Turning a surveillance signal into a decision

The first time I watched a cluster of cases surface on a line list, I assumed the difficult work was over. We had detected something. In reality, detection was the easy part. The signal sat there — a handful of rows, a suspicious rise — waiting for someone to decide whether it meant anything.

Detection is cheap, interpretation is expensive

Event-Based Surveillance trains you to cast a wide net: rumours, media reports, community alerts. Integrated Disease Surveillance and Response gives you the structure to verify and escalate them. But between the alert and the action lies a translation problem. A district health officer does not have time to read a dataset. They need one sentence: what is happening, how confident are we, and what should happen next.

So I started writing that sentence before I wrote anything else. If I couldn't compress the signal into a single, verifiable claim, I didn't understand it well enough yet. That discipline changed how I cleaned data, how I framed tables, and how I flagged uncertainty — because every choice had to survive being reduced to a decision.

What I carry forward

Surveillance is ultimately a communication system wearing a statistics costume. The maths matters, but the value is created in the last hundred metres — when a signal becomes a sentence, and a sentence becomes a response that reaches people in time.