John Allen Paulos in his always excellent Who’s Counting? column at ABC News:
News story after news story repeats the statistic that out of 140 or so human cases of avian flu reported so far in Southeast Asia, more than half have resulted in death. The reporters then intone that the mortality rate for avian flu is more than 50 percent.
This, of course, is a terrifying figure. But before examining it, let’s first look for a bit of perspective. The standard sort of influenza virus, it’s believed, infects somewhere between 20 million and 60 million people in this country annually. It kills an average of 35,000, and it thus has a mortality rate that is a minuscule fraction of 1 percent. The Swine flu in the 70s killed a handful of people, more of whom may have died from the vaccine for it than from the disease itself. And the Spanish flu of 1918 to 1919 — the deadliest pandemic in modern history and also an avian flu — killed 500,000 to 700,000 people here and an estimated 20 million to 50 million people worldwide. Most assessments of its mortality rate range between 2 percent and 5 percent.
If the avian H5N1 virus mutated so that human-to-human transmission was as easy as it is with the normal flu, and if the mortality rate of more than 50 percent held, the U.S. alone would be facing tens of millions of deaths from the avian flu.
There is one glaring problem with this purported mortality rate of more than 50 percent, however: It is based on those cases that have been reported, and this leads to an almost textbook case of sample bias. You wouldn’t estimate the percentage of alcoholics by focusing your research on bar patrons, nor would you estimate the percentage of sports fans by hanging around sports stadiums. Why do something analogous when estimating the avian-flu mortality percentage?
More here.