A public-health claim may reach you without a date, source or denominator. Investigating one does not require special tools. It requires patience and a consistent routine. This guide uses the same routine we follow on the Global Signal Desk, described in full in our methodology.
1. Start with a narrow question
“Is the outbreak getting worse?” is too vague. Try “How many confirmed cases did WHO report through its latest cutoff date, and how does that compare with its previous report?” A good question names a place, a time window and a measurable thing.
2. Go to the primary source
Find the agency that produces the data: WHO or ECDC for international reports, CDC or FDA for the United States, a state or local health department for local events, and the company itself for a recall. News stories and social posts are useful for noticing an event but should lead you back to one of these pages. Open the agency page, not a screenshot of it.
3. Check the dates
Record three dates: when the agency last updated the page, the cutoff date of the data, and the date you looked. A count “through September 23” published on September 25 is two days old before you read it.
4. Read the definitions and denominators
- What counts as a case: confirmed, probable or suspected?
- What is the percentage a percentage of? Interviewed people and all cases are different groups.
- Is the number cumulative, or for one reporting period?
5. Separate locations and totals
A list of states where people became ill is not the list of states where a product was sold, and a national total tells you little about risk in your town. Name the geography every time you write a number down.
6. Compare independent sources
If two agencies report different figures, find why: different cutoff dates, definitions or counting methods are common. Report both and explain the gap. Do not average them. Two articles that both quote the same agency are one source, not two.
7. Keep an evidence log
Write down every claim with its source URL and the date you accessed it. The table below shows a simple format. Rows are illustrative examples of how to fill it in.
| Claim | Source | Source states | Accessed | Status | Note |
|---|---|---|---|---|---|
| Illnesses linked to recalled alfalfa seed | FDA outbreak investigation page | 76 ill, 16 states; updated Oct 1, 2026 | 2026-10-02 | Confirmed by agency; investigation ongoing | Illness states differ from distribution states |
| Ebola case total in the DRC | WHO Disease Outbreak News | Count through a stated cutoff date | 2026-10-02 | Confirmed cases only | Do not mix with probable or suspected counts |
| Viral post says a product is unsafe | None found; checked FDA recalls page | No matching notice | 2026-10-02 | Unverified | Absence of a notice is not proof of safety |
8. Say how sure you are
Use plain labels: confirmed by an agency, reported by one source, conflicting, or unverified. Write what you still do not know. “I could not find an official notice” is a legitimate finding and does not mean the claim is false.
9. Update and correct
Agencies revise counts. Return to your sources, note what changed and when, and fix any earlier statement that no longer holds. Our signals show visible update notes for this reason.
10. Share responsibly
- Link to the primary source, include the date, and quote numbers exactly.
- Do not identify patients, families, small-town victims or private individuals.
- Do not give medical advice. Point people to their clinician or health department.
- Do not call something an outbreak, a cause or a cover-up unless a primary source does.
Turn it into a project
Students can track one recall or outbreak through its updates for two weeks, then write a short report that covers the timeline, definitions, what changed and what remains unknown. Compare your log with the Signal Desk's October 2 scan and explain any differences. Projects like this build the skills used in the roles described in our careers guide. Ideas for classes are collected on the projects page.