View Single Post
  #31  
Old November 12th 04, 06:14 PM
Just zis Guy, you know?
external usenet poster
 
Posts: n/a
Default

On Fri, 12 Nov 2004 16:32:17 GMT, "Steven M. Scharf"
wrote:

Yes, it's every bit as wrong as trying to imply causation between
helmet wearing and presentation in emergency rooms - both completely
ignore the actual mechanisms of cause and effect in favour of
pretending helmets make all the difference.


Not really.


No, you're right - those who draw invalid conclusions from
observational studies use them to try to ram through legislation,
whereas I am not trying to compel anyone to do anything beyond think
for themselves.

Your implication of causation is propaganda, pure and simple.


Actually no: it is non-existent, pure and simple.

You claim that you were just stating a fact, but the reality is
that you want people to make a connection based on correlation, and many
people do confuse correlation and causation.


No, what I want people to do is start thinking for themselves. That
means raising little red flags in their minds that, hey, all those
studies which claim vastly inflated benefits, there isn't actually any
real-world evidence to back them up.

The emergency room statistics are valid in the context in which they are
presented, in bicycle accidents involving impact to the head, helmeted
cyclists fare far better in terms of the seriousness of injuries. The
MHL proponents focus on this fact, ignoring the bigger picture.


But the ER studies completely ignore too many other factors. Such as:
what if only risk-averse cyclists are wearing helmets? what if
helmets cause some increased rotational forces which turn some
injuries into deaths while reducing the severity of other injuries?

All this is speculation. Observational studies are speculation, too.
They are not clinical trials, or randomised controlled studies (that
would be unethical for any notionally protective device). They are
some researcher's best guess at how the intervention might affect the
outcome based on his best guess at the effect of the various
confounding factors. The confidence intervals are absurdly large -
the 95% confidence interval for TR&T's 88% is, from memory, between
60% and 98%, and the range of efficacy estimates from observational
studies ranges from about -10% to about 90%, with a pretty wide
distribution. Does 88% sound like a precise figure to you? You know
about the concept of meaningless precision? In context, 88% is like
saying something which ranges between 60c and $1 costs "approximately
88c" - it's absurd. And don't forget that this figure has never been
repeated since, even by the original authors whose later estimates are
much lower. Don't forget that word estimates, either.

And then we compare the observational studies with real-world figures
based on whole populations, millions of cyclists as opposed to the few
thousands (or in some cases tens) in the observational studies. What
do these much larger data sets tell us? They tell us that the effect
of helmets on serious and fatal head injuries is officially
unmeasurable.

That's not a surprise. Your computer came packed in more foam than a
helmet contains, and you wouldn't expect that to survive the kind of
impact which produces life-threatening injuries, like being hit by a
truck. The real surprise is that anyone would think otherwise. But
they do. They see 88% reduction and think, hey, if I wear a helmet I
can do what I want because if I get hit by a truck there's an 88%
chance I'll be fine! And the result of that is risk compensation,
which blows away what little benefit helmets might otherwise provide.

The way to get benefit out of helmets is to sell them quietly, without
hype. To say that cycling is very safe, but, hey, nobody likes to get
a headache if they fall off so you might want one of these. That's
what the evidence says they're good for, and perversely if they were
sold that way the benefit might even become measurable at a population
level, because without the hype the risk compensation might reduce.

Instead we have characters like these pushing helmets:
http://www.bikebiz.co.uk/daily-news/article.php?id=4886

Now we know who ate all the pies, anyway.


I have yet to see any evaluation of the relative merits of different
cycle safety interventions which puts helmets anywhere other than
last.


This is not the point though. The MHL proponents will correctly point
out that the other safety interventions cannot practically be
implemented, or that even if they were, they should not be exclusive.
You'll never win the debate based on the relative effectiveness of the
different interventions.


It's a matter of context and priorities. Remember again, I am not the
one selling something, the Liddites are pushing their plastic
prophylactics, so they have all the burden of proof. Their usual
tactic seems to be to find the biggest number they possibly can and
scream "BIKE DANGER!!!".

When was the last time you saw a large-scale cycle safety campaign
that focused on anything other than helmets?

Here is a little exercise for you. Plot on a graph the percentage of
injuries which were head injuries for pedestrians and cyclists in New
Zealand. Start five years before their mandatory helmet law, when
helmet use was under 45%, and plot through to a couple of years after,
when helmet use was over 95%. See if you can tell which line is which
without looking at the legend. Now co-plot the helmet wearing rate.
Correlate the helmet wearing rate with the relative %HI of cyclists
and pedestrians. Ponder for a while what this tells us about the
efficacy of helmets against recorded injuries and fatalities.


Again, you are confusing correlation with causation. You'll lose every
time with that approach.


Nope. Here the lesson is absolutely valid. It goes to the heart of
the weaknesses in observational studies, because they, too, confuse
correlation with causation. It's science. You have a hypothesis -
helmets prevent deaths and serious injuries. You devise a test - make
everyone wear helmets. You have an outcome - no change in the number
of deaths and serious injuries. Where I come from that would be
counted as proof that the hypothesis was wrong. The Liddites
disagree: they think people are just doing it wrong. Presumably out
of spite for being forced to protect themselves against "BIKE
DANGER!!!"

Helmet promotion is justified with scare stories about death and
serious injury. We know, because every single large scale study has
told us so, that helmets make no measurable difference to death and
serious injury rates. We also know that the risk of cycling is small
to start with, and outweighed by the benefits. So the logical next
step is for the handwringers to butt out and leave us alone.


No argument there. But if you read the articles and letters in the
Toronto newspaper, you'll see that this is barely mentioned. They are
concentrating solely on the reduction in severity of injuries when
accidents occur, not on the fact that the accidents are so rare as to be
inconsequential.


This only works if you think that helmets not only reduce the severity
of injuries, but radically change the distribution curve, so that
injuries which were previously high on the severity scale are now low
on the severity scale, injuries which were low on the severity scale
are unchanged, but fatal injuries are also unchanged - because the
proportion of head injuries does not change, and in order to get
recorded they generally have to be serious enough to make it to
hospital. I suppose it is theoretically possible that helmets have
this extreme differential effect, but William of Ockham would have
suggested an alternative hypothesis.

So you say. On the other hand, countering propaganda is usually seen
as a Good Thing; I have already played a significant role in stopping
one helmet law.


Countering propoganda is a good thing when you do it with facts, rather
than with more propoganda. Another good approach is to try to make the
MHL proponents look ridiculous.


Having already played a significant role in defeating one helmet law I
know exactly what to do thanks. And I think you will find that your
fundamental problem is simply that you don't like some of the facts on
offer.

Guy
--
May contain traces of irony. Contents liable to settle after posting.
http://www.chapmancentral.co.uk

88% of helmet statistics are made up, 65% of them at Washington University
Ads
 

Home - Home - Home - Home - Home