A methodology using fatal crash records would compare BACs found in motorcyclists involved in fatal crashes to BACs found in the public-at-large. The fatal crash data would come from FARS. Because of the relative infrequency of fatal motorcycle crashes, several years of data would be used. Also, because collecting comparison data may need to be done in certain locations where a sufficient number of motorcyclists can be found, consideration needs to be given to restricting FARS cases to those same areas. As with the Injury Crash Records methodology, this methodology provides data only on crashes, so consideration must be given to the issues, advantages, and disadvantages that apply to the method used to collect comparison data. Thought also would need to be given to whether to include riders involved in crashes where the rider was not fatally injured (e.g., passenger or pedestrian was fatally injured).
In general, fatal crashes are the best documented crashes, and FARS data are relatively complete with respect to BAC data. For cases in which there is no BAC, it is possible to impute BACs. In the past, this has been accomplished using the Klein (1986) imputation procedure. Currently, the Subramanian (2002) imputation method is being used. In the opinion of some, fatal crashes are the most important to prevent. To be able to say "chances of being killed in a crash are increased x percent at a given BAC" would be more persuasive for these people than to talk about injuries. For this reason, study results that are generalizable to fatal crashes may be more valuable than alternativesfor such audiences.
Data are limited to fatal crashes only. There may be a gap in time of several years between a crash case and the comparison case. Because there are relatively few fatal crashes, the sample size would be much smaller than for using injury cases from an equivalent time period. Where BACs are missing, there is likely to be a systematic reason (rather than missing BACs randomly). This may reduce validity of findings. As with the Injury Crash Records methodology, this methodology only provides part of the data needed. It would be necessary to collect the comparison data, which is a considerable task. There is a possibility that confounding variables linked to alcohol use might obscure the correlation between BACs and crash risk. For example, drinking riders have been shown to be less likely to wear helmets. This affects their likelihood of being included in the FARS dataset. There is also some evidence that alcohol itself may increase the likelihood of dying for a given level of injury (Waller et al., 1986; Luna et al., 1984).