AAA surgery--chance result?
This is an interesting article; at least empirically
it would make sense to screen older men for AAA.
Indeed, a control chart shows that there was a
significant decrease in emergency surgery in 2002
although not in other years.
The comparison of mortality, however, is not
statistically significant, assuming independence.
True, the sample percentage fell in the second
period, but the p-value of 0.17 means that this
result would be expected to have occurred by chance 1
time in 6 in repeated samples of similar size.
The error is compounded further by stating that the
mortality "fell 42%," when again the statistical
significance at usual levels was not present.
It is akin to throwing a coin 10 times and obtaining
6 heads, a probability about one in five. One could
throw the same coin 10 more times and obtain 4 heads
with similar probability. The number of heads has
fallen a third, but the coin hardly has ability to
think about how to land.
I hope the authors continue tracking their cases,
perhaps looking at it on a semi-annual or annual
basis, where the chances of showing a difference
sooner could be greater. Again, it's a good idea--it
just needs to be proved.
Michael S. Smith, MD MS (Stat.)
Tucson, AZ USA
Competing interests: No competing interests