This is an edited version of a comment I made elsewhere.
A new study has been published recently on the impact on head injury rates of cyclists by the compulsory helmet laws in New South Wales, Australia. This claims there’s a positive effect on injury rates from such helmet laws, even adjusting for the reported fall in cycling rates due to the introduction of the law. This is being reported by some as evidence that compulsory helmet laws work or even that they lead to a 29% drop in head injuries. However, that seems a misleadingly simplistic view of it.
A careful look at the study shows there appears to be a significant benefit only over a short-period of time, across the few months where the law is passed. While beyond that the helmet law has managed to turn a decreasing head injury rate into an increasing head injury rate (fig 3). This is despite accident rates themselves being somewhat stable (fig 2). Within a year and a half of the law coming in, the head injury rate is almost back to the same level as before! Given the increasing trend, it is likely that not long after the study period head injury rates would have been worse than before!
So what is happening? There is a divergence between the arm injury rate and the leg rate in fig 3. The arm injury rate has increased, mirroring the head injury rate, while the leg rate stays flat – possibly reflecting the stable-ish overall accident rate. This suggests the nature of accidents may have changed.
The study itself notes there are some possible limitations to its findings:
- Injury rates are seasonal, and they have only very limited amount of data (less than a year) on the pre-law rates.
- They assumed exposure to potential injuries was identical for head, leg and arms. While this seems a reasonable assumption, as they say, their own data shows this assumption may not be entirely safe.
- They have no data on cyclist types or behaviour, so they can’t factor out things like proportionally more cyclists in riskier environments having given up, or cycling in safer environments having increased (e.g. commuter v recreational cycling).
One possibility is risk compensation, that the extra perception of safety from the wearing of helmets leads to drivers and/or cyclists taking on more risks. E.g. a study on the behaviour of motorists when passing cyclists has shown that motorists pass closer if the cyclist is wearing a helmet. Another possibility is that injury rates just happened co-incidentally to be at a high prior to the law, and that the law has had little causal effect. Further, the very proposal of the law no doubt lead to a lot of media exposure of the issues around cycle safety. That media exposure no doubt had a strong impact on awareness, even prior to the law’s passing – awareness which faded in time after it was all done and dusted. Or some combination thereof.
That means, so far as the study is accurate, that beyond a short-period, the study actually shows an apparent detrimental impact on safety in terms of head injury rates, and only a transient impact on head injury ratios relative to arm and leg injury rates (the metric they relied on to normalise out changes in cycle use).
Wearing a helmet may well protect your head from serious injury if you get into an accident, the problem is that in wearing one, it you may also be more likely to get into an accident and that accident may even be more serious. Which of those is better is a hard question, but it should probably be an individual choice. Interestingly, some of the safest countries for cycling have very low rates of cycle helmet use – almost no-one wears them in the Netherlands, except for serious sports cyclists.
Update: dg01d linked me to this interesting discussion on cycle helmet safety in the BMJ. Also fine-tuned the language in the conclusion on the possible trade-offs in risk.
The case against bicycle helmet advocacy: a quick guide « Paul Jakma's Blog said
[…] the same at the end of the study period as before the helmet law, and trending to surpass it! (My Blog: study-shows-australian-cyclist-helmet-law-leads-to-increasing-head-injury-rates). Helmet use also appears to induce risk-compensation behaviour in motor vehicle drivers – they […]
Ian Brett Cooper said
Every requirement to wear helmets will increase head injury rates. Why? Because the helmet tends to turn fatalities into survivable head injuries. The British found this out after introducing the Brodie helmet in WW1. This effect applies to modern cyclists too and must be taken into account when calculating the effect of helmets on safety.
Paul Jakma said
From the study: “head injury admissions were defined as all injuries to the skull, face and scalp”. To my understanding this means “head injury” is based on the specific injuries presented on admission – and so does _not_ signify the outcome for the patient. That is, “head injury” includes those who died and had a head injury (potentially with other factors), I think.
Admittedly it’s not explicitly clear in the paper, but I would be very surprised if a state-level admissions system and a peer-reviewed paper in reasonably good journal both managed to completely ignore injuries which resulted in death. I suspect it’s not terribly explicit in the paper because this is taken as given by those working in the field. I’m ready to stand corrected though.
Ian Brett Cooper said
But are people who die at the scene of an accident admitted to hospital and listed in hospital records based on their head injuries? Or do they bypass the hospital and go straight to a morgue admission?. It’s not the hospital outcome I’m concerned with – it’s the admission procedure. It’s very easy for studies to miss this kind of issue.
Paul Jakma said
I’ve gotten a reply from one of the authors. The data does not include accidents where the cyclist was declared dead before hospitalisation and never admitted. It does include fatalities where the cyclist was declared dead after admittance & registration.
Ian Brett Cooper said
So the study completely ignores data where the victim died at the scene of the crash or on the way to the hospital.
Thus, helmet use may be turning what were helmet-less instant fatalities into helmeted survivable injuries, thus increasing the apparent head injury rate.
If this is the case, a decrease in head injuries is very likely to appear to be an increase in head injuries.
But there’s no way to tell, because the data is missing. I think more rigorous studies need to be done.
Paul Jakma said
That depends on how many fatalities-at-scene do not get admitted. Paramedics can not declare people dead, such fatalities may still be admitted and declared dead there-after. It all depends on procedures, and how common it is in Australia to have doctors at the scene and have patients declared dead before hospital, I guess.
Another point, made in the comments on the other discussion on this paper, the link to which I have added in the comments here, points out that procedures may have changed significantly over time, further adding wrinkles to being able to do comparisons.
Paul Jakma said
Also, this Brodie helmet effect doesn’t explain the significant decrease and then the increase there-after does it? I’ve read elsewhere that was a significant dip in cycling rates when the law was introduced, with some amount of recovery thereafter, which might explain it a bit. I still wonder how media exposure affected things (was there significant media discussion of cycle helmets and cycle safety around that time)?
Paul Jakma said
There was a rebuttal to this paper published in the same journal a few months later. “The impact of compulsory cycle helmet legislation on cyclist head injuries in New South Wales, Australia: A rejoinder”, by Chris Rissel:
Why I Started This Blog | Injury Stats said
[…] follows is an email I wrote to Paul Jakma, the author of a blog post that was critical of one of my papers (he mentions he made a similar response here). As I pointed out to him, much of what he wrote was […]
Paul Jakma said
Jake’s account potentially may cause a reader to think I ignored his comments. That wasn’t the case. The opposite in fact. Firstly, it was *I* who emailed *him* asking for clarifications, which led to the email he sent me, now posted (in part) on his blog.
Secondly, I had asked him if I could use his replies, but he specifically told me he didn’t wish me to, as he didn’t want to be seen getting involved in a public helmet debate. Strangely, he didn’t include that part of his email to me on his blog post.
Given Jake’s request, there’s little I could have done other than have taken his points, and have rewritten and presented them myself in such a way as to avoid giving any impression these comments were from Jake. This would have required my digging further into the technicalities of his response, the statistics and data, and validating them for myself, then writing that up. This would have been nice to do of course, but I did not find the time unfortunately.
Of course, nothing would have prevented Jake from posting a correction himself. Either by email to myself, giving permission to quote his email, or by simply leaving a comment on this blog (all comments are published, other than spam and abuse). Had Jake notified me (e.g. by email) that he had a newer academic paper addressing some or all of the points in this blog, I would have gladly linked to it – I wasn’t aware of it.
I am somewhat disappointed that, rather than first sending a simple email, Jake chose instead to publish a blog with accusations of “[Paul does not] seem interested in getting his facts correct about our research”. I think it would be more fair to say that Jake did not seem interested in availing of any of the fairly straight-forward options to have his view be added to this blog!
Here’s our full correspondence, unedited other than the portion Jake has posted publicly on his blog:
My initial emails to Jake:
Jake’s first reply:
My reply, asking for his person to quote his comments:
I’d forgotten to mention why I was asking, so I quickly followed up my email to Jake:
Jake then sent me a detailed reply, most of which he has now published on his blog. However, he (presumably deliberately) has not posted the beginning of it, where he asks me not to quote him:
Paul Jakma said
On the point about risks exposure and types of accidents, a recent paper has some interesting points about assuming these are equal between helmet wearers and non-helmet wearers: “Overestimation of the effectiveness of the bicycle helmet by the use of odds ratios”.