Much of the progress that has been made in impaired driving in the last decade or more has been facilitated by lessons learned from other countries. For example, the United States drew valuable lessons regarding deterrence from analyzing the results of the British Road Safety Act of 1967. Similarly, we have learned about alcohol policy and serious enforcement and penalties from some of the Scandinavian countries. The Australian experience with random breath testing has influenced some of our own enforcement efforts. It is therefore both timely and appropriate for the National Highway Traffic Safety Administration to sponsor a systematic effort to gather information about impaired driving in countries around the world. In particular, this report summarizes information gathered on the measurement of alcohol involvement in countries around the world. This report is a follow-up to a previous phase of this effort. The report for that phase is entitled Literature Review of DWI Laws in Other Countries (NHTSA 2000). That project collected information on laws and policies related to drinking and driving in countries around the world. The intent of this entire effort is to contribute to our understanding of impaired driving countermeasures and their impact and of how the current situation in the United States compares to other countries.
Previous work has pointed out the difficulties of making comparisons internationally. Ross (1993) compared roadside surveys and fatal crash studies in various countries, but noted the methodological and definitional difficulties. Voas (1993) pointed out that while national crash record systems appear to provide an attractive method of comparing the effectiveness of traffic safety programs, differences in definitions, data collection methods, and file structure impair the ability to make meaningful analyses.
Other researchers have examined the accuracy of national data systems. For example, Oestroem and colleagues (1993) compared police assessment of alcohol impairment with blood alcohol analysis on a large sample of driver fatalities in Sweden. Alcohol was detected in twice the proportion of fatally injured drivers as had been identified by police. The authors recommended that Sweden change its national traffic crash record system to reflect blood alcohol analysis rather than police assessment only. To date, the system has not been changed.
This report provides an updated analysis of these and other issues. The parameters of the report are described below.
The primary purpose of this project is to provide comparisons with the United States, and therefore possible guidance in the development and implementation of impaired driving policies in this country. Therefore, the main focus of data collection was on countries that would be considered most directly comparable to the United States, economically and demographically, as well as those countries with which we have the most direct dealings. These countries include:
Most of the information for this report was gathered through inquiries from key informants identified in each of the countries of interest. Most informants were from government transportation agencies. Some informants were from relevant university departments. In some cases, available information was collected from other published or unpublished sources. Some countries did not respond despite repeated requests for information. Appendix A indicates the source(s) of information for each country.
Several parameters enter into the measurement of alcohol involvement in fatal crashes. Appendix B displays these parameters for each of the countries in the study. Key parameters are described below.
Countries vary in their definition of what constitutes an alcohol-involved traffic crash. In some countries, a crash is defined as “alcohol-involved” if any alcohol is detected in a driver.1 In some countries, this definition is extended to include alcohol detected in a pedestrian involved in a crash. The National Highway Traffic Safety Administration (NHTSA) in the U.S. defines a fatal crash as alcohol-related if either a driver or a non-motorist had a measurable or estimated BAC of 0.01 g/dl or above. Other countries define a crash as alcohol-involved if a driver has a blood alcohol content (BAC) over the illegal limit for that country. Of course, illegal limits vary from country to country. In some cases, this definition is extended to include alcohol over the illegal limit for drivers in a pedestrian2 involved in a crash. That is, though there is not a legal BAC limit for pedestrians, if a pedestrian involved in a crash has a BAC over the legal limit for drivers (e.g., .08 or .05, depending on the country), the crash would be considered alcohol related.
In some countries, alcohol involvement is not defined by actual measurement of the presence of alcohol, but rather on police reports that alcohol was involved. NHTSA uses this definition for nonfatal crashes. These reports may be based on police suspicion due to a variety of cues or circumstances.
There is some variation in how fatalities are defined. Most countries define a time limit following a crash during which a death must occur in order for the fatality to be considered caused by the crash. Thirty days is the most common limit. France has a time limit of eight days and Canada's limit is 12 months. Some countries do not have a specific time limit. Obviously, the longer the time limit, the more deaths will be included, although it is unknown how significant the resulting differences would be.
In cases where blood tests are taken on drivers or pedestrians who are still living, the BAC changes as alcohol is digested and metabolized. Clearly, the amount of time that passes after a crash before the driver or pedestrian is tested will have an effect on the accuracy of the test in determining the extent to which alcohol impairment played a role in the crash. Countries surveyed do not report having specific time limits for testing.
Clearly, the accuracy of estimates of alcohol involvement in fatal crashes is dependent on the degree to which reports of alcohol involvement are free of bias. In general, the higher the proportion of drivers (and pedestrians) tested, the less bias will be present. When only a small proportion of drivers are tested, the chance that bias will be introduced is significant. Frequently, for example, living drivers involved in fatal crashes are difficult to test because conditions at the crash site are dangerous and the primary concern is safety and the swift treatment of injuries. Many surviving drivers are not tested because of these difficulties, because they are taken to the hospital, or for other logistical reasons. It is also the case that some countries may not commonly test fatally injured drivers because there is no legal reason for testing, since they are beyond legal reach. In some countries, testing occurs only when police suspect the presence of alcohol. In this case, certain types of drivers may be less likely to be tested (e.g., women or the elderly), thus introducing bias into the estimates of alcohol involvement.
Rates of testing vary significantly from country to country in this study. Some countries were unable to report testing rates. The lowest reported testing rate for drivers was Spain, which reported that 17.5 percent of drivers were tested in 1998. The highest rates were reported in Canada (83.1%) and France (approximately 90%). Approximately 63% of fatally injured drivers are tested in the United States. This rate, however, varies greatly from state to state.
In some countries, autopsies or post-mortems are routinely performed on a significant proportion of people killed in traffic crashes. In two of the countries studied (Norway and the United Kingdom), the results of these examinations are reported to traffic safety agencies and used in the calculation of alcohol involvement in fatal crashes. In other countries (e.g., Australia and Sweden), while autopsies are performed on 90% or more of fatally injured drivers or pedestrians, these results are not used in the calculation of official statistics.
Table 1 summarizes the reported rates of alcohol involvement in fatal crashes as well as several parameters for measurement. As can be seen from the table, reported alcohol involvement in fatal crashes varies widely. Five countries report alcohol involvement rates of less than 10 percent (based on either illegal alcohol levels or the detection of any alcohol). By contrast, five countries report alcohol involvement rates hovering between 27 and 41 percent. This variation does not fit any easily discernable pattern. It appears, however, that some of the lowest alcohol involvement figures are in countries with very low or unknown testing rates (i.e., Austria, Belgium), whereas some of the highest alcohol involvement figures are found in countries with relatively high testing rates (e.g., Australia, Canada, New Zealand, and the United States).
Notes for Table 1:
Reference to Column Heading 3 - Definition of Alcohol-Involved
Countries could select from the following categories:
More than one category may apply and may be reported in different circumstances.
|Percent of Alcohol Involvement
|Definition of Alcohol-Involved
|Percent of Drivers Tested
|Percent of Pedestrians Tested
|28% of all drivers and motorcyclists killed at .05 or higher; 37% of pedestrians 16 and older at .05 or higher (1997)
|75% of all drivers or motorcyclists involved in fatal crashes; 90% of those fatally injured
|88% of fatally injured pedestrians
|8.5% at .05 or higher (1998)
|unknown - no systematic testing of drivers
|8.9% had any alcohol (1998). Illegal BAC is .05
|24.7% of drivers and pedestrians total
|38.6% (1997) had any alcohol. Illegal BAC is .08
|20.2% (1995) at .05 or higher
|49% of drivers in fatal accidents (1996 data); 75% of fatally injured drivers
|47% of pedestrians in fatal accidents; 49% of fatally injured pedestrians. 28% of cyclists in fatal crashes; 31% of fatally injured cyclists
|24% of fatally injured drivers at .05 or higher
|19% at .05 or higher (1998)
|17% at .03 or higher (1997) The illegal BAC is .05.
|Unknown - each State may determine testing rules. Testing only takes place if alcohol is suspected by police.
|Unknown - not obligatory, State may determine
|7.8% had any alcohol (1998) Illegal BAC is .05. An
|68.3% (mostly non-injured drivers, some injured drivers, very few dead drivers)
|Few cyclists; no pedestrians
|27% had any alcohol or drugs (1998). Illegal BAC is .08
|70-75% killed in fatal crashes, Approximately 33% of all drivers in fatal crashes
|Approximately 33% of pedestrians and cyclists over 15 years of age
|8.8% - multi-vehicle; 32.9% - single-vehicle at .05 or higher (1990 - more recent data not available because of privacy restrictions)
|< 60% (from autopsies)
|41% had any alcohol. 29% over illegal limit (.08) (Jan. and Feb. 1998)
|3.3% were suspected by police of alcohol involvement (official statistic). 18% had alcohol based on fatally injured drivers autopsied (1998)
|> 90% autopsied. Official statistics based on police suspicion only
|10% of motorcyclists; 19% of cars and other motor vehicles at .08 or higher (1998)
|68% (48% by police, 20% by coroner's courts)
|39% of pedestrians; 39% of cyclists
|38.6% at .01 or higher, 30% at .10 or higher. Illegal BAC .08 or .10, depending on state (1998)
|Testing rate varies from State to State. Nationwide, 63% of fatally injured drivers were tested in 1998.
Inconsistencies in measurement and reporting parameters impair our ability to make comparisons from country to country regarding alcohol involvement in fatal crashes. In addition, the accuracy of reports, even within the definitional and measurement parameters provided, is questionable. The DG VII Working Group on Alcohol, Drugs and Medicines of the European Union has addressed the issue of accuracy among member countries. In its current report (in press), it points out that the number of recorded traffic crashes in national databases does not reflect the true level of alcohol related crashes because in each member country there is an element of underreporting. Under-reporting can be very high in some southern European member countries where the results of police breath tests or blood tests undertaken by the appropriate authorities are only partially recorded in national databases and where the level of enforcement is relatively low.
Only one member country, Great Britain, makes any statistical adjustments to correct for under-recording. Several countries do not publish any statistics on alcohol involvement at all.
The EU Working Group report makes several suggestions for increasing the comparability of statistics among EU member countries. Many of these same suggestions could apply to worldwide comparisons. Suggestions include:
Other conditions that would improve the ability to make international comparisons worldwide include reliance on testing rather than police suspicion to establish alcohol involvement, a higher proportion of testing in fatal crashes, and common definitions of fatalities.