Seizure Diaries Don't Work Very Well. Here's What the Evidence Says.

Seizure diaries are the standard outcome measure in epilepsy research. They're also unreliable, and we've known it for almost twenty years.

The foundational study is Hoppe, Poepel & Elger (2007) in Archives of Neurology. They recorded 91 inpatients on video-EEG and compared what the unit captured against what patients wrote down. About half of seizures went undocumented, including most nocturnal events and most focal impaired-awareness seizures. The follow-up review by Elger and Hoppe (2024) says it hasn't changed: undercounts of 40% to 60% are still typical, and worse for nocturnal and focal aware types.

It isn't a patient problem

The easy reaction is to blame patients. They forget, they fill in the diary at the last minute. That framing misses the mechanism.

A person having a focal seizure often doesn't know it's happening, and can't recall it afterward. Nocturnal seizures happen during sleep. Even for generalized tonic-clonic events with full recovery, memory for the event decays fast and gets rebuilt at diary-completion time instead of event time. Patients aren't bad reporters because they're bad patients. They're bad reporters because you're asking them to report events their brain never encoded.

So the fix isn't reminding patients harder. It's to stop leaning entirely on post-hoc self-report for an endpoint the brain is built to forget.

What the undercount does to your endpoint

Here's the part that matters for protocol design.

If your primary endpoint is a 50% or greater reduction in seizure frequency, and baseline diary accuracy sits around 50%, your baseline count is already a coin flip away from reality. So is your follow-up count. You're taking a ratio of two noisy numbers and treating a threshold cross as a treatment effect.

That doesn't make every past epilepsy result wrong. Big effects show up even through noisy instruments. It does mean your study is far less sensitive to small real effects than your power calculation assumes, and responder-rate analyses are the most exposed. A 5-point responder-rate gap between arms is easy to wave away when the instrument is wrong half the time.

There's a subtler issue too, from Fisher et al. (2016). Patients are hyperaware when a trial starts and report faithfully. Vigilance fades as the study runs on, and documentation drops. In the data, that looks like a treatment effect.

What real-time capture changes

The case for logging seizures as they happen, instead of reconstructing them on Sunday night, isn't about technology. It's about the data you get back.

Log an event in the moment and you get a timestamp, not a day of the week. That timestamp opens up things a weekly count can't touch: medication-to-seizure interval, sleep the night before, clustering, circadian pattern. All of it feeds secondary analyses and label expansion.

For nocturnal events, smartphone accelerometer data can flag probable seizures overnight that a caregiver reviews in the morning. It's not continuous EEG, but it beats a memory check at breakfast, and it goes straight at the part of the undercount caused by the patient being asleep. A 2024 study in Scientific Reports found app diaries reported seizures with much better temporal resolution than paper, and a richer dataset downstream.

Burden is the wrong objection

The usual pushback is that event-driven capture is more burdensome than a weekly diary. Building Forma, and consistent with the EMA literature back to Shiffman, Stone & Hufford (2008), we've found the opposite. A 60-second post-event prompt is lighter than a 10-minute weekly reconstruction. The work lands when it's cheapest, right after the event, when memory is fresh. That's also how you hold off survey fatigue: small asks at the right moments, not one big obligation at an arbitrary time.

This is the heart of Forma's thesis. Keep the patient's existing validated instrument, and add continuous real-world data around it. Voice-first daily logging captures the seizures and the context the diary misses between visits. Additive, not a replacement.

If you're designing a protocol

1. Keep your validated instruments, add real-time capture. QOLIE-31 and NDDI-E have psychometric properties and regulatory acceptance. Run them on schedule. Layer real-time logging on top as the primary efficacy capture.
2. Design for the caregiver. In pediatric epilepsy and refractory epilepsy, the caregiver is often the better reporter. Build that into consent and data structure from day one.
3. Pre-specify how timestamps map to your endpoint. Real-time logs produce a richer structure than a weekly count. Write the analytical crosswalk at the protocol stage so reviewers aren't surprised.
4. Don't over-promise. Real-time capture beats paper, but it doesn't reach video-EEG fidelity. Say so. Reviewers respect a clear statement of instrument limits more than a completeness claim the literature contradicts.

The accuracy problem has been documented since Hoppe 2007, and it hasn't gone away. The tools to reduce it exist. See how we put them to work in our studies, including rural pediatric epilepsy where the diary gap is widest. The friction is in adopting them, not building them.