How to extract named unmet needs from public conversation

Patient forums are rich and messy. The goal is not to summarise everything people said — it is to name the recurring challenges that should change a programme, message, or research priority, and to size them honestly.

1. Scope the indication before you collect

Define the condition, therapeutic area, project phase, and keyword set: clinical terms, lay language, brand and class names where relevant, and community-specific slang. Without that scope, collection either misses the real conversation or floods analysts with noise.

2. Collect from public sources only

Stick to open platforms and forums your client’s DPA and optics can defend. Closed or login-gated health communities need explicit legal and controller sign-off. Public does not mean anonymous — treat handles and health discussion as personal data with erasure paths.

3. Code for named needs, not vague themes

Prefer specific labels patients would recognise: “insurance delays before treatment starts,” not “access issues.” Specificity is what makes frequency counts actionable and what medical reviewers can challenge or accept.

4. Size by frequency within the indication

A need mentioned once is a signal to investigate. A need that recurs across authors and weeks is a design input. Frequency is not truth — but it is better than ordering insights by how memorable the last quote felt.

5. Keep the quote trail

Every retained insight should point to source evidence: platform, date, and verbatim text. Paraphrased themes without provenance fail the first serious compliance review.

6. Human review before the deliverable

Automation accelerates coding; humans still validate labels, reject junk, and protect against overclaiming. Regulated clients do not want “the model said so.” They want a reviewed evidence pack.