Analytics & surveillance

Statistical outbreak detection: EWMA and CUSUM for laboratory data

The authoritative outbreak signal comes from verified lab data and transparent statistical models, reviewed by a human epidemiologist - not from a model declaring an outbreak on its own.

6 min read All articles

Deterministic statistics, not a black box

Salus's surveillance layer uses EWMA (exponentially weighted moving average) and CUSUM (cumulative sum) control-chart methods, the same class of statistical process control used across public health surveillance, to flag when a case count or signal has moved outside its expected range. These are transparent, well-understood methods with tunable parameters (EWMA's lambda/L, CUSUM's k/h), not a proprietary model whose reasoning can't be inspected.

As of this writing those thresholds are the published starting values from Salus's technical design work, not values calibrated by a working epidemiologist against a specific jurisdiction's baseline. Treat a fresh deployment's alert thresholds as a starting point to tune, not a validated clinical instrument out of the box.

Where the human stays in the loop

A statistical flag becomes a surveillance signal for a human epidemiologist to review, not an automatic case declaration, an automatic classification change, or an automatic notification to an outside agency. AI personas may retrieve, summarize, and correlate signals; they do not independently declare an outbreak. That boundary is architectural, not a policy promise layered on top.

See it in your laboratory's context.

Explore the live demo, no login and no request form.