to be 2 hours, rather than 1.

This desire for a longer time period in the presence of a paramedic may be explained by the different epilepsy profile of the PWE known by the SOs. For example, they were more likely to have an intellectual disability. This may lead to slightly different care preferences since some epilepsy syndromes with unique presentations are more common in those with an intellectual disability (e.g. Lennox–Gastaut syndrome). 206 The presence of an intellectual disability may also be important because it can take more time to be confident if a person with one is beginning to recover to their baseline health following a seizure (e.g. due to possible communication difficulties). 207 Another characteristic of the PWE known by the SOs that may also be important is they were more likely to be prescribed a rescue medication. Administration of benzodiazepines following a seizure is associated with longer recovery times. 208,209 The aforementioned explanations do, however, remain speculative since we did not have the sample size to formally test them (see Strengths and limitations).

The other notable difference between PWE and SOs in their preferences was for the attribute ‘What happens next’ when imagining an ‘Atypical seizure’. PWE preferred to ‘Stay where they were’, with going to an UTC or an ED associated with increasing reductions in utility from their perspective. For SOs, staying at home was associated with disutility (i.e. preference to avoid). They instead preferred conveyance to a UTC, followed by ED.

This may again be explained by the different epilepsy profiles of the PWE they knew. Rescue medication was more highly prescribed amongst this group, and they reported having had more contact with the ambulance service in the prior 12 months. Ambulance guidelines40 recommend conveyance to a health facility for all those treated with benzodiazepines (unless they have care plan that states otherwise). Awareness of this amongst the SOs might explain their preference for a UTC.

Knowledge exchange event

Three KE events were run with 27 stakeholder representatives (ambulance clinicians, neurologists, epilepsy nurses and commissioners). Having shared the key DCE findings with them, we asked the stakeholders to reflect on which attribute levels represented the optimum balance between service user preference and NHS feasibility.

Stakeholders were found to be broadly of the view that the attribute levels favoured by service users for the different seizure scenarios were largely feasible. Importantly, there was also consensus amongst workshop participants that there was sufficient consistency in the preferences of service users across the seizure scenarios to focus on developing a single CP for epilepsy.

That workshop participants considered service users preferences to be largely feasible may be attributable to the detailed formative work we completed for the DCE survey. WP1c helped ensure the attribute levels and combinations of them presented within the DCE were anticipated to all be within the bounds of realism and likely to be safe.

Across the three KE events, the CP configuration which stakeholders said should be prioritised for implementation was comprised of ambulance clinicians having access to medical records; the person largely staying where they are (if judged to be clinically appropriate); the time being < 6 hours (be it 1, 2 or 3 hours); for crews to have access to specialist advice during the episode; for the GP to be notified of the incident; and for the episode to generate a follow-up appointment with an epilepsy specialist (either within 1 or 2–3 weeks).

Based on stakeholders’ judgement regarding the optimal attribute levels, 18 possible CP configurations can be constructed. The DCE evidence permitted us to estimate their utility and where they would rank out of 288 possible CP configurations (with rank 1 being the one best aligned with service users’ preferences). It was found all 18 of the CPs would be expected to carry high utility. The configurations would be estimated to have a median rank for PWE of 9.5 (range 1–19) for an ‘Atypical seizure’, 30.5 (range 1–71) for a ‘Public typical seizure’ and 42.5 (range 1–60) for ‘Home typical seizure’. The configuration representing current care was ranked by PWE as low as 230/288 for a ‘Public typical seizure’, 247/288 for a ‘Home typical seizure’ and 248/288 for an ‘Atypical seizure’. Thus, any of the 18 CPs should mean that users would be being offered a preferable CP.

Stakeholders also offered other important insights which further underlined the value of the KE exercise. One of these was that while they said the attribute levels so far noted were optimal and, on the whole, would be NHS-feasible, there were circumstances during a calendar year or location when these targets might not be achievable. Thus, flexibility for some of the attributes might be required to maintain deliverability. Namely, during periods of intense pressure on the NHS and if a specialist known to the patient being cared for is not available to provide advice to crews.

To support subsequent discussions about what to do during these periods and what impact the changes could have, we estimated the utility and rankings of the amended CP configurations. The changes to attribute levels did reduce expected utility and as the DCE findings discussed earlier would suggest, the impact of these changes was not equal.

Of most concern was the reduction in utility of the CPs seen when restrictions were placed on attributes so they assumed the levels expected to be needed during periods of ‘winter pressure’. One of these related to increased ‘Time’ for assessment, monitoring and treatment. The change in attribute levels in this circumstance were sufficient to mean seven of the CPs that could be generated could be perceived as ‘worse’ than current practice by service users. This finding highlights the disutility (utility loss) attributable to the increase in time (6 + hours) experienced during ‘winter pressures’. The reduction in utility was less pronounced if the only attribute restriction was that the CP was not permitted to involve a specialist being available to advise attending paramedics.

In addition to helping understand optimal attribute levels, stakeholders also provided granular information that could be helpful for implementation of a new CP configuration (e.g. how best to ‘brand’ it to ambulance crews, what grades of clinicians could to use it and which ambulance performance measures might need to evolve to ensure they support CP use). They also identified which attribute levels likely require the most attention in terms of work and investment to make them a reality. Namely, seizure care plans/access to information about the person’s medical history and how best to use existing specialist capacity.

Survey of service providers

This survey received an excellent response rate of 82%. This is likely attributable to the importance of the problem and concern about the inadequacy of current practice. What also might be important was the survey’s short duration, the manner of invitation and that the survey was carefully piloted. Another strength of the survey is that we asked responding services who reported no planned or recent CP changes about what ‘usual care’ looked like for them. This helped ensure we identified already embedded changes made by the more innovative service providers prior to the 5-year period asked about.

The survey also asked respondents to report on changes their service or ones they work with had or were planning to make. We took this comprehensive approach since we were aware some of the CP changes might lead to service providers other than those primarily responsible for instigating the CP change being asked to deliver them or support it (e.g. for epilepsy specialists to receive referrals from the ambulance service). A potential limitation to this approach is that we cannot rule out ‘double-counting’ of changes (e.g. multiple neurology centres or EDs might report a change being instigated by one regional ambulance service). If this did occur, our estimate on the number of services making or considering a CP change might be exaggerated.

Qualitative interviews with service users and ranking exercise

Going into the DCE phase, the qualitative interviews presented the best available evidence on attributes of importance to service users. This was preferable to relying on expert judgement or researcher opinion.

For the interviews a purposive sample was sought to try to obtain the full range of views about the topic. It is probable that the sample we eventually recruited did not include the full range of views about the topic. Not only were participants drawn from only one English region, but they were also well educated (72% had post-secondary level qualifications compared to ~44% in the UK general population170). This may explain why our findings showed participants’ decision-making was mostly concordant with medical guidelines.

Discrete choice experiment survey

A key strength of the DCE was its rigorous development and thorough description. 173

The strengths of the formative work include (1) the systematic approach to attribute development and refinement that ensured those selected were important to service users, but also NHS-feasible and capable of experimental manipulation, (2) a mixed-methods approach to attribute selection and scenario development, that enabled greater interpretation of the relationship between attributes and the decision-making process and maximised face validity of the vignettes and (3) that its conduct aligned with good practice guidelines. 149,173

An indicator of the strength of the formative work is that having read the seizure scenarios vignettes, ~70% of the PWE said they were familiar with them. A potential limitation of the formative work was that the sampling frames used to recruit PWE and SOs for the qualitative interviews was subject to bias.

When the COVID-19 pandemic began to impact the UK, rapid revisions to the recruitment methods for the formal DCE were required to ensure it could continue. We described in detail what changes were made, their rationale and how the DCE survey was ultimately administered. In providing this information, we conform to reporting guidelines. 191

Our revised approach enabled us to secure usable stated preference data from approximately 600 PWE who had recent contact with the ambulance service and their SOs – a population that is challenging to identify and potentially vulnerable. The sample size was sufficient to examine preferences for all seizure scenarios and for all participant types. It was though insufficient to allow us to complete all the subgroup analyses of interest.

A key strength to our DCE was the representativeness of the sample we recruited. This is likely in part because SOs could still participate in the DCE even if the PWE they knew could not (e.g. due to a lack of capacity). When we compared our sample’s headline characteristics to the target population, we found limited differences. Nevertheless, given how important sample representativeness is to the generalisability of our findings, in the subsequent paragraphs, we consider ways in which our methods might have led it to differ from the parent population.

Firstly, under the revised recruitment and survey approach, all participants were asked to complete the DCE online by themselves. Online self-completion is the most common approach used for DCEs. 134 Ninety-six per cent of households in Great Britain report internet access. 210 Nevertheless, there is the possibility that the revised approach could have excluded from participation a minority of people who could not or did not want to participate in this way. Digital exclusion is linked to wider inequalities in society and is more likely to be faced by people over 65, those on low incomes, and disabled people. 211 Our sample was certainly 8 years younger than the target population, and this could be due to the manner of DCE completion. It is hard to know how the slightly younger age of our participants impacts on preferences. With regard to the other factors that can be associated with digital exclusion, it is pleasing to note our sample was at least comparable to the target population in terms of social deprivation.

Secondly, rather than recruiting most participants directly from the ambulance service, we, in the end, recruited most of our DCE participants by public advertisement. This included the adverts being circulated by epilepsy user groups to their affiliates. The approach enabled recruitment of a sufficient number of participants during the COVID-19 pandemic and permitted us to capture the views of people from across England. Nevertheless, it has been contended that people affiliated with user groups might differ in important ways from those who are not. It is important to note that our DCE subgroup analyses did not detect any significant differences in the care preferences of participants recruited by public advert and those recruited directly from the ambulance service.

Thirdly, participants completed the DCE within the pandemic. Concerns regarding acquiring an infection and requests to ‘socially distance’ could have influenced the preferences people expressed. Our results do not support this possibility. While the pandemic had changed some participants willingness to access certain NHS services at the time, this did not for the most part alter their preferences for postseizure care during ‘normal times’.

Finally, one notable difference between our sample and the target population was the large proportion (86%) of our participants that reported contact with an epilepsy specialist in the 12 months prior to recruitment. The NASH-3 initiative reported only 52% of PWE attending ED would be expected to have had such contact. 44 This may suggest those who took part in the study represented those who are more supported from the target population. However, it may also reflect a difference in methods.

Specifically, we asked participants whether ‘they had seen or spoken to a doctor or nurse that specialises in epilepsy? (such as a neurologist or epilepsy nurse specialist)’. NASH, in contrast, relied on what, if anything, was recorded within the ED records regarding the extent of contact the person had had. It is also possible that during the period of recruitment, specialist services were initiating more contact with PWE in the communities they served due to the pandemic (e.g. to inform them as to how get in contact during periods of COVID-19-related restrictions, to rearrange booked appointments or offer more general advice regarding how to manage epilepsy in the context of the pandemic). We know some services were doing this.

Knowledge exchange events

For WP3, we developed and then used a novel approach that allowed us to work, during a global pandemic, efficiently and collaboratively with representatives from key stakeholder groups.

Strengths include the standardised approach used to share the DCE evidence and the efforts given to ensuring the information was presented in an accessible way. The composition of the groups at the events was also a strength. Different stakeholder groups and persons from different regions of the country sometimes expressed different priorities and experiences. It was important that other stakeholders were aware of these prior to the group making their judgements as to which CP(s) represented the optimal configuration.

The use of well-established NGTs was also a strength. Despite the short duration of the events and that participants were not familiar with one another, each participant shared and discuss their views openly and constructively. Moreover, there were no instances of particular participants dominating the conversation. 200

It was reassuring that the stakeholders at the event largely identified the attribute levels preferred by service users as all being deliverable and realistic. This is likely to be attributable to the robust formative works that informed the DCE.

The KE workshops built on the evidence obtained by the preceding WPs. They enabled the project to identify the configuration/s that appear to strike the optimal balance between service user preference and NHS feasibility and are the strongest candidates for implementation and evaluation. This addressed an important information gap which determined the need for the current project. It is important to acknowledge, though, that their identification is ultimately based on an informed judgement made by stakeholders. It remains to be determined how well they will work in practice. Evidence from the different WPs emphasised how a range of contextual factors might influence whether a CP would operate effectively. Realist evaluation methods could ultimately help understand for whom the candidate CPs do work, to what extent, in which circumstances and how and why. A final, potential limitation of WP3 is that due to the need to restrict participant numbers, there were no representatives at the events from general practice or emergency medicine. They might have shared views and experiences that could have changed the judgement of the group with regard to optimal CP.

Design of the research

The project arose as the research team heard of moves towards alternative CPs for ambulance users and were aware that the views of PWE were not being considered. Epilepsy Action – the largest UK user group associated with epilepsy – were also aware and concerned. We thus held a PPI event with the support of Epilepsy Action and the ‘Brain and Spine Foundation’. The context and proposed study were shared with 23 PWE and their SOs (family, friends). Ten of the representatives had had seizures in the prior year and eight had visited a hospital ED. They gave feedback on a draft of the project and suggested changes and ways to ensure it would answer the questions they had, including on ways to capture some indication of the views of PWE who could not themselves due to a severe learning disability. Users were disappointed by the current trajectory and welcomed the current study. Asked how important the proposed project was on a scale from 1 to 10 (10 = extremely important), the average rating was 9. After the project was commissioned, PWE and SOs continued to be actively involved at every stage of its completion.

Developing participant information resources

In the study development stages, we worked with PPI users from Epilepsy Action to develop recruitment materials.

Management of the research

The study was supported and directed by an Independent Steering Committee and a DCE development panel. Both included two service users. A study PPI group of 15 volunteers whose lives were impacted in some way by epilepsy was also informed. They were all supported in their role by Epilepsy Action who have an active PPI scheme. Service users were reimbursed for travel and time according to the INVOLVE framework. One of the key roles of the PPI group was scrutinising the draft version of the DCE survey and sharing their views on its ease of understanding.

Analysing the research and dissemination of research findings

As the results of the study were secured, the service users on the study Steering Group and PPI group were all actively engaged with and helped with our interpretation of the results. Having now finished the study and as we are moving towards dissemination, we shall be seeking to work with users to help us share the findings, including co-presentation at conferences and meetings.