STRENGTHS AND LIMITATIONS OF THIS STUDY

We use methodology for specifically estimating paediatric hospital catchment populations to calculate rates of unplanned emergency admissions for epilepsy at Trust level, which is a strength of this study.

Hospital Episode Statistics, which contains high-quality data on most of the hospital care for children and young people in England, was used and this is a strength for this study.

This study is the first to use Epilepsy12 audit data to investigate whether wide variation in epilepsy admissions among children and young people can be explained by variation in new diagnosis of epilepsy at the Trust level, after adjusting for age, sex and deprivation.

Catchment populations in this study were calculated based on data on total emergency admissions for children and young people; therefore, community Trusts that did not have emergency departments were not included in our new epilepsy diagnoses analysis, which is a limitation.

Introduction

Epilepsy is estimated to affect over 112 000 children and young people (CYP, 0–24 years) in the UK and is one of the most significant neurological conditions affecting CYP.1 Quality and coordination of healthcare are important determinants of outcomes for CYP with epilepsy; failure to provide consistently high-quality care for CYP with epilepsies has been linked to high rates of overdiagnosis and underdiagnosis of epilepsy and wide geographical variation in epilepsy admission rates and deaths.2 In the UK, some epilepsy services are world leading, but the quality of care is geographically variable, and patients in many areas do not have access to optimal monitoring and treatment.3 Geographical variation in paediatric epilepsy admissions in England is significant even after adjusting for factors such as deprivation and ethnicity, and the causes for variation remain largely unknown.4 5 Factors that are likely to be important include incidence of epilepsy, prevalence, quality of epilepsy management and access to epilepsy services, all of which require further investigation.1 4 There is a lack of evidence in the literature on the causes of geographic variation of unplanned hospital admissions for epilepsy among children and young people, and no studies were identified that have previously looked at the relationship between new diagnosis of epilepsy rates and unplanned hospital admissions for epilepsy.

We hypothesised that the incidence of new epilepsy diagnoses can significantly impact the variability in hospital admissions for paediatric epilepsy cases. An upsurge in newly diagnosed epilepsy cases may result in an increased demand for emergency hospitalisation among affected CYP, particularly as there is an increased likelihood of children having further seizures during the first year since diagnosis, when treatments are typically being adjusted.4 Consequently, regions or Trusts with higher rates of new epilepsy diagnoses may observe a higher frequency of unplanned emergency admissions for paediatric epilepsy. Conversely, areas with lower incidence may tend to report fewer emergency admissions for paediatric epilepsy.

In addition, the diagnosis rate of epilepsy could be an indicator of availability of paediatric neurology expertise at Trust level. If epilepsy is consistently underdiagnosed in a particular Trust, it may indicate that healthcare professionals are not adequately identifying and evaluating individuals with epilepsy symptoms.6 This could be due to lack of awareness, limited access to diagnostic tools, or misinterpretation of symptoms.6 7 In such cases, the quality of care may be suboptimal because some individuals who need treatment are not receiving it. On the other hand, overdiagnosis of epilepsy among children and young people that have syncope or psychogenic attacks has been reported in previous studies.7 8 In a study by Smith et al, the frequency and causes of an erroneous diagnosis of epilepsy were investigated for children referred with ‘refractory epilepsy’.7 An overall misdiagnosis rate of 26.1% is reported in this study with incomplete history taking and misinterpretation of investigations such as EEGs being largely responsible, which indicates a lack of clinical expertise.6 Overdiagnosis of epilepsy can lead to unnecessary treatment and potential harm to patients.

Epilepsy12 (the National Clinical Audit of Seizures and Epilepsies for Children and Young People) uses data collected from hospital and community services to support further quality improvement in paediatric epilepsy services.9 The recent Epilepsy12 report highlighted the presence of local and regional variation in socioeconomic distributions for epilepsy incidence.9 A key recommendation from the audit was that Epilepsy12 should develop an ascertainment measure based on estimated population sizes, sociodemographic factors and estimates of epilepsy incidence by National Health Service (NHS) Trust.9 This would allow data on variation in new diagnoses of epilepsy to be used to interpret the data on variation in epilepsy admissions and other measures of healthcare utilisation, and thus, help inform appropriate national and local responses.

This study therefore aimed to investigate the relationship between emergency admissions for epilepsy and new epilepsy diagnoses at Trust level. We aimed to investigate unwarranted geographical variations in emergency admissions for epilepsy and in new diagnoses of epilepsy to identify areas that have higher/lower expected epilepsy admission and new diagnosis rates than expected, which can be targeted for further quality improvement.

MethodsAim

To investigate the relationship between emergency admissions to hospital for epilepsy among CYP in England and new diagnoses of epilepsy at Trust level.

Objectives

To use local catchment populations for CYP to calculate expected rates of new epilepsy diagnoses and epilepsy-related admissions for each Trust.

Compare expected rates with observed rates of new epilepsy diagnoses and epilepsy-related admissions at Trust level.

Investigate the relationship between the number of new diagnoses of epilepsy and emergency admissions for epilepsy at Trust level.

Identify hospital Trusts with higher and lower than expected rates of epilepsy admissions and diagnosis and identify areas to target further work.

MethodsCalculating hospital catchment populations

Hospital Episode Statistics Admitted Patient Care (HES-APC) data were accessed for all emergency inpatient admissions for CYP from birth to 18 years for 365 days from 1 April 2018 until 30 March 2019. For each admission, we analysed:

Geographical data on the admitting NHS Trust and the lower layer super output area (LSOA).10 These are the smallest geographical unit for which Census data are available and can be derived from the patient’s postcode.

Sociodemographic data on sex and age. Adjustment for these variables is needed as inpatient activity during childhood is known to vary markedly with age and between males and females.11

ICD-10 codes of the primary diagnosis and any secondary diagnoses.

HES-APC data were then linked to the CYP population within each LSOA, disaggregated by single year of age and sex (based on ONS 2012 mid-year estimates by age, sex). Deprivation quintile was obtained from HES-APC data for the Index of Multiple Deprivation (IMD).

Hospital catchment populations were estimated using a proportionate flow method described by Arora et al.5

Derivation involved four steps as described below:

HES data were used to count the number of patients in each age, sex and deprivation quintile group admitted from each LSOA to (1) any provider (NHS Trust) and (2) each individual provider. HESIDs (unique patient identifiers within the HES data) were used to ensure that each patient was only counted once; if a patient had multiple admissions, we selected only the first admission.

Within each age (0–3, 4–7, 8–11, 12–15, 16–18 years) group and sex group, we calculated the proportion of patients from each LSOA that was admitted by each provider out of the total of patients who used any provider.

For band i (age, sex and deprivation quintile) and LSOA j , the number of patients admitted to provider a is denoted by . The total number of patients admitted from LSOA j in band i across all providers b is given by:

The proportion of patients in band i from LSOA j admitted to provider a is then calculated as:

Where represents the proportion of patients that went to provider a .

3. For each LSOA j , this proportion was multiplied by the LSOA resident population in the specific age-sex-deprivation group to give the LSOA catchment population for each provider. The catchment population for provider a is calculated as:

4. The provider-specific catchment populations for each LSOA were then summed across all LSOAs to give the total catchment population for provider a in band i (age, sex and deprivation). Thus, the total catchment population for provider a and band i is: C for provider a, band i (age and sex):

Use of hospital catchment populations to investigate variation in epilepsy admission rates for CYP across England

HES-APC data for emergency admissions between April 2018 and March 2019 were accessed to measure the total number of emergency admissions for epilepsy for CYP aged 18 and under and to calculate Trust level admission rates per 100 000 CYP for admissions with a primary diagnosis of epilepsy. ICD-10 codes G40 and G41 were used to identify admissions related to a diagnosis of epilepsy.

Indirectly standardised ratios for observed to expected emergency epilepsy admissions, for each Trust, were calculated using the following steps:

Calculate the national emergency admission rate for epilepsy between April 2018 and March 2019 for CYP.

The adjusted expected admission rates for each Trust were calculated by multiplying the national epilepsy admission rate per thousand CYP by the total catchment population for that Trust within each age-sex-IMD group.

The expected total number of admissions for each Trust was the sum of expected admissions within each age-sex-IMD group.

Divide the observed numbers of emergency epilepsy admissions for each Trust by the age-sex-IMD adjusted expected admissions to obtain standardised ratio of emergency admissions.

We then derived the ratio of observed to expected epilepsy admissions, highlighting those Trusts with admission rates that differed by more than two SD from the expected rate for the population served. All analyses were performed using SAS statistical software.

The standardised ratio for epilepsy admissions compares the observed number of admissions at each provider to the expected number, adjusted for the population’s characteristics. The expected number is calculated by applying a standard rate (the rate across all providers nationally) to the provider’s catchment population. The standardised ratio is then the observed admissions divided by the expected admissions.

Use of hospital catchment populations to investigate variation in new epilepsy diagnoses for CYP across England

Unit-level Epilepsy12, round 3, cohort data (July–November 2018) were accessed to investigate geographical variation in new epilepsy diagnoses. Epilepsy12 audit data was used to calculate observed vs expected new diagnosis of epilepsy for each Trust. Expected rates of a new diagnosis of epilepsy were calculated based on the total number of new diagnoses of epilepsy nationally, which was obtained from the Epilepsy12 audit, and adjusted for each Trust based on the catchment population they serve. We adapted a similar approach and derived the ratio of observed to expected new epilepsy diagnoses, highlighting those Trusts with new diagnosis rates that differed by more than 2 SD from the expected rate for the population served. Only epilepsy services linked to an acute trust could be included in the analysis as hospital catchment populations were based on emergency attendances to each Trust by CYP.

Investigating the relationship between epilepsy admissions and new epilepsy diagnosis at trust level

Epilepsy12 is an audit of new diagnoses of epilepsy and the audit is carried out in set cohorts. We chose the July–November 2018 cohort for new diagnoses. Unplanned admissions for epilepsy by Trust are usually analysed over a 12-month period April–March. This allows comparison with other studies and reduces the impact of seasonal variation on comparison.12 HES data for emergency admissions between April 2018 and March 2019 were therefore analysed in this study, as this was the closest match to answer our research question. From a clinical perspective, we believe that it is reasonable to use admissions 3 months before the diagnosis period for two reasons: first, a diagnosis of epilepsy is typically made after two or more seizures; it is not uncommon to have another hospital admission in the months before the second seizure,13 which triggers investigations and leads to a diagnosis. Second, there is typically a delay of weeks/months between the second seizure and formal diagnosis, while investigations such as EEG and MRI are performed, and the results reviewed.

The relationship between emergency epilepsy admissions and new diagnoses at Trust level was investigated by plotting standardised ratios of observed to expected epilepsy admissions against standardised ratios for observed to expected new diagnosis of epilepsy in a scatter plot. The Pearson correlation coefficient was then calculated.

Missing data

There was a low level of missing postcode LSOA data (<1%) for Trusts, and as we collected data spanning 1 year, we did not have the issue of people moving during our study. There was a low level of missing data on unplanned emergency admissions for epilepsy. We only excluded patients with missing or invalid age (<0.1% of case) when calculating rates of unplanned hospital admissions for epilepsy using HES-APC data. There was a higher level of missing data, where Trusts had not submitted data to the Epilepsy12 audit. Only 74 out of 134 Trusts have submitted new diagnosis data to the Epilepsy12 July–November 2018 audit. The new diagnosis of epilepsy rates was calculated only on Trusts where data on new diagnosis of epilepsy has been submitted.

ResultsEpilepsy admissions

There were 937 520 emergency admissions accessed for CYP under the age of 19 years who lived in England. There were 9246 emergency admissions for CYP with a primary diagnosis of epilepsy, and there were 22 386 emergency admissions with a diagnosis of epilepsy in any field, in England between April 2018 and March 2019. In total, there were 134 NHS Trusts in England with emergency admissions for epilepsy.

Online supplemental figures 1 a and b represent the emergency admissions for epilepsy for CYP in England between April 2018 and March 2019. Those living in the most deprived quintiles had 1.82 times greater number of emergency admissions for epilepsy and had 1.96 times greater number of emergency admissions for any diagnosis of epilepsy, when compared with those living in the least deprived areas.

Online supplemental table 1 presents admission rates by Trust for a primary diagnosis of epilepsy, any diagnosis of epilepsy and all-cause admissions per 100 000 CYP between April 2018 and March 2019. Emergency admission rates with a primary diagnosis of epilepsy ranged from 12 to 171 per 100 000 (14.3-fold variation). The median value was 69 per 100 000, with an IQR of 56 to 91. Excluding the five Trusts with the highest and lowest admission rates, which is a common form of sensitivity analysis,14 the range reduced to 35 to 120 per 100 000 per year (3.4-fold variation).

All-cause admission rates in Trusts providing acute paediatric services ranged from 4050 to 12 171 per 100 000 per year (threefold variation), with a median of 7468 and an IQR from to 5983 to 8878 per 100 000. Excluding the highest and lowest five Trusts, the range was from 4306 to 10 842 per 100 000 per year (2.5-fold variation).

Figure 1 is a funnel plot of adjusted emergency admission ratios for a primary diagnosis of epilepsy by Trust. 31 hospital Trusts (23%) had admission ratios for epilepsy ≤2 SDs than expected. The emergency epilepsy admission rate for Trusts that had lower than expected admission ratios ranged from 12 to 59 per 100 000 population. 29 hospital Trusts (21.6%) had admission ratios for epilepsy ≥2 SDs than expected. The emergency epilepsy admission rate for Trusts that had higher than expected admission ratios ranged from 90 to 171 per 100 000 population.

Figure 1

Funnel plot showing the standardised ratio (observed/expected) for paediatric epilepsy admissions in English Hospital Trusts between April 2018 and March 2019.

New Epilepsy Diagnoses Results

There were 960 new diagnoses of epilepsy audited between July and November 2018 for CYP in England, for 74 Trusts where the catchment population could be calculated and who submitted clinical data to the Epilepsy12 audit. Table 1 presents new epilepsy diagnoses rates by Trust between July and November 2018. The median new diagnosis rate of epilepsy was 13.6 per 100 000 with an IQR from 10.9 to 16.6.

Table 1

Number of new epilepsy diagnoses and epilepsy diagnoses rates per 100 000 by Trust in England from Epilepsy12 audit data, July–November 2018

Figure 2 is a funnel plot of standardised ratios for observed vs expected new diagnoses of epilepsy by Trust. 8 (12.2%) hospital Trusts had new diagnosis ratios for epilepsy ≤2 SDs than expected. The diagnosis rate for Trusts that had lower than expected diagnosis ratios ranged from 3 to 8 per 100 000 population. 6 (8.1%) hospital Trusts had new diagnosis ratios ≥2 SDs than expected. The diagnosis rates for Trusts that had a higher-than-expected new diagnosis ratio for epilepsy ranged from 21 to 27 per 100 000 population. As the funnel plot is adjusted for age, sex and deprivation, it shows that relatively little of the observed variation can be explained by random chance among relatively large units.

Figure 2

Funnel plot showing standardised ratios for observed vs expected new diagnoses of epilepsy by NHS Trust. NHS, National Health Service.

Relationship between emergency admission ratios of epilepsy and new diagnoses of epilepsy by trust

Figure 3 is a scatter plot of the standardised ratio for emergency epilepsy admissions against the standardised ratio for new epilepsy diagnoses for the 74 NHS trusts in England that we had diagnoses data for. There does not appear to be any correlation between emergency admissions and new diagnoses at Trust level (Pearson correlation r −0.06, p 0.63, R2 0.003). These standardised ratios are in the range from 0 to 250 due to variations in observed vs expected admissions and variations in observed vs expected rates of new diagnosis of epilepsy. The ratio is 100 if the Trust has the same observed vs expected rate of new diagnosis of epilepsy/unplanned admissions. In some cases, a Trust has up to 2.5 times the observed rate of admissions, because of an unusually high number of unplanned admissions for epilepsy relative to the size of its catchment population, giving a standardised ratio of 250 of observed/expected unplanned hospital admissions for epilepsy, for example.

Figure 3

Scatter plot of the standardised ratio for emergency epilepsy admissions against the standardised ratio for new epilepsy diagnoses.

Discussion

We found widespread unexplained variation in unplanned epilepsy admissions, which cannot be explained by geographical variation in new diagnoses of epilepsy among CYP aged 0–18 in England. This raises concerns about the equity, quality and accessibility of epilepsy services nationally. A similar analysis in 2014/15 on primary admissions for epilepsy reported 11-fold variation in emergency admissions among CYP.5 This is concerning as variations in epilepsy admissions have not improved between 2014 and 2018, despite several national quality improvement programmes.9 Healthcare service factors such as access to paediatric epilepsy specialist nurses, and access to tertiary neurologists in areas such as the Northwest of England with higher-than-expected emergency admissions for epilepsy, should be investigated. Areas with fewer than expected emergency admissions for epilepsy may reflect better access to higher-quality epilepsy services. The fact that wide variation in admission rates is largely not accounted for by diagnosis rates, reinforces this point (ie, variation appears to reflect Trust and other NHS factors rather than patient need). Further work comparing differences between Trusts that have higher and lower than expected epilepsy admissions may shed more light on the causes of this wide variation in emergency hospital admissions for epilepsy seen in England.

When interpreting variation in new diagnosis of epilepsy among CYP in England, it is important to consider that epilepsy is a clinical diagnosis and there is no single diagnostic test to confirm a diagnosis of epilepsy.15 Previous studies have reported a 25%–30% misdiagnosis rate of epilepsy among CYP,6 7 with a lack of awareness of common conditions in childhood that mimic epilepsy, and overinterpretation of EEG findings reported as common reasons for misdiagnosing epilepsy.6 In this study, there was less variation in new diagnosis of epilepsy in England when compared with variation in emergency admissions. Where variation exists, Trusts that had lower than expected rates of new diagnosis of epilepsy, after adjusting for age and deprivation, could reflect reduced access to diagnostic services such as access to EEGs and reduced quality of EEG reporting.9 However, Trusts with a higher-than-expected rate of new diagnosis of epilepsy could also reflect a lack of expertise and thus a lack of access to clinical training, variations in diagnostic protocols and a higher misdiagnosis rate of epilepsy.7 Other studies have found geographical variation in incidence of epilepsy among CYP that could be explained by ethnicity and deprivation.16 We were unable to account for ethnicity in our analysis of variation in new diagnosis of epilepsy at Trust level as Epilepsy12 data does not collect ethnicity data, which is a limitation of this study. We have however accounted for deprivation in our analysis and the fact that wide variation in new diagnosis of epilepsy at Trust level exists after accounting for deprivation, indicates that there are likely other factors, such as expertise at Trust level which may be having an impact.

There are several factors that should be considered when interpreting the geographical variation in emergency admissions and new diagnosis of epilepsy in this study. First, variations in emergency admission rates for CYP with epilepsy can reflect differences in emergency management of seizures, availability of community-based support, such as epilepsy specialist nursing services, effectiveness of ongoing seizure control, thresholds for seeking admission and admission criteria of local departments. The occurrence of seizures in childhood epilepsy can be unpredictable. For a few children, long-term seizure control can be difficult. These children could significantly influence the number of emergency admissions at Trust level. The ratio of observed to expected emergency admissions due to epilepsy was divided by the observed to expected ratio for all types of emergency admissions at Trust level. There was still a significant range of variation between Trusts following this analysis, which indicates that bed capacity and thresholds for admissions cannot fully explain the wide geographical variations in epilepsy admissions across England (online supplemental material). Second, only 74 Trusts out of 134 reported clinical data to Epilepsy12 in the July–November 2018 round, which could have led to some selection bias. Third, catchment populations in our study were calculated based on data on total emergency admissions for CYP. Therefore, trusts that did not have emergency departments were not included in our diagnoses analysis. This could have affected our new epilepsy diagnosis case ascertainment analysis and potentially underestimated the relationship between new diagnosis and unplanned hospital admissions, in our study. Fourth, there is known to be variation in coding of epilepsy in CYP, particularly for the under-five age group where febrile convulsions can be wrongly coded as epilepsy.17 18 This could have potentially overestimated unplanned admissions for epilepsy, particularly for this age group. However, this variation in coding is unlikely to be patterned at a Trust level.

Further deep-dive work on Trusts that have either higher or lower than expected epilepsy admission rates may be beneficial in identifying potential reasons for wide variation. Specific deep-dive work could include looking at the availability of epilepsy specialist nurses, access to tertiary neurologists and local epilepsy diagnostic pathways. Local and regional teams may find unit-specific benchmarking data useful in guiding quality improvement and service development work.19 At a national level, benchmarking data at a unit level could also inform development of guidance on access to diagnostic and treatment services for epilepsy and on setting best practice tariffs that aim to incentivise high-quality epilepsy care for CYP. It is vital that the data from this study is complemented with views of the clinical teams providing the services and CYP to make future service improvements.20

Data availability statement

Data may be obtained from a third party and are not publicly available. Accessibility statement: Unit level Epilepsy12 data on new admissions are publicly available. The pseudonymised patient data (Hospital Episode Statistics data) that were used for this study can be accessed by contacting NHS Digital (see https://digital.nhs.uk/services/data-access-request-service-dars). Access to these data is subject to a data sharing agreement (DSA) containing detailed terms and conditions of use following protocol approval from NHS Digital.

Ethics statementsPatient consent for publication

Not applicable.

Ethics approval

We had approval from the Secretary of State and the Health Research Authority under Regulation 5 of the Health Service (Control of Patient Information) Regulations 2002 to hold confidential data and analyse them for research purposes (CAG ref 15/CAG/0005). We have approval to use them for research and measuring quality of delivery of healthcare, from the London Southeast Ethics Committee (REC ref 20/LO/0611).

Acknowledgments

We would like to thank the Royal College of Paediatrics and Child Health for funding this work. We would also like to thank the Dr. Foster Unit at Imperial College London who helped with data access. The Dr Foster Unit is an academic unit in the Department of Primary Care and Public Health, within the School of Public Health, Imperial College London. The unit received research funding from Dr Foster Intelligence, an independent health service research organisation (a wholly owned subsidiary of Telstra), until September 2021. The Dr Foster Unit at Imperial is affiliated with the National Institute of Health Research (NIHR) Imperial Patient Safety Translational Research Centre. The NIHR Imperial Patient Safety Translational Centre is a partnership between the Imperial College Healthcare NHS Trust and Imperial College London. The Department of Primary Care & Public Health at Imperial College London is grateful for support from the NW London NIHR Applied Research Collaboration and the Imperial NIHR Biomedical Research Centre.