Reading an Outbreak: Case Definitions, Attack Rates and the Epidemic Curve

Field epidemiology reads an outbreak with three linked tools. A case definition decides who counts, an attack rate measures illness among the exposed, and the epidemic curve plots onsets over time to suggest point-source versus person-to-person spread. Each is powerful, and each carries a built-in way to mislead if read alone.

Who counts: the case definition

A case definition is a set of standard criteria for deciding whether a person has the condition under investigation. The U.S. Centers for Disease Control and Prevention, whose Principles of Epidemiology course teaches these methods, frames it as clinical criteria combined with restrictions on time, place, and person: a specific set of symptoms or laboratory findings, within a defined period, in a defined area, among a defined group. Definitions are usually tiered by diagnostic certainty. A confirmed case has laboratory verification. A probable case has typical features or an epidemiologic link to a confirmed case. A suspected (or possible) case meets a looser clinical bar.

That tiering exists because a case definition is a trade-off between sensitivity and specificity. Early in an investigation, epidemiologists deliberately cast a wide net, counting confirmed, probable, and suspected cases together to gauge the size and reach of the problem. Later, when they test hypotheses about cause, they often tighten the definition to confirmed cases, because false positives can wash out a real association.

Both moves can mislead. A loose definition inflates the count and can pull in unrelated illness, diluting the very exposure signal the investigators are hunting for. A tight, laboratory-anchored definition can undercount the tail of milder cases and skew the sample toward people with the access and severity to get tested. Change the definition midway, and the outbreak's apparent trajectory changes with it, for reasons that have nothing to do with transmission.

Measuring risk: the attack rate

The attack rate is an incidence proportion dressed for the outbreak setting: the number of new cases divided by the population at risk over the period in question. In a foodborne investigation, the workhorse is the food-specific attack rate, the proportion of people who ate a given item and became ill. The CDC's teaching example is a gathering where 30 of 99 people who ate potato salad fell ill, an attack rate of 30.3 percent.

The rate earns its keep by comparison. Investigators compute an attack rate among those exposed to a suspect item and another among the unexposed, then take the ratio. A much higher rate among the exposed points toward that exposure. For conditions that spread between people, the secondary attack rate, cases among contacts divided by the number of contacts, captures onward transmission.

The pitfalls live in the denominator and the comparison. Attack rates depend on knowing who was truly at risk, which relies on people accurately recalling what they ate days earlier. Exposures travel together, so the salad and the dressing served with it can be statistically inseparable. And with small gatherings, a difference of two or three cases can swing a rate dramatically, making a chance pattern look like a culprit.

Reading the epidemic curve

Plot cases by date of onset and the histogram takes on a shape that constrains the possible stories. A point-source outbreak, where everyone is exposed at roughly one time and place, produces a tight cluster with a steep climb and a more gradual fall, with cases spread across a single incubation period. A continuous common-source exposure, such as an ongoing contaminated water supply, flattens into a plateau. An intermittent source gives an irregular, jagged curve. A propagated outbreak, spreading person to person, is classically drawn as a series of progressively taller peaks spaced one incubation period apart, though the CDC notes this textbook pattern rarely appears cleanly in real data.

The curve also does quantitative work. In a point-source outbreak with a known agent, epidemiologists can count backward from the peak by one average incubation period, and backward from the earliest case by the minimum incubation period, to bracket the likely window of exposure.

Where the curve deceives

The shape is only as good as the onset dates behind it. Reporting delays truncate the right side of a curve, making an ongoing outbreak look as if it is already fading. Recall imprecision blurs onset dates. Weekend and holiday reporting gaps carve artificial dips that mimic real fluctuations in transmission. Mixed patterns, a point source that then seeds secondary person-to-person cases, can blend into a single ambiguous hump. And because the curve is built from cases, the choice of case definition silently shapes it.

What the three tools do together

The strength of an outbreak investigation is triangulation. The case definition fixes the population the other two tools operate on. The attack rate quantifies risk within that population. The epidemic curve places it in time and suggests a transmission mode to test. Read in isolation, any one of them can point the wrong way. Read together, with their limits in view, they turn scattered reports of illness into a defensible account of what happened and why. This article is educational and is not medical advice.