Dorothy,

Thank you for a summary of your past work on mandatory helmet laws. However, it is important to point out our paper isn’t just another rehash on the subject, but it is also a paradigm shift in the way MHL data is analysed.

As mentioned above and worded more succinctly in our paper, our analysis is “exposure free” meaning neither the exact number of cyclists on the road nor the number of cyclists wearing helmets is needed nor does it affect our results. This is based on an assumption of proportional exposure to head injury and to arm injury. Our data suggests this is reasonable since the monthly change in the head/arm ratio before the law is 0.997 (95% CI: 0.978-1.016) and 1.006 (95% CI: 0.960-1.055) after the law. As you can see, there is no statistically significant “trend” effect during either of those periods and the ratio is quite stable. However, we did notice a statistically significant and immediate drop in the ratio at the time of the helmet law (0.725, 95% CI: 0.539-0.974). Among pedestrians, we also found a non-statistically significant “trend” in the periods before and after the law and, by comparison, did not find a statistically significant drop at the time of the legislation. So any attributable factor to the immediate decline in bicycle related head injury was associated with cyclists only and also indicates the drop for cyclists was not due to general road safety improvements. In terms of actual numbers of head injuries, we noted 1288 bicycle related head injuries in the 18 month period before the law and 866 in the 18 period thereafter. That is a relative decrease in bicycle related head injury of 32.8%. Combining this with our analysis means that about 84% (0.275/0.328) of the decline in head/arm injuries during the 36 month study occurred immediately after the law and did not occur gradually.

To summarise our results, we detected a statistically significant decline in the ratio of bicycle related head to arm injuries immediately after mandatory helmet legislation that is independent of exposure (e.g., cycling rates), is a benefit in addition to general road safety improvements and is directed at bicyclists only while not noticing any other statistically significant declines during the entire 36 month period. What else happened at the same time as MHL that meets that criteria?

Dorothy,

The models we have chosen are in fact "exposure free" and the drop that we attribute to MHL is in addition to general road safety measures. Regardless of the fluctuations in the number of cyclists of different age groups post MHL, the drop in bicycle related head injury would still be 0.725 (95% CI: 0.539-0.974). A quick computation of the maximum likelihood estimates taken from the negative binomial generalised linear regression would demonstrate that. We discuss the conflicting reports on the number of cyclists who stopped at MHL and our choice of statistical model in our AAP paper. Please feel free to download it at http://handle.unsw.edu.au/1959.4/50858.

Another of your comments suggests that arm injuries have increased. As I mentioned in my last response, I now posit to you "What event happened at the same time as MHL that is independent of exposure, is a benefit in addition to general road safety improvements and is directed at bicyclists only while not noticing any other statistically significant declines during the entire 36 month period that would have increased arm injuries but had no effect on head injuries?"

Jake,

Dorothy and you are arguing different issues, both of which could be true. Your argument is that the ratio of head to leg injuries improved. You argue this is a paradigm shift, unfortunately this is not exactly true as similar analyses have been performed in New Zealand in an attempt to remove the "exposure" factor (probably as few serious people would today argue that MHLs do not decrease bicycling due to a combination of factors). The NZ results did not produce any conclusive magic bullet in favour of MHLs.

Dorothy is arguing that overall cycling participation dropped but injury rates did not show a corresponding drop. Again in NZ we know bicycling participation has dropped without a matching drop in injuries.

How can both be true? Let's use made up numbers to explain. Before the MHL we'll say there were 100,000 km cycled, 10 leg injuries and 6 head injuries. After the MHL we'll say there were only 50,000 km cycled, 7 leg injuries and 3 head injuries. Total injury rates have increased (from 8/50K to 10/50K), head injury rates have stayed the same (3/50K), and head/leg ratio has improved (from 6:10 to 3:7).

Those numbers are made up, but it is well accepted that injury rates are not linear with participation and the numbers follow that trend. As you know only one counter-example is required to disprove the theory. So your claim of "exposure free" would appear to not only fail to disprove Dorothy's argument, it would seem to question the adequacy of your methodology in determining the benefit or otherwise of MHLs.

Furthermore there appears to be a more fundamental issue which throws into question the usefulness of your result, I'll address that in a separate post.

Harvey,

There was a decline in bicycle related head injury hospitalisations in NSW which I mentioned above, “In terms of actual numbers of head injuries, we noted 1288 bicycle related head injuries in the 18 month period before the law and 866 in the 18 period thereafter.” There was a decline in bicycle related head injury hospitalisations for both children and adults.

Dear Harvey C.,

How about revealing who you are (like all of us have) so that we can identify what if any conflicts of interest you have. It is obvious from your recent numerous postings that you have decided to flood the website with comments which clearly favour anti-helmet law lobbyists. This is quite mischievous of you.

Edward, firstly we did not do the survey, so it is not ‘our’ survey. Let's get that right. The citation that Jake quoted is from the Cycling Promotion Fund and the National Heart Foundation survey. Their survey included more than 40% of respondents who do not own or have access to a bike and said they prefer other forms of recreation and exercise activities to cycling or don’t like cycling at all. Just under a quarter (I presume of 40% non-cyclists respondents) felt cycling was not a safe form of transport (which was also a common reason for cyclists not riding more often).

The study also highlighted that a common theme for not cycling is due to road and safety issues. Unsafe road conditions, volume of traffic and general safety were key reasons for holding back people. Weather conditions were more likely to be rated as a factor than personal matters such as lack of time or motivation.

In regards to risk, a previous study I was involved in provided information concerning 41 deaths and 1776 hospital separations arising from collisions between pedestrians and cyclists and between cyclists and motor vehicles over the 5 year study period July 2000 to June 2005 in NSW, i.e. approximately over 350 incidents per year http://www.sciencedirect.com/science/article/pii/S0001457509002140. Hence the risk is many, many, orders of magnitude higher than being struck on the head by an asteroid (not that it would matter whether or not you were wearing a helmet, if you somehow were freakishly hit by one) you would be pulverised anyway and want something more substantial than a helmet to protect you. While we did not look at head injuries specifically in these incidents I am sure there would be a substantial number per year and worth further investigation.

Other studies have been carried out around Australia. The data and information is out there. You just have to look for it. See for example Garrard et al, Cycling injuries in Australia: Road safety’s blind spot? Page 37 of http://www.acrs.org.au/srcfiles/ACRSVol21-3-WebLR.pdf. Numbers quoted are 1 in 40 traffic crash fatalities and about 1 in 7 serious injuries. They state in their paper that for the Sydney area relative risk of injury on a bicycle is around 13-19 times higher than in a car. To further quote their work: “While sample sizes preclude a direct comparison with other locations, the cyclist fatality rate of between 4 and 7 per 10^8 km in Sydney is several times greater than in the Netherlands (1.1 per 10^8 km), Denmark (1.5) and Germany (1.7), though comparable to the USA (5.8) [6]. The cyclist serious injury rate in Melbourne of between 124 (police data) and 315 (hospital data) per 10^8 km cycled is very much greater than in the Netherlands (14), Denmark (17) and Germany (47), though, once again, comparable to the USA (375). A recent analysis reported a killed or seriously injured cyclist casualty rate of 54 per 10^8 km in Britain in 2008 based on police crash reports. The ‘safety in numbers’ theory has been proposed as a possible explanation for these large international differences in cycling fatality and injury rates, but countries and cities with high levels of safe cycling also have far better conditions for cycling. It is likely that good cycling infrastructure, policies that treat cycling as a legitimate form of transport, lower urban speed limits, national driver and cyclist education, skills and training programs, and stricter levels of liability for drivers in car/cyclist interactions all contribute to improved cyclist safety.”

The risk is high and the cost of burden to tax payers significant which is why MHL also helps reduce the tax payout. But really what we all should be doing is focussing all our passionate energy on convincing governments to fund better infrastructure for cyclists.

Jake,

I am sorry; but there is no misconception, at least here, on this issue. As I explained earlier by providing a simple counter-example canceling out the exposure does not work. Repeating the flawed assertion does you no service.

While at any *particular* given unit of time the exposure cancels, you cannot compare the resulting ratios at *different* points of time unless you can show that there is zero correlation between exposure and injuries.

Not only have you not done this, but much research has shown that "safety in numbers" applies, which is related to exposure. You also have to contend with the research showing that the health benefits of cycling outweigh the costs, so reducing participation produces a health cost.

In other words your theory is relying on an assumption of "all other things being equal" in a situation where it is *known* that all other things are *not* equal, and a counter-example to your theory can be trivially constructed. Let's return to it:

Let's say that before the MHL there were 100,000 km cycled (the "exposure"), 10 leg injuries and 6 head injuries; and after the MHL there were only 50,000 km cycled, 7 leg injuries and 3 head injuries. Now total injury rates have increased (from 8/50K to 10/50K), head injury rates have stayed the same (3/50K), but head/leg ratio has improved (from 6:10 to 3:7).

So by your theory the situation has improved, while in reality it has stayed the same for head injuries, become worse for arm injuries, and become worse for participation.

(This is not of course to say this example shows that the MHL is the only factor affecting these variables, and certainly for participation it is but one negative contributor. The requirement of a counter-example is simply to disprove the theory, not produce a replacement one.)

Given the above your categoric conclusion is unfortunately without any merit whatsoever, you have proved nothing. Or, allowing politics/religion to enter (and sadly this debate is dominated by this and not science), you have re-enforced the anti-MHL argument by being so categoric without basis.

You need to re-visit the subject and produce a theory which accounts for the known variables.

I apologise, Nigel, you are correct in that I did not clear up all your misconceptions. Instead, I only cleared up those relating to our paper and not your misconceptions regarding statistical science. Counterexamples are useful in deterministic fields such as pure mathematics or computer science. Statistical science is governed by the laws of probability and is the basis for most scientific fields such as biology, psychology, medicine, geology, economics, political science, public health, population health, pharmacology, and pretty much anything that ends in -ology. The role probably has in statistics is to separate probably events from those that are improbable. What you have laid out lies squarely with the improbable. You have not presented any data that supports any of your hypotheses. Picking numbers out of the air is not a valid argument.

The Conversation is meant to be a collegial forum to express intellectual ideas and for debate. Since you have resorted to baseless accusations that our original, peer-reviewed research is politically or religiously motivated, the debate is over. I will not engage with you any longer.