Seizure diaries remain a key source of patient-reported information for epilepsy management, but patient participation is often intermittent. The authors argue that sustained engagement with a diary app materially affects the completeness and reliability of self-reported seizure data. To address this, they developed a structured, multidimensional measure to quantify how users interact with the Helpilepsy seizure diary application and to classify engagement patterns.
The analysis used de-identified data provided by Neuroventis under data-sharing agreements with the Basic and Clinical Neuroscience team at King’s College London. Because the dataset comprised existing de-identified records and did not involve recruitment, intervention, or access to identifiable personal information, the authors report that additional Research Ethics Committee approval was not required under applicable UK research governance regulations. The raw data are not publicly available without explicit permission from Neuroventis.
To characterise user interaction with the app, the authors combined elements commonly used in business application engagement metrics with clinical variables familiar in epilepsy care. While precise feature definitions and weighting used to generate the composite metric are not detailed in the abstract, the general approach integrated multiple engagement dimensions rather than relying on a single measure (for example, seizure uploads alone). This multidimensional approach aims to better reflect the range of diary behaviours that contribute to useful, reliable self-reported data.
Using the multidimensional engagement metric, the authors applied clustering to separate users into groups according to their engagement patterns. Three distinct clusters were identified and labelled as low, medium, and highly engaged users. The clustering permitted comparison of demographic and clinical characteristics across engagement levels to explore drivers of sustained diary use.
Further analysis of the clusters produced a profile for the highly engaged group. Compared with lower-engagement users, highly engaged users tended to:
Be older at the time of epilepsy diagnosis.
Have a longer duration of epilepsy.
Have been prescribed a greater number of anti-seizure medications.
Receive higher doses of commonly used anti-seizure drugs.
Use more medications typically prescribed for refractory epilepsy.
These associations suggest that patients with more longstanding or treatment-resistant epilepsy—and thus greater therapeutic complexity—are more likely to sustain comprehensive interaction with an e-diary.
The findings carry implications for the design and deployment of electronic seizure diaries. Key takeaways reported by the authors include:
Onboarding processes should prioritise patients who are newer to epilepsy, since engagement appears lower in those with shorter disease duration or more recent diagnosis.
App design should recognise that meaningful engagement includes more than the act of uploading seizure events; other diary features contribute to the overall engagement profile and completeness of clinical information.
Targeted strategies—such as enhanced onboarding, tailored reminders, or feature education—may increase sustained participation among users at risk of low engagement.
The authors present these conclusions as design considerations rather than clinical practice recommendations; the manuscript is a preprint and has not undergone peer review.
The study used existing de-identified data supplied by Neuroventis. According to the authors, the nature of the dataset meant additional Research Ethics Committee approval was not required. Data are not publicly available without Neuroventis permission.
Competing interest disclosures include that two authors (L.A. and J.S.) are employees of Neuroventis. M.P.R. reported speaker and consulting fees and receipt of funding support; P.F.V. reported travel and consultancy fees from an industry source but declared no nonfinancial competing interests. The remaining authors declared no conflicts.
Note: This report summarises results as described in the preprint abstract posted on medRxiv. The preprint has not been peer reviewed and should not be used to guide clinical practice. Specific methodological details, statistical parameters, and full feature definitions of the engagement metric were not reported in the abstract and require consultation of the full manuscript or underlying data.