Discussion

The data presented here confirmed that older drivers are more vulnerable to the traumatic effects of crashes. We found a positive relationship between age and fatal injuries, after controlling for gender, BAC, restraint use, crash direction, traveling speed, and vehicle wheelbase and model year. Whereas the odds of a fatal injury among young drivers (<30 years) were less than 80% that of reference drivers aged 40–49, the odds of a fatal injury for an 80+ driver were five times that of drivers aged 40–49. These results confirmed and expanded on the findings of others by controlling for driver, crash, and vehicle characteristics.

Wisconsin data showed that adults aged 85 years and over were at three times greater risk to be hospitalized or fatally injured in a crash compared with adults aged 16–64 . Our unadjusted OR for drivers aged 80+ was similar at 3.35, but it climbed to 4.98 after controlling for other variables. This difference attests to the importance of controlling for confounding variables. These statistics are alarming given that increasing numbers of older adults will be driving in coming years.

Partner Links:
* Nutsforcars.com: Offers information about upcoming auto shows, lowrider car clubs, as well as Import, Custom, and Cadillac car clubs throughout the US.
* Wrapvehicles.com: Auto wrap company offering vehicle wraps including, cars, trucks, vans, buses, and trailers.

Consistently with others, data on gender showed an overall 50% increased risk for fatal injuries in women compared with men. However, in both this study and a previous one, this effect appeared mostly limited to younger women. Proportionally fewer women among older age groups in this study were fatally injured compared with men of the same age, results once again consistent with the literature. The reasons for this age by gender interaction are not clear as our findings were obtained after controlling for driver and important crash and vehicle characteristics. Possibly, overall health, which cannot be determined with the FARS database, may explain gender differences. As women typically live 6 years longer than men, it is possible that older men are less healthy than women of the same age.

Anecdotally, alcohol is believed to exert a protective effect against fatalities; higher fatality rates among intoxicated drivers compared to sober ones are often attributed to the reckless driving behavior of drunk drivers. However, intoxicated drivers and passengers may be at increased risk of fatal injuries because of adverse physiological effects of alcohol on the body.

Our analyses on BAC and fatalities revealed a ‘U'-shaped relationship, supporting both possibilities. The odds of a fatal injury were 50% lower among drivers with alcohol concentrations at or below the typical legal limits (<0.10) than among sober drivers. Hence, anecdotal support for a protective effect of alcohol may be real. However, at higher BAC (>0.19) alcohol had a detrimental effect on fatality risk. At a BAC of 0.30 or greater, drivers were three times more at risk of a fatal injury than sober drivers. Because the analyses controlled for other driver behavior and crash variables, the lethal effect of high alcohol concentrations possibly represents a weakened physiological response to the crash trauma as suggested by others. However, not all drivers were tested for alcohol. Survivors who did not appear impaired may not have been tested, thus possibly biasing these results.

Waller et al. also studied the impact of BAC on fatality risk. Using three categories (BAC=0.10–0.14, 0.15–0.19, and =0.20), they found the lowest injury rates for the 0.15–0.19 category, higher rates for the 0.10–0.14 category, and much higher rates for the =0.20 category. Further epidemiological studies, and studies using animal models are desirable to understand the physiological mechanism underlying the effect of alcohol on the human body's ability to survive serious trauma.

In spite of the results presented here, alcohol use continues to be a great source of excess traffic fatalities. Although drivers with low BAC may be less at risk to die in a crash, once it has occurred they are more at risk of committing driving errors that may lead to fatal crashes. We found in a recent study that a BAC greater than zero but less than 0.05 was associated with a 45% increased OR of making a driving error compared with sober drivers. When the BAC exceeded 0.20, the OR of a driving error increased by more than 700%. Thus, alcohol consumption is not protective. A high BAC is doubly lethal because both the risk of a driving error and fatal injury are highly elevated compared with sober drivers.

Restraint use (manual and automatic seatbelts), has been promoted as the most important behavior to prevent fatalities. Evans proposed that more than 40% of all fatalities incurred by non-belted drivers and passengers could be avoided with seatbelts. At the very least, researchers estimated that we could expect a reduction of 15% in all severe and fatal injuries if all drivers and passengers were belted, even after controlling for age and vehicle deformity.

Consistently with other reports, our analyses on restraint data showed an unequivocal protective effect of three-point belts. Wearing a seatbelt halved the risk of a fatal injuries in the dataset analyzed. Two-point belts had OR in the expected direction but sample sizes were too small to provide precise estimates. However, most vehicles on the roads at the present time are equipped with three-point belts.

Given the strength of the association between restraint use and non-fatal injuries, a causal link is highly probable. Furthermore, there was a gradient (dose-related response) between restraint use and fatal injuries, such that the risk of a fatal injury decreased from no restraint to two-point restraints to three-point restraints. These facts reinforce the belief that seatbelts save lives. Even assuming a 14% over-reporting of belt use among survivors, the odds of a fatality would have been 23% lower among our sample of belted drivers compared with non-belted drivers. However, the protective effect of belts may not generalize well to all crash conditions outside those studied here. The majority of the crashes studied involved front impacts, for which belts are best suited ; seatbelts are reportedly less effective in struck vehicles than striking vehicles.

This problem is especially relevant for older drivers because they are more likely to be struck by incoming vehicles at intersections than other age groups. The fatality risk of struck drivers is five times that of the striking drivers when hit on the right side, and ten times more when struck on the left side. However, our data did not support such a difference across points of impacts. Drivers hitting a fixed object on the left side had twice the odds of a fatal injury compared with drivers who experienced front impacts. However, our reliance on single-vehicle crashes and control for variables such as age may explain this difference.

Air bag data did not show a protective effect using the current paradigm. This result is surprising given that air bags are expected to perform well in front impacts. Other researchers reported reductions in fatalities ranging between 16 and 28%. However, their analyses provided less control for driver, crash, and vehicle characteristics, and early models equipped with air bags may have been luxurious, large vehicles. It is also unclear what additional benefit, if any, air bags provided to seatbelts, and whether specific groups of drivers may have been adversely affected by air bags; some data point to a fatality risk for shorter drivers. Nonetheless, air bag data were available for fewer observations than other variables, warranting caution in the interpretation of this finding. Further studies are required to evaluate the effect of air bags.

Data on speed as a fatality risk factor are less contentious. According to a Canadian study, the odds of a fatality in crashes that occurred in 70–90 kph zones were almost six times those of crashes occurring in zones with slower posted speed limits. Although these data did not rely on the speed of vehicles, it is unlikely that this limitation would have biased the effect considerably. In our regression model, traveling at a speed of 112 kph (70 mph) or more was independently associated with a 164% increase in the odds of a fatality compared with speeds of less than 56 kph (35 mph). Although our results are not as dramatic as those presented by Zhang and colleagues, they attest to the safety issue related to speed. Dischinger found that larger decelerations (i.e. larger traveling speeds at impact) were associated with more post-injury medical complications, independent of age and injury severity. Because the impact of speed may be related to trauma sustained with the sudden deceleration experienced, future research would benefit from the inclusion of traveling speed indices.

Possibly the vehicle attribute most closely related to injury severity is size. Although weight and length are highly correlated, weight is the most studied characteristic. Evans reported that in any two-vehicle crash, the passengers of the heavier vehicle fared better than those of the lighter vehicle. Yet, drivers of two 900 kg (2000 lbs) vehicles were at twice the risk of fatal injuries compared with drivers of two 1800 kg (4000 lbs) vehicles reported that every 454 kg (1000 lbs) increase in vehicle weight was equivalent to the driver's ability to withstand front impact crashes of 10 more kph (6 mph) before being fatally injured. Using wheelbase length, we substantiated these findings after controlling for other critical variables. A 25 cm (10 inches) increase in wheelbase translated into a 10% reduction in the odds of a fatality. This finding supports the protective value of larger vehicles independent of their drivers.

On the other hand, our findings regarding model year are not consistent with others. We found that recent model year vehicles were associated with an increased risk of fatalities of 5% for each 5 years. Others reported that recent models were safer and others that there was no relationship. Furthermore, whereas the present analyses did not find an association between vehicle age and fatalities, others reported such an association. Discrepancies between these studies are likely explained by crucial methodological differences and samples. Nonetheless, the model year effect we reported was extremely small. We suggest caution in the interpretation of this finding.

Possibly the main limitation of this study is its generalizability to all types of crashes. To answer our research question, we focused the analyses on a very controlled set of observations. Although we believe this strengthens inferences that can be made, these are limited to the type of crashes studied. Further studies will be required to determine whether our findings also apply to other crash situations (e.g. multi-vehicle crashes).