About Outbreaks
How Germwatch detects illness activity across communities and regions.
Illness spreads in different ways - sometimes slowly across a region, sometimes rapidly within a neighbourhood. Germwatch uses two complementary detection systems to capture both patterns. Together, they provide a complete, real‑time picture of emerging illness activity.
• System A - Statistical Outbreak Detector (SOD)
• System B - Syndrome Cluster Detector (SCD)
Each system looks at the same underlying symptom reports, but through a different lens. This dual approach allows Germwatch to detect early warning signals and real‑world clusters with high accuracy.
System A - Statistical Outbreak Detector (SOD)
Detects unusual regional activity compared to baseline using 28‑day trends, recent averages, and z‑scores.
System B - Syndrome Cluster Detector (SCD)
Detects localised clusters of real symptom reports using geographic density and syndrome patterns.
System A - Statistical Outbreak Detector
System A monitors illness activity across large regions (using WHO regional boundaries). It looks for statistical deviations – unusual rises in symptoms compared to what’s normally expected for this time of year.
It analyses:
- 28‑day baselines
- recent 7‑day averages
- z‑scores (a way of measuring how unusual something is compared to what’s normal)
- syndrome‑specific trends
- regional behaviour over time
System A answers: “Is this region behaving unusually compared to its normal baseline?”
Where System A Is Used
System A powers Germwatch’s regional intelligence layer. It does not appear on the homepage or Cluster Intelligence page, because those focus on localised clusters. Instead, System A is used for early warning, trend detection, and official outbreak reporting.
- Daily Outbreak Snapshots - regional anomaly levels, z‑scores, baseline trends.
- Outbreaks.json - the official daily list of statistically significant regional outbreaks.
- Alert.json - a single regional outbreak summary for partners and public‑health organisations.
- Public‑health integrations - councils, researchers, and organisations use System A for early warning.
- Seasonality & trend analysis - long‑term modelling and baseline behaviour.
System B - Syndrome Cluster Detector
System B focuses on geographic clustering - groups of people in the same area reporting similar symptoms at the same time.
It analyses:
- raw symptom submissions
- latitude and longitude
- DBSCAN‑style clustering (a clean, practical way of finding groups of points that are close together)
- cluster size and density
- syndrome mix
- cluster severity
System B answers: “Are multiple people in the same area getting sick at the same time?”
Why Germwatch Uses Two Systems
Outbreaks don’t always start the same way.
Early stage outbreaks
- A region’s baseline rises
- Cases are scattered
- System A detects it
- System B does not
Localised outbreaks
- Several people in one neighbourhood report symptoms
- Region baseline may still be normal
- System B detects it
- System A may not
Established outbreaks
- Both systems detect activity
- One sees the regional trend
- One sees the localised cluster
What Appears on the Homepage?
The homepage uses System B (Syndrome Cluster Detector) only. It is based on real people, is geographically precise, avoids false positives, and is easy for users to understand.
What Appears on the Cluster Intelligence Page?
The Cluster Intelligence page shows System B exclusively:
- emerging clusters
- active clusters
- cluster severity
- syndrome mix
- cluster timelines
- cluster maps
How Germwatch Defines a Cluster
A cluster forms when:
- 3 or more reports
- within 10 km
- within the same time window
- with valid location data
Cluster severity is based on size:
- 3-4 - Emerging (Low)
- 5-7 - Emerging (Low–Moderate)
- 8-11 - Active (Moderate)
- 12-19 - Active (Strong)
- 20+ - Active (Severe)
Germwatch detects outbreaks using two complementary systems – one finds early regional statistical signals, and the other finds real‑world localised symptom clusters. Together, they provide fast, accurate, and clinically meaningful outbreak intelligence.