A couple of weeks ago the American Journal of Public Health published a paper by Janessa Graves of the Harborview Injury Prevention & Research Centerthat asserted that public bike sharing programs (PBSP), like Capital Bikeshare, were "associated with increased risk of bicycle related head injuries." Primarily this claim was based upon an analysis of "trauma center data for bicycle-related injuries from 5 cities with PBSPs and 5 comparison cities." This analysis showed that in PBSP cities, head injuries increased from 42.3% to 50.1% after implementation, while in non PBSP cities the ratio remained basically the same, 38.2% before implementation in those cities and 35.8% after. The study has several potential problems with it's methodologies, conclusions and how it was reported in the media. But mostly I want to focus on what we can get from this study and how we can move forward.
The biggest problem I have with the study is the way that it buried the lede. The most interesting thing about it was not that, in PBSP cities, head injuries increased as a percentage of all injuries; but that in cities that launched PBSP, traumatic bicycle-related injuries (including head injuries) decreased more than in cities that didn't. And as a result, many news organizations mis-reported the conclusions, NPR entitled their article "Head Injury Risk Rose In Cities After Bike-Sharing Rolled Out", when in fact it went down. The Post's Lenny Bernstein had to change his title after criticicm of it, but the article still says that claims that "bike share programs are very safe, much safer than, say, crusing around on your own bicycle" are wrong, when, in fact, the conclusion one would make based on this study is that those claims are correct.
The study has many limitations that bring even its central result into question. When I talked to Graves about the study and some of its limitations she leaned heavily on prior studies showing helmet efficacy and seemed to imply that this was no longer in question. In the paper, she cites the controversial results of Thompson's studies that continue to show much higher efficacy than other studies. Of course if the thesis is that Thompson has showed helmets to be very effective, and we need to do more to make them available to bike-share riders, then why do we need this study? The conclusion has to be able to stand on its own.
But the real shame is that the results, even if true, do not sufficiently support the paper's conclusion that "steps should be taken to make helmets available with PBSPs. Helmet availability should be incorporated into PBSP planning and funding, not considered an afterthought following implementation...Promotion of helmet use through integrated rental programs should be a critical element of all PBSPs. This would provide users with the opportunity to reap the benefits of the PBSP while contributing to their safety."
The idea that - all other things being equal - if cyclists wore helmets more often, there would be fewer head injuries is not particularly crazy. If people wore helmets more often in general there would likely be fewer head injuries. But you can't get to the conclusion that bikeshare should do more to promote helmet use through integrated rental programs until there is data that shows that integrated rental programs reduce head injuries - and that they do so without causing more injuries elsewhere, or reducing bike-share use and its many health benefits. Instead, she cites a news story that Citibike is offering limited helmet rentals in New York City.
The best conclusion one can make from this study is that bike-sharing makes biking safer, but the low-rate of helmet use might represent an opportunity to build on those gains if an effective way of encouraging more helmet use can be found. But mostly this study's results leaves me with the sense that more research is needed. Here are some examples of what could be done next:
- If the Citibike helmet rental program is a model to be followed, has it increased helmet usage? Has the increase in helmet usage reduced the head injury ratio without increasing total injuries? How did ridership change? Are there other programs that are working?
- Is the injury ratio result valid?
- Using only 1 year's post-PBSP data means that we only have 1 winter's worth of data, if that winter was more harsh than the prior two, perhaps that caused more head injury crashes.
- The study compares 5 cities to 5 others. But one of the controls - NYC - made up 60% of the population in the control cities. And while in each group there was one sunbelt city, in the PBSP group it is Miami Beach, which made up only 2% of the PBSP population, but in the control was LA which made up 27% of that population. More data from more cities would help.
- 2 of the 5 control cities have mandatory adult helmet laws, but none of the PBSP cities do.
- The pre and post implementation dates aren't exactly right, as in DC, for example, post-implementation starts on May 1, 2010, but the program didn't start until 5 months later. The study should be reworked to tease out the correct time period.
- We need better data on the head injuries themselves. Did skull fractures go up in PBSP cities, because that's the kind of head injuries we'd expect to see a helmet prevent. If instead there were more broken jaws or concussions, then it might be something else that is causing this?
- What happened to other kinds of injuries. For example, what if broken collar bones plummet in PBSP cities? Perhaps the change in injury ratios is the result of slower riding. Or fewer rear-end collisions because the bikes all have lights. Or perhaps it is in part caused by an decrease in neck injuries in PBSP cities since some studies indicate helmets can cause more neck injuries. More data on other injuries broken out by group would help to eliminate the possibility that PBSPs are just better at reducing other injuries than they are at reducing head injuries.
- How does the population of PBSP users differ from other cyclists and can any differences be controlled for?
- How does the population of PBSP trips differ from other bicycle trips and can any differences be controlled for?
- What percentage of total miles biked in each city is by shared bike, and is that enough to account for the difference in ratios?
These are not meant to be criticisms. This wasn't a study of the things I listed. Science is an incremental, collaborative and iterative endeavor, and the Graves paper wasn't meant to be a comprehensive study. When I talked to her she was clear that she saw this as a starting point, and I think as such, it's valuable. I disagree with the conclusions, and I'm not 100% confident that the results will stand up, but I want to see more.
If it turns out that there is a cost-effective way to get more bikeshare riders to wear helmets, that would be a good thing, because as Graves pointed out in our conversation, there are likely people who are choosing not to use bikeshare because they won't ride without a helmet and can't get one at the time they want one. If it turns out that a method for making rental helmets available improves public health, all the better. If we can improve public health and still reap all of the land use, mobility, pollution reduction and congestion relief of bike share, who would possibly oppose that? Not me.