07 March 2012

The dangers of statistical correlation

The reduction in premature births is very welcome but the attribution of cause is a bit iffy.  The BBC website reports:
Scotland was the first country in the UK to ban smoking in public places, followed by Wales, Northern Ireland and England in 2007.
After the legislation was introduced in Scotland, fewer mothers-to-be smoked - 19% compared with 25% before.
At the same time there was a significant drop in the number of babies born prematurely or with low birthweight.
The investigators believe both are linked to the smoking ban, even though these rates started to go down some months before the ban was introduced and smoking incidence started to creep up again shortly after the ban. 
The facts that premature birth rates started to go down before the ban and that smoking crept up after the ban cast doubt on the link to the smoking ban.  As does this additional finding:
... the reduction in premature births was both in non-smokers and women who continued to smoke when pregnant, which they say suggests passive smoke exposure is likely involved.
Is passive smoke exposure involved?  Evidence?

Or are the researchers so determined to link the health improvement to the ban that their own evidence to the contrary is simply dismissed?  More research needed, I suggest.

And, yes, as a smoker, I'm biased.  But non-smokers may also be biased.

No comments: