The idea
A mobile app for people managing a chronic condition — migraines, IBS, eczema, endometriosis flares — where the trigger is rarely obvious from memory alone. The user logs symptom severity a few times a day alongside quick entries for food, sleep hours, stress, weather, and menstrual cycle if relevant. After a few weeks of data, the app runs correlation analysis across all logged variables and surfaces the ones that actually move the needle: "flare severity is 40% higher on days following under 6 hours of sleep" or "symptom onset correlates with dairy within an 18-hour window." No diagnosis, just a ranked list of candidate triggers to discuss with a doctor or test by elimination.
Why build this
Elimination diets and trigger diaries are standard advice from doctors for a huge range of chronic conditions, but the tooling patients get is a paper log or a generic notes app — no one runs the actual statistics, because doing so by hand across weeks of multi-variable data is tedious and error-prone. Existing symptom trackers (Bearable, Flaredown, MyPainDiary) log data well but mostly stop at charts and calendars; they don't run correlation or lag analysis, which is where the actual signal is since triggers often show up 12–48 hours after exposure, not same-day. A tool that does the statistics automatically turns months of tedious logging into something actionable.
Stack sketch
- Mobile: Flutter or React Native for a fast daily-logging UI — this only works if entering a day's data takes under 30 seconds
- Local-first storage: SQLite on-device, since symptom and health data is sensitive and most users won't want it leaving their phone
- Analysis: on-device statistics using a lightweight stats library (simple-statistics in JS, or a small Python analysis service if the app opts into optional cloud sync) — Pearson correlation plus lagged correlation across a 0–72 hour window per variable pair
- Optional sync: end-to-end encrypted backup to a small self-hosted backend (Postgres + FastAPI) for users who want cross-device access, encryption keys never leaving the client
- Notifications: local reminders to log, tuned to the user's usual symptom-check times rather than a fixed schedule
Scope for v1
- In: daily quick-log flow for symptom severity (1–5 scale) plus 4–5 configurable variables (food categories, sleep hours, stress level, weather auto-pulled by location, cycle day)
- In: a results screen after 14+ days of data showing top 3–5 candidate correlations with lag windows and a plain-language explanation of what the number means
- In: CSV export for sharing with a doctor
- Out: any medical claims or diagnosis language — this is explicitly a data organization tool, not a clinical one
- Out: multi-user or family accounts, integrations with wearables (tempting, but adds scope before the core loop is proven)
Where it could go
The clearest expansion is wearable integration — pulling heart rate variability, sleep stages, and activity automatically from Apple Health or Google Fit instead of manual entry, which would remove the biggest source of logging fatigue and let the correlation engine work with much richer sleep and stress data than a 1–5 self-rating. A second path is condition-specific templates: pre-built variable sets and known trigger categories for migraines vs. IBS vs. eczema, built from published clinical literature, so a new user isn't guessing what to track on day one.
Watch out for
The single biggest risk is a false-positive correlation with too little data leading someone to eliminate a food or habit unnecessarily — the app needs to be conservative about what it calls a "candidate" (minimum sample size, confidence intervals, explicit "correlation isn't causation" framing) rather than optimizing for exciting-looking results. Any language that reads as medical advice also needs a hard legal review before shipping in app stores that scrutinize health apps closely.