Wednesday, July 18, 2012

Epidemiologists call them "absolute risk"; you and I could call them real numbers

In a recent Huffington Post blog post, Gil Welch Dartmouth approached a old topic, our pet, "the problem is relative."  Snippet:



Numerous studies have shown that the general public has exaggerated perceptions of health risks they face — as well as exaggerated expectations of the provision of medical care.


It is because they are stupid? No instead, the problem concerns how various sources of health information — researchers, doctors, reporters, web designers, advertisers, etc. — often frame their messages: using the relative change.


"Forty percent more high" and "low" are 50% more variation on declarations. While they are easy to understand, they are also incomplete. Relative change drastically can overstate the underlying effect. It's a great way to scare people.


For example, survey earlier this year found that women with migraine had a 40 percent greater chance of developing multiple sclerosis. That looks scary.


But the researchers were careful to add some important context: multiple sclerosis is a rare disease. In fact, for women with migraine, the chance of developing multiple sclerosis over 15 years was considerably less than 100 — 1:0 pm each only 0.47%. To be sure, which is about 40% higher than the analogous to women without migraines — 0.32% — but it's much less scary. More importantly, it is a much more complete piece of information.


What makes it more complete is the context of the two additional numbers: the risk of developing multiple sclerosis in women with and without migraine. Epidemiologists call these risks "absolutes". You and I could call them real numbers.


Relative change also exaggerate the effects in the other direction. It's a great way to make people believe that there was a real breakthrough.


When we HealthNewsReview.org, generally grade stories in give them a poor score for the criteria of "quantification of benefits" or "quantification of damages" if they use only values of relative risk.


 

0 comments:

Post a Comment

Followers