Visit-based EHR
Rich at the encounter, empty between them. Hard to see nocturnal clustering, rescue use, or caregiver burden.
Pharma & research partners
Real-world epilepsy evidence, without patient files
Epilog is a privacy-first diary used by people living with epilepsy and the caregivers who log for them. With research sharing on, we produce coded weekly snapshots — seizure patterns, ASM mix, mood, and age band — that partners receive only as aggregates.
Why this conversation
Clinic notes miss the days between visits
Most epilepsy “real-world data” is sparse EHR, claims, or paper diaries. The clinically useful interval is daily life: events, rescue, ASMs, and mood as they happen — including children logged by a parent.
📄
Paper / memory diaries
Unstructured, incomplete, and rarely coded. Side effects and time-of-day are lost in free text.
📱
Epilog’s interval
Coded logging at the moment of care: type, duration, severity, rescue, Watch vs phone, ASM mix, mood scales, age band.
Product
A complete tracker for patients, caregivers, and doctors
Built by LND Tech after lived experience with paediatric epilepsy. Clinical records stay on device unless the person opts into a feature — research sharing, encrypted Pro sync, or share-with-doctor.
⏱️
Seizures
One-tap timer, 11 recognised types, severity, duration, triggers, recovery, Watch capture.
💊
Medications
ASM tracking, frequency, dose, predefined side effects, currently taking vs stopped.
🌤️
Diary & mood
Tags, 1–5 scales for stress, anxiety, energy, sleep, and an elevated-risk mood flag.
🩺
Clinic workflow
On-device doctor reports and optional live summaries — a separate product from research data.
Doctors buy time and signal quality in clinic. Research partners buy population signal. We do not mix the two datasets.
The wedge
Privacy is the product, not a footnote
A small EU company cannot out-scale a claims warehouse. We can be the partner who will never hand over a re-identifiable file.
- ✓ Local-first clinical vault — we cannot read on-device records
- ✓ Research payload is coded fields only; no free text, names, or IDs
- ✓ At most one snapshot per week while sharing stays on
- ✓ User can switch research off at any time
- ✓ Partners get aggregates and k-anonymous cuts, not Firebase access
Legal basis (current app)
Anonymous research sharing is on at first use unless turned off (GDPR Art. 6(1)(a) consent). For a named pharma or academic study we add a second, explicit opt-in — IRBs typically will not accept default-on as study consent.
Recipients: Google Firebase (processor) and LND Tech. Aggregated, non-identifiable datasets may be shared with healthcare research partners only.
Schema v6 snapshot
What a weekly research snapshot contains
Same coded fields the privacy policy lists. Month-year only — never clock time, GPS, notes, or identity.
⚡
Seizures
- Type, time-of-day bucket
- Severity 1–5, duration
- Rescue used, Watch vs phone
- Predefined triggers & place
- Coarse region (city / area)
- Video attached (yes / no)
💊
Medications
- Name as entered, category
- Frequency, currently taking
- Dose unit + daily total
- Predefined side effects
- ~Months on therapy
💜
Mood
- Tags & 1–5 rating
- Stress, anxiety, energy, sleep
- Elevated-risk flag
- Time-of-day + month-year
👥
Cohort (banded)
- Age band, never exact age
- Relationship: Self, Child…
- Illness-episode duration
- Coverage gaps kept visible
Hard exclusions
What never enters a research partnership
Never in the snapshot
- ✕ Names, emails, photos, profile IDs, device IDs
- ✕ Exact birthdates, timestamps, or GPS coordinates
- ✕ Free text: notes, custom triggers, custom places
- ✕ Doctor, hospital, or emergency-protocol details
- ✕ A stable ID to follow one person across weeks
Never the default SKU
- ✕ Raw Firestore dumps or Firebase Auth user lists
- ✕ Unique-patient counts from repeated weekly rows
- ✕ Causal claims (“drug X reduces seizures”) without protocol + stats
- ✕ Advertising audiences built from health snapshots
- ✕ Clinic share-with-doctor summaries (patient-initiated, not research)
Evidence questions
What medical affairs and HEOR can actually ask
| Partner question | What Epilog can show |
|---|---|
| Which ASMs are people logging? | Medication mix, currently-taking %, approximate duration on therapy, age-band slice |
| What side effects show up in the wild? | Predefined side-effect frequencies, cut by drug name / category / age band |
| When do events cluster? | Time-of-day buckets and month-year — not clock time |
| How severe or long are events? | Severity 1–5 and duration buckets (<30s through 5m+) |
| Rescue use in the community? | Share of seizures with rescue marked as taken |
| Paediatric vs adult patterns? | Age bands including caregiver-logged Child profiles |
| Wearable capture? | % of seizures flagged as Apple Watch recordings |
| Geography? | Coarse region mix plus GPS / geocode coverage quality |
Methodology
How to read the dataset — before anyone quotes an N
🔁
Weekly repeats
Each opted-in device uploads a full snapshot about once a week. The same seizure or ASM row can appear again. Use percentages and per-snapshot averages, not pooled totals as unique events.
🧩
No longitudinal ID
By design there is no way to follow one person. That is a privacy feature. It rules out individual trajectories; it still supports population mix and sliced distributions.
🔤
Names as typed
Medication strings are user-entered. Mapping to ATC / RxNorm is a paid deliverable. Free-text “Other” is excluded on purpose.
Coverage gaps (unknown age, no GPS, missing triggers) stay in the dashboard. We treat missingness as a finding, not something to hide. Small cells are suppressed (k-anonymity) before a partner ever sees a cut.
Commercial packages
Three ways to partner — none of them are a data dump
📊
Insight brief
Quarterly aggregate dashboard: ASM mix, seizure type, time-of-day, rescue, Watch share, age bands, side effects. Fastest path. Sample of the same views our research admin already produces.
Medical affairsPatient orgs
🎯
Custom cut
A defined slice — e.g. focal seizures, age 0–17, one region or ASM class — delivered as k-anonymous tables. Cells below threshold are dropped.
HEORRegistries
🔬
Study module
You fund a protocol. We add optional in-app questions behind a named, explicit opt-in. You never receive UIDs. Best fit for observational / RWE programs.
Study opsAcademic PI
Guardrails
How a partnership actually runs
| Step | What happens |
|---|---|
| 1. Scope | Written questions, population, geography, and whether we need a study-specific opt-in |
| 2. Paper | DPA, purpose limitation, no re-identification, no onward sale, EU-friendly processing terms |
| 3. Ethics | Your IRB / ethics path. We will not claim default-on app sharing as protocol consent |
| 4. Delivery | Aggregates or k-anonymous tables. Optional live dashboard. No raw snapshot JSON as the default |
| 5. Review | You see methodology notes (repeat bias, missingness) in the same pack as the charts |
Stage honesty
The instrument is live. Volume is early. That is the founding-partner window.
We will not sell this as a finished claims-scale RWE asset. We will sell a working consent path, a locked schema, and a dashboard that already folds snapshots into the questions on the previous slides.
What we are asking
- ✓ One named cohort (e.g. paediatric epilepsy, one country, one ASM class)
- ✓ Co-recruitment or clinic recommendation so N grows on purpose
- ✓ Ethics letter + DPA so the first cut is publishable-quality
- ✓ A sponsored insight brief or study-module fee — not a seven-figure licence
How this scales
Now — pilot insights, proof of instrument.
Hundreds of weekly opt-ins — recurring aggregate dashboard.
Thousands + optional modules — prospective observational add-ons, still aggregate and protocol-bound.
LND Tech · Epilog
Let’s pick one question and one cohort
A first call should leave with a draft question list, whether you need a study opt-in, and whether the first deliverable is a brief, a custom cut, or a protocol.
Contact
Leandro Barreto · LND Tech
Suggested first 30 minutes
- 1 Your evidence gap (HEOR, medical, registry, or PI)
- 2 Geography and age band that matter
- 3 Brief vs custom cut vs study module
- 4 Ethics / DPA path on your side