The best that could be said about this so-called report is that maybe it was a joke or a demonstration of how dopey a "study" of bicycling risk could be constructed. Then again, PT Barnum had a theory about the kind of people who might believe the integrity of any study that reaches the "correct" conclusion.

This isn't data dredging, but you knew that.

BTW, has anyone noticed? These days, EVERY time you go to the doctors office you have to wait for HOURS while a bunch of lycra-clad weenies and hipsters in commuter Levis get the doctor to put a Fred Flinstone band-aid on their pinky and kiss it and make it feel better?

Oh, wait.

-mr. bill

What kind of weenie goes to the doctor's office, let alone often enough to say "EVERY time"?

Bear in mind that the upper bound 5% is the absolute maximum, under the most generous assumptions possible, and includes every injury including minor scrapes and bruises.

It's based on USA data, so the helmet use of particular localities don't factor into it. However, if you could check the accident rate in your locality with 80% helmet use - injury accidents on the street per total cyclists in a year - it would serve as a reality check. I strongly suspect that it would still be close to 1%, just like the overall data.

If you have a 1%, or 5%, as the percentage of cyclists who have an injury accident in a year, now look for the percentage of the injuries that involve a head injury. Where a helmet might help. It's not all of them, not even close. We do have to go with Hospital Emergency numbers for this, no choice. It won't give us exact numbers because of the well known under-reporting problem, but it does provide a decent look at the proportions.

This is a table "Ten most common injuries in bicycle injury related ED visits, 2009" from a statistical brief from the National Center for Biotechnology Information .

I'll post the table here:
[TABLE="class: no_bottom_margin"]
[TR]
[TH="bgcolor: #F0F0F0, align: center"]Principal diagnosis[/TH]
[TH="bgcolor: #F0F0F0, colspan: 3, align: center"]Bicycle-related ED visits[/TH]
[/TR]
[TR]
[TH="bgcolor: #F0F0F0, align: center"]Rank[/TH]
[TH="bgcolor: #F0F0F0, align: center"]Number[/TH]
[TH="bgcolor: #F0F0F0, align: center"]Percentage[/TH]
[/TR]
[TR]
[TD]All bicycle injury related visits[/TD]
[TD="align: center"]--[/TD]
[TD="align: center"]418,700[/TD]
[TD="align: right"]100%[/TD]
[/TR]
[TR]
[TD]Superficial injury; contusion[/TD]
[TD="align: center"]1[/TD]
[TD="align: center"]108,200[/TD]
[TD="align: right"]26%[/TD]
[/TR]
[TR]
[TD]Fracture of upper limb[/TD]
[TD="align: center"]2[/TD]
[TD="align: center"]66,500[/TD]
[TD="align: right"]16%[/TD]
[/TR]
[TR]
[TD]Open wounds of head; neck; and trunk[/TD]
[TD="align: center"]3[/TD]
[TD="align: center"]50,600[/TD]
[TD="align: right"]12%[/TD]
[/TR]
[TR]
[TD]Sprains and strains[/TD]
[TD="align: center"]4[/TD]
[TD="align: center"]40,800[/TD]
[TD="align: right"]10%[/TD]
[/TR]
[TR]
[TD]Open wounds of extremities[/TD]
[TD="align: center"]5[/TD]
[TD="align: center"]28,400[/TD]
[TD="align: right"]7%[/TD]
[/TR]
[TR]
[TD]Intracranial injury[/TD]
[TD="align: center"]6[/TD]
[TD="align: center"]21,300[/TD]
[TD="align: right"]5%[/TD]
[/TR]
[TR]
[TD]Fracture of lower limb[/TD]
[TD="align: center"]7[/TD]
[TD="align: center"]14,400[/TD]
[TD="align: right"]3%[/TD]
[/TR]
[TR]
[TD]Joint disorders and dislocations; trauma-related[/TD]
[TD="align: center"]8[/TD]
[TD="align: center"]9,500[/TD]
[TD="align: right"]2%[/TD]
[/TR]
[TR]
[TD]Skull and face fractures[/TD]
[TD="align: center"]9[/TD]
[TD="align: center"]6,600[/TD]
[TD="align: right"]2%[/TD]
[/TR]
[TR]
[TD]Crushing injury or internal injury[/TD]
[TD="align: center"]10[/TD]
[TD="align: center"]3,800[/TD]
[TD="align: right"]1%[/TD]
[/TR]
[/TABLE]


At most, 19% of the injuries are what you're concerned about. That's considering ALL of the "Open wounds of head, neck, and trunk" to be only "head" and of course they're not.

So this maximum of 19% head related injuries, times your maximum of 5% chance of injury, and you're back to under 1% again. Again, these are upper bound maximums with crazy generous assumptions just to see how high the number can get, and it's still under 1% chance of a head injury in a year. A more reasonable estimation is surely much less than this.

That's still not the end of it. The under 1% is only the chance of having a (head-related) injury accident. For a picture of how much a helmet helps those odds we have to multiply this <1% by some percentage representing helmet effectiveness in head injuries. That's where it gets murky (and, again, why I wouldn't let the other guy just combine "effectiveness" with the probability of an accident). The oft-quoted 85% figure has of course been discredited. I'm not going to cite sources on this part, because I hope to avoid the straw-man arguments that always pop up on both sides when this comes up, but there is a range of credible estimates on what percentage of injuries are mitigated or prevented by helmets. The broadest range of even remotely credible estimates are approximately from 40% to 65%.

So the helmet risk mitigation is multiplying 1% by 40% to 65%, or the maximum is under 0.65%, which is the chances that a helmet will help a member of "all cyclists" during a year of riding. With every assumption made in order to maximize this risk number, unreasonably generous to the risk side. Is that still over the top risk for you?

OK I think I understand now... Less than 1% chance of being in an accident, less than 1% chance the helmet helps/needed if in an accident. So Less than 0.01% chance the helmet is useful for the average cyclist.? But once in an accident where head hits the pavement the helmet can/could be beneficial 65% of the time...

That is my reasoning through the numbers. But it was 5% chance of injury accident (adjusting for extra helmeted accidents not reported), and 19% chance that a helmet would be needed, which by chance .05 x.19=.0095, close enough to 1% for that final number (not .01%) Then that times .65 just as you reasoned.

edit: I'd like to point out that this makes sense if you're looking at total injuries for everyone, for example you want to make something mandatory, or you're trying to decide if it's worth the effort of advocacy. This particular calculation is rather less useful for an individual.

I don't want a mandatory helmet law, was just trying to point out that over all, for people trying to decide whether to wear a helmet or not, that 1% not including instances where the helmet helped seemed wrong... Deciding to not to wear a helmet at 1% is a totally different than deciding to not wear a helmet at 5%... It is for me anyways...

I may be missing something. If you want everyone to wear a helmet, why do you not want a mandatory helmet law?

That may be more of a political question than a helmet question. :)

I never said people should wear a helmet... What I was getting at is they need to decide whether to wear a helmet or not on all the numbers available, not just, oh it wont happen cause... I think 1% chance of having an accident is small, I also think 1% chance of if you do have an accident and hitting your head is very small, BUT I wear a helmet not because I think I will ever need one... But because I want my brain to have that 65% chance of the helmet helping if I ever do hit my head. ;)

Still haven't gotten to see the full paper, but if (as the wording seemed to imply) the association listed was not something they initially set out to test but only did so after gathering their data and then seeing such an association then it is indeed data dredging as defined in the wikipedia article I cited. As stated, such associations only hypothesized once the data is in hand are good for directing future studies but should not be used to draw conclusions based on the initial study data.

The basic problem is that the usual statistical test for significance is at the 0.05 level - i.e. only one chance in 20 that it happens by chance. But when you first gather data and then look for associations there are a very large number of possible ones to consider. For every 20 possible associations one of them may well appear to be "statistically significant" by the usual test even if the data is completely random.

And no, I haven't noticed your claimed behavior of patients at doctors' offices. But then I haven't been to one in a few years except when visiting my daughter and stopping by her office. Most of her patients are wearing diapers rather than lycra, but she does give the older ones Disney and Superhero band-aids when they get their shots. Was there a point to your tale?

Makes sense. You're like me, in that you wear a helmet based on your own judgment, and you respect others' decisions based on their own judgments. We do have a mandatory helmet law where I live, but I would wear a helmet with or without an MHL. :)

Point taken. I will not bring this up again. :)

Your *IF* is not true, so then your dependent clause is not true either. They are not data dredging.

(Now *IF* they noticed a correlation between explorer, safari, chrome, firefox and opera and injury rates, *AND* published a browser correlation, that would be data dredging. But they did not publish such a correlation (and I don't know if they found any such correlation), so they are not data dredging.)

-mr. bill

I'd actually use the 0.13% chance (of an accident injury helped by a helmet in a year) instead of the 0.65% chance because it's more likely to be closer to reality. I'd never wear a helmet if those were my odds, it would be silly.

But those aren't my odds - those are considering the whole population of people who say they ride a bike in a year. Maybe they ride a couple of miles two or three times in the summer, they're included. So it's still silly to say "everyone" should wear a helmet all of the time when there's a tiny chance that it will be useful, because we'd be generalizing to the whole group. Most of what I see in helmet advocacy is misguided in my opinion, for this very reason. Put simply, wearing a helmet is of almost no benefit to the large majority of these people.

I wish helmet advocates would restrict themselves to saying, "for 'this' type of riding (explaining why), accidents are more likely and you should wear one" and acknowledge that "for most casual riding, there is little danger and use your best judgment." There wouldn't be so much pushback from the gross exaggeration of the dangers.

I probably ride ten times or a hundred times as much as the median of that group, in all kinds of weather, and probably a different kind of riding. My individual risk is far greater. For my own risk I look at accidents per mile, or per hour if I can reasonably relate it. I try to weigh in aspects of the specific routes that I take, against the relative frequencies of accidents (intersections, type of road, rural vs urban, time of day, etc), and I judge that against my personal capabilities. We all do that to some degree I think, and call it "judgment". It, the risk, is enough that I don't feel foolish at all wearing a helmet most of the time, nor do I have the slightest qualm about riding bare-headed for a short trip.

While I agree the risk of crashing on a bike is very low, some here claim it is a good excuse not to wear a helmet. However risk is a random thing, and while you may not crash for 7 years yet, you could crash tomorrow. So----------why not be prepared and wear a helmet?

Apparently the helmet-hating Dutch people need medical attention for a bicycle accident on average once every 93 years and 9 months.
They end up in the E.R. every 1125 years and 2 months
And are admitted to a hospital every 5625 years and 8 months.

Fietsongevallen algemeen

Fun Fact, Dutch helmeted riders end up in the hospital at a rate that is 25+ times higher than the bareheaded.

You may have linked to the wrong article?

At a quick glance, it says that 75,000 people are treated by Emergency Department visits each year, so if we accept that everyone in the country bikes (16 million), and guessing that only a relative few have multiple visits, we have about 0.5% of them in the emergency department each year. Not very different from the stats I found for the USA.

For them, same as in the USA it would be silly to advocate that they all wear helmets with such a tiny chance that they'd need it.

You could slip and fall tomorrow while walking, so.....

Helmet? What helmet? :D

http://home.comcast.net/~daihard/images/no_helmet.png

You're right it was sloppy number crunching from my part it is more one ER visit per 225 years and, a hospital admission per 1125 years. The amount of regular cyclists seems to be somewhere between 80-90% of 16,8 million people.so you get very roughly once per 2 centuries and once per millennium respectively. Which is still a magnitude lower than mr bill's hapless helmeted heroes.

I understand and I think we're agreeing - I just don't want anyone mis-using my calculation to think that a helmet would be only a 0.13% help for them in a year. Even in the Netherlands, the guys hitting it hard on racing bikes wear helmets.

I DO want people to think about the actual risk from the perspective of the whole cycling population, if they're arguing or proposing or even advocating for something that impacts the whole group. 0.13% - that's .0013 - is a valid number I think whenever we think of a mandatory law, or requiring helmets on paved trails or government installations, or bicycle safety groups talking about how important helmets are. They need to acknowledge the actual risk levels.

I'll let CanarusMalmari and wphamilton do the maths over and over and over again - and still get the answers wrong.

I hope to ride in the Netherlands someday. Meanwhile, where I do ride is not the Netherlands.

-mr. bill

If you've spotted a mistake in my math, then serve it up.

You asked what's wrong with your math. Let's start with your numerator. (Hint - FARS.) How's it going finding out what your numerator means?

-mr. bill