Why helmet laws fail

You won't spend long looking at the helmet issue before someone tells you that helmets prevent 85% of head injuries and 88% of brain injuries.  This is a litmus test: anybody who says this is either ill-informed or deliberately trying to mislead you.  Here's why:

The 85% solution

A long time ago in a city far far away (1988 in Seattle, actually) there was a group of doctors who were helmet enthusiasts.  They set about conducting a study they could cite as part of their agenda to make helmet use compulsory (no I am not making this up, as far as I can tell they had already decided back then).  They conducted what's called a case-control study, based on people presenting at the emergency room of a large hospital.  At the same time one of them was conducting street counts of helmet use to publish a paper bemoaning the low level of helmet use, and thus "proving" that a law was necessary.

The paper was published in 1989 in the New England Journal of Medicine, and is still far and away the most widely cited helmet study.  It is usually cited as saying that "helmets prevent 85% of head injuries and 88% of brain injuries".  It is the sole source for these figures, which have never been reproduced in any other study.

Over the years a number of people have analysed this paper and its conclusions, and this is what they have found:

• Head injury was counted as any injury of the head or face, including areas not covered by helmets
• Brain injury mainly means simple concussion (when liddites talk about scary brain injury figures they almost always mean concussion)
• The "case" and "control" groups are completely different.  The "population control" group of members of a single health co-operative were mainly white, had a higher average income, were more likely to be female, and were riding mainly on bike track in family groups.  The cases were more often male, of black or Hispanic descent, riding alone on city streets.
• The figure is based on assuming that the population control is representative of the general population in the following respects:
  • Helmet use rates
  • Numbers and severity of injury sustained
  • Likelihood of seeking treatment for injury
• None of the above is likely, and in respect of the helmet wearing rate, the street counts conducted at the same time show that the helmet wearing rate among the cases was not significantly different from the general wearing rate - using the street counts instead of assuming that the group health co-operative was representative reduces the effect to zero within the bounds of experimental error.
• The "case" group were also:
  • More likely to have been hit by cars
  • More likely to have suffered other injuries
  • More likely to have had their bikes damaged or destroyed in the crash
• The figure measured was actually around 75%, but the authors chose to assume it was really 85% after "correcting" for differences between the case and control populations
• The figure of 85% is an odds ratio, so cannot be stated as a proportion of injuries prevented.
• The 95% confidence intervals are absurdly wide, indicating that the case sample sizes are very small

There's plenty more, but in the end what it boils down to is this: the authors took two completely different groups of cyclists, with completely different injury profiles, compared them, assumed that helmet wearing rates were the same even though one of the authors had counts which showed they were not, and then attributed all the difference to helmets.  And then added a bit for good measure. 

But above all the figures don't prove that helmets prevent any injuries at all - they simply show that the kinds of cyclists who wear helmets, are less likely to suffer severe crashes.  One statistician obtained the source data and calculated that, using the same logic, helmets prevent over 75% of broken legs!  But on this tenuous foundation is built the whole house of cards.  I have yet to hear of a helmet law which was not proposed on the basis of 85% injury reductions.

Just Like Eddie

These influential authors are still among the leading publishers of helmet research.  They recently published a review of helmet research in the Cochrane library, a very important source.  This report was dominated by their own work, used the 85% figure despite its documented shortcomings, and counted one of their data sets twice.  Their work infects every corner of the helmet debate.

So: helmet laws fail because they are based on absurdly optimistic predictions of injury reductions.

<< Now you see me... | Why helmet laws fail | How injured is injured? >>