A study looking at Emergency Department attendances at NHS hospitals by people with epilepsy in the SAFE Trial

Epilepsy is the recurring tendency to have unprovoked seizures. With a prevalence of ~1%, epilepsy is the second most common serious neurological disorder in the UK. As well as having potentially important life implications for patients and families, epilepsy also has important societal impacts. One is the cost of providing emergency care. In the UK, 20% of people with epilepsy (PWE) visit hospital Accident and Emergency Departments (A&E’s) each year for seizures. In England alone, there are around 100,000 visits to A&Es each year. The cost of this in 2015/16 was ~£70 million.

One reason costs are so high is because half of the PWE visiting A&Es are admitted to hospital; indeed, 85% of admissions for epilepsy occur on such an unplanned basis. Readmissions further drive costs up; ≥60% of PWE re-attend A&E within 12 months. This rate of return is higher than seen for other long-term conditions with episodic relapse, like asthma, and diabetes.

Seeking emergency care for epilepsy can be appropriate, important, and even life-saving. Evidence from projects, such as the research team’s recent UK-wide National Audits of Seizure Management in Hospitals, now show though that most persons attending A&E do not attend for such reasons. Instead, most have known, rather than new epilepsy and present with non-emergency states which do not require the full facilities of an A&E. One of the reasons driving this use it that patients and their family members frequently lack the confidence and knowledge to manage seizures by themselves.

The research team, based at University of Liverpool, has developed seizure first aid training for this part of the epilepsy population and has recently completed a pilot randomised trial of it, called the Seizure First Aid Training for Epilepsy, SAFE trial. The SAFE trial focused on the 60% of this group (and their informal carers) who make multiple attendances in a year and together account for ~90% of all A&E visits made for epilepsy. The trial compared receipt of the intervention to usual care alone. The trial was completed with NHS ethical approval and HRA approval; was sponsored by the University of Liverpool and publicly registered (ISRCTN13871327), and was funded by the National Institute of Health Research (Health Services and Delivery Research programme (Project Reference No:14/19/09).

A pilot randomized trial is not designed with the aim to prove the superiority of one treatment over another, but rather to try out aspects of the larger trial and address design uncertainties that exist (Whitehead et al. Clin Trials 2014; 38: 130–133). The size of a pilot trial is rarely adequate to conduct statistical hypothesis tests as would be the case for the main definitive trial. Despite this, pilot trials have an important role in health care and benefit the health and social care system since they help us to understand how best to complete a full trial so that it can be well positioned to generate the scientifically rigorous evidence required to inform care and maximize patient outcomes. Moreover, pilot trials can help society avoid wasting finite resources on trials that are unfeasible or poorly designed. To this end, major funding bodies such as the UK’s National Institute for Health Research (NIHR) and the Medical Research Council expect pilot evidence on the feasibility of a trial before large amounts of money are released for a large trial to be completed.

A pilot trial was necessary as a range of uncertainties existed as to how to conduct a definitive trial. Uncertainties pertinent to this request for data from the HES A&E system were:

• The absence of an initial estimate for the sort of effect the seizure first aid intervention had on the proposed primary outcome for a definitive trial – namely participants subsequent use of A&E – and lack of an estimate of the annual rate of A&E use in the control arm and its dispersion. Without this information it is difficult to know what sort of sample size would be required for a definitive trial to ensure it was adequately powered to detect an effect if one existed. Going ahead to main trial without this information would have risked recruiting too few or too many participants. If too few were recruited, the probability of finding a clinically relevant difference would have been low and therefore, the chance of providing an inconclusive result high. Conversely, if too many participants were recruited then resources would be wasted, more patients than necessary could be given a treatment which will later be proven to be inferior; or an effective treatment may be delayed from being identified.

• The second uncertainty concerned how best to measure/ capture information on a person’s A&E use in a trial. For example, one could ask participants to self-report on their use, but this presumes all participants are well enough to answer the question and can provide an accurate answer. Memory impairment and mood disturbance are common in epilepsy and may impair recall. Given these uncertainties, and since the HES system provides the only comprehensive record of a person’s use of all NHS A&Es across the country, HES A&E data is required for the participants in the trial relating to the 12 months before and after they entered the trial. All participants in the trial provided explicit consent for their HES A&E data to be obtained. Having access to their HES A&E data would allow the above noted uncertainties to be addressed in the following ways:

• By enabling a description of the use of A&E by participants in the two treatment groups before and after entering the trial in order to generate an initial estimate of any change that occurred in A&E use in the two treatment groups and determine the annual rate of ED visits in the control group and its dispersion parameter. All this information could then be factored into a sample size calculation for a future definitive RCT.

• Participants in the trial were asked to self-report on their use of A&E during the 12 months before and after entering the trial. Having HES A&E data for these participants for the same periods of reference would enable a comparison of patients self-reported use of A&E against objective data on their A&E use. This comparison would allow measurement of the extent of agreement between the two measurement approaches and inform discussions about how best to measure A&E use within a future definitive trial, and indeed any other similar trials.

A multi-centre, external, pilot randomised controlled trial (RCT) was conducted with PWE aged ≥16 years who visited the A&E of one of three NHS hospital trusts in the NW of England (namely; Aintree University Hospital, Royal Liverpool University Hospital, Wirral University Teaching Hospital), in the prior 12 months for epilepsy on ≥2 occasions who could independently complete questionnaires in English, along with one of their family members or friends who have an informal caring role.

Ostensibly eligible patients were identified and invited to participate in the trial by their NHS A&E consultant who sent them an invitation letter in the post, along with a Participant Information Sheet. Persons who were interested in taking part were in turn contacted by a GCP-qualified, postdoctoral study researcher who confirmed patient eligibility, provided information and answered any questions the patient had. The researcher also provided the patient with a further copy of the Participant Information Sheet. Participants had a minimum of 24 hours to decide whether they wanted to take part or not.

As part of the consent process, 58 participants were recruited between May and December 2016. Participants provided informed written consent to participate and for the research team to access identifiable data from the HES A&E system on the number of times they had visited an NHS A&E in the 12 months prior to entry into the trial and then during the time period they were enrolled in the trial. Data on participants’ use of A&E before coming into the trial is required to permit adjustment for potential differences in baseline use of A&E between the two trial arms (i.e. those who took the training and those who did not) .

Participants taking part were put into one of two groups at random by a computer. The first group is called Group A and the second Group B. People who are put in Group A get the Seizure First Aid Training course (treatment) straightaway and people in Group B continue to receive their normal medical care (treatment as usual (TAU). The health of the people in the two groups will be compared to see if the Seizure First Aid Training was helpful or not. After the two groups’ health has been compared, people in Group B then get to go on a Seizure First Aid Training course if they want it.

Over the course of the trial, patient participants were followed-up and required to each complete three sets of questionnaires, either in a face-to-face interview with a research worker (at baseline and at 12-month follow-up) or through the post (at 6-month follow-up). At these assessment points, participants were asked to self-report on the number of times they had visited any NHS A&E. At baseline they reported upon A&E use in the previous 12 months. At follow-up they reported on A&E use since their prior assessment.

Prior to each assessment point, participants were contacted and asked whether they wanted to continue to participate in the trial or whether they wanted to withdraw their consent. During the course of the trial 5 patient participants formally withdrew and so HES A&E data will not be requested for them.

The size of the project’s sample size does not impact on the ability of the project to achieve its aims. Sample sizes between 24 and 50 have been recommended as ‘adequate’ for pilot trials (e.g., Sim & Lewis, J Clin Epidemiol 2012 65: 301-8; Julious, Pharml Stat 2005 4: 287-91).

Data management and analysis is to be conducted within the Clinical Trials Research Centre (CTRC) at the University of Liverpool. The specific methodological activities involved in the processing of data are as follows:

The NHS Digital HES A&E data will be requested for all trial participants who provided consent and who have not subsequently formally withdrawn from the trial. A request for data will be made for all applicable participants on one occasion only.

The time period for getting data on participants’ use of A&E relates to the 12 months prior to them entering the trial and the 12 months following their enrolment. The research team will send NHS number, the unique (pseudonymised) study ID, and the date of recruitment to the study to NHS Digital.

Within the data file sent back to the University of Liverpool from NHS Digital, the University of Liverpool need to know the study ID associated with each individual’s A&E visit that occurred within the time period of interest by the University of Liverpool’s list of patients. This is necessary so data from the HES A&E system can be allocated to the correct participants in the trial and to enable a comparison of the use of A&E of participants in the two trial arms. In order to marry the detail provided from participants the date of A&E visit is required from NHS Digital to ensure accurate matching.

The resulting file generated by NHS Digital would include any A&E visits by patients who participated in the trial captured by the HES system that occurred within the relevant time. The data provided would include the date of the visit. The data file would be securely transferred back to the CTRC at the University of Liverpool, again using NHS Digital’s Secure Electronic File Transfer SEFT system.

Analyses will not be completed using the identifiable data set and to ensure as few can access this file a structured process, used previously by the research group when using HES data will be followed. Specifically, having received the data file from NHS Digital the trial’s postdoctoral research fellow will, within the confines of CTRC, work with the identifiable dataset to create a pseudonymised version. To do this, the research fellow will attribute all A&E visits captured by the HES system to the appropriate participants in the trial. The resulting file will only contain participants Unique Study Number and the number of A&E visits that they made during the relevant time periods. The patients NHS number will not be included in this file. The data will then be linked, using the patients Unique Study Number, with the main trial database and the self-report data provided by participants in the trial. Following this process, the data set will then be accessible to the study team members involved in the analysis, including the Clinical Investigator and the trial statistician.

The primary outcome by which treatment effect will be estimated is defined as the number of epilepsy-related A&E visits made by patient participants over the 12 months following randomisation measured by HES data. The number of A&E visits at the end of the 12-month follow-up period will be presented, in addition to the change in the number of A&E visits at the end of the 12 months compared to the number of A&E visits in the 12 months prior to baseline. Results will be presented as mean and standard deviations if data are normally distributed and median, IQR and range if data are skewed. Results will be presented overall and by treatment group.

The difference between the treatment group compared to the TAU control group will be expressed as a mean difference and 95% confidence interval and statistically tested according to a 5% level of significance by an independent t-test if data is normally distributed. In addition to aid interpretation, 90% and 80% confidence intervals of the mean difference will be reported. The difference between the groups will be tested with a Mann-Whitney U test if the data are skewed.

To maintain the original blinding of the trial statistician during the SAFE trial, anonymised HES data without any details of intervention will first be analysed and overall results presented. Subsequently, intervention allocations of individuals within the HES data will be made available to the trial statistician and results will be presented by treatment group and statistical testing will be performed.

For the secondary objective, self-reported numerical results will be compared to those of the results of epilepsy-related A&E visits calculated from HES data. If possible, Bland-Altman agreement statistics will also be calculated to determine the agreement of the two measurement methods of recording A&E visits. The term “if possible” is used not because the sample size might preclude completion of these analyse, but rather to account for the fact that without having seen the HES A&E data and there was the possibility, albeit unlikely, that the researchers could not calculate these statistics because the HES data received was is in a different format to the self-reported data and so the two cannot be directly compared.

All data received will be stored using the University of Liverpool Research Data Management Datastore (https://www.liverpool.ac.uk/csd/records-management/storage-and-disposal/). Data will be stored electronically on University of Liverpool central servers, located in an access-controlled server room and connected to the main University network, located behind a firewall. Physical access is limited to Computer Services Department staff. Data will be encrypted using industry standard techniques meeting the Information Governance Toolkit standard (8HN20). The data will not be transferred to an additional location. The SAFE trial CI will act as data custodian (https://www.liverpool.ac.uk/library/research-data-management/storing-your-research-data/). The University of Liverpool ‘Information Security Policy’ and ‘Research Data Management Policy’ provide further information.

The pseudonymised dataset will be accessed by specific members of the SAFE trial research team based in the University of Liverpool. All outputs will contain only aggregate with small numbers suppressed in line with the HES analysis guide.

All data will be stored and accessed at the University of Liverpool at all times. All personal data in this trial is kept strictly confidential and is being handled, stored and destroyed in accordance with GDPR.

Only individuals substantively employed by the University of Liverpool and are part of this study will have access to the data. No other collaborator of this study will have access to the data received from NHS Digital.

All organisations party to this Agreement must comply with the Data Sharing Framework Contract requirements, including those regarding the use (and purposes of that use) by “Personnel” (as defined within the Data Sharing Framework Contract - i.e. employees, agents and contractors of the Data Recipient who may have access to that data).

The data provided by NHS Digital will directly inform the outputs of the NIHR (HS&DR) funded project and allow the research team to achieve several of the key objectives.

First and foremost, the data will allow an estimate of the effect of the training intervention. This will help understanding around whether the seizure first aid intervention developed is likely beneficial in reducing ED use.

Secondly, the data provided by NHS Digital will enable understanding of how data on A&E use captured by the HES system compares to patient self-report. The results from these analyses will provide knowledge on how to conduct the definitive trial if it is deemed appropriate.

Findings will be published in the form of a publicly available report to the NIHR and within a peer-reviewed publication. In no publication will the identify of participants be identified.

Final results from this trial and the associated outputs are expected by December 2018 in line with the project’s completion date. All presented and published findings will be anonymised and compliant with NHS Digital's operating procedures in relation to presentation and publication. Non-identifiable aggregate data will be used in presentations and publications. Outputs will consist of descriptive statistics and statistical measures of agreement between data retrieved from the HES A&E system and participants’ self-report.

The HES Analysis Guide rules will be complied with. Only the mean/ median number of A&E visits by participants in the two groups and the dispersion parameters shall be described. The change in A&E use between the two groups will be reported and compared. In no instance shall data where cell counts are less than 5 as specified in section 5.1 of the HES Analysis Guide be published. Moreover, the maximum number of visits by an individual shall not be reported since this is information that can relate to an individual.

The primary focus here is to ascertain the agreement and additional benefits of data from the HES A&E data set and not to report on specific clinical criteria. It is not the intention to present record level data in any report. In all outputs data will be de-identified and all measures taken to ensure that individuals cannot be identified. For example, information regarding geographic location, timing and gender will be omitted as well as explicit clinical / personal details. Furthermore, both epilepsy and the clinical event of A&E attendance(s) under assessment are very common.

This project will have both direct and indirect outputs.

The pilot trial will include the analysis of individual participants’ use of A&E at baseline and over the 12 months of follow-up. This will assist in determining whether incorporating A&E attendance from electronic medical records in place of patient self-report records provides a more rigorous data set. This information will be included in the analysis on completion of the pilot trial. This output will take the form of reports and presentations.

This project’s findings will also inform the wider epilepsy research community contributing to the development and improvement of efficient future trial design. In particular, information from HES data will provide evidence on how accurate people with epilepsy are at self-reporting on previous emergency department use.

Using HES data from the electronic system to provide the primary outcome data would extend the timescale of a future trial and increase costs. At present, no evidence exists therefore, to be able to inform a future trial about how best to measure A&E use, the coverage and accuracy of patient self-report will be compared to data from the HES system. This will help determine whether the expense associated with use of the HES system as the primary means of measuring A&E use is warranted.

These outputs will be disseminated to clinicians and academics involved in the conduct of clinical trials and research, concerning clinical trials methodology, within the epilepsy population. Dissemination of findings will be presented at academic conferences (potentially; 13th European Congress on Epileptology http://epilepsyvienna2018.org/scientific-programme/) and in peer-reviewed journals (potentially; Journal of Neurology, BMJ Open, Epilepsy and Behavior) and will take the form of a narrative assessment of:

- the methods and feasibility of access to electronic medical records

- the agreement and reliability of data from routine sources

- the benefit / limitations of data from electronic medical records

Whilst the trial will not be statistically powered to detect a clinically meaningful difference in outcome between treatment groups, summary statistics will be conducted to measure the effect of the intervention on the proposed primary and secondary outcome measures (outlined previously) and the precision of such estimates at the post-treatment time points. As such, it will directly inform the methodology employed in the data collection and analysis of a possible future definitive trial. This benefits health and social care by informing the methodology to be employed in data collection and analysis of a definitive trial, therefore strengthening the evidence base which underpins the data and the results. This output will be measurable based on the methods subsequently employed in the definitive trial.

Indirect benefits to health and social care will be achieved through output via presentation and publication to the research community involved in clinical trials and trial methodology. The output of this trial aims to inform the implementation of data from electronic HES records in a prospective definitive trial. Resultantly, RCT’s will use electronic medical records where a benefit is offered over self-report methods of data collection. This will result in improved efficiency of RCT’s, frequently funded through public sources and improved participant experience.

Benefits include having significant implications for the lives of patients and reducing unnecessary emergency admissions is a key factor in helping to relieve financial pressure on healthcare services.

Another major social issue is the indirect cost of epilepsy due to lost employment. The health and social costs could be reduced, and quality of life improved via better outpatient management. However, around 40% of those diagnosed have poorly-controlled epilepsy and continue to have two or more seizures per year, despite antiepileptic drug treatment. These findings highlight missed opportunities for epilepsy self-management. Guidelines are clear that, with the correct training, such seizures can be safely managed by patients and their families within the community. Evidence indicates that people with epilepsy that frequently visit the A&E might benefit from a self-management intervention that improves their own and their informal carers’ confidence and ability in managing seizures and empowers them to be able to tell others from their wider support network about first aid.

It follows therefore, that indirectly the assessment of data from the electronic HES records in a definitive randomised controlled trial to evaluate the effectiveness of seizure first aid training intervention for people with epilepsy will provide the best possible information in relation to A&E attendance by people following a seizure. This data will subsequently be used to inform service commissioning decisions. Commissioners planning health and social care services need good information about the experience of people with epilepsy and the intervention(s) they receive, as well as the result of that intervention as a means to ensuring that people are getting the services that are right for them. Reducing unnecessary emergency visits to hospital by people with epilepsy is identified as one way that resource limited health services can generate savings. In addition, reducing emergency visits is also important for service users; not least because emergency department visits can be inconvenient, distressing and do not typically lead to extra support.

Yielded Benefits is not a requirement for new applications.