Public bikesharing schemes are sprouting up in towns and cities worldwide. The bikes are generally provided without helmets, and this has led to concerns regarding the risk of serious head injuries.
It has been shown that the users of these bike hire schemes are less likely to wear helmets, high-visibility clothing or specialist cycling Lycra than people riding their own bikes. We’ve argued this is a good thing, as it helps normalise the image of cycling away from a specialist past-time, reducing the perception that riding a bicycle is a risky activity or only for super-sporty people.
But a recent study by Janessa Graves and colleagues published in the American Journal of Public Health concluded that there was a link between the introduction of bikesharing schemes in North American cities and the risk of bicycle-related head injuries. So it was argued that helmets should be incorporated into the schemes as standard from the outset.
We suggest that these concerns are misplaced, and agree with the many other commentators who have argued that the study’s data don’t justify the authors’ conclusions. In fact, the paper’s data could be reasonably interpreted to argue the opposite – that the take-up of bikesharing schemes leads to lower risk of injury.
Crunching the numbers
The authors compared the number of injuries (head and non-head) in five cities with bikesharing schemes, and five without, over a 12 month period. Their result is the percentage of cyclist injuries that are head injuries, and this does significantly increase in cities with bikesharing schemes, from about 42% pre-bikeshare scheme to 50% afterwards. In the control cities, the figure dipped from 38% to 36%.
But the result does not represent the risk of head injuries; that is, they did not divide the number of head injuries recorded by a measure of the total amount of cycling being done in the city, to get the “risk per trip” or “risk per kilometre cycled”.
As the study’s data don’t include the amount of cycling trips in each city, we’re also unable to calculate the risk per kilometre cycled. But we can examine what happens to the absolute numbers: the number of head injuries falls in absolute terms in cities with bikesharing schemes relative to the baseline. Crucially, it also falls relative to the control cities. The only reason the percentage of head injuries rises is that the number of non-head injuries fall even faster, leaving head injuries as a greater proportion – but not greater in absolute numbers.
This can be represented as the change over time in the likelihood of injury in cities with a bikesharing scheme versus those without. On this measure of relative change (technically, an odds ratio) all values are less than 1, which indicates that things have got relatively safer in the bikesharing cities.
- for non-head injuries, an odds ratio of 0.59, equivalent to a 41% relative reduction in the odds of an injury
- for head injuries, an odds ratio of 0.89, or an 11% reduction relative reduction in the odds of injury
- for all injuries, an odds ratio of 0.70, equivalent to a 30% relative reduction in the odds of an injury
So a reasonable interpretation would be that cities with bikesharing schemes experienced decreases in all types of injury, including a trend towards a decrease in head injuries, relative to those cities without. In other words, the data suggest injury rates seem to have gone down in bikesharing cities. It’s quite plausible that this actually underestimates the true decrease in risk, since introducing bikesharing schemes is likely to have led to an increase in total cycling.
Other evidence disagrees
There are other reasons to be sceptical of the study’s conclusions. One of the surprising findings in the study is that the percentage of head injuries rises fastest among children. This is precisely the opposite of what one would expect if bikesharing schemes were to blame, since children are less likely to use bikesharing schemes – the bikes are adult-sized, and riders typically need a credit card. So blaming a rise in head injuries among children on a lack of helmets among bike hire users doesn’t seem consistent, as researcher Kay Teschke has pointed out.
The study’s conclusions are also not supported by other evidence pointing to lower injury rates on bikeshare bikes than on personal bikes. The Graves study doesn’t record who is being injured – that is, whether or not they are helmet wearers and whether or not they are bikeshare users. Our recent modelling study of the London bikesharing scheme did have these details, and found that injury risks were lower among bikeshare users compared with the average for cyclists in the same area. This was despite the fact that those hiring bikes were much less likely to wear helmets or hi-viz than other cyclists. Initial reports from New York also suggest that risk of injury among bikesharing users is lower than expected.
Gaps in the data
It’s important to be cautious before attributing either more or fewer injuries to bikesharing schemes. For a scheme to be responsible for the 41% reduction of non-head injuries reported by the Graves paper, the scheme would have needed to have replaced a substantial proportion of existing cycling trips and carry a very low risk of non-head injuries among users (for example, equivalent to a 41% replacement of existing cycling combined with a zero risk of non-head injuries among users).
Such replacement of existing cycling levels are just implausible, because hire bikes are usually only available in a relatively small, downtown part of a city, and represent only a small proportion of cycle trips even in those areas. Our study in London also suggested the majority of trips are new cycling trips.
As an example, bike hire trips in London make up only around 4% of all cycle trips. The study’s authors do not estimate the proportion of bikesharing trips in their cities, but do show that the average number of shared bicycles is three per 1000 population, which is only around 2.5 times higher than in London. So a plausible estimate of the amount of cycling due to bikesharing (2.5 x 4%) is around 10% of cycling journeys.
In other words, far smaller than the 41% replacement of existing cycling needed to be responsible for the change in injuries observed. It’s more likely that the introduction of a bikeshare scheme was one of a number of policies and changes that, taken together, led to safer cycling in these cities.
So a broad ecological study such as the paper from Graves and her co-authors tells us little about the effect bikesharing schemes have on injury rates amongst users. To truly understand this we need direct evidence and individual-level statistics about users and non-users and injury rates. To understand how cycle hire may affect risks for all cyclists we are going to need evidence from different sources, such as qualitative data, policy analysis, and systems models.
What this study does tell us fits with the possibility that injury risks may actually be lower when using hire bikes, and that the introduction of these schemes may go hand-in-hand with a general lowering of risk. So for now, calls for bikesharing schemes (or all cyclists) to require helmets are not supported by the evidence available.
Hard Evidence is a series of articles in which academics use research evidence to tackle the trickiest public policy questions.