The warning is dire: Up to 90,000 "possible" deaths from a potential swine flu outbreak.
But how did the president's science advisers, who came up with the number, reach that estimate?
It is based on complicated models of how such illnesses are transmitted, how they mutate, how prepared the healthcare system is, and what's known from the patterns of past flu epidemics, such as the ones in 1918, 1957, and 1968.
It even taps the so-called "virus detectives," a branch of the Centers for Disease Control and Prevention (CDC) that tracks, in the best Sherlock Holmes manner, the threats to America's health – often traveling to remote parts of the world to study outbreaks.
The problem with dire warnings, though, is that the complacency scientists are in part trying to break may be caused by the very studies they tout – the crying wolf syndrome. The avian bird flu predictions in 2005 included estimates of millions dead. Worldwide, 282 people died.
Accordingly, after swine flu fears abated this spring, only 1 in 8 Americans is now worried "a great deal" about the virus, according to an ABC News-Washington Post poll this week.
"Scientists walk a very fine line in how much information do we give: How much do we want to make people aware versus how much do we want to downplay it?" says Sarah Bass, an expert in health-risk communication at Temple University in Philadelphia.
Understanding the past to predict the future
In Tuesday's report, released by President's Council of Advisors on Science and Technology, the 90,000 deaths figure was couched as a "possibility" rather than a prediction. Annually, about 36,000 Americans die of the regular flu, according to CDC estimates.
The CDC said Tuesday that the numbers should be approached cautiously. "We don't necessarily see this as a likely scenario," Anne Schuchat, director of the National Center for Immunization and Respiratory Diseases, told the New York Times.
Modern models are, in large part, based on data gleaned from past flu pandemics, including how viruses mutate to deadlier forms. But an August article in the Journal of the American Medical Association points out that past assumptions about, for example, the 1918 Spanish flu pandemic may be overstated.
Scientists have suggested that the 1918 flu was particularly deadly because it mutated late in the season, and flu models have taken that into account. But the authors of the August article – virologists at the National Institute of Allergy and Infectious Diseases in Bethesda, Md. – pieced together the genetics of the 1918 flu and found that that model might not be correct.
Instead, they say that there's cause to hope for "a more indolent pandemic course and fewer deaths" than in previous outbreaks.
" I think it's very difficult and perhaps a disservice to assume that a new pandemic is going to behave in a way like 1918," National Institute virologist Jeffrey Taubenberger told the Canadian Press.
A good, but flawed, model
One problem for today's flu detectives is that worldwide data is not complete since the event is ongoing. That means they have to rely on older data to fill in the blanks.
"What you have is probably a very good model," says Walter Dowdle, the former acting director of the CDC in Atlanta. "The problem is, does it apply to the present? There is a dilemma in trying to create a model at the same time an event is happening."
But he adds: "I don't think people are exaggerating the problem to create demand. When people make these predictions, it's based on what they consider genuine science."
Follow us on Twitter.