Prior research has established that there is a significant alcohol-crash problem in the United States. At the millennium, we need to know the size and nature of that problem and its etiology. This chapter is concerned with the general magnitude of the alcohol-crash problem nationwide as can be estimated from (1) studies of the crash involvement of drinking drivers, pedestrians, and bicyclists; and (2) studies that have examined the involvement of such persons relative to that of others. The alcohol-crash problem as it involves specific groups is discussed further in Chapter 4.
Fatal Crashes. The 1978 report noted the lack of any national study methodically investigating the many variables that describe highway crashes, and had to piece together bits of information from separate studies never intended for global application. Despite this, the report found a "remarkably consistent" picture. Some 40% to 55% of all fatally injured drivers in the studies had a blood alcohol concentration (BAC) of .10 or more. Of the drivers killed in single-vehicle crashes, 55% to 65% had a BAC of at least .10.
By the time the next update was completed (U.S. Department of Transportation NHTSA, 1985), initial data from NHTSA's Fatality Analysis Reporting System (FARS) had become available. These data came from 14 to 17 states that had the "most complete" BAC data on fatally injured drivers in the years 1980, 1981, and 1982. They showed that about 50% of drivers in fatal crashes had a BAC of at least .10; this result was "essentially in accord" with that from the 1978 update where the midpoint for this group of drivers was 47%.
More useful data from FARS became available for use in the next update which was published in 1989 (Jones and Lacey, 1989). The newly available data revealed a steady decline in the percentage of drivers with BACs exceeding .10 since 1980. The decline amounted to about 11 percentage points in absolute terms, or a relative percentage of 22%. A similar decline was reflected in NHTSA 's expansion of the data from 15 states that measured the BAC of a high percentage of fatally injured drivers to all states1, and there was also a decline in the percentage of fatally injured drivers at the higher BACs. The next update (Jones and Lacey, 1998a) reported a continuing downward trend in the percentage of drivers in fatal crashes who had been drinking, both for drivers with any measurable amount of alcohol in their blood (.01+) as well as those at or exceeding most states' illegal limit2 for blood alcohol concentration (BAC) of .10. For the latter group, this percentage had gone from 30% to just under 19%, a decrease of 37%.
The latest FARS data available for this review (U.S. Department of Transportation NHTSA, 1999) adds another year (1998) to the alcohol-related fatal crash time series. Figure 2-1 plots annual number of fatalities in fatal crashes involving a driver with .01+ BAC and with .10+ BAC. The continuing downward trend for both measures over the entire range of years is apparent, but there is also a distinct flattening out of the .01+ series starting in about 1992. The same effects are present in series for the alcohol-related fatalities as a percent of all fatalities (Figure 2-2) and for the fatality rate per 100,000 population (Figure 2-3).
Figure 2-1: Alcohol-Related
Fatalities with a Driver
at BAC .01+ and .10+, 1982-1998
Figure 2-2: Alcohol-Related
Fatalities as a Percentage
of All Fatalities, 1982-1998
Figure 2-3: Alcohol-Related Fatality
per 100,000 Population, 1982-1998
The 1999 FARS report of 1998 data also shows that 39% of all fatal crashes involved a driver or non-occupant with a BAC of .01+ but did not report the percentage of all fatal crashes that involved a driver or non-occupant with a BAC of .10+. However, analysis of the FARS data for 1998 by the authors of this report indicated that about 30% of the fatal crashes involved at least one driver or non-occupant with a BAC of .10+. This is about 60% of the percentage estimated in the 1978 and 1985 state of knowledge reports.
We also examined 1998 FARS data on the BACs of fatally injured drivers. We used the data from the 15 "good reporting states" that have been studied in past examinations of the alcohol-crash problem (Fell and Nash, 1989). The 15 states were selected for these studies because the states measured the BACs of at least 80% of fatally injured drivers. Figure 2-4 shows the percentage of the fatally injured drivers that exceeded various BACs, and includes a plot of the midpoints of data from various early studies circa 1970 analyzed in the 1978 update. Three points from NHTSA's 1999 FARS report of 1998 data are included on the graph.
Figure 2-4: Percentage of Fatally
Exceeding Specified BAC by Data Source
Figure 2-4 is interesting in several ways. First, it illustrates very clearly the large reduction during the past 30 years in the percentage of fatally injured drivers at all BAC levels. Percentage reductions from 1970 were in 30%-35% range for all BACs except the very highest, which appear somewhat lower. Second, the percentages from the FARS report (which used NHTSA's algorithm for filling in missing data) are a bit lower than those from the fifteen "good reporting states." This suggests that the percentages of all fatally injured drivers at given BACs in the 15 states really were lower than those for the U.S. as a whole, or that the missing data from the 15 states were from drivers with low BACs.
Figure 2-5 shows the BAC distribution of the fatally injured drivers with non-zero BACs. The largest percentage of such drivers (24%) had BACs in the .15 - .19 range. Only 10% were in the .01 - .04 range, and 11% were in the .05 - .09 range. Seventy-nine percent had BACs of .10+, and 63% had BACs of .15+. Other researchers (Simpson and Mayhew, 1991; Simpson and Mayhew, 1992) have reported similar results for high-BAC drivers in earlier examinations of FARS data, labeling such drivers as "hard-core" drinking drivers.
Figure 2-5: BAC Distribution of
Fatally Injured Drivers
in 15 States, 1998
Non-Fatal Crashes. The 1978 report reviewed four U.S. studies dating back to 1938, concluding that 9-13% of drivers in injury crashes and about 5% of drivers in property damage crashes had a BAC of .10+. The 1985 report estimated that, circa 1980, 18% of injury crashes and 5% of property damage crashes involved drivers with a BAC of .10+.
Recently, non-fatal crash data from NHTSA's General Estimates System (GES) became available. The GES contains data obtained from a national probability sample of traffic crashes for which a police accident report (PAR) was prepared. About 50,000 such PARs are collected and coded each year by NHTSA. The GES includes data on the investigating officers's judgment of whether alcohol was involved in the crash. The latest reported data from GES (U.S. Department of Transportation NHTSA, 1999) indicate that 9% of the injury crashes and 5% of property-damage-only crashes involved alcohol in the officers's judgment. These percentages are remarkably close to those in the1978 state of knowledge report quoted above; however, the 1978 estimates were for crashes involving drivers with a BAC of .10+.
Objective measurement of the drinking behavior of drivers using the roads who are not involved in crashes has been examined in a number of studies. Some of these have been studies of the relative risk of a crash. Relative risk is defined here as the probability of a crash involving a driver at a given non-zero BAC divided by the probability of a crash at a BAC of zero. Citing an earlier article by Hurst, the 1978 update showed that relative risk defined in this way can be calculated by dividing the percentage of crash-involved drivers at a given non-zero BAC by the percentage of crash-involved drivers at zero BAC, and then dividing that number by the percentage of non-crash involved drivers at the given BAC divided by the percentage of non-crash involved drivers at zero BAC. Thus, the percentage of non-crash involved drivers at any given BAC was a by-product of the relative-risk studies.
The 1978 update presented data from five such studies dating back to 1938, indicating that some 1%-3% of the non-crash involved drivers had a BAC of .10+. All but the 1938 study selected the non-crash involved drivers from drivers of vehicles traveling at the same times and places as those of the crashed vehicles. The "roadside survey" method was used for collecting the data in all of these studies of non-crash involved drivers. This method involves stopping drivers selected from the traffic stream. The BACs of those agreeing to cooperate are measured, and they are asked to respond to a short survey.
In the early 1970s, more than 100 roadside surveys were performed to help evaluate the effectiveness of the 35 sites that participated in NHTSA's Alcohol Safety Action Projects (ASAP). For the most part, these were conducted during nighttime hours during weekends. In 1973, the University of Michigan conducted a NHTSA-sponsored nationwide survey of 24 cities and counties with populations over 20,000. Its data were collected on Friday and Saturday nights between 10:00 p.m. and 3:00 a.m. These studies found that 5%-6% of these nighttime weekend drivers had a BAC of .10+, compared to the 1%-3% percent of the drivers sampled around the clock in the relative risk studies.
Since 1973, two other nationwide roadside surveys have been conducted, the first in 1986 and the second in 1996. The first survey (Lund and Wolfe, 1989) was conducted by Mid-America Research Institute under the sponsorship of the Insurance Institute for Highway Safety. Its locations were designed to match the 1973 locations as closely as possible. It found that 3% of the drivers had a BAC of .10+, compared to 5% in the 1973 survey, a reduction of about 40%. The reduction in the percentage of drivers with a BAC of .05+ was also nearly 40%.
The 1996 survey was conducted at locations chosen to produce data that would be comparable to that from the 1986 survey and was carefully designed to do so (Voas, Wells, Lestina et al., 2000). It found that the percentage of drivers at .10+ to be 3% (all percentages rounded to the nearest integer), and the percentage at .05+ to be 8%. Both figures were unchanged from 1986. However the percentage of drivers in the .01-.05 range changed considerably, from 18% to just 9%.
In recent years there have been several nationwide telephone surveys querying a sample of drivers in general about their drinking-driving and related topics. These surveys provide another perspective on the alcohol-crash problem. NHTSA has sponsored such a survey every two years since 1991. Respondents have been a nationally representative sample of persons age 16 or older.
The latest NHTSA survey was conducted in 1999, but its results were not available at this writing. Balmforth summarized the results of the 1997 survey (n=4,010) and compared some of its results with those from prior surveys (Balmforth, 1998). She found that 24% of the 1997 respondents said they had driven within two hours after consuming alcoholic beverages at least once during past year. These individuals were defined as "drinking-drivers" and, on average, consumed 2.5 drinks prior to driving. The drinking-drivers also said they made an average of 1.7 drinking-driving trips in the past 30 days. The report by Balmforth also estimated the BACs of the drinking-drivers, finding that 13% had a BAC of .05+ and that 5% had a BAC of .08+.
In comparing the 1997 results with those from prior surveys, data from respondents of age 16-64 were used. The percentage of drinking-drivers in this group changed very little in the four surveys, amounting to 28% in the first two surveys and 24% and 25%, respectively, in the last two surveys. With respect to number of self-reported drinking-driving trips made in the past 30 days, there was a significant decline during the first three surveys (from 2.3 to 1.5), followed by a small increase to 1.6 in the 1997 survey3. No comparisons of BACs among the four surveys were included in the report.
The National Household Survey on Drug Abuse (NHSDA) conducted by the Federal Government is another source of self-reported information on drinking and driving in the United States. This survey has been conducted periodically since 1971, and the 1996 survey contained a special Driving Behaviors Module funded by NHTSA. A summary of the design and findings of the 1996 survey drawn from this module is contained in a government report (Townsend, Lane, Dewa et al., 1998).
The Driving Behaviors Module involved 11,847 personal interviews in a nationally representative sample of households. The respondents were individuals age 16 and older reporting that they had driven a motor vehicle in past 12 months, and whether they had driven within two hours after drug or alcohol use. The findings relative to alcohol did not differ greatly from those of the NHTSA survey discussed above. Twenty-seven percent of the respondents reported that they had driven within two hours of alcohol use, including 4% who had used both alcohol and other drugs. The report also presented estimated BACs of those who had driven within two hours after alcohol use, indicating that 8% had a BAC of .08+ and that 30% had BACs in the .02 - .079 range.
One way of assessing the role of alcohol in causing a traffic crash under a given set of conditions is to estimate the relative risk of a crash, where relative risk is defined as in the above discussion of non-crash involved drivers. This approach recognizes that crashes are probabilistic events and that one must express the role of alcohol in probabilistic terms. Prior studies of relative risk have relied on the case-control method to obtain data for estimating relative risk. The term derives from the study of diseases, and MacMahon and Pugh (1970) describe the method as:
" . . . an inquiry in which groups of individuals are selected in terms of whether they do (the cases) or do not (the controls) have the disease of which the etiology is to be studied, and the groups are then compared with respect to existing or past characteristics judged to be of possible relevance to the disease." (p. 241)
Most important, MacMahon and Pugh point out that such studies are better described as case-comparison studies since they do not incorporate the type of control that may be obtained in experimental studies that are conducted in the laboratory. This distinction has to be kept in mind when interpreting the results of so-called case-control studies in the field of traffic safety. Note that the case-control method has many variants and does not require pair-wise matching of cases and controls.
Using data from several studies of relative risk, the 1978 update concluded that crash risk increases as driver BAC increases. The relative probability of a crash was found to begin to increase "precipitously" as the driver's BAC approached .08. At a BAC of .10, the probability of a fatal or serious-injury crash was estimated to be 6 to 12 times that of a driver with no alcohol. The relative probability of a fatal crash was said to be much higher at higher BACs, over 20 at a BAC of .15.
Prior reviews found no rigorous studies of relative risk nationwide, nor has this review found any such study that was designed a priori to measure relative risk nationwide. However, a very careful "case-control" analysis of relative risk was performed recently using 1995 and 1996 data from FARS for the "case" component and data from the 1996 roadside survey for the "control" component (Zador, Krawchuk, and Voas, 2000).
The Zador study matched the FARS cases to the roadside survey cases as closely as was possible under the constraints of the study. Two notable exceptions were made to help increase the sample size of the FARS drivers, (1) retaining crashes that occurred during the midnight and 1:00 a.m. hours and (2) accepting crashes for the Friday and Saturday nights for the whole year rather than for just the period during which the roadside surveys were conducted. Drivers of four-wheeled passenger vehicles who were in the FARS group were classified by the number of crash-involved vehicles (one, two, and any number of vehicles). Both fatally injured drivers and drivers involved in fatal crashes were studied for each classification of number of crash-involved vehicles, resulting in a total of six groups. Relative risk was calculated for each group as a function of driver age, sex, and BAC. In calculating relative risk, Zador used the odds of a crash at a given BAC relative to the odds of a non-crash at the same BAC, the resulting unadjusted odds ratio being equal to the relative risk calculated as described above.
Figure 2-6 shows the results of the analysis for two groups of drivers of age 21 and over, (1) drivers involved in single-vehicle fatal crashes and (2) drivers involved in all fatal crashes. Risk curves of both groups are roughly the same for both groups up to a BAC of about .05, and then start to diverge. At BACs in the .080-.099 range, single-vehicle drivers in fatal crashes had a relative risk of about nine, compared to a relative risk of about six for drivers of all vehicles.
Figure 2-6: Relative Risk of Fatal
Crash Involvement for Drivers
Age 21 and Over (Zador, et al., 2000)
Of interest is the reduced relative of risk below 1.0 (≈.2) at BACs in the .001-.019 range, implying a lower risk than at BAC=0. This is remindful of the infamous "Grand Rapids Dip" (noted in the 1978 update) found in the well-known 1963 case-control study, but since discredited as being due to disproportionate representation of demographic subgroups in different blood alcohol concentration class intervals (Hurst, Harte, and Frith, 1994). Hurst and associates found no such dip after applying a statistical model to the data that accounted for such differences, finding that relative risk increased monotonically with BAC regardless of self-reported drinking frequency. (An earlier hypothesis for the dip was that the persons at low BACs were more often higher frequency drinkers who, for some reason, were safer drivers at low BACs.)
Relative risk did not vary by driver sex for these two groups of age 21+ drivers. Relative risk rose to over 80 (not shown on the graph) at BACs of .15+, but with large variances, apparently due to the small number of drivers at these BACs. The relative risk of drivers under the age of 21 also had high variances at the higher BACs and is discussed later in Chapter 3.
We found no new studies of the relative risk of non-fatal crashes. The state of knowledge in this area is essentially unchanged from that reported in prior updates. Data from the 1960s and1970s indicate increasing relative risk as BAC increases, but much lower levels of risk at any BAC (in the 2 to 4 range for all but the very high BACs) than for fatal crashes.
The 1978 report found a paucity of studies on the magnitude of the alcohol-pedestrian safety problem, but nevertheless ventured an estimate that about one-third of all fatally injured pedestrians had a BAC of .10 or more at the time of their death.
One interesting finding discussed in the 1985 report (from a study in New Orleans by Blomberg, Preusser, Hale, et al.) was that the relative risk of involvement in a fatal pedestrian crash did not begin to rise until the pedestrians reached a BAC of .15 to .20. This is consistent with the hypothesis that safe walking is generally easier than safe driving, since the relative risk curve for fatal motor vehicle crashes starts to rise at a much lower BAC.
FARS data indicate that the pedestrian component of the alcohol-crash problem has also been decreasing since 1982, but with some flattening out in recent years Figure 2-7. There were 855 fewer fatalities at BACs of .10+ in 1998 than in 1982, a percentage decrease of 36%.
Figure 2-7: Number of Pedestrial
Alcohol-Related Crashes at Two BAC Levels by Year
However, there was a much smaller decrease in the percentage of pedestrian crashes that were alcohol-related Figure 2-8, amounting to about 16% from 1982 to 1998 for those with a BAC of .10+.
Figure 2-8: Percentage of Pedestrian
Alcohol-Related Crashes at Two BAC Levels by Year
The latest FARS report did not present data on the role of alcohol in bicycle crashes5. Li, Baker, Sterling, et al. (1996) analyzed medical examiner data on all fatally injured bicyclists aged 10 years or older from 1987 to 1994 in Maryland (fatal cases, n = 63) and compared the data with trauma registry data on all injured bicyclists who were treated at a regional trauma center during the same time period (nonfatal cases, n = 253). Variables studied were those related to BAC, demographic characteristics, and injury circumstances. The researchers found that fatal cases were more likely than the nonfatal cases to have positive BACs (30% vs. 16%, p < .01) and to have a BAC of .10+ (22% vs. 13%, p < .01).
Applying the percent of bicyclists with BACs of .10+ to nationwide fatality data from FARS and injury data from NHTSA's General Estimates System for 1998, provides a rough order of magnitude estimate of the size of the alcohol-bicycle crash problem in the U.S., amounting to some 200 fatalities and 7,000 injuries. Li and associates also found that bicyclists who died at the scene were four times as likely as those who died at hospitals to be legally intoxicated (35% vs. 9%, p < .02). Given a serious bicycling injury, intoxication was associated with significantly increased likelihood of fatality, with an adjusted odds ratio of 2.8 (95% confidence interval, 1.3 to 6.3).
Finally, we note a Finnish case-control study that estimated the relative risk of and alcohol-related bicycle crash (Olkkonen, 1993). The study involved 200 bicycle victims who were injured fatally in road traffic accidents during the years 1982-1988, and 700 cyclists who were used as unmatched controls for these cases. The study found that alcohol was involved in 25% of the collision accidents and in 63% of the single accidents involving cyclists aged 15 to 64 years and whose blood alcohol was measured. Only 4% of the controls were under the influence of alcohol. A relative risk was of the order of 3 over all, and 58 for the collisions related to alcohol use.
Not all of the alcohol-related fatal pedestrian and bicyclist crashes involved drinking walkers or riders. FARS 1998 data indicate that 11% of over 5,000 fatal pedestrian crashes occurring in 1998 involved a driver with a BAC of .10+, and that nearly half of the 11% involved a pedestrian at zero BAC (Table 2-1).
|No Driver Alcohol Involvement||Driver Alcohol Involvement BAC .01-0.09||Driver Alcohol Involvement BAC .10 or Greater||Total|
|No Pedestrian Alcohol Involvement||55%||3%||5%||3,264|
|Pedestrian Alcohol Involvement, BAC .01–.09||5%||1%||1%||324|
|Pedestrian Alcohol Involvement, BAC .10 or Greater||23%||3%||5%||1,573|
At the millennium, 1998 data indicate that about 12,500 fatalities in the U.S. traffic crashes involve a driver with a BAC of .10+. This amounts to about 30% of the 41,4716 traffic crash fatalities that occurred in 1998. Further, an estimated 30% of all fatal crashes involve a driver or non-occupant with a BAC of .10+, and also, about 30% of all fatally injured drivers have a BAC of .10+. Finally, the fatality rate of persons in crashes with a driver at .10+ was 4.6 per 100,000 population in 1998.
Crashes involving pedestrians and bicyclists are not as well defined as those involving only motor vehicles. In 1998, over 1,500 fatally-injured pedestrians had a BAC of .10+, a decrease of about 36% since 1982. The percentage of fatally-injured pedestrians has also declined, but by a much smaller percent (about 15%). In addition, an estimated 5% of fatal pedestrian crashes involved a driver with a BAC of.10+ and a pedestrian of zero BAC. The precise number of fatally-injured bicyclists nationwide at BAC .10+ is not known, but is estimated to be in the hundreds.
All of these measures of the alcohol-crash problem have declined since objective data on the problem became available. The fatality rate, an especially important measure since it accounts for population growth, has declined nearly 50% since 1982. Nevertheless, there has been a distinct "flattening out" of these measures in recent years, suggesting that the problem needs increased emphasis to maintain the overall downward trend.
The situation with respect to non-fatal crashes involving alcohol is less clear. In recent years, NHTSA's General Estimates System (GES) has helped clarify this component of the alcohol-crash problem, providing data on the investigating officers's judgment of whether alcohol was involved in the crash. Data from GES (U.S. Department of Transportation, NHTSA, 1999) indicate that 9% of the injury crashes and 5% of property-damage-only crashes involved alcohol by this criterion.
Obviously, not all drinking-drivers become involved in alcohol-related traffic crashes. Surveys of driving behavior help define the percentage of non-crashed drinking drivers using the roads at given times. Roadside surveys suggest that some 3% of drivers on the road during nighttime hours during weekends have a BAC of .10+, a somewhat higher percentage than found in earlier studies conducted around the clock during all days of the week. Other surveys of drivers in general have asked questions about their respondents's driving after drinking. NHTSA's 1997 nationwide telephone surveys found that 24% of the 1997 respondents said they had driven within two hours after consuming alcoholic beverages at least once during the past year, and it was estimated that 5% of these had a BAC of .08+. A 1996 survey, involving personal interviews in households, obtained similar results, estimating that 8% had a BAC of .08+.
The above estimates have used a BAC level of .10+ in defining the general magnitude of the alcohol-crash problem in the United States. This is because the fatal-crash risk is so high at that level as to become societally unacceptable. This has resulted in extensive societal pressures (including criminal sanctions) to prohibit driving at a BAC of .10+, and at this writing, even at .08+ in 29 states and the District of Columbia and Puerto Rico. The greatly increased risk at BAC= .10+ over that at BAC=.00 is evident simply by comparing the percentage of .10+ drivers in fatal crashes (about 30%) to the percentage of .10+ non-crashed drivers on the road (roughly 3%). This suggests alcohol-crash over-involvement by a factor of ten7.
This is, of course, only a rough estimate. More careful analyses of the relative risk of a fatal crash due to alcohol have been performed over a period of more than 40 years, attempting to account for factors other than alcohol that can influence risk estimates. The latest estimate of relative risk for drivers of age 21 and higher of is of the same order of magnitude as the rough estimate, and also indicates significant risk at lower BACs. In addition, one study found that fatally injured drivers with BACs of .10+ were about 50% more likely to have been responsible for their crash than were drivers at zero BAC8.
Other levels of risk could be used for defining the size of the alcohol-crash problem. For example, defining acceptable risk at the .01+ BAC level would add another 4,300 fatalities to the estimate of approximately 12,500 indicated on page 7, while defining it at the .20+ level would subtract 6,300.
In closing this chapter, we repeat a comment from the 1978 update of the state of knowledge of alcohol and highway safety:
"The above figures, while indicative of a large-scale national problem, do not, of course, prove that alcohol caused the crashes in which drinking was involved. Traffic accidents are probabilistic, with many factors entering into the probability equation. The most that can be said on the basis of epidemiologic evidence is that, on the average, alcohol beyond a certain amount, is associated with increased crash risk." (p. 32)
Thus, the magnitude of the alcohol crash problem at the millennium depends on the level of alcohol-crash risk an informed public is willing to tolerate, given available alternatives to reducing that risk. In any case, the inescapable conclusion is that alcohol-related crashes are a much smaller societal problem at the millennium than they were 20, or even 10, years ago.