Vector 1

Demo · Clinical alert math

How many alarms does it take to catch one patient?

Sepsis, deterioration and stroke alerts share the same math. A model's accuracy, how common the condition is, and where you set the alert threshold decide how many alerts fire, how many are real, and how many patients are missed. Move the sliders to see the tradeoff your nurses live with.

Your unit

Presets are illustrative round numbers, not figures from a specific hospital or vendor model. AUC is the chance the model ranks a random true case above a random non-case.

What a day looks like

The four numbers behind every alert

Sensitivity and specificity come from a standard binormal model of a classifier with the AUC you set: scores for true cases and non-cases follow two bell curves, and the threshold is placed to catch the share of cases you chose. Real models differ in shape, so treat this as a planning tool for the tradeoff, not a validation of any product.