Randomized trials and self-reported accidents as a method to study safety enhancing measures for cyclists

User avatar
Aushiker
Posts: 22207
Joined: Tue Jun 05, 2007 1:55 pm
Location: Fremantle, WA
Contact:

Randomized trials and self-reported accidents as a method to study safety enhancing measures for cyclists

Postby Aushiker » Tue Mar 13, 2018 3:45 pm

Abstract
A large number of studies show that high visibility in traffic is important in the struggle of getting the attention from other road users and thus an important safety factor. Cyclists have a much higher risk of being killed or injured in a traffic accident than car drivers so for them high visibility is particularly important. A number of studies have examined the effect of high visibility, such as reflective clothing, but most studies have been primitive, the data limited and the results very uncertain.

In this paper we describe the safety impact of increased visibility of cyclists through two randomised controlled trials: permanent running lights on bicycles and a yellow bicycle jacket, respectively.

The effect of running lights was studied through a trial where the lights were mounted to 1,845 bicycles and 2,000 others comprised a control group. The bicycle accidents were recorded every two month in a year through self-reporting on the Internet. Participants were asked to report all cycling accidents independently of severity to avoid differences between participants as regards to which accidents were reported. They reported a total of 255 accidents i.e. 7 accidents per 100 cyclists. The results showed that the incidence rate for multiparty bicycle accidents with personal injury was 47% lower for cyclists with permanent running light. The difference is statistically significant at the 5% level.

The effect of a yellow bicycle jacket was examined through a trial with 6,800 volunteer cyclists. The half of the group received a bicycle jacket and the other half comprised a control group. Both groups reported every month all their bicycle accidents independently of severity on the Internet. They reported a total of 694 accidents i.e. 10 accidents per 100 cyclists. The treatment group was asked each month if they carried the jacket on their last cycling trip. The results showed that on a random day the treatment group carried the jacket or other fluorescent cycling garment on 77% of their cycle trips. The incidence rate for multiparty accidents with personal injury was 38% lower than the control group. The difference is statistically significant at the 5% level.

The trials were not blind and it seems that the lack of blinding has influenced the level of the groups accident reporting. To address this bias we used a correction factor formed by the difference in the number of single accidents of the two groups.

The experiences with self-reporting of accidents via a web based questionnaire sent by e-mail with one respective two month intervals were very good; in both trials more than 80% answered all questionnaires whereas less than 2% did not answer, and the quality of the self-reported accident was considered high.
[My bold]

The paper is currently available at https://www.sciencedirect.com/science/a ... via%3Dihub but may go behind a paywall in the near future.

AdelaidePeter
Posts: 264
Joined: Wed Jun 07, 2017 11:13 am

Re: Randomized trials and self-reported accidents as a method to study safety enhancing measures for cyclists

Postby AdelaidePeter » Tue Mar 13, 2018 4:03 pm

Thank you. I will have to take a closer look, but on the surface the evidence looks pretty strong.

human909
Posts: 8669
Joined: Tue Dec 01, 2009 11:48 am

Re: Randomized trials and self-reported accidents as a method to study safety enhancing measures for cyclists

Postby human909 » Tue Mar 13, 2018 4:21 pm

AdelaidePeter wrote:Thank you. I will have to take a closer look, but on the surface the evidence looks pretty strong.


Agreed. Though lets not forget this is a Danish study.

I would suggest that the percentage of legitimate "sorry I did not see you accidents" in Denmark are much higher than in Australia.

The two collisions with cars that I have been in have been "I saw you but I though I had enough room" and "I'm a $#%& and will cut in front of you brake heavily and drive off". I can't see how visibility differences would have influenced those incidents.


All that said even if my hypothesis that legitimate "sorry I did not see you accidents" are less in Australia, that would simply imply the effects observed in the study are reduced but certainly not challenge the conclusion that visibility measure reduce accidents.

User avatar
Thoglette
Posts: 3842
Joined: Thu Feb 19, 2009 1:01 pm

Re: Randomized trials and self-reported accidents as a method to study safety enhancing measures for cyclists

Postby Thoglette » Tue Mar 13, 2018 4:37 pm

AdelaidePeter wrote:Thank you. I will have to take a closer look, but on the surface the evidence looks pretty strong.


Given they don't talk about the weather or lighting conditions and it's a meta analysis of work they did a while back there's some "eh, hang on" here. But 300 seconds with my favourite search engine suggests they are mainstream safety researchers.

They're up front about some other issues.
Lahrmann et al. wrote:... is likely that the apparent effect on single accidents in both trials actually reflect a systematic under-reporting of accidents in the treatment group due to an inherent bias in favour of the measure amongst the members of the treatment group.


I'll read it in detail later.
Stop handing them the stick! - Dave Moulton
"People are worthy of respect, ideas are not." Peter Ellerton, UQ

eldavo
Posts: 1655
Joined: Tue Sep 01, 2009 12:21 am
Location: Perth, Western Australia
Contact:

Re: Randomized trials and self-reported accidents as a method to study safety enhancing measures for cyclists

Postby eldavo » Tue Mar 13, 2018 4:41 pm

Yes weather and lighting are significant factors for visibility :D

RobertL
Posts: 582
Joined: Wed Feb 17, 2016 3:08 pm
Location: Brisbane

Re: Randomized trials and self-reported accidents as a method to study safety enhancing measures for cyclists

Postby RobertL » Wed Mar 14, 2018 4:08 pm

The trials were not blind


I'm trying to work out how you could "blind" a trial with people wearing different coloured jackets...
Image

User avatar
Thoglette
Posts: 3842
Joined: Thu Feb 19, 2009 1:01 pm

Re: Randomized trials and self-reported accidents as a method to study safety enhancing measures for cyclists

Postby Thoglette » Wed Mar 14, 2018 4:50 pm

RobertL wrote:
The trials were not blind

I'm trying to work out how you could "blind" a trial with people wearing different coloured jackets...


You can't. It's a bit like economics or political history: proper experiments are impossible.

Then there's the whole issue of defining "visibility" (in one of their earlier papers they rely on) which, to a large degree, is the ultimate in victim blaming. After all, every car is equipped with a set of rather bright headlights. If you "can't see" a human sized object you are driving too fast for the conditions.
Stop handing them the stick! - Dave Moulton
"People are worthy of respect, ideas are not." Peter Ellerton, UQ

fat and old
Posts: 3296
Joined: Sat Aug 30, 2014 12:06 pm
Location: Mill Park

Re: Randomized trials and self-reported accidents as a method to study safety enhancing measures for cyclists

Postby fat and old » Thu Mar 15, 2018 6:36 am

Thoglette wrote: If you "can't see" a human sized object you are driving too fast for the conditions.


In most, if not the vast majority of cases. Certainly not all cases.

RobertL
Posts: 582
Joined: Wed Feb 17, 2016 3:08 pm
Location: Brisbane

Re: Randomized trials and self-reported accidents as a method to study safety enhancing measures for cyclists

Postby RobertL » Thu Mar 15, 2018 1:39 pm

fat and old wrote:
Thoglette wrote: If you "can't see" a human sized object you are driving too fast for the conditions.


In most, if not the vast majority of cases. Certainly not all cases.


Yeah - I had a "coming together" with a car who was pulling out of a t-junction at walking pace. I was turning into the t-junction at not much faster. He just did not look properly, but he could not have been going much slower without being stationary.

I call it a "coming together" because he was going so slowly I was able to shout at him and make him stop. He stopped about 2cm away from me - so he never actually made contact. Then I clipstacked and fell onto his bonnet. By this stage he was very apologetic. I was telling him that it was OK and that he should just stay still (so that I could extricate myself). He then started reversing the car out from under me so I yelled at him to "STAY STILL!!". He finally got the message.
Image

human909
Posts: 8669
Joined: Tue Dec 01, 2009 11:48 am

Re: Randomized trials and self-reported accidents as a method to study safety enhancing measures for cyclists

Postby human909 » Thu Mar 15, 2018 2:42 pm

Yep. Plenty of incidents caused largely by the motorist not looking for cyclists or other smaller objects and just looking for motor vehicles.

Like I said earlier if you have an environment where the users ARE looking out for cyclists. Then collisions are much more likely to be influenced by the actual visibility of the cyclist than in environments where the users aren't looking for cyclists or simply don't care.


The reality is that we are constantly filtering out irrelevant information from our vision. When it comes to driving we are busy filtering out most things except for road boundaries and other road users. For most Australian drivers this can include pedestrians and cyclists being filtered out.

(Just because I'm a cyclist doesn't make me completely immune from this learnt behavior. That is why I try to be quite diligent looking for cyclists.)

Jmuzz
Posts: 347
Joined: Mon Jan 29, 2018 1:42 pm

Re: Randomized trials and self-reported accidents as a method to study safety enhancing measures for cyclists

Postby Jmuzz » Thu Mar 15, 2018 5:43 pm

RobertL wrote:
The trials were not blind


I'm trying to work out how you could "blind" a trial with people wearing different coloured jackets...


You have to take the accident determination out of their hands.
Log sensor data and use camera or radar imaging to determine near misses.
It's all possible with current tech, but multiple cameras becomes too heavy and expensive to be viable.
Or you use basic 360 view camera and outsource manual review to a $2 an hour country.

If all vehicles ever get black boxed (basically always on Strava) then all the near misses would show up very easily. So would all the bad/hoon drivers.
Would change the road landscape overnight. But has big privacy implications (though Google already has it in most cases, and CIA/ASIO etc have anything they have).

User avatar
Thoglette
Posts: 3842
Joined: Thu Feb 19, 2009 1:01 pm

Re: Randomized trials and self-reported accidents as a method to study safety enhancing measures for cyclists

Postby Thoglette » Thu Mar 15, 2018 6:09 pm

Jmuzz wrote:Or you use basic 360 view camera and outsource manual review to a $2 an hour country.

These days, only for the training data. There's a couple of companies in .au who can get a machine to do it.
Stop handing them the stick! - Dave Moulton
"People are worthy of respect, ideas are not." Peter Ellerton, UQ

Who is online

Users browsing this forum: HenryCharlie