Abstract
BACKGROUND Self-management is increasingly recognised as essential for improving outcomes for people living with epilepsy, yet existing self-management technologies (SMTs) for epilepsy remain limited. Recent advances in wearable sensing, mobile health technologies, and artificial intelligence (AI) offer opportunities to develop data-driven SMTs, including seizure forecasting functions. However, these approaches remain underexplored in epilepsy and require user-centred design to ensure they are clinically meaningful and have real-world usability. OBJECTIVE To evaluate the acceptability, engagement and usability of a real-time data collection prototype designed as the foundation for a data-driven epilepsy SMT, and to explore users’ perspectives on future data-driven functionalities, including seizure forecasting. METHODS We conducted a mixed-methods, in-the-wild study between June 2024 and February 2025. The prototype was a smartwatch device and smartphone application capturing data related to seizures and seizure precipitants. Fourteen adults with epilepsy used the prototype for 30 days. Quantitative outcomes included device wear time and EMA completion rates. Qualitative data were collected through semi-structured interviews and analysed using thematic analysis. Findings informed iterative prototype refinement throughout the study, with three iterations of version 1 (1.0-1.2) and one iteration of version 2. RESULTS Quantitative data: Data completeness in Version 1 exhibited a bimodal pattern, with most participants (n=9) demonstrating high levels of device and EMA usage (EMA daily completion >75% and device daily wear >80%), while a subgroup showed minimal to no data. Post-hoc analyses suggested that the missing data observed in Version 1 was likely attributable to technical issues. In contrast, Version 2 (n=5), which addressed the technical problems, demonstrated high levels of device and EMA usage; across all participants, EMA daily completion exceeded 75% and daily device wear was 100%. Qualitative findings: The prototype was reported as acceptable, engaging and usable, facilitated by the smartwatch’s discreet and comfortable design, the ability to integrate the technology into daily routines, and the value of data entry in providing health insights and prompting behaviour change. Considering data-driven self-management features, participants expressed largely positive views toward future seizure forecasting, highlighting potential benefits including reduced uncertainty, improved autonomy, and quality of life. Additional desired functionality included insights into seizure triggers, access to holistic health data, self-management guidance and data sharing with carers and clinicians. Findings raised concerns around the potential for data misinterpretation. CONCLUSIONS This study indicates that a data-driven SMT for epilepsy, incorporating seizure forecasting, is both desired by users and that a prototype of the underlying data collection system was acceptable, and supported sustained engagement. Future data-driven SMTs should prioritise seamless integration into users’ everyday lives and incorporate flexible design to reflect the heterogeneous and individual nature of users’ needs. Through this study, we also propose methodological approaches to support the design and development of emerging AI-enabled SMTs. INTERNATIONAL REGISTERED REPORT RR2-10.2196/60129