By randomly sampling outwardly healthy people for a disease like Ebola, public-health officials can substantially improve their estimates of epidemic likelihood.
By Nathan Collins
A Doctors Without Borders health worker carries a child suspected of having Ebola in Paynesville, Liberia. (Photo: John Moore/Getty Images)
By the time Americans heard about Ebola cases showing up in West Africa two years ago, health officials already had an inkling that the situation was about to get very, very bad. Indeed, it turned out to be the worst outbreak of the disease in history, killing more than 11,000 people and infecting nearly 30,000. But could anyone have seen that coming? Likely not, according to a new analysis: without knowing who was infected, but not yet symptomatic, it’s too difficult to say.
“During the earliest stages of infectious disease outbreaks, two main questions are i) will a major epidemic occur, and ii) what will the final size of the outbreak be?” write Robin Thompson, Christopher Gilligan, and Nik Cunniffe in PLoS Computational Biology. Thompson, Gilligan, and Cunniffe show for the first time that epidemiologists can’t answer those questions with data on symptomatic cases alone—they also need to know who’s infected but not yet sick.
If conventional estimates put the likelihood of a major epidemic at 50 to 60 percent, the true likelihood could be anywhere between 23 and 83 percent.
The usual, back-of-the-envelope way to estimate the likelihood of any major epidemic relies on two numbers: the “basic reproductive number”— epidemiologist-speak for the average number of healthy people who will be infected by one sick person—and the number of people currently infected. The bigger either number is, the more likely there’ll be a serious outbreak. Recent estimates put Ebola’s basic reproductive number (in West Africa) at upwards of 1.5. If only one person is infected, that translates to a one-in-three chance of a major outbreak. With two infections, the likelihood jumps to 55 percent; with three infections, it goes to 70 percent.
The problem is, that calculation assumes we actually know how many people are infected, but that’s nearly impossible, because there’s always a delay—the incubation period—between infection and the first outward signs that someone is sick. That means knowing two or three people are sick isn’t that informative. Maybe no one else is infected, or maybe half the population is, and knowing whether it’s the former or the latter makes a difference.
A huge difference, in fact. Using computer simulations, the researchers show that, if conventional estimates put the likelihood of a major epidemic at 50 to 60 percent, the true likelihood could be anywhere between 23 and 83 percent.
“Our work shows rigorously, for the first time, that no matter how accurately disease transmission parameters [such as the basic reproductive number] are estimated, precise estimates early in outbreaks of whether a major epidemic will occur will remain unavailable without data about presymptomatic infection,” the researchers write.
Fortunately there may be a way around those problems: testing asymptomatic people. By randomly sampling outwardly healthy people for a disease like Ebola, the authors write, public-health officials can substantially improve their estimates of epidemic likelihood, though exactly how much their estimates improve depends on how many people they test and how accurate the tests are.