Induced Exposure Study
In automobile studies, an Induced Exposure study gets crash case data from drivers who are judged culpable in a multi-vehicle collision (Cerrelli, 1973). Nonculpable drivers are the comparison group. Differences between BACs in the culpable group and the nonculpable group would be considered indicative of the extent to which BAC contributed to the crash, as opposed to being generally present in the population at large. The primary reason for conducting such a study is to take advantage of BACs recorded as part of the crash report. If this methodology were to be conducted exactly the way in which it is done for four-wheeled vehicles, it would be necessary to find crashes in which two motorcyclists, ideally operating independently of each other (as opposed to riding together in a group), collided. This occurrence is extremely rare. An alternative would be to use crashes between motorcycles and four-wheeled vehicles and compare culpable versus nonculpable riders. Using this method, comparisons would no longer be from the same crash and would, therefore, not be automatically matched for location, time of day, and so on.
Such a study could be done using archival data (e.g., FARS data). Determining culpability could be attempted by using information already in the FARS file. In the case of multiple vehicle crashes, cases in which the rider was speeding or otherwise operating unsafely could be considered rider-culpable cases. In the past, studies have been conducted in which vehicles that were struck from behind or were not moving when struck were deemed nonculpable.
There are several ways in which single-vehicle crashes could be handled. The first would be to exclude all single-vehicle motorcycle crashes. Because a high percentage of total motorcycle crashes are single-vehicle (Shankar, 2001), this would result in a large reduction in the number of cases. A second technique would be to try to find a way to separate single-vehicle crashes into rider-culpable and rider-nonculpable cases. To our knowledge this system would have to be developed, because no system to identify nonculpable riders in single-vehicle crashes currently exists. A third system, and one that has been used before, would consider all riders in single-vehicle crashes culpable, the idea being that, in the absence of another motorist, the rider would be the only person who could be culpable. This ignores the possibility that another motorist may have caused the rider to crash and then left the scene. The crashed rider may be unable to report, or police may be unable to verify, the influence of the other motorist.
Another way to determine culpability would be to use data that contains police determinations of culpability (acknowledging that police determinations seek culpability from a legal perspective) or to have research staff determine culpability from crash reports. FARS does not have this information, though it may be possible to identify cases of interest through FARS and go back to Traffic Collision Reports (TCRs) in the States’ files to obtain enough information to determine culpability. At the panel meeting, mention was made of work done by Terhune (1983) that dealt with “Responsibility Analysis” – in which objective criteria were developed to assign culpability in a crash. A team of raters would assign culpability based on reading crash reports. The team would first establish inter-rater reliability by rating the same cases and comparing results. After a sufficiently high level of reliability had been established, the raters could split the remaining cases between them (Terhune, 1992).
There may be an advantage to doing a preliminary study to determine what demographic and driver-history differences exist between culpable and nonculpable samples. If the samples differ for age, gender, previous violation history, and other factors determined to be correlated with crash likelihood, this would indicate that any differences in BACs between them cannot solely explain differences in culpability.
The sole advantage to this methodology is that it could be conducted entirely using existing data. It is, nevertheless, a very significant advantage.
This methodology would require crashes in which motorcycles crash into each other (probably not feasible) or would compare one motorcycle/auto crash with another (which loses the advantage of having culpable and nonculpable participants matched for time and place).
It is likely that alcohol increases rider’s risk of being involved in a crash in which the rider is deemed nonculpable. For example, an impaired rider may be less able to avoid a collision with a car that pulls into his path. To the extent that this is true, it would tend to mask the effects of alcohol’s contribution to crashes of both culpable and nonculpable riders. Subsequent comparison of BACs of “crashing” (culpable) and “comparison” (nonculpable) riders would result in an underestimation of alcohol’s contribution to the crash.
A problem with using single-vehicle off-road crashes as indicators of motorcyclist culpability is that it is possible that another vehicle could have forced the motorcycle off the road.
While it may be possible to determine whether the crash and comparison riders are similar with respect to age, gender, and other demographic variables that could be related to crash risk, there are other variables, such as miles ridden per year, that could be different between the two samples, which would affect crash likelihood, and which could not be determined from the data. Another potential problem is that BAC measures are not taken in all cases and, when they are taken, are often taken for the purpose of convictions. Therefore, this dataset may be skewed.
The fact that an Induced Exposure study can be conducted using entirely archival data makes it relatively inexpensive to do. However, such a study relies heavily upon the accuracy and reliability of the data collected within the TCR and the FARS. Any extra effort expended to judge the culpability of riders in crash records will increase the costs of performing the study, but probably not so much that it would put it in a higher cost category.