Developing patient responsive information for epilepsy

<h3>Objectives</h3> Technology allows adaptation to changing patient needs, expectations and access to health services. Standardised websites provide information often in the ‘frequently asked questions’ format. There is little research about whether these answer the questions of people with epilepsy (PWE). <h3>Materials and Methods</h3> This study used secondary datasets and social media to ask PWE ‘what would you ask a neurologist/epileptologist’ and ‘do you have any questions relating to your epilepsy’. Questions were themed using an unsupervised topic modelling algorithm to process and categorise the data into an aggregated question set. <h3>Results</h3> The study generated 2752 questions from PWE in Europe, North America and Australia, in the raw data format of natural language. Questions centred on restrictions and fears about lifestyle in 70% of cases. 85% of the questions are not currently answered by Epilepsy charity and medical websites. <h3>Conclusion</h3> In this sample, PWE presented questions framed in terms of lifestyle restrictions and risks. This study is the first stage in developing a Virtual Assistant/Chatbot for the use of PWE to provide accurate and interactive responses to their real questions. This generation of supervised topic classification provides a list of questions which will be used to generate answers required for a Virtual Assistant.