Seizure first aid training for people with epilepsy attending emergency departments and their significant others: the SAFE intervention and feasibility RCT

Evidence from adult and paediatric literature at the time of project design

Evidence from the literature on self-management interventions for people living with other chronic conditions such as diabetes,82,83 arthritis84–86 and asthma87–89 indicates that it is possible to elicit improvements in self-management. Consequently, in these fields self-management support has become well established. 90–96 In the UK, self-management courses have become freely accessible to people with diabetes. 97,98 The evidence base on improving self-management skills in PWE is more limited.

A Cochrane review by Bradley et al. 99 examined care delivery and self-management strategies for adults with epilepsy. It found four self-management studies focusing on adults with epilepsy. 100–103 None of the trialled programmes focused exclusively on seizure management or those attending EDs or systematically involved SOs, and none had been trialled in the UK. However, the review did conclude that there was tentative evidence that such interventions can improve epilepsy self-care skills.

Helgeson et al. ’s100 small US RCT is noteworthy because it evaluated an intervention that included some discussion of seizure first aid. The intervention, the Sepulveda Epilepsy Education programme, is a psychosocial 2-day group course. 100 At follow-up at 4 months, participants receiving the intervention demonstrated a statistically significant increase in their understanding of epilepsy, a decrease in seizure fear and a decrease in hazardous medical self-management practices compared with wait-list controls. However, health service utilisation was not measured and no significant changes were found in relation to anxiety or confidence managing epilepsy.

As the number of trials conducted with adults with epilepsy is small, Lindsay and Bradley’s104 Cochrane review of self-management interventions for children with epilepsy and their parents is instructive. Lindsay and Bradley identified two RCTs of interventions that contained modules on seizure first aid.

The first, conducted by Tieffenberg et al. ,87 evaluated a Spanish group-based programme, the Civil Association for Research and Development in Health (ACINDES in Spanish), for children aged 6–15 years and delivered by teachers to educate children and parents about epilepsy. It consists of five weekly 2-hour sessions, followed by a reinforcement session 2–6 months later. At follow-up at 12 months, children in the intervention arm showed statistically significant improvements in epilepsy knowledge compared with children in the control arm. There was also a significant reduction in ED visits. The mean number of ED visits made by children in the intervention arm in the 12 months prior to randomisation was 0.90. This reduced to 0.22 at follow-up at 12 months. For the TAU arm the mean number of ED visits made by children went from 0.83 to 0.46.

The second RCT was by Lewis et al.,105,106 and examined the effect of another Spanish intervention called the Children’s Epilepsy Programme. This intervention taught children and parents about seizures, living with epilepsy, and communication. Children and parents separately attended four 1.5-hour group sessions and met together at the end of each session to share learning. At follow-up at 5 months, children in the Children’s Epilepsy Programme arm had significantly improved epilepsy first aid knowledge compared with children in the control arm. 105 Their parents also showed significantly greater reductions in anxiety than parents in the control arm. 106

That training in self-care is associated with reduced service utilisation in the paediatric epilepsy population, without compromising patient outcome, concords with the larger, higher-quality evidence base on interventions to reduce ED use by children with asthma, which is another chronic, relapsing condition. 88,89,107 Boyd et al. ,89 for example, completed a Cochrane review of 17 RCTs of educational interventions for children (and their parents) at risk of asthma-related ED attendances. Data from > 3000 children who presented to an ED for acute exacerbations and were then followed for an average of 10 months were included. Educative interventions led to a 37% reduction in the relative risk of reattendance at an ED in the treatment arm compared with the control arm, and a 21% reduction in the relative risk of subsequent hospital admission in the treatment arm compared with the control arm.

Conclusions from trials regarding the ability of self-management interventions to elicit behaviour change in PWE do, however, need to be tempered, because most trials have been found to be of low methodological quality. 108–110 Such trials are at a higher risk of bias and potentially overestimate effects (see Savovic et al. ). 111 In Helgeson et al. ’s100 trial of the Seizures and Epilepsy Education programme, for instance, no information was provided regarding random sequence generation or allocation concealment, only 38% of those randomised to the intervention completed it and outcome assessments were not blinded. 110

The course in its original form

The ES’s course was delivered to groups of 10–20 people. It lasted ≈ 3 hours, with breaks included. Educational aims for course recipients are specified and outlined in Table 1. Materials for the course included slides, professionally produced video clips of seizure types and first aid and additional information booklets on topics such as risk management and emergency medication. An information pack provided participants with a permanent record of the material and included space for notes to promote active processing of material as well as participation.

The intention was for learning to be elicited rather than taught, with the behaviour of the educational facilitator seeking to promote a non-didactic approach. Course participants were encouraged to share experiences and ask questions. To some extent there is therefore meant to be tailoring of the information that is presented so that it aligns with the needs of the group being taught (i.e. is patient centred). This, and that the course consists of a number of interacting components that may act both independently and interdependently, means that it is what the Medical Research Council refers to as a ‘complex intervention’. 121,122

Justification for developing seizure management training: a policy and service user response

The ES course is not informed by detailed theoretical modelling or a clear behaviour change model. Having reviewed and observed the intervention, our multidisciplinary research team nevertheless identified several ways by which receipt of the intervention (or a suitable adaption) could plausibly support PWE to make fewer ED attendances and improve patient and SO outcomes. These are detailed in full in Appendix 1. In brief, it was considered that it might increase patients’ and SOs’ practical understanding of seizure management, increase their knowledge of how to make appropriate care and lifestyle decisions (including the need for ED attendance) and reduce fears about risk, thereby increasing self-confidence and empowerment.

Importantly, the aims of the programme also broadly aligned with what PWE and their SOs had generally said they wanted and how they wanted to receive it. For example, studies consistently show that PWE and their SOs want more information about living with epilepsy. 27,31,123–129

Adapting SAFE

To maximise acceptability, benefit and behaviour change potential, it was recognised that the ES’s course would probably require adaptation for the target audience. To this end, the ES agreed for their course to form the basis of a new, adapted course, entitled SAFE, for PWE who frequently attend EDs and their SOs. To identify the changes required we utilised a collaborative framework underpinned by a philosophy of experience-based co-design. 130 This is an approach to improving health-care services that combines participatory and user experience design and processes to bring about quality improvements in health care. 130

Stage 1: recruitment of health-care professional representatives

Different health disciplines can be involved in the care of PWE. Some patients will identify a general practitioner (GP) as the main provider of their ambulatory care whereas others will identify a specialist, such as a neurologist or epilepsy. The voluntary sector is also an important support structure for many PWE. When someone seeks emergency care for a seizure, other parts of the health system come into contact with PWE, including paramedics and ED staff. All parties were considered as being positioned to be able to offer insights into the support needs of PWE who attend for seizures. We therefore chose to adopt purposeful sampling for the identification and selection of information-rich cases136 from the main parts of the care pathways encountered by PWE. This involved identifying and selecting individuals who were especially knowledgeable about or experienced with the phenomenon of interest. 137 In addition to knowledge and experience, participant availability and willingness to participate were key considerations. 138,139

Seven professional organisations (Table 2) identified a representative potentially interested in participating. Each representative was sent a participant information sheet (see Appendix 2), and those taking part signed a consent form. Interviews were conducted by a qualitative researcher (Adwoa Hughes-Morley; see Acknowledgements) and took place at a time and in a format convenient for the representative, be that via telephone, Skype™ (Microsoft Corporation, Redmond, WA, USA) or face to face at the representative’s office. Each representative was offered a consultancy fee of £200.

Ultimately, a consultative group comprising nine health-care professionals from seven different disciplines was established to provide feedback on the content of the ES course. This group included two consultant neurologists (one with a specialist interest in epilepsy), two consultants in emergency medicine, a consultant paramedic, a epilepsy nurse specialist, a GP with a specialist interest in epilepsy, a service commissioner with a health-care background and an educational representative from an epilepsy charity other than the ES.

Stage 3: recruitment of service user representatives

Feedback on the intervention was obtained from a purposive sample of service user representatives. PWE were eligible to participate in SAFE training if they met the following inclusion criteria:

• established diagnosis of epilepsy (≥ 1 year)

• currently prescribed antiepileptic drug(s)

• age ≥ 16 years (no upper age limit)

• visited an ED for epilepsy at least once in the past 2 years (as reported by the patient)

• living in the north-west area of England

• able to provide informed consent and participate in SAFE course in English.

People with epilepsy were excluded from participating in SAFE training if they reported:

• acute symptomatic seizures

• severe current psychiatric disorders

• life-threatening medical illness.

Patient participants could take part with or without a SO. SOs with the following characteristics were eligible to participate in SAFE training:

• a SO to the patient (e.g. family member, friend) who the patient identifies as providing informal support

• age ≥ 16 years (no upper age limit)

• living in the north-west area of England

• able to provide informed consent and participate in SAFE course in English.

Significant others were excluded from participating in SAFE training if they reported:

• severe current psychiatric disorders (e.g. acute psychosis)

• life-threatening medical illness.

The service user representatives were identified via adverts circulated (via newsletters and websites and at meetings) to the affiliates of user groups, including the Mersey Region Epilepsy Association (Liverpool, UK), the Brain & Spine Foundation (London, UK), NeuroSupport Services Ltd (Nottingham, UK) and the ES. This approach enabled the recruitment of ‘experts’, defined as informed individuals with knowledge or experience of a specific subject. 140,141 Eligible patients and SOs interested in taking part were sent a participant information sheet (see Appendix 3) and those taking part signed a consent form.

A total of 23 service user representatives were ultimately recruited, comprised 13 PWE (seven men and six women) and 10 SOs (four men and six women). Each received a £10 shopping voucher.

Intervention development panel members

The process was overseen by an intervention development panel. The panel considered and discussed the findings from the interviews and FGs and made required adaptations to the intervention. The panel included patient and SO representatives [Mike Perry (MP) and Linda Perry (LP)], a psychologist (AN), a neurologist (LR), a medical sociologist (MM), a research nurse with specialist qualitative research training (DS) and a representative from the ES’s training division. Patient and SO representatives were active in all aspects of the decision-making and were reimbursed in line with guidance. 142

Stage 1: consultation with health-care professional representatives

Consensus exists regarding what constitutes appropriate seizure first aid. 143–145 Thus, the purpose of this stage was to interview representatives from the main professional bodies caring for PWE to ascertain whether or not the medical information presented by the programme was correct and whether or not SAFE could be an intervention they could, in the future, support. It was considered important to seek feedback from professionals in the first instance to prevent PWE (and SOs) being exposed to possibly incorrect information. Moreover, it would allow us to identify from the start what sort of seizure first aid intervention was considered feasible for delivery in the context of the NHS. The health-care professional representatives each conducted a baseline document and audiovisual review of the course materials for the ES’s existing intervention.

In advance of their interviews, each representative was provided with the course materials. Approximately 2 weeks later, data from each of the nine representatives were collected via audio-recorded semistructured interviews. A topic guide was developed to reflect the intended purpose of the stage and on the basis of the literature (see Appendix 4). It was refined through the iterative process of the interviews. 146,147 The exact questions varied depending on the representative’s area of expertise, but all health-care professional interviews included key discussions on:

• identifying inaccuracies in the content of the existing ES programme

• likes/dislikes and the appropriateness of the current content and delivery

• suggestions about how to make the programme more helpful

• how SAFE might be best rolled out in the NHS if a future trial found it to be effective.

Stage 2: optimisation of the intervention’s behaviour change potential

A significant component of the ES’s intervention consisted of health-related information provision. It was anticipated that the provision of such information could reassure service users and increase seizure management confidence and competence. However, for some PWE the information might actually highlight that their prior use of ED conflicted with medical guidance. From a psychological perspective, this could be construed as a threat to self-integrity. As a consequence, these people might be at risk of rejecting or denigrating the information provided by the SAFE intervention, which may compromise its ability to elicit behaviour change. According to self-affirmation theory,148 people are fundamentally motivated to preserve a positive, moral and adaptive self-image and to maintain self-integrity. Consequently, health messages that threaten one’s sense of self-image can be subject to defensive processing (e.g. motivated scepticism, unrealistic optimism). 149

To mitigate against this and maximise the behaviour change potential of SAFE, the intervention development panel decided to introduce a self-affirmation exercise, namely Reed and Aspinwall’s150 kindness questionnaire, which individuals would complete at the start of SAFE training. The rationale was that evidence indicates that having a person complete an exercise such as recalling one’s acts of previous kindness prior to receipt of health risk messages reduces resistance to threatening or dissonant health risk information. This is because theory indicates that a person’s self-image can be maintained by self-affirming in one domain (e.g. recalling one’s acts of kindness) even if one is being threatened in another domain (e.g. health), because people can defend their global sense of self-worth rather than (for example) against the threat directly attributable to health risks. 151,152

The kindness questionnaire was considered ideal because it is brief (taking ≈ 5 minutes to complete) and effective150,153 and does not need to be delivered by specialists. It consists of 10 questions that participants work through by themselves, including ‘Have you ever been concerned with the happiness of another person?’ and ‘Have you ever forgiven another person when they have hurt you?’, with yes/no response options. The intention was that PWE and SOs would each complete it at the start of SAFE. Their questionnaires were not to be collected by any member of the intervention team and the participant would not be asked how they answered.

Stage 3: consultation with service user representatives

To allow service user representatives to critically feed back on SAFE, two practice courses were run in November 2015 using the intervention that resulted from the first two stages. These were followed by FG discussions.

The courses took place in a local hospital’s education centre and were delivered on weekdays to groups of ≈ 10 patient–SO dyads. A facilitator from the ES (Juliet Bransgrove; see Acknowledgements), who was an epilepsy nurse specialist with experience of delivering the ES’s course, delivered the courses. She underwent a period of familiarisation with the adapted intervention by reviewing the new materials and a trainers’ manual (see Report Supplementary Material 1) and meeting the intervention development panel.

The FGs, which lasted ≈ 60 minutes each, were conducted by a trained qualitative researcher (DS) to explore participant views. The researcher observed each course and recorded impressions of participants’ engagement with the materials, the group and the facilitator. A topic guide (see Appendix 5) reflecting the discrete sections of the course guided the FGs. Participants were asked about issues related to the course content and delivery as well as for views around its perceived strengths and barriers to its successful implementation.

Initial impressions

There was consensus on the need for such an intervention and its potential for cost-effectiveness. As one representative noted:

[I]t will be a powerful tool to upskill and reassure [. . .] The clearer they are about what constitutes that person’s normal epileptic fit and what maybe a bit unusual, the more [. . .] appropriate the response the better their [. . .] overall experience, privacy, dignity and everything to follow [. . .] [I]t’s a fairly easy case to make for the commissioners [. . .] Resources directed by commissioners to pay for this really unnecessary attendance could be redirected [. . .] win–win.

ED consultant

The existing intervention was seen to provide a useful starting point for adaption and professionals liked the videos and associated information booklets:

There’s a solid foundation in there which you can build on to then create the course to make sure it reaches the kind of outcomes that you are after [. . .] It’s just making sure that it hits all those objectives along the way.

User group representative

The practical challenges of hosting a group-based course were highlighted by many; it was thought that ‘getting people together may be a difficulty’ (epilepsy specialist nurse) that would require careful consideration, especially in relation to ‘distance and timing’ (epilepsy specialist nurse). It was important, therefore, to ‘think about local delivery and minimising travel time’ (epilepsy nurse specialist).

It was also felt that substantial changes were needed to make the intervention appropriate for its new audience and to achieve the aim of attenuating unnecessary/avoidable ED use.

Areas of intervention in need of revision

It was deemed important to better emphasise the benefits of the course to PWE and SOs with a more focused and ‘clear message’ at the start on the need or not for ED attendance after a seizure. Language level was highlighted as an area that required revision in line with the average UK literacy level. There were also suggestions, as presented below, to revise the style of presentation so that it was less for people who might be involved in epilepsy because of their profession:

[I]t’s written in quite a wordy way and probably a few more pictures and a few less words [. . .] would be better. This is probably a very good presentation to new health-care professionals [. . .] but probably not great for people that [. . .] have no training whatsoever.

Consultant paramedic

It was considered that a behavioural change focus (emergency medicine consultant 1) should be brought to the fore and that the benefits to the participant and SO of avoiding unnecessary ED visits should be emphasised. As explained by the GP representative:

It’s about helping people manage their epilepsy better. What will the course do for you? What can you get out of the course? [. . .] The course should focus on behavioural change, but the impression should not be given that the focus is about reducing A&E [accident and emergency department] admissions – that would be counterproductive. Rather, it is about highlighting the benefits, such as better management will reduce inconvenience in having to go to A&E.

GP

A recurring theme was the need to better elicit and address patient concerns and that patient and SO participation should be promoted through more interactive exercises, the inviting of questions and the discussing of fears, because this was when ‘true education happens’. This was summarised by one representative:

[S]eizure management, what you should do and what you wouldn’t do [. . .] I’d make that part interactive [. . .] get them to tell you, get their opinion and their views before and then you show them what they probably should do afterwards but find out. Because you’ll probably [. . .] challenge them [. . .] that’s the point where you really get them I think.

Consultant neurologist

A number of participants pointed out that, when seizures happen in public places, the decision to seek emergency care is not necessarily the patient’s or their SO’s. Therefore, it was recommended that the intervention should support patients to develop and carry with them personalised care plans on paper or on their smartphones, which could be used inform decision-making by paramedics:

[I]f we [the ambulance service] tip up and we don’t know anything about you we are taking you to hospital [. . .] We’re tipping up to what we consider to be a first-time fit until proven otherwise [. . .] Have a care plan written or something on you that says ‘I’m an epileptic and this is what happens to me normally’.

Consultant paramedic

Course delivery

Participants observed that, to make the course suitable for delivery in the NHS and to promote quality and consistency among trainers, the intervention should become fully standardised and a trainer’s manual, including recommended times for each topic/activity, developed.

With respect to attributes and skills of an ideal facilitator to deliver the course, some identified epilepsy nurse specialists. However, others felt that the epilepsy voluntary sector was well developed and, therefore, related commissioning organisations could help avoid shortfalls where specialist staff were not available. It was considered that the following should be true in any circumstance:

The facilitator is someone who should have knowledge of epilepsy and be good at leading groups. They need the ability to keep the course on track and to time, especially when participants may want to talk a lot!

Epilepsy nurse specialist

The need to know

All patient and SO participants identified the need for such a course, with lack of prior support in self-management as a recurring topic of discussion. For example, participants explained as follows:

It can be quite overwhelming I think for partners. It’s like ‘all on their shoulders’ what happens. I think your carer needs a lot of support too.

Patient 3, male (M), FG 1

The consultants they just presume that you know [about epilepsy] [. . .] but for all the years he got put on tablet or tablets [. . .] you didn’t see or hear any of this [information presented on the course]. So it’s suddenly all of an eye-opener and talking here you realise we are not on our own.

SO 1, female (F), FG 1

Similarly, another SO noted:

I always think of epilepsy as the poor relation [. . .] not much on TV or adverts about epilepsy support [. . .]. A course like this helps to develop that sense of support as well as improve knowledge.

SO 2, M, FG1

Concerns expressed by SOs centred on the ‘need to know I’m doing the right thing’ (SO 4, F, FG 1). PWE expressed concerns about disclosure and how best to tell others around them how they should help if a seizure happened; they wanted information and advice on how best to manage this. Overall, service user participants described three areas of perceived need: knowledge acquisition around epilepsy, emotional and/or practical support, and dealing with isolation and stigma.

Taken as a whole, the content of the revised course (version 1.1; see Table 3) was felt to be ‘excellent’ (patient 3, M, FG 2) and appropriate. Of particular importance to the users was the straightforward guidance that an ambulance was required when seizures lasted for 5 minutes or longer. This information alone was found to be helpful and reassuring, and some said that they would no longer always call immediately for an ambulance: ‘I think I will wait longer [to call an ambulance] than I did before picking up the phone’ (SO 3, F, FG 2); ‘I will wait [to call an ambulance] rather than when he has a seizure going for the phone straight away’ (SO 7, F, FG 2).

Concerns with regard to how to tell others about epilepsy and how to help if a seizure happens were expressed:

Like when you go somewhere you need to remember to like tell people that you have seizures.

Patient 1, M, FG 1

I need to know how best to share with others [family/friends/colleagues] the implications of having epilepsy.

Patient 2, F, FG 1

To this end there was consensus, for the most part, that the need to feel informed and reassured on what to do when seizures occurred had been met. Participants expressed how they had ‘learned a lot’ (SO 5, F, FG 1) from the session.

The balance between taught and interactive components was felt to be appropriate. The provision of information was viewed as reassuring and the opportunity to practise the recovery position was valued. As one patient asserted:

Watching a video would just go straight over me head, but actually putting [patient name] on the floor and putting him in the right position will help. For me that’s very useful [. . .] something like that I won’t forget.

Patient 5, F, FG 1

The training session was considered to cover more than implied by its title. Participants said that the ‘wider remit’ (patient 10, F, FG 2) was desirable but that a more accurate title was needed to engage future service users. ‘Managing seizures: epilepsy first aid training, information and support’ was identified as more suitable for the purposes of advertising.

Barriers to and drivers of effective participation in training

Service users’ perceptions of barriers to and drivers of successful training were explored. One was the self-affirmation kindness questionnaire. 149 Its positioning and purpose in the session were not understood by most participants: ‘[. . .] just coming into the session the questionnaire seemed inappropriate’ (SO 3, F, FG 1). It was also found by some to be threatening: ‘It felt like a test and a bit off-putting’ (SO 4, F, FG 1).

Some service users reported that another barrier was that there was ‘a lot of information to take in’ (patient 4, M, FG 1), and issues relating to memory difficulties were highlighted. Therefore, participants supported the use of handouts and requested an online copy of the materials that they could access and share with others.

With respect to content, important feedback from service users was that they appreciated that attention was given to the different types of seizures and how to manage them and that the focus was not simply on ‘grand mal seizures’ (SO 5, F, FG 1). However, they suggested that more time be given to exploring triggers and auras and to explain that not everyone has triggers, which in itself is a potential risk. It was also suggested that new sections should be included to discuss the risks associated with post-ictal states and how best to deal with them. Finally, some suggested that ‘dealing with an injury as well as dealing with the seizure can be difficult’ (SO 7, F, FG 2), and, therefore, information and advice on how to deal with common seizure injuries were needed.

Course delivery

The size of the group (up to 20 people) was considered to be appropriate and encouraged discussion. Indeed, peer support and a sense of feeling ‘less isolated’ was highlighted as a key training experience. Many patients had not previously discussed their epilepsy with people other than their immediate family members and/or health-care professionals. They valued the group format and described how they appreciated being able to meet others who were ‘in the same boat’ and ‘realising you are not on your own with it’ (patient 6, M, FG 1). One SO noted:

[R]eally glad I came. I think it’s just being around people who know exactly, exactly how you feel and that’s why this should have been put on a long, long time ago.

SO 1, F, FG 1

In terms of who would be best to facilitate the course, many felt that it should be a health-care professional because they believed that this would make the course ‘credible’ (patient 8, F, FG 2) and promote uptake. Others, however, argued that it could be facilitated by a representative from a user group because what was most important was that the trainer was knowledgeable and empathetic and had the skills to facilitate discussions. Either way, it was argued that standardised training for the facilitators was important.

Stage 1

There is no central, ‘live’ system that can be accessed and searched to identify PWE who have made visits to NHS EDs in England. Well-documented challenges to information sharing between NHS services also mean that specialist services and GPs are not necessarily aware of ED attendances made by PWE for whom they care. 20,21 However, hospitals in the UK do maintain local electronic records of attendances at their EDs. For each attendance, a local record contains the patient’s contact details and a number of searchable fields, including date of attendance, patient age, home postcode, presenting complaint and, if provided, a discharge diagnosis. These attendance systems of the EDs were used to identify PWE who had attended the EDs in the last 12 months for invitation to the trial.

Searches of the local systems for PWE were completed by the business intelligence units of the hospitals. We had envisaged that an independent expert panel of neurologists and ED clinicians would determine the search criteria to be used (e.g. presentation and diagnosis codes). During study set-up it became apparent that the architecture of the systems and the coding processes at each of the sites were sufficiently different that local knowledge of the system was required to have confidence that the criteria used would allow for a sufficiently sensitive and specific search to occur. Therefore, we instead worked with the local principal investigators at each site to identify the optimal search strategy. The search terms used are detailed in Appendix 6.

Stage 2

The ‘triage cards’ of the patients captured by the searches during stage 1 were reviewed by the local principal investigator and their team to identify persons who were and were not eligible to invite. This level of detailed screening was necessary because (1) no symptom presentation is pathognomonic of epilepsy (e.g. a presentation coded as ‘blackout/faint’ or ‘fit/seizure’ could be attributable to syncope, diabetes, head injury or stroke rather than epilepsy) and (2) the discharge diagnosis field was often empty. In 2016/17, ≈ 35% of ED attendances in England were not allocated a valid primary diagnosis code. 161

Those patients who, following review of their triage card, were considered ostensibly eligible by the local principal investigator were sent an invitation pack, which included a covering letter from the ED consultant and a patient participant information sheet (see Appendix 7). The letter informed the patient that, unless they opted out of further contact within 3 weeks or sent notification that they were ineligible, they would be telephoned by the research team with more information about the study. Patients could opt out by e-mail, telephone or by returning a Freepost slip. Those opting out were encouraged to detail any reasons. Those interested in taking part were asked to await contact by the research team and consider whether or not they would like to take part with a SO.

Stage 3

Interested patients were telephoned to answer questions the patient had, confirm whether or not the patient wanted to participate and verify their eligibility (including that they had made two or more ED visits for epilepsy in the last 12 months). If a person was confirmed as eligible and wanted to participate, consent in principle was taken over the telephone and the research worker arranged a baseline/enrolment appointment at which informed signed consent was to be obtained from them and from their SO if they were taking part with one.

Multiple telephone calls were attempted with patients who had not opted out. Attempts to call participants were made on different days of the week and at different times. A total of three calls per unresponsive patient were attempted, typically one in the morning, one in the afternoon and one in the evening. It was important to do this to minimise the influence of work status on ability to consider participation in the study. If a person could not be contacted (e.g. a correct number was unavailable or they did not answer), they were posted a letter asking them to contact the team if they were interested in participating. The sending of a letter was not in the original protocol; this was a substantial amendment and, with funder and governance approval, incorporated into protocol 1.2 (9 June 2016).

Derivation and statistical analysis

The eligibility rate is defined as the percentage of patients screened that satisfy eligibility criteria (see Chapter 3). The consent rate is defined as the percentage of eligible patients who provided informed, written consent for randomisation. The recruitment rate is defined as the number of participants recruited per calendar month.

Eligibility and consent rates are presented as percentages with 95% CIs. Recruitment rates are presented as the actual number of participants recruited per calendar month presented alongside the expected number of participants recruited per calendar month on recruitment graphs.

The age, sex and social deprivation profiles (available from ED records) of eligible individuals who did and did not consent to take part in the trial are presented side by side in table columns for visual comparison to evaluate representativeness of the trial sample.

Retention rate is defined as the percentage of randomised patient and SO participants completing 3, 6 and 12 months of the study without withdrawing (i.e. without formally withdrawing from any further data collection). Reasons for withdrawal, where known, are presented.

The completion rate of study assessment tools (i.e. the percentage of patient and SO participants completing each measure outlined in Table 4) is presented as an indicator of retention and participation in the trial.

The unblinding rate is defined as the proportion of correct treatment allocation guesses made by the research worker at the end of the trial (12 months) or at withdrawal of the participant from the trial. The unblinding rate is presented with 95% CIs. A Cohen’s kappa statistic for agreement (i.e. the correct guess of allocation) and 95% CIs are also presented.

Derivation of outcome

The proposed primary outcome of a future trial is defined as the number of epilepsy-related ED visits made by patient participants over the 12 months following randomisation measured by HES data. As a secondary outcome, the number of self-reported epilepsy-related ED visits made by patient participants over the 12 months following randomisation is also presented.

Statistical analysis

Epilepsy-related ED visits are presented at baseline (T0) and at 12 months (T3), and the number of ED visits made by the end of the 12-month period compared with the number of ED visits made in the 12 months prior to baseline is also presented. Results are presented for all patient participants and by treatment arm as mean and SDs, in addition to the median, the minimum and the maximum number of epilepsy-related ED visits. Completeness of data for self-reported epilepsy-related ED visits is presented and no imputation of missing data was performed.

The difference between the SAFE plus TAU arm and TAU only arm is compared at 12 months with and without adjustment for baseline ED visits via negative binomial regression (NBR) models for count data. Overdispersion (i.e. variance larger than the mean) and an excess number of zeros for individuals who did not visit an ED in the last 12 months was anticipated; therefore, zero-inflated NBR models were also applied. The preferred model (negative binomial or zero-inflated negative binomial) was determined by Vuong’s test. 180

Between-arm differences are presented as rate ratios with 95% CIs and statistically tested according to a 5% level of significance. In addition, as per guidance for pilot trials,181 90% and 80% CIs are also presented to aid interpretation of between-arm differences.

Epilepsy-related ED visits, measured by HES data, were compared with self-reported epilepsy-related ED visits and Bland–Altman plots and limits of agreement statistics182 were calculated to determine the agreement of the two measurement methods. A Bland–Altman plot shows the mean number of ED visits in self-reported and HES data for each patient participant (x-axis) and the difference in the number of ED visits in self-reported and HES data for each patient participant (y-axis). Bland–Altman limits of agreement (mean ± 2 SDs) of the difference in the number of ED visits are also shown on the plot. These limits can be interpreted as the level of agreement that is ‘acceptable’ between the two measurements of recording ED visits.

Estimation of sample size of a future definitive trial

The average annual rate of ED visits in the SAFE plus TAU and TAU only arms and the likely dispersion parameter estimated from the preferred NBR model are used to estimate the sample size of a future definitive trial according to the methods outlined by Keene et al. 183 for a negative binomial regression. The number of patient participants required per arm in a definitive trial to detect the size of the effect shown in the pilot study is:

where Z_{1 – α/2} and Z_1 – β are critical values of the normal distribution for specific values of α and β (typically, α = 0.05 for a 5% significance level and β = 0.2 or 0.1 for 80% or 90% power, respectively), µ₁ and µ₂ are the estimated ED rates from the two treatment arms and k is the negative binomial shape parameter from the associated gamma distribution, which explicitly represents variability between subjects.

Derivation of outcomes

Details of the secondary outcome measures are described in Table 4.

Total scores for Quality of Life in Epilepsy Scale-31 item-Patient-weighted (QOLIE-31-P) (PWE participants only), burden (SO participants only), confidence in managing seizures/epilepsy (SO participants only), mastery (PWE participants only) and fear of seizures (PWE and SO participants) are presented at baseline (T0), 6 months (T2) and 12 months (T3). The changes in total scores at 6 months (T2) and 12 months (T3) from baseline (T0) are presented. Burden categories (‘little or no burden’, ‘mild to moderate burden’, ‘moderate to severe burden’ and ‘severe burden’) are presented.

Total scores for stigma (PWE participants only) and distress (anxiety and depression scores; PWE and SO participants) are presented at baseline (T0) and at 12 months (T3). The changes in total scores at 12 months (T3) from baseline (T0) are also presented and stigma categories (‘no stigma’, ‘mildly to moderately stigmatised’ and ‘highly stigmatised’) and anxiety and depression categories (‘normal range’, ‘suggestive of anxiety/depression’ and ‘probable anxiety/depression’) are presented.

The ‘knowledge of what to do where faced with a seizure’ score is presented as the number of questions answered correctly (out of 13) at baseline (T0) and at 12 months (T3), in addition to the change in knowledge score at 12 months (T3) from baseline (T0).

Self-reported seizure frequency is presented at baseline (T0), at 6 months (T2) and at 12 months (T3) according to Thapar’s seizure frequency scale. 169 The total numbers of seizures between baseline (T0) and 6 months (T2) and between baseline (T0) and 12 months (T3) are also presented.

Statistical analysis

Total scores and the change from baseline scores of all measures, plus the total number of seizures, are presented as mean and SD in addition to the median, the minimum and the maximum scores for each scale. Burden, stigma, anxiety and depression categories are also presented as the number and percentage of participants in each category. Results are presented for all participants and by treatment arm and separately for PWE and SO participants where applicable.

Completeness of data for all self-reported scales is presented. Individual missing values in given scales of the QOLIE-31-P184 and the Hospital Anxiety and Depression Scale (HADS) scale were imputed according to recommended rules. 185 No imputation was performed for other scales. Total scores are presented separately for all individuals (including those with missing data) and for individuals with complete data for the scale.

Visual comparison of differences between treatment arms only are made for secondary outcome measures; no formal statistical testing is performed.

Search period

The stage 1 searches were originally set up to identify PWE who had made an ED visit between 1 January 2015 and 31 December 2015. However, recruitment proved more challenging than anticipated and, with funder and governance approval, the period in which people could be identified as having attended the EDs was extended by 7 months to 31 July 2016 (minor amendment, incorporated into protocol version 1.2 on 9 July 2016). A total of 417,381 attendances for any reason were recorded at the EDs between 1 January 2015 and 31 July 2016.

Eligibility

The stage 1 searches indicated that 6359 (1.5%) of the ED attendances were attributable to epilepsy or a suspected epileptic seizure. These visits had been made by 4016 individuals. The Consolidated Standards of Reporting Trials (CONSORT) flow diagram is presented (Figure 3). Following stage 2 review of the triage card associated with these visits, 1220 individuals were considered to have visited for established epilepsy, with 555 of them being considered eligible for participation in the trial and sent an invitation.

FIGURE 3.

The CONSORT flow diagram of eligibility screening for the trial.

Having received the letter, no patient opted out, but 122 patients did send notification that they were ineligible (Tables 5–7). For nine patients the postal address was incorrect or the letter was not sent, in error. The remaining 424 patients were telephoned to determine their interest in the study and verify eligibility (stage 3). Based on these figures, the eligibility rate (424 eligible participants out of 4016 unique individuals screened) was 10.6% (95% CI 9.6% to 11.5%).

It took the local principal investigator at each site ≈ 3 full days to complete the stage 2 screening.

Formal withdrawals from trial

Among the 51 randomised patient participants, only three (5.9%) withdrew consent to participate and for routine data to be collected on them over the course of their 12 months in the trial; two withdrew from the TAU condition and one from the SAFE arm. This meant that consent remained in place for securing routine outcome ED data for 48 (94.1%) patient participants. The reasons for participant withdrawal are described in Appendix 11. No withdrawals were initiated on the patient’s behalf by a health-care provider or a member of the research team. Among the 37 randomised SO participants, five (13.5%) withdrew.

Participation in questionnaire-based follow-up assessments

A total of 37 (72.5%) of the patient participants and 21 (56.8%) SOs attended their scheduled post-randomisation primary outcome assessment at 12 months (T3). The 37 patient participants attended their post-randomisation follow-up visit at 12 months at a median of 286 days (range 252–365 days) post randomisation. Appointments often took place before 12 months because slower recruitment meant that time for follow-up within the life of the project was reduced.

With respect to the interim questionnaire assessments, 43 (84.3%) patients returned the 3-month (T1) follow-up questionnaire and 39 (76.5%) patients and 26 (70.3%) SOs returned their 6-month (T2) follow-up questionnaire.

Primary outcome: emergency department data from Hospital Episode Statistics system

Secondary outcome data

The extent to which the secondary outcome measures were fully completed by the patient participants who participated in the follow-up assessments varied by assessment point, by measure and by participant type (Tables 9 and 10). Self-reported ED use at 12 months post randomisation (T3) by patient participants was the secondary means by which ED use was captured and so, arguably, the most important of the secondary outcome measures. Self-reported ED use data at 12 months (T3) were available for 34 (66.7%) out of the 51 randomised patient participants.

Patient participants

Hospital Episode Statistics emergency department data

Hospital Episode Statistics ED data for the 48 patients for whom consent were maintained over the trial show that they together made 122 ED visits in the 12 months before randomisation. Frequency of ED use among them was positively skewed (Figure 5). The median number of visits was 2 (range 0–12) (see Table 12).

FIGURE 5.

Histogram of the number of ED attendances in the previous 12 months by patients according to HES system.

As described in Chapter 3, patient identification centred on the use of NHS ED attendance records and only patients who, when telephoned, said that they had made two or more ED visits for epilepsy in the prior 12 months were recruited. Despite this process, data from the HES system indicated that four (8.3%) patient participants had not made any ED visits during the 12 months prior to recruitment and a further 19 (39.6%) were noted to have made only one ED visit.

Self-reported emergency department data

Self-reported data were available for 45 (88.2%) patient participants. The median number of visits was four (range 0–107) (see Table 12). Again, despite the screening processes, four (8.9%) patient participants responded on the questionnaire completed subsequent to telephone screening that they had not made any visits during the prior 12 months and 3 (6.7%) reported that they had made only one visit.

Relationship between Hospital Episode Statistics and self-reported data

There were 42 patient participants who had self-reported ED data at baseline (T0) and for whom consent remained in place for obtaining their HES ED data. Bland–Altman plots of the agreement between self-reported and HES data on ED visits were produced.

Primary focus is given to the plot shown in Figure 6. This is based on data from 41 rather than 42 patient participants. This is because the distribution of self-reported ED use was particularly influenced by one patient participant who self-reported 107 visits in the prior 12 months. We can confirm that that report of 107 visits was checked against the participant’s answer on the questionnaire and did not reflect a data entry error. Given that only one ED visit in the 12 months was recorded for this patient participant in HES data, we considered the self-report of 107 ED visits to be an outlier and excluded this in the plot shown in Figure 6 (Appendix 14 presents a Bland–Altman plot without any exclusions).

FIGURE 6.

Bland–Altman plot of agreement between self-reported and HES data on ED visits at baseline (n = 41; excludes outlier).

Figure 6 shows that disagreement existed between self-reported and HES data for most patient participants. Most (76.2%) participants self-reported more ED visits than were indicated by the HES system. Only three (7.1%) participants reported the same number of ED visits as were recorded by HES. Figure 6 shows that the limit of agreement was –5.0 visits (i.e. five fewer visits in self-reported data than in HES data) to 7.7 visits (i.e. 7.7 visits more in self-reported data than in HES data).

Psychosocial measures

Compared with a maximum possible score of 100, the mean QOLIE-31-P score was 48.3 (SD 17.3), with a large range (17.1–79.5). Twenty-six (50.9%) participants had ‘probable’ anxiety and 11 (21.6%) had ‘probable’ depression. Assessment of self-stigma revealed that most (n = 42; 82.3%) felt at least some stigma because of epilepsy and 15 (29.4%) felt highly stigmatised.

Comparability of treatment arms, including in emergency department use

Considering the small sample size, the process of randomisation was largely successful in generating two treatment arms broadly similar in demographics and scores on the baseline assessment tools (see Tables 11 and 12). Some differences were apparent in their ED use prior to randomisation (as measured by HES and self-report); ED use was slightly higher in the TAU arm than in the SAFE plus TAU arm (see Table 12).

Representativeness

The age, sex and social deprivation profile of the randomised patient participant sample was compared with that of the patients who were eligible to participate but who declined to participate or were not contactable. This showed that the two groups were comparable in age and social deprivation profile but that females were slightly over-represented in the recruited sample (Table 13).

Significant other participants

Most (72.5%) patient participants took part in the trial with a SO. SOs were typically a partner or spouse (43.2%), a parent (21.6%), or a son or daughter (16.2%). Most (75.5%) SOs lived in the same household as the patient and had contact with them every day of the week (89.2%). Their mean age was 43.2 years (SD 17 years, range 17–79 years) and most (59.5%) were female. Among the SOs, 32.4% reported having a medical condition themselves.

Anxiety levels among SOs were high, with 15 (40.5%) of the SOs having a score indicating ‘probable’ clinical anxiety. Only three (8.1%) had a score indicating ‘probable’ clinical depression. The mean Zarit caregiver burden score170 was 18.9 (SD 12.51), with most being categorised as reporting either little or no burden (n = 23; 62.2%) or mild to moderate burden (n = 11; 29.7%). Further characteristics of the SOs, including by treatment arm, are provided in Appendix 15.

Adherence scale: inter-rater reliability

For 96% of adherence items, the raters made the same judgement about the extent to which the item was delivered but the weighted kappa statistic was only 0.66 (95% CI 0.50 to 0.83). This was a consequence of prevalence bias (–0.83; bias index 0.06), with 94.6% of the raters’ ratings using the category ‘fully delivered’. Given this, the PABAK-OS statistic of 0.91 (95% CI 0.85 to 0.97) (which equates to ‘substantial agreement’) provided a more accurate reflection of rater concordance.

Didacticism measure: inter-rater reliability

With a coefficient of 0.96 (95% CI 0.78 to 0.99), the intraclass correlation coefficient indicated that the two raters’ judgements regarding didacticism were highly correlated.

Adherence

Adherence was high. Among the 259 items meant to be delivered across the seven courses, 228 (88.0%) were ‘fully delivered’. Only eight (3.1%) were ‘not delivered’ (Table 14). The average adherence rating (with a maximum of 2) given to the items across the SAFE courses was 1.88 (SD 0.11, range 1.65–1.97) (see Appendix 16).

The mean and range of adherence scores given to the individual intervention items showed that no item proved too challenging to be fully delivered at least once. The mean score of only one intervention item (namely, requiring the facilitator to inform the participants about when the demonstrated recovery position should and should not be delivered) fell below 1 (i.e. to 0.79).

Didacticism

The mean percentage of facilitator speech across the courses was 55% (SD 5.4%), with a range of 49–64% (see Table 14).

Proposed primary outcome: emergency department use measured by the Health Episode Statistics system

Routinely collected data on ED use from the HES system were secured for 48 (94.1%) of the randomised patient participants. The mean number of visits recorded over the 12 months following randomisation to the SAFE plus TAU arm was 1.8 (SD 3.14) and the median was 1 (range 0–12). For the TAU arm, the mean was 3.4 (SD 4.78) visits and the median was 2 (range 0–20) visits. Compared with the 12 months prior to randomisation, mean ED use over the follow-up period reduced for the SAFE plus TAU arm by 0.3 (28%) visits. For the TAU arm, it increased by 0.4 (11%) visits over the follow-up period (Table 15).

Emergency department use was overdispersed. However, there was not an excess of zeros (Vuong’s test:180 z = –0.17; p = 0.87). For this reason, NBR compared the effect of treatment arm on ED use, which showed that the estimated rate of ED visits following randomisation was around 46% lower in the SAFE plus TAU arm than in the TAU arm (rate ratio = 0.54). This difference was not statistically significant at the 5% alpha level (95% CI 0.24 to 1.18), nor at the 10% level (90% CI 0.28 to 1.04). The difference was statistically significant at the 20% level (80% CI 0.32 to 0.90; p = 0.12) (Table 16).

Owing to a slight imbalance between treatment arms in ED use in the 12 months prior to randomisation (i.e. ED visits at baseline), we also applied an adjusted NBR, including SAFE plus TAU arm and baseline ED visits. The results showed that an increased number of ED visits pre randomisation was associated with an increased number of post-randomisation ED attendances (rate ratio 1.33, 95% CI 1.18 to 1.52). With the adjustment, the estimated effect size for treatment arm reduced, with the rate of ED visits for the SAFE plus TAU arm being around 38% lower than that for the TAU arm (rate ratio 0.62). In this adjusted analysis, treatment arm was not significantly associated with ED use at the 5% (95% CI 0.33 to 1.17) or 10% (90% CI 0.36 to 1.06) level; the treatment arm was statistically significant at the 20% level (80% CI 0.41 to 0.94; p = 0.14).

Secondary outcome data

Relationship between Hospital Episode Statistics and self-reported data

Among the 34 patient participants who had self-reported ED data at 12 months (T3) and for whom consent remained in place and HES data were secured, 11 (32.4%) had the same number of ED visits recorded in HES and self-reported data: seven (41.2%) from the TAU arm and four (23.5%) from the SAFE plus TAU arm. A total of 13 patient participants (38.27%) had more ED visits recorded in HES data than in self-reported data and 10 (29.4%) self-reported more ED visits than were recorded by HES data; proportions were similar across the two treatment arms (Table 18).

TABLE 18 - Comparison of HES-recorded and self-reported ED visits during the 12 months post randomisation

HES-recorded vs. self-reported ED visits	Treatment arm, n (%)		Total, n (%)
	SAFE plus TAU	TAU
More HES recorded than self-reported	7 (41.2)	6 (35.3)	13 (38.2)
Equal	4 (23.5)	7 (41.2)	11 (32.4)
More self-reported than HES recorded	6 (35.3)	4 (23.5)	10 (29.4)

Figure 7 shows a Bland–Altman plot of the agreement between the mean number of self-reported and HES-recorded ED visits for the 34 patient participants; the limits of agreement were –5.3 visits (i.e. 5.3 fewer visits recorded in self-reported data than in HES data) to 4.4 visits (i.e. 4.4 visits more recorded in self-reported data than in HES data). Compared with baseline (see Figure 6), this Bland–Altman plot demonstrates a relatively even spread of self-reported versus HES-recorded ED visits.

FIGURE 7.

Bland–Altman plot of agreement between self-reported and HES-recorded ED visits at 12 months (n = 34).

Staff time and costs

Time was split into preparatory work (time in meetings, designing the SAFE intervention, administration of SAFE’s content, planning and organising practice sessions, attending practice sessions, reviewing and editing SAFE’s content) and programme delivery (time spent inviting participants, organising and attending SAFE sessions, packing up, and administration of participant expenses and the website). Staff recorded their time against each task for each SAFE session.

Travel time and costs

Staff involved in SAFE’s delivery recorded estimates of time spent travelling to sessions and any travel costs. Travel costs were recorded for participants and taken from the study finance records as the total claimed per group.

Other costs

The two academic staff central to adapting the intervention (AN and DS) recorded all resources and items required to complete the activities identified in stage 1. This included office space for staff involved in course preparation and administration, telephone calls, stationery, printing and postage, leaflets for patients/SOs, advertising, participant remuneration (practice sessions only), venue hire, refreshments, equipment, and website domain and administration costs.

Participant uptake

Recruitment of people with uncontrolled epilepsy into evaluative studies has previously been reported to be challenging. In the SMILE (UK) trial,208 PWE living in the London area who had experienced at least two seizures in the prior year were invited to a trial of self-management that had a 12-month follow-up period;212 37% of those invited took part. In a non-randomised trial of self-management for PWE who had attended ED on one or more occasions in the prior 12 months, the uptake rate was 26%. 128 Challenges can arise owing to the nature of the condition itself (e.g. unpredictable but frequent seizures) and its consequences (e.g. anxiety about travelling alone, inability to drive, memory impairment). A range of evidenced-based strategies were therefore used in the pilot RCT to try to maximise recruitment. These included participant vouchers, first-class postage for invitations, flexibility regarding when and where research appointments took place, the use of an ‘opt-out’ approach to contacting patients about the study, multiple contact attempts and all patient-facing documents being developed collaboratively with a patient and public involvement group. Despite this, a low recruitment rate was experienced.

The characteristics of the recruited patients offers one potential suggestion for the particularly low participant rate seen in the pilot RCT. Specifically, 82% of participants felt stigmatised by their epilepsy, 21.6% highly stigmatised. This is higher than among those with epilepsy generally, including those with uncontrolled seizures (63%). 212 Epilepsy, for reasons rooted deep in its history, has been called a ‘stigmatising condition par excellence,’213 and negative beliefs and lack of knowledge about the condition continue to be present. Consequently, PWE can feel ashamed of their diagnosis, guilty and reluctant to talk about it. 214 Those attending ED appear to be particularly at risk of this. This could be a barrier to participating in research. It is currently unclear how recruitment could be altered to mitigate the feeling of stigmatisation.

It is also worth considering what changes might be possible to who approaches patients regarding participation. Local ED clinicians invited PWE to participate in the pilot RCT. In trials it is usually preferable for a usual care provider with whom a patient has an established relationship to do the inviting. This was not considered ideal for the pilot RCT because national data indicated that most from the target population do not have a specialist usual care provider for their epilepsy (at least not one they have seen recently). Moreover, it was apparent that GPs were often not informed of ED visits made by PWE on their practice registers. Therefore, neither specialists nor GPs are positioned to reliably identify PWE for invite. Given the low participant uptake rate, a modification that a definitive trial could consider making would be for the EDs to identify ostensibly eligible patients from their records and then request that the identified patients’ GPs do the inviting. What impact this would have on uptake is not clear. It would require additional NHS support costs and additional approvals to permit EDs to communicate with general practices on what would be a research matter.

One of the leading reasons that patients gave for not wanting to take part was lack of interest. Some of this lack of interest might have arisen because of the time that had elapsed between the ED visit that led to their identification and them receiving the study invite (the median was 9 months). This could be addressed by shortening the time period between identification and invite, or, even more drastically, recruiting patients prospectively instead. The latter would have significant implications for the resources required by a definitive trial. In addition, the pragmatics of recruiting PWE directly from the ED requires clarification. For example, patients may often be in a post-ictal state and, furthermore, a stay in the ED is typically far shorter than a stay on a hospital ward.

We note that it is possible that the serial assessment of participants and the need for them to complete multiple secondary outcome questionnaires at the assessments might have been seen as overly burdensome by people considering the invitation to participate in the trial. It might also have negatively affected retention of the participants who were recruited. If a definitive trial committed itself to using routine ED data as the source of data for the primary outcome measure, that trial could theoretically reduce or even totally drop the use of the secondary assessment measures. This might increase uptake and improve retention. In considering how dramatic the improvements might be, it is perhaps worth noting that none of the participants who did take part in the trial and who completed the trial participation feedback questionnaire identified the questionnaire packs as being burdensome or as an area of the trial that required improvement.

Identifying eligible patients

Identifying eligible patients for the pilot RCT from ED attendance systems required substantial time and effort on behalf of the local principal investigators at the EDs. This was because of the nature of the condition, because the ED administrative systems did not code visits in a sufficiently granular way and because many of the data on patients that was needed to determine eligibility (e.g. presence of certain comorbidities) was recorded in an unstandardised way in non-searchable, free-text fields. Consequently, searches of the attendances systems were limited to simply identifying persons who had attended for a ‘suspected seizure’. The triage cards of these identified persons needed to be accessed and screened by local clinicians to find people suitable for invitation. This activity could not have been delegated to research staff, including research nurses, since it is not ethically permissible for a person outside the patient’s care team to access information on their medical history for the purposes of research without prior consent. 215

Since our pilot RCT took place, a new coding system for EDs – the Emergency Care Data Set216 – has been phased in. From the perspective of a definitive trial, it brings some advantages because it seeks to standardise and increase the granularity with which presentations and comorbidities are recorded. For example, a specific presentation code for seizure caused by a neurological condition now exists. 216 The system also asks for diagnosis to be recorded using detailed Systematised Nomenclature of Medicine (SNOMED) clinical terms; however, the fields ‘discharge diagnosis’ and ‘comorbidity’ in the Emergency Care Data Set are not mandatory. 217,218 Therefore, it remains to be seen how well completed these fields are before it is possible to conclude that the presence of the new system means that PWE can be more readily identified for a definitive trial, and demands on local sites reduced to a reasonable level.

Securing routine outcome data on emergency department use

For a definitive trial of SAFE to be well positioned to be funded and for it to be designed optimally, stakeholders would want to be confident that primary outcome data could be secured and a fairly precise estimate would be needed as to when the data would be available. Our pilot demonstrated that the provision of such data is not a given and that the timeline is hard to estimate. Bodies such as the NIHR have endorsed the use of routine data sources in clinical research. It is possible that, as routine data are requested more often for trials, NHS Digital’s application processes will be revised and become more suitable for the research environment. It is notable that our experience is not unique. The challenges in securing routine outcome data from NHS Digital were described in detail by Powell et al. 174 in 2017, and recommendations for improvements with the system were made.

Effect of the SAFE intervention

Efficacy was not the primary focus of our pilot RCT. No power calculation was completed and, with such a small sample, imbalance in pre-randomisation covariates between the treatment arms is possible. The estimated modest effect of SAFE on subsequent ED use – namely, reducing it from 2.1 visits over a 12-month period to 1.8 (compared with a slight increase in the TAU only arm from 3.0 to 3.4 visits) – was used to inform the sample size calculations for a definitive trial. Therefore, it is important to consider how realistic this estimate is because a larger effect would probably make a definitive trial cheaper (i.e. because fewer patient participants and thus recruitment sites would be required).

When the pilot RCT was being designed the empirical evidence on the ability of self-management interventions to change behaviour in PWE was positive but inconclusive. An updated systematic review by Luedke et al. 219 has since been published. It considered evidence from 13 randomised trials and two non-randomised trials of self-management interventions for epilepsy published before April 2018. The review concluded that interventions showed clinically important benefit on only a limited number of outcome measures. Interventions that primarily sought to increase knowledge changed self-management behaviours moderately, whereas interventions that included techniques such as cognitive behavioural therapy or problem-solving treatment improved quality of life only modestly. Overall, self-management interventions were not found to decrease seizure rates.

One RCT from the USA220 identified by Luedke et al. ’s review219 is particularly instructive because it is the only one to have also considered change in ED use. It focused on what its authors, Sajatovic et al. ,220 labelled an ‘at-risk’ population, defined as adults with established epilepsy who had had at least one negative health event within the past 6 months, be it a seizure, an accident, a self-harm attempt or an ED visit. The RCT compared the effects of an educational intervention, called ‘self-management for people with epilepsy and a history of negative health events’, to a wait list control. It comprised eight sessions, addressed myths surrounding epilepsy and introduced participants to key self-management tasks. A total of 120 patients were recruited and followed up for 6 months. The trial found no statistically significant differences between arms in terms of subsequent ED/hospitalisation use. For the intervention arm it reduced by 0.44 visits from 1.22 visits. For the wait list control arm it reduced by –1.26 visits from 2.4 visits.

The findings of the updated review suggest that the relatively small estimated effect of SAFE is in the region of what might be expected of a self-management intervention and thus the estimate from our pilot trial constitutes a reasonable basis for the sample size calculations for a definitive trial. That a self-management intervention reduces ED visits by a only small amount also makes sense from a clinical perspective because we now know that factors beyond low seizure confidence can influence whether or not an ED visit occurs. One is that studies with the ambulance service show that peripheral issues, such as performance targets and difficulties accessing information on patient medical history to determine normality or otherwise of the seizure presentation, can influence whether or not a person is conveyed to ED following an emergency call. 77,78

Progression, feasibility and optimal design of a definitive trial

Thabane et al. 132 provide a framework for judging whether or not to proceed to a definitive trial following a pilot. They describe the options as (1) ‘continue without modification’ (feasible as it is), (2) ‘continue without modifications, but monitor closely’ (feasible with close monitoring), (3) ‘continue, but modify the protocol’ (feasible with modifications) and (4) ‘stop’ (main study is not feasible). In satisfying only one of the pre-specified progression criteria, a definitive trial based on our pilot trial’s design is not feasible. On the basis of the discussion above, we do not consider that any options are currently available to modify the protocol in such a way for it to become feasible. Therefore, we recommend not proceeding to a definitive trial.