Health data anomaly review prompt

By HealthData Prompt • Updated July 2026 • 5 min read

When something “feels off,” you want the model to find outliers relative to your period — not scare you with population cutoffs. This prompt emphasizes confidence levels and co-moving metrics.

Best Apple Health metrics to include

• A multi-day window (ideally 14+ days) so “unusual” has a baseline
• Sleep, HRV, resting heart rate, steps/workouts
• Any vitals you track: SpO₂, respiratory rate, blood pressure, temperature
• Optional typed context: illness, alcohol, travel, new medication, heat wave

Copy-ready prompt

You are scanning personal Apple Health data for anomalies relative to the rest of the provided period. Informational wellness use only — do not diagnose or estimate disease risk.

I will paste a structured summary. Using only that data:

1. Define what “baseline” means here (period average / typical range visible in the paste).
2. List the strongest outliers: dates + metrics + how far they sit from the rest of the window.
3. Group outliers that landed on the same day or consecutive days (clusters matter more than isolated blips).
4. For each cluster, note which other metrics moved with it (or failed to move).
5. Separate likely data-quality issues (missing sleep, single noisy SpO₂ sample, incomplete logging) from multi-metric shifts.
6. Assign a plain confidence label to each finding: high / medium / low — based on sample size and co-movement.
7. End with: top 3 anomalies worth a second look, and 2 questions that could explain them without assuming illness.

Use only the data provided. State uncertainty clearly. Do not diagnose.

Personal context (optional): ________

--- PASTE APPLE HEALTH SUMMARY BELOW ---

Follow-up questions

1. Which anomaly has the highest confidence, and why?
2. Are any “weird” days actually just missing data?
3. Generate 3 non-medical hypotheses that could explain the biggest cluster.

Related prompts

• Weekly health summary prompt
• Vitals trend review prompt
• HRV & resting heart rate recovery prompt

Medical note: Anomaly detection in consumer wearables is not triage. Seek urgent care for severe symptoms regardless of what the chart shows.