surge · EARS / Farrington
Is a symptom rising here?
Weekly counts in an area are compared with that area's own recent past, so a busy city is not mistaken for an outbreak.
Anpheros Research · Global Health Signals
What people report, counted together.
Never one person. Never fewer than 20.
When thousands of people note their symptoms, patterns appear that no single record can show: a symptom rising in one area, a difference between countries, combinations that keep returning. Global Health Signals measures them for research, from people who chose to take part, and only ever as anonymous groups.
Status: research preview · Groups of at least 20 · Aggregates only
Fig. 0 · Synthetic region, weekly count of one symptom. Bars: people reporting it. Dashed line: the top of the expected range. The last three weeks exceed it.
01 · What it measures
Four signals, counted across people.
Each signal answers one question with one statistical method. Every method has a version number, so a result can be reproduced and checked.
cluster · space-time permutation scan
Is it gathering in one place and time?
Areas and weeks where reports bunch together more than chance allows, tested by shuffling places and dates thousands of times.
compare · age and sex standardised
Do countries differ?
Rates are compared after adjusting for age and sex, so a country with an older population does not simply look sicker.
sequence · cohort
What tends to follow what?
Symptom combinations that are later followed by a recorded condition, counted only when at least 20 people share the path.
Many tests, one correction
q < 0.05 after Benjamini–Hochberg
Scanning many areas, weeks and symptoms makes chance findings likely. All tests of one kind are corrected together, and a signal is reported only when it survives the correction and its effect is large enough to matter.
02 · How people are protected
Nobody is visible. Only groups.
20
The smallest group we compute. Below 20 people, a number is not shown, published or handed to anyone.
- 01
A separate choicePeople take part only if they tick research consent. It is apart from using the app and can be withdrawn at any time. - 02
Location blurred on the phonePosition is rounded to about 10 km on the device, before it is sent. Withdrawing consent deletes it. - 03
Aggregates onlyNo record, name or row of data leaves Anpheros. Researchers receive counts and rates for groups. - 04
Only the account holderFamily members kept in someone's account are not included, because they did not consent themselves. - 05
Study by studyAccess is granted for one study at a time, under a data-use agreement, and every release is recorded.
03 · Limits
What it will not tell you.
a diagnosisA thousand people with headaches do not share one illness. Signals describe groups; causes are for doctors and researchers to study.a country's populationThe data come from people who use Anpheros and chose to take part. They are not a representative sample of any country.a confirmed illnessSymptoms are what people noted, coded to ICD-10. They are not confirmed by a clinician.a public alertThis is a research service. It does not replace national public-health agencies, ECDC or WHO.
04 · Validation
Tested where the truth is known.
Before any real data, each method ran on simulated populations with patterns we planted ourselves, and on populations with nothing planted at all.
Fig. 1 · Share of planted patterns found, synthetic populations
An early version raised false alarms in 42% of empty runs, because effect sizes were filtered before the correction for many tests. The order was reversed before anything shipped. Real reports are noisier than simulations; real results will be published with their own error rates.
05 · Access
Two ways in.
For universities, public-health teams, drug-safety researchers and health-tech developers.
Researchers
Request research access
Tell us about your study. Approved researchers sign a data-use agreement and receive aggregate data for their question. The first releases follow the privacy review now under way.
Request received.
We read every request and answer by email, usually within a week.
Sending…
Please fill in the required fields and tick the agreement.
Wait a moment for the spam check above, then send again.
Too many requests in a short time. Please try again in an hour.
The request could not be sent. Please try again, or write to contact@anpheros.com.
Everyone
Public reports
Weekly counts by country, open to everyone, published only where a group reaches 20 people.
- Weekly signals by countryfirst release pending
- Methods and thresholdson this page
- Validation on synthetic dataon this page
No country has enough participants for a weekly report yet. The first report will appear here as soon as one does.
Global Health Signals · research preview. It describes patterns in anonymous groups; it does not diagnose or recommend treatment.