📄 health informatics

Quantifying User Engagement with the Helpilepsy Seizure Diary

This paper introduces a multidimensional engagement metric for the Helpilepsy seizure diary, using clustering analysis to identify distinct user groups and revealing that highly engaged patients are typically older with longer epilepsy histories and more complex medication regimens, thereby highlighting the need for targeted onboarding strategies for newer patients and the importance of utilizing diverse diary features beyond simple seizure logging.

Original authors: Davies, J., Biondi, A., Viana, P. F., Ampe, L., Schreiber, J., Richardson, M. P.

Published 2026-08-07

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Davies, J., Biondi, A., Viana, P. F., Ampe, L., Schreiber, J., Richardson, M. P.

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ⚕️ This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer

Imagine you are trying to keep a secret diary about your life, but instead of just writing down what happened, you are tracking something unpredictable and tricky, like a storm that sometimes hits your neighborhood. In the world of medicine, this "storm" is epilepsy, a condition where the brain has sudden, unexpected electrical storms called seizures. For doctors to help patients calm these storms, they need a clear map of when and how often they happen. This is where the "seizure diary" comes in. Think of it as a digital notebook where patients record their seizures, how they slept, how they felt, and what medicines they took.

But here is the catch: just like anyone might forget to write in a diary or get bored after a few weeks, patients often stop using these digital tools. If the diary is empty, the map is useless. Scientists call this "engagement." It's not just about opening the app; it's about how deeply and consistently a person uses all its features. The big question researchers are asking is: Who actually sticks with these diaries? Is it the same kind of person who writes every day, or does it depend on how long they've had the condition or how complicated their treatment is? Understanding this is crucial because without a full diary, doctors can't see the full picture, and patients might miss out on the best care possible.

The Digital Detective Work

In this study, a team of researchers acted like digital detectives, looking at the "Helpilepsy" app, a popular electronic seizure diary used by thousands of people. They didn't just count how many people opened the app; they wanted to measure how engaged the users were. To do this, they built a special "Engagement Score," kind of like a video game score that adds up points for different actions.

Imagine the app is a video game with different levels. To get a high score, a player needs to:

Fill out their profile completely (like setting up a character).
Log seizures regularly (not just in a big burst, but consistently).
Log sleep scores regularly (even though sleep isn't a seizure, it's a key part of the game).
Use the app's features to set medication reminders.

The researchers gave each user a score based on these habits. Then, they used a computer trick called "clustering" to sort the 17,151 users into three distinct teams: the Low Engagement team, the Medium Engagement team, and the High Engagement team.

Who Are the Super-Engaged?

The results revealed some surprising patterns about who makes up these teams.

The Low Engagement group was the largest, with about 9,800 users. These players tended to use the app mostly as a simple seizure counter. They would log a seizure when it happened, but they rarely filled out the other parts of the diary, like sleep or mood scores. It was like someone who only writes the date in their diary and nothing else.

The Medium Engagement group had about 6,300 users. These folks were better at filling out their profiles and setting medication reminders, but they still struggled to keep up with logging sleep or mood scores regularly. They were using the app as a tool, but perhaps didn't see the value in the extra features yet.

The High Engagement group was the smallest, with only about 1,006 users, but they were the most dedicated. These users filled out everything: profiles, seizures, sleep, and mood, and they did it consistently over time.

The "Experienced Veteran" Profile

When the researchers looked closely at the High Engagement group, they found a very specific "profile" of a typical super-user. These weren't necessarily the youngest or the newest to the game. In fact, the data suggested the opposite.

Age of Diagnosis: The highly engaged users were, on average, diagnosed with epilepsy at an older age (around 18 years old) compared to the low engagement group (around 11 years old).
Experience: They had lived with epilepsy for longer. The highly engaged group had an average of 33 years since their diagnosis, while the low engagement group had only about 24 years.
Medication Complexity: This group was taking more medicines. On average, they were on 4 different anti-seizure medications, compared to just 1 or 2 for the other groups. They were also taking higher doses of common drugs and using more specialized, "second-line" medications often reserved for difficult-to-treat cases.

The researchers suggest that this pattern makes sense. Someone who has been dealing with epilepsy for a long time, or who was diagnosed later in life, might have a more mature understanding of their condition. They have likely spent more time with doctors, tried many different treatments, and perhaps found that the only way to get relief is to be very active in tracking everything. They aren't just logging seizures; they are actively managing a complex health puzzle.

What This Means for the Future

The study suggests that engagement isn't just about having a cool app; it's about the user's journey. The highly engaged users seem to be those who have realized that the more information they give, the better their treatment can be. However, the researchers also noted a challenge: the highly engaged group didn't stay on the app as long in terms of total time as the others. This might mean that once they get the information they need and find the right treatment, they don't need to use the app as much anymore.

The paper concludes that to help more people, app designers might need to focus more on "onboarding" (teaching) new users, especially younger ones or those who have just been diagnosed. These users might not yet see the value in logging their sleep or mood, so the app needs to show them why it matters. The study didn't prove that high engagement causes better health outcomes, but it strongly suggests that the people who are most engaged are those with the most complex medical needs and the strongest motivation to find answers.

In short, the paper paints a picture of the "super-user" not as a tech wizard, but as a seasoned veteran of their own health, who knows that in the game of epilepsy management, every piece of data counts.

Technical Summary: Quantifying User Engagement with the Helpilepsy Electronic Seizure Diary

Problem Statement
Seizure diaries are critical for the long-term management of epilepsy, yet patient engagement with these tools is often sporadic. While previous research has established the feasibility of electronic seizure diaries and measured general adherence (retention rates), there is a significant gap in understanding the multidimensional nature of user engagement. Existing studies often focus on whether a user continues to use an application rather than how they interact with its various features over time. This lack of granular understanding limits the ability to design tools that effectively support sustained self-management, particularly given the heterogeneity of epilepsy patients in terms of seizure frequency, disease duration, and treatment complexity.

Methodology
The authors conducted a secondary analysis of de-identified longitudinal data from 25,782 unique users of the Helpilepsy® mobile application (Neuroventis BV), collected between 2017 and 2026. The dataset included demographic information, seizure reports, sleep scores, mood scores, medication details, and profile completion data.

To quantify engagement, the authors developed a composite Engagement Metric defined as a weighted sum of five components:

Profile Completeness Ratio (PCR): The percentage of available profile features (e.g., gender, birth year) filled out by the user.
Seizure Upload Regularity ( $R_{seiz}$ ): A measure of the consistency of seizure reporting, calculated using a normalized "burstiness" score (standard deviation vs. mean of inter-upload times) combined with a saturation function to account for low-frequency users.
Sleep Upload Regularity ( $R_{sleep}$ ): Calculated similarly to seizure regularity but applied to sleep score uploads.
Tenure Usage Ratio (TUR): The ratio of unique days a user uploaded sleep scores to the total number of days they had the application installed. This serves as a proxy for sustained engagement with optional, non-biologically triggered features.
Medication Reminder (MR): A binary indicator (0 or 1) of whether the user has set up medication reminders.

The metric is defined as:
$Engagement = w_{PCR} \cdot PCR + w_{Rseiz} \cdot R_{seiz} + w_{Rsleep} \cdot R_{sleep} + w_{TUR} \cdot TUR + w_{MR} \cdot MR$
Weights were assigned as follows: $w_{PCR}=0.20$ , $w_{Rseiz}=0.25$ , $w_{Rsleep}=0.25$ , $w_{TUR}=0.20$ , and $w_{MR}=0.10$ .

After calculating scores for 17,151 users (after outlier removal), the authors applied K-Means clustering to identify distinct user groups. The optimal number of clusters was determined using an elbow plot and silhouette scores, both of which supported a three-cluster solution.

Key Results
The clustering analysis identified three distinct engagement groups: Low (9,807 users), Medium (6,338 users), and High (1,006 users).

Behavioral Differences:
- Low Engagement: These users primarily used the app as a simple seizure diary. They had the highest relative seizure upload regularity but near-zero scores for profile completeness, sleep upload regularity, and tenure usage.
- Medium Engagement: This group showed high profile completeness and medication reminder usage but low regularity in uploading sleep scores and seizures. They appeared to use the app for seizure tracking and medication management but did not consistently engage with optional features like sleep or mood tracking.
- High Engagement: These users exhibited high scores across almost all metrics, particularly in sleep score upload regularity and tenure usage. They were the most likely to report mood and sleep scores.
Clinical and Demographic Correlations:
- Age and Diagnosis: Highly engaged users were significantly older at the time of diagnosis (median 18 years) compared to medium (14 years) and low (11 years) groups. They also had a longer time since diagnosis (median 33 years vs. 27 and 24 years, respectively).
- Medication Burden: The highly engaged group reported a higher number of anti-seizure medications (ASMs) (median 4) compared to medium (2) and low (1) groups. They were more likely to be prescribed second-line and refractory epilepsy medications (e.g., brivaracetam) and received higher doses of common first-line drugs (e.g., levetiracetam, lamotrigine).
- Temporal Stability: While engagement levels remained relatively stable within groups over time, the highly engaged group showed a shorter average duration of app usage (fewer 3-month intervals) compared to the other groups, suggesting they may achieve their clinical goals and discontinue use sooner.

Key Contributions

Multidimensional Engagement Metric: The paper introduces a novel, composite metric that moves beyond simple retention rates to capture the breadth and consistency of user interaction with specific app features (seizures, sleep, profile, reminders).
User Segmentation: By applying unsupervised clustering, the study empirically demonstrates that seizure diary users naturally segregate into low, medium, and high engagement groups, each with distinct behavioral patterns.
Clinical Profiling: The study links engagement levels to specific clinical characteristics, identifying that highly engaged users are typically older, have longer time since diagnosis, and face greater treatment complexity (refractory epilepsy).

Significance and Claims
The authors claim that understanding these engagement profiles is vital for improving the design of digital health tools. The findings suggest that:

Onboarding Strategies: Newer patients (younger age at diagnosis, shorter time since diagnosis) may require more targeted onboarding and education to understand the value of tracking non-seizure variables like sleep and mood.
Feature Importance: Engagement is not solely driven by seizure reporting; optional features like sleep tracking are critical differentiators for highly engaged users.
Clinical Integration: The profile of the highly engaged user (older, treatment-resistant, high medication burden) suggests that engagement is driven by a mature understanding of the disease and a need for active self-management. This implies that digital tools should be designed to support these specific needs, potentially through closer integration with clinical care.

The paper concludes that while data sparsity remains a limitation, the proposed framework provides a robust method for characterizing user engagement, which can inform the development of more effective, personalized epilepsy monitoring tools.

More like this