A Pandemic of Confusion About Flu's Death Rates The CDC's Figure of 100 Influenza Deaths a Day Is Flawed, but That Doesn't Stop People From Spreading

Excerpt:


Dr. Shay and colleagues attempted to compensate for this perceived underreporting of flu deaths by looking for clues in seasonal mortality patterns. If flu's human toll is uniquely resistant to measurement, it is also uniquely measurable by statistical sleuthing, according to Dr. Shay. That's because flu's activity varies each year, in time of onset, duration and severity.

More interactive graphics and photos In a paper earlier this year, he and his co-authors unveiled four different statistical models for measuring flu deaths. Some of these models made use of lab tests from around the country, correlating the fluctuation in positive flu-test results with the waxing and waning in respiratory and circulatory deaths. All four models came up with similar estimates: Tens of thousands of people dying each year from the flu. One of the models produced a 10-year average through the 2002-2003 flu season of 36,171, the basis for the 36,000 figure now ubiquitous in public-health campaigns and media coverage.

This estimate has its skeptics. It's based on lab testing of people who primarily are ill, and not a representative sample of the population, Jimmy Efird, a statistician with the University of North Carolina's Center for the Health of Vulnerable Populations, points out. And it makes use of the same death-certificate data whose flaws necessitated the research, potentially replicating the errors.

Peter Doshi, a graduate student at the Massachusetts Institute of Technology, has gone further in his critique. In a 2005 article in the British journal BMJ provocatively titled "Are US flu death figures more PR than science?" Mr. Doshi criticized the CDC for sometimes confusing deaths caused by the flu with those where the infection was one contributing factor. In an article last year, Mr. Doshi compiled recorded deaths from the flu for the past century and found that the numbers in years of notable outbreaks within the past 60 years weren't that different from years just before and after them.

Dr. Shay defends his measurements but acknowledges they aren't exact, noting, for instance, that his models don't take into account temperature and air pollution, which themselves can affect the incidence of respiratory illness.

If people make decisions based on faulty information, says Mr. Doshi, "then this is important and needs fixing."
-snip-
hat-tip Pixie