- Bikesafer Home
- About Us
- Crash Crisis
- Crash Study Info
- Study FAQ
- Study History
- Pilot Study
- Refer a Friend
The OECD methodology explained.
The OECD methodology is mentioned a lot in discussions of motorcycle crash causation studies, because it is the one which is being used in the new US DOT Motorcycle Crash Causation Study.
Organization for Economic Co-Operation and Development
OECD, for short, is an international think tank with 30 countries, including the US. It does research, often in the economic area, but is also interested in infrastructure and technology and raising sustainable living standards in general. The bike study methodology was developed in the OECD's Directorate for Science, Technology and Industry.
OECD called in experts from the Head Protection Lab, Harry Hurt's organization, including Terry Smith, Jim Ouellet and Stein Husher to work on the methodology. The Hurt study was used as a model in this process.
Although the OECD methodology was claimed to be used in the Thai study, Dr. Kasantikul's report does not mention the OECD methodology by name, though it acknowledges the work of Head Protection Lab researchers, many of whom were working on the OECD methodology around the same time as the Thai study waw being done.
A methodology is a body of practices, procedures, and rules used by those who work in a discipline or engage in an inquiry, or a set of working methods. Think of it as a recipe for a complicated activity.
Maids European Study
When they were getting ready to do the Maids bike crash causation study, the European motorcycle makers' group, ACEM, joined the OECD RS 9 group sometime around 1997 and US NHTSA participated. The methodology they developed was based on Hurt's landmark US study from 1981, which was the only major study of its kind at the time, so there are a lot of similarities.
All this initialese sounds boring, but the idea of not reinventing the wheel makes sense. The methodology has been used in Thailand and as well as the EU Maids project.
The easiest way to look at how the methodology works is to use Maids as an example. The new US study should be very similar. You could get on the Maids site and see for yourself, but the site is password protected. They'll let you register and get your own password for the asking, but we went ahead and summarized their account of the methodology.
The first step is to decide on the area to sample. In the case of Maids, five locations were chosen. In the case of the US study, the research is to be done in Los Angeles. Clear criteria are laid down to decide which crashes to study, basically any injury-causing crash that a research team can respond to.
Exposure Data / Control Group
Exposure data is collected by the creation of a control group. The methodology recommends a control group the same size as the accidents sampled. The purpose of the control cases is to show when a factor is different in the crash group. The control group is created by interviewing bikers who are in the area of a crash at the same sort of time. They might come back to the scene of a crash at the same day and time a week later, and interview a biker at a filling station near the crash site. They also examine his motorcycle. The idea here is that the control group represents a cross-section of bikers. An example of how this works might be, if ten percent of the bikes in the control group are scooters, and twenty percent of the crashes involve scooters, then riding a scooter might be a factor that makes you more likely to have an accident. This data in general is called exposure data because it is a measure of how a biker is exposed to crash risk factors.
The general idea of a control group is similar to how they do epidemiological and drug research. It is good science in this case, because by definition you can't cause an accident to do an experiment.
Crash Data Collection
The data on the crashes is collected by the researchers responding to crash scenes. Agreements are made with police, first responders and 911 lines. The accident is fully reconstructed, including all contributing human, environment and vehicle factors. This includes the initial conditions of the accident (e.g., vehicles, travel direction, roadway alignment, lighting, traffic controls, etc.), as well as to the pre-crash motions of all involved vehicles. Intended motions (e.g., turning, negotiating a bend, etc.) as well as collision avoidance maneuvers are investigated and coded. Detailed post-crash vehicle inspections provided investigators with information regarding the condition of the vehicle as well as evidence of contact damage, use of lights and tire marks from braking. The scene is diagrammed, sight lines of all participants photographed, and all pre-and post-crash vehicle movements plotted. Obstructions are noted, participants and witnesses interviewed, driver and rider licensing and training is determined. (quotes from Maids).
The rider is followed to the hospital or morgue, where all injuries and road rash are located and the severity documented. Rider equipment is examined and the bike followed to where it is towed, and evaluated for impact damage and the mechanical condition. The combination of vehicle damage and rider injuries helps to determine the movement and points if impact during the crash. About 2000 items of data are captured during this process.
After the initial data is collected, the evidence is used to reconstruct the crash, including all vehicle movements prior to impact, rider and motorcycle trajectory, using standard crash reconstruction techniques. The biker's body, gear and bike, and impact marks at the scene are coded and the effectiveness of protective gear is evaluated. This information is used to list the crash causation factors and their level of contribution to the crash.
The process includes detailed skill and training requirements for the researchers, quality control procedures based on a detailed third-party check of ten percent of the cases, which ensures consistency in the data. At the end of the reconstruction page, the data is anonymized to remove any way to identify the crash participants, and entered into the database, and an archive is created with maybe 50 digital photos, evidence from the site and the biker's gear, reconstruction diagrams etc.
Presentation and statistics.
The methodology also provides for standard methods of presenting the data, and a statistical method called the 'chi-square' test. It's complicated, but it produces a number which is a measure of the probability that the control numbers and the crash sample numbers are the same within the margin of error. If they are not the same, then the factor being measured may be significant as a crash causation factor.
Best to have an example:
Maids found that 5.1 percent of the bikers in crashes did not have the required license, and only 1.4 percent of the riders in the control sample. The chi-square test was 18.1 with a probability of less than .0001, so they concluded that the probability that this factor was unrelated was less than one-hundredth of one percent. In other words, not having a license made you something like four times more likely to be in a crash. They also found that 66% of bikers in crashes had the right motorcycle license, while 76% of riders in the control group had the endorsement. As training is mandatory for a bike license in the EU, they inferred that more training makes you less likely to be in a crash.
Another example of this type of result is when Hurt found that the control sample for motorcycles with headlights n during the day far exceeded the number in the crash group with lights on. This led to the removal of the headlight switch in bikes, and a probable improvement in bike visibility and accident frequency.
The data also supports other types of analysis - e.g. analysis of fatal versus non-fatal accidents revealed in Maids that bike defects were more likely to result in rider death even when they did not cause the crash.
What we'd expect to find in the new study.
If we were guided by science, we'd wait until the study was done, but we blogged this issue in September. Of course, this is speculation. We think that driver right of way infractions will be still present and a major problem, but much less than what Hurt reported. We think speed and alcohol may be more of a factor. And rider training-related issues, skill levels and riding strategies will be major. We think that single-bike accidents may increase, and causes like bikes rear-ending other vehicles.
We are really interested in items which might indicate successful crash avoidance measures that individual bikers are using. This would be like Hurt finding safer bikers with their lights on during the day. This might include
- ABS (see blog)
- Running lights
- Other lighting changes
- Loud pipes (who knows)
- bright-colored or contrasting gear
- Rider training level - basic and advanced
- Rider perception problems, e.g. eyesight.
- Rider cognitive issues
This list is the crux of why we want the 900 to 1200 crash sample. It is very unlikely that we'll get enough ABS or modulators in a 300-crash sample to get significant data on these issues.
And who knows what other factors might emerge from the full study. There might be a few bikers in South California, waiting to be interviewed, that have found the Holy Grail of the bike safety world, the key to saving thousands of bikers, and without a decent-size study, we might never know.
With a study of sufficient size, the OECD methodology will probably find good safety practices, if they are out there.
Photos and information courtesy of maids-study.eu