Comments | m_francis: You Can't Argue with SCIENCE

m_francis

You Can't Argue with SCIENCE

Oct 22, 2009 10:49

This is SCIENCE, Man!

From CBS: Swine Flu Cases Overestimated?
www.cbsnews.com/stories/2009/10/21/cbsnews_investigates/main5404829.shtml

With most cases diagnosed solely on symptoms and risk factors, the H1N1 flu epidemic may seem worse than it is. For example, on Sept. 22, this alarming headline came from Georgetown University in Washington D.C.: " ( Read more... )

science, statistics

Comments 7

New paint... jjbrannon October 23 2009, 02:49:56 UTC

...on an old wagon.

It's Spanish influenza.

JJB

Epidemiology is science... geeklady October 23 2009, 13:01:07 UTC

...even if it feels more like horse racing.

The problem with measuring cases based on those who seek medical care is that it is a biased sample. As a rule, only the very ill and very health-conscious patients will come for voluntary testing. This measure doesn't acknowledge mild cases of swine flu, or cases where the patients are unlikely to seek medical care, it's not getting an accurate picture of the disease in the population. But accurate pictures of disease in the US population is what the CDC is for, so in July they transitioned to a mathematical model. It's not published yet, to my knowledge, or I'd provide you with a link to the PDF.

What the CDC is still collecting are confirmed cases of H1N1 that are severe enough to require hospitalization or cause death, AND total cases of confirmed flu and pneumonia that also require hospitalization or cause death. They are trying to get a picture of the impact of the most severe impact of H1N1 compared to the most severe impact of similar diseases ( ... )

Re: Epidemiology is science... m_francis October 23 2009, 20:10:03 UTC

The whole idea of surrogate measures can be tricky. It's like detecting the presence of a bear by finding the footprints of a rabbit thought to have been frightened by the bear. (That was a description of the detection of neutrinos!) Validating a model is not a trivial task, even in the inanimate sciences. You can get away with measuring Rockwell hardness in place of tensile strength, but it matters which alloy you're checking. In sciences where some of the variables have minds of their own, a model with twenty variable factors in it stumbles because the twenty-first went unaccounted for!

Agreed, regarding the newshounds tendency toward scaring the readers.
http://www.stats.org/stories/2009/lack_stats_affects_media_oct8_09.html

Re: Epidemiology is science... geeklady October 24 2009, 15:30:10 UTC

Ah, but epidemiology isn't the study of disease spreading between individuals, it is the study of disease spread through a population. Individuals are measured to define the population, and not as variables in and of themselves.

It is not an exact science, more of an extended application of probabilities. Look at the first ever study of disease spread, John Snow's study linking cholera outbreaks to the Broad Street pump water. Not EVERYONE who drank from that water got cholera, but the majority of the cholera cases during the outbreak did drink from the Broad Street pump. The link wasn't made less likely because some drank from the pump and didn't get cholera.

Re: Epidemiology is science... geeklady October 24 2009, 15:31:15 UTC

This was me, livejournal keeps logging me out for some reason. Maybe it's jealous of my real blog?

Thread 6