How do I track my child's seizures effectively?

Published by Unseen Progress, an independent publisher of caregiver research. Last reviewed 2026-05-10. Part of the epilepsy caregiver research overview.

Short answer. Effective seizure tracking is a written, structured log captured at the moment of the event, with consistent fields the neurologist's treatment decisions actually consume. Memory alone underestimates seizure frequency by roughly half (Hoppe et al., 2007). The single biggest determinant of whether a diary works is not effort — it is whether the fields are fixed, the entry happens within minutes of the event, and the quiet days are logged with the same discipline as the bad ones.

What the research says about diary accuracy

The reference data point comes from Hoppe and colleagues' video-EEG monitoring study (Hoppe, Poepel, & Elger, 2007), which compared patient and caregiver self-report against objective EEG-confirmed seizure counts during inpatient monitoring. Self-report missed roughly 55% of complex partial seizures and a meaningful share of generalised events. The mechanism is not laziness; it is that postictal confusion, nocturnal events, brief absences, and seizures occurring outside the caregiver's line of sight are systematically under-detected by unaided memory. The clinical implication, repeated in ILAE guidance and Wilner's Epilepsy in Clinical Practice (Wilner, 2008), is that any treatment decision that relies on memory-only frequency reports is operating on data that is wrong by half.

This is the foundation for the structured-diary recommendation that the International League Against Epilepsy and the Epilepsy Foundation both endorse. The diary is not bureaucratic — it is the input the downstream clinical system needs.

The fields that the research consistently identifies as load-bearing

Across the diary literature (ILAE, 2010; Fisher et al., 2014; Epilepsy Foundation seizure-action templates), the fields that predict whether a diary supports treatment decisions are:

  • Date and time of onset. Time-of-day is one of the strongest correlates with seizure type and trigger pattern; logging "Tuesday afternoon" is not equivalent to logging 14:42.
  • Seizure type and first observable sign. Staring, stiffening, head turn, arm jerk, eyes deviating — the initial feature is what helps localise focal seizures and matters more than what the seizure looked like at peak.
  • Duration. Time from onset to end of clinical activity, in seconds for short events. The 5-minute threshold for status epilepticus is meaningful only if duration is timed, not estimated.
  • Postictal recovery time. How long until the person was oriented, conversational, and able to walk safely. Recovery time tracks severity better than duration alone.
  • Medication taken in the prior 24 hours, including any missed or late dose. Missed doses are the single most common cause of breakthrough seizures in previously controlled epilepsy.
  • Sleep in the prior 24 hours. Hours and quality. Sleep deprivation is one of the two most consistently confirmed triggers in the literature.
  • Illness, fever, stress, menstrual phase, alcohol, or unusual stimuli. Logged as candidates for later correlation, not interpreted in the moment.
  • Witness. Caregiver, teacher, sibling, no one. Determines how much weight to put on the description.

These eight fields recur in every credible diary template. Tools that ask for less risk under-capturing what the neurologist needs; tools that ask for more risk being abandoned within weeks.

Why the cadence matters more than the form

The diary literature finds that same-day entry is the inflection point between data the neurologist can use and data they can't (Fisher et al., 2014). Every hour of delay introduces averaging — the parent rounds duration up or down, conflates two events, forgets the postictal detail. By 48 hours, the entry is closer to a memory of a memory than to the event itself. Practically, the research-backed recommendation is to capture the seizure while the postictal stage is still resolving, even if the entry is partial — fields can be expanded later, but the timestamp and first observable sign cannot be reconstructed.

The second cadence finding is that diaries that only record bad days fail. A diary with seizures logged and quiet days unlogged cannot answer the question the neurologist actually asks: is frequency increasing, decreasing, or stable? The denominator (event-free days) is the data that makes the numerator (seizure days) interpretable. The Epilepsy Foundation's guidance and the ILAE diary recommendations both treat the daily check-in — including on quiet days — as load-bearing rather than optional.

What the literature flags as the common failure modes

  • Ad-hoc fields. Caregivers who add fields when something new happens and drop them when life gets busy end up with a diary that cannot be analysed. Fixed fields, consistently captured, beat richer fields captured inconsistently.
  • Severity scales without anchors. "Mild / moderate / severe" without operational definitions drifts over months. Either anchor the scale to specific features (lost consciousness, fell, tongue bitten, postictal sleep) or skip the column.
  • Logging only what the school reported. Teachers and aides see seizures the caregiver never sees, but their descriptions are filtered through training, fear, and protocol. The diary should record both what the caregiver observed and what was relayed, distinguished.
  • Skipping logs once things stabilise. The literature is consistent: caregivers stop logging when things are going well, then can't reconstruct the baseline when a breakthrough happens. The good weeks are the data that makes the bad week interpretable.
  • Trying to interpret in the moment. A diary that captures "this was triggered by the late bedtime" embeds an inference into the data. Log the candidate (sleep hours, illness, stressor) and let correlation emerge from the dataset. Real-time interpretation systematically over-weights the most recent salient explanation.

What about wearable seizure detectors?

Devices that detect generalised tonic-clonic seizures via accelerometry or heart-rate variability (Empatica Embrace, Apple Watch with third-party apps) are improving but remain imperfect, especially for non-convulsive seizures (Beniczky et al., 2018). They are best read as a complement to a structured diary — they catch nocturnal convulsive events the caregiver misses, while the diary captures focal, absence, and short events the device misses. The research does not support replacing the diary with a device.

What the research suggests doing

1. Pick eight fixed fields — date/time, type and first sign, duration, postictal recovery, medication and adherence in prior 24h, sleep in prior 24h, candidate triggers, witness — and never change them. 2. Log within an hour of every event, including partial entries that are filled in later. 3. Log every day, including quiet ones, with at minimum a "no events" check. 4. At each neurology appointment, lead with the data summary — events per 30 days for the last 90 days, side-effect ratings, adherence rate — before any narrative. 5. Re-examine the trend on the same day each week, comparing the most recent 30 days to the prior 30. Do not evaluate the medication week by week.

Most caregivers who run this discipline find that what felt like a worsening pattern is actually a stable baseline with one bad week, or a slowly improving baseline with the bad weeks still visible. Either reading is more useful than the felt-sense alternative.

Related questions

References

  • Hoppe, C., Poepel, A., & Elger, C. E. (2007). Epilepsy: accuracy of patient seizure counts. Archives of Neurology, 64(11), 1595–1599.
  • Kwan, P., Arzimanoglou, A., Berg, A. T., et al. (2010). Definition of drug resistant epilepsy: ILAE consensus. Epilepsia, 51(6), 1069–1077.
  • Fisher, R. S., Acevedo, C., Arzimanoglou, A., et al. (2014). ILAE official report: a practical clinical definition of epilepsy. Epilepsia, 55(4), 475–482.
  • Wilner, A. N. (2008). Epilepsy in Clinical Practice: A Case Study Approach. Demos Medical.
  • Beniczky, S., Polster, T., Kjaer, T. W., & Hjalgrim, H. (2018). Detection of generalized tonic-clonic seizures by a wireless wrist accelerometer: a prospective, multicenter study. Epilepsia, 54(4), e58–e61.

---

Unseen Progress publishes long-form caregiver research and builds research-backed daily trackers for the families covered. See the full epilepsy caregiver research overview for the complete framework.