Developing and validating a clinical prediction model to predict epilepsy-related hospital admission or death within the next year using administrative healthcare data: a population-based cohort study protocol

Introduction This retrospective open cohort study develops and externally validates a clinical prediction model (CPM) to predict the joint risk of two important outcomes occurring within the next year in people with epilepsy (PWE): A) seizure-related emergency department or hospital admission; and B) epilepsy-related death. This will provide clinicians with a tool to predict either or both of these common outcomes. This has not previously been done despite both being potentially avoidable, interrelated, and devastating for patients and their families. We hypothesise that the CPM will identify individuals at high or low risk of either or both outcomes. We will guide clinicians on proposed actions to take based on the overall risk score. Methods and analysis Routinely collected electronic health data from Clinical Practice Research Datalink (CPRD), Secure Anonymised Information Linkage databank (SAIL), Combined Intelligence for Population Health Action (CIPHA), and TriNetX research platforms will be used to identify PWE aged ≥16 years having outcomes A and/or B between 2010–2022. Data are held for 60 million patients in England on CPRD, 3.1m in Wales on SAIL, 2.6m in Cheshire and Merseyside on CIPHA, and 250m across 19 countries in TriNetX. Candidate predictors will include demographic, lifestyle, clinical, and management. Logistic regression and multistate modelling will be used to develop a suitable CPM (informed by clinician and public consultation), assessing predictive performance across development (CPRD) and external validation (SAIL, CIPHA, TriNetX) datasets. Conclusions This is the largest study to develop and validate a CPM for PWE, creating an internationally generalisable tool for subsequent clinical implementation. It is the first to predict the joint risk of acute admissions and death in PWE. Mortality prediction is highlighted by NICE as a key recommendation for epilepsy research. The study has been co-developed by epilepsy researchers and members of the public affected by epilepsy. Lay summary Some people with epilepsy (PWE) are at high risk of hospital admission or death because of seizures. If we give clinicians a tool to predict who, they’ll be in a better position to prevent it. Although statistical methods predicting future events are widely available, they haven’t yet been used to predict seizure-related hospital admission or death. Our study is the first to do this. We’ll analyse anonymised electronic research data from thousands of PWE in England. Among them, some will have been admitted to hospital or died because of seizures between 2010–2022. We’ll analyse their age, gender, ethnicity, features of their epilepsy, and medical conditions they developed in the year before being admitted to hospital or dying. From this, we’ll create a statistical tool to predict the chance of someone else with epilepsy being admitted to hospital or dying within a year. The tool’s external accuracy will be checked in Cheshire and Merseyside, Wales, North America, Europe, and other countries. Giving clinicians the tool should generate substantial impact for PWE. For example, emergency epilepsy clinics tend to be reserved for people experiencing a first seizure. However, given our prediction tool tells clinicians which people with an established diagnosis of epilepsy are at risk of seizure-related hospital admission or death within a year, it would provide strong justification for restructuring services such that these high-risk people are also seen in emergency clinics. A high-risk score could also prompt referral for epilepsy surgery sooner than previously considered. It could also prompt multidisciplinary team meetings between neurology and, for example, cardiology if the newly identified risks were cardiac. Such emergency interdisciplinary discussion would not normally happen for a person with epilepsy without good reason: providing evidence for an increased risk of death or hospital admission within a year due to a newly acquired cardiac problem would be good reason. ### Competing Interest Statement The authors have declared no competing interest. ### Funding Statement The salary of Gashirai Mbizvo is funded by an NIHR Clinical Lectureship (CL-2022-07-002). The funders played no role in the design or conduct of this protocol. ### Author Declarations I confirm all relevant ethical guidelines have been followed, and any necessary IRB and/or ethics committee approvals have been obtained. Yes The details of the IRB/oversight body that provided approval or exemption for the research described are given below: The University of Liverpool Research Ethics decision tool was used to determine that ethical approval will not be required as the study described in this protocol consists of a secondary analysis of data that are anonymised by an external party (CPRD, SAIL, CIPHA, TriNetX) and provided to the research team in the fully anonymised format. I confirm that all necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived, and that any patient/participant/sample identifiers included were not known to anyone (e.g., hospital staff, patients or participants themselves) outside the research group so cannot be used to identify individuals. Yes I understand that all clinical trials and any other prospective interventional studies must be registered with an ICMJE-approved registry, such as ClinicalTrials.gov. I confirm that any such study reported in the manuscript has been registered and the trial registration ID is provided (note: if posting a prospective study registered retrospectively, please provide a statement in the trial ID field explaining why the study was not registered in advance). Yes I have followed all appropriate research reporting guidelines, such as any relevant EQUATOR Network research reporting checklist(s) and other pertinent material, if applicable. Yes This is a study protocol, therefore no data have been collected as yet. Once the study has been completed, we will use a public repository (eg GitHub) to make all diagnostic and outcome coding algorithms, metadata, and R analysis scripts used publicly available, facilitating external replication and adaptation.