CHARACTERISTICS OF DWI/DUI DRIVERS; DRIVING PERFORMANCE

Transcription

1 CHARACTERISTICS OF DWI/DUI DRIVERS; DRIVING PERFORMANCE

2

3 Patterns of Traffic Violations with Special Emphasis to DUI in Tennessee, USA H.N. Mookherjee Department of Sociology, Tennessee Technological University, Cookeville, Tennessee, U.S.A. Keywords Drinking-driving, DUI offences, DUI offenders, Traffic violations Abstract The primary objective of this paper is to study the driving patterns of the Tennessee DUI drivers, in order to understand alcohol-related crashes and fatalities in Tennessee. Data were obtained from the Tennessee Department of Highway Safety. Driving records were analyzed for drivers who received a citation for a driving violation in 1999, since they received their driving licenses. Out of a total 4,355,230 valid licensed drivers in 1999, 671,544 drivers were cited for traffic movement violations. These cited drivers were recorded to have a total of 989,848 traffic movement violations since they received their driving license. The analyses revealed that among these drivers 68% were males, 78% were whites, and about 15% (99,388) were cited for DUI offences. In addition to the typologies of the drinking-drivers traffic violations, this study explores the possibility of detecting indirectly the drivers driving skills and attitudes toward drinking-driving. Introduction Studies describing the characteristics of drinking-drivers and DUI offenders (1,3,4,7,12) reveal that unemployed, single, separated or divorced individuals, with a lower level of formal education and lower socioeconomic status are over-represented for drinking-driving and DUI offences. Most are males between the ages of However, the results on lifestyle behavior studies of DUI offenders are not consistent. Some have indicated a relationship between higher rates of traffic movement violations and automobile crashes among crash involved and/or DUI offenders (2), but others have found no relationship between drinking-driving and auto accidents (6). Studies also identify a higher level of tobacco and drug use among the crash involved drivers (2,8), where the events occur mostly in relation to sports, bar attendance, spending time with friends and other social activities (10). Studies on motivations to drink and drive indicate that these activities are affected by sub-cultural norms, expectations, and values implemented by individuals (10). Some studies have also revealed that work-related stress might have been the pre-conditional factor in drinking-driving, which then leads to auto-crashes. Attitudinal studies, suggest that personal attitudes and decisions made well in advance of drinking-driving are highly predictive of impaired driving and crashes (5)

4 Most of the above studies are derived from the implied assumption that the impaired drivers are mostly involved in auto-accidents, and their impairments are directly related to their alcohol addiction. In short, impaired drivers are alcoholics. As a result, researchers have developed typologies of impaired drivers (11,12), where the delinquency group (11) or the hard core drinking-drivers (1,3,9) are found to be more likely to be involved in crashes or drinkingdriving crashes. This delinquency group consists of mostly male drivers who are less likely to be heavy drinkers with less driving exposure. They exhibit the highest drinking-driving problems, even more than those of the heavy drinkers (12:61). These findings suggest that the heavy drinkers are not always the problem drivers, who are involved in drinking-driving crashes. We believe that attitudes of the drivers are the important factor, which lead to their drinking-driving behavior and auto- crashes. We assume that the drivers driving attitudes and skills are reflected through the patterns of their driving movement violations. We will explore the patterns of driving violations of Tennessee DUI drivers, as a test case, to better understand the alcohol-related motor vehicle crashes and fatalities in Tennessee. Methods The data were collected from the Tennessee Department of Highway Safety. Driving records of the drivers who received a driving citation in 1999 for driving violation, as recorded, since they received their driving license, were analyzed. Driving violations were categorized according to (1) minor traffic violations, e.g., reckless driving, speeding, driving through red light, and related violations; (2) driving under the influence (DUI) and related violations; (3) accidents- property damage; (4) accidents- bodily injuries; and (5) fatal accidents. During 1999 in Tennessee, out of a total 4,355,230 valid licensed drivers, 671,544 drivers were cited for traffic movement violations. These cited drivers were recorded to have a total of 989,848 traffic movement violations since they received their driving license. Among these 671,544 cited drivers about 15% (99,388) were cited for drinking-driving (DUI) offences. Our main interest was to find out the patterns of driving offences among these DUI offenders. We analyzed the driving records of these DUI offenders (99,388), since they received their driving license. Results The analyses revealed that among these 99,388 DUI offenders in Tennessee during 1999, 78.5% were males and 77.4% were white adults. Comparatively, more non-white males than white males (93% vs. 74%), but more white females than non-white females (26% vs. 7%) were convicted for a DUI offence. However, regarding the total number of traffic violations in 1999, both whites and non-whites received almost equal proportions of citations in DUI and related offences (15% vs. 15%), accidents-bodily injuries (3% vs. 3%), and fatal accidents (0.2% vs. 0.1%), while in accidents-property damage whites (10%) received more citations than that of the non-whites (7%). Interestingly, for minor traffic violations non-whites (75%) received more citations in comparison to whites (72%)

5 When considering the age of the DUI offenders, it was noted that the majority of the offenders were from the age-group (38%), and next to that were the age-group (33%). The DUI convictions were lower among the higher age-groups (40-49 =14%; =5%; 60 & over =3%). However, the DUI related traffic violations among the age-group 21 & below (9%) were lower in comparison to that of the other age-groups. Table 1 presents the different types of traffic violations committed by the 1999 DUI offenders, since they received their driving license. The severity of drinking-driving in connection with accidents is vividly revealed in this table (Table 1). The record shows that these 99,388 DUI offenders were cited with a total of 394,283 traffic violations. Among these DUI offenders, 59% were cited with one-dui offence, 21% were charged with two-dui offences, and the remaining 20% were multiple-dui offenders. The record also shows that the one-time DUI offenders were cited for a total of 152,630 traffic movement violations, indicating their somewhat reluctant driving pattern. On the other hand, these one-time DUI offenders had only about 38.4% of the total traffic violations in DUI offences, while the two-times or multiple-dui offenders had the most traffic violations in DUI offences (45% - 75%). Again, these one-time drinking-driving offenders minor traffic violations were only 33% of their total traffic violations, the remaining 67% of the traffic violations were serious offences, including accidents- property damage, bodily injuries, fatal accidents, and drinking-driving. In addition, these one-time DUI offenders committed on an average 1.6 other traffic violations, while the repeat DUI offenders committed on an average more than 5.2 other traffic violations. These records of traffic violations not only reflect these drivers pattern of driving, they also suggest to take caution about the repeat DUI offenders

6 Table 1: DUI offenders number of DUI offences by all traffic violations Number DUI Minor traffic Accidents- Accidents- Fatal Total of DUI Offenders violations property bodily acci- violadamage injuries dents tions One 58,639 50,369 27,473 15, ,630 (38.4%) (33.0%) (18.0%) (10.0%) (0.5%) (100%) Two 21,103 25,016 14,015 6, ,720 (48.1%) (28.5%) (16.0%) ( 7.3%) (0.09%)(100%) Three 9,929 19,095 6,052 3, ,754 (50.7%) (32.5%) (10.3%) ( 6.5%) (0.01%)(100%) Four 4,261 8,817 3,198 1,987-31,046 (54.9%) (28.4%) (10.3%) ( 6.4%) (100%) Five 2,485 9,532 1, ,070 (51.6%) (39.6%) ( 5.3%) ( 3.5%) (100%) Six 1,391 7,154 1, ,645 (47.2%) (40.5%) ( 8.7%) ( 3.5%) (100%) Seven 696 2, ,785 (62.6%) (36.4%) - ( 1.0%) (100%) Eight ,206 (75.3%) (13.6%) ( 5.2%) ( 5.9%) (100%) Nine & more 488 3, ,427 (55.0%)(32.1%)( 9.4%) (3.3%) (0.3%) (100%) Total 99,388(182,219) 126,739 54,729 29, ,283 (Total DUI offences committed by DUI offenders are presented in italics)

7 Table 2: DUI offenders and their total traffic violations Number Total # Number of traffic violations committed by DUI offenders Total # Mean # of DUI of DUI of viola- of violaoffences offenders (1) (2) (3) (4) (5) (6) (7 & +) tions tions One 58,639(59.0%)12,314 21,755 13,370 6,216 2,580 1,114 1, , (100) (21.0) (37.1) (22.8) (10.6) ( 4.4) ( 1.9) ( 2.2) Two 21,103(21.2%) 3,482 5,529 5,529 2,912 1,583 2,068 87, (100) (16.5) (26.2) (26.2) (13.8) ( 7.5) ( 9.8) Three 9,929(10.0%) 1,708 1, ,574 58, (100) (17.2) (18.7) (18.7) ( 9.4) (36.0) Four 4,261( 4.3%) ,065 2,131 31, (100) ( 5.4) (19.6) (25.0) (50.0) Five 2,485( 2.5%) ,107 24, (100) ( 6.1) ( 9.1) (84.8) Six 1,391( 1.4%) 1,391 17, (100) (100) Seven 696( 0.7%) 696 7, (100) (100) Eight & 884( 0.9%) , more (100) (100) Total 99,388(100%) 12,314 25,237 20,607 13,832 8,336 4,921 14, , (100) (12.4) (25.4) (20.7) (13.9) ( 8.4) ( 5.0) ( 14.2) (Percentages are presented in parentheses and in italics.) Table 2 presents the total number of traffic violations committed by the 1999 DUI offenders. Among all these DUI offenders (99,388) about 59% were first-time DUI offenders, and their total traffic violations numbered 152,630, which is an average of 2.6 traffic violations committed per offender until Data revealed that the convicted second, third, fourth, fifth, and sixth-time DUI offenders, on an average, were involved in 4, 6, 7, 10, and 13 traffic violations per person, respectively. The convicted eighth-time or more DUI offenders were involved in 14,633 or an average of 17 traffic violations per person, including the eight or more DUI offences. The data further indicated that among the first-time DUI offenders, 12,314 were involved only in DUI offence, where the remaining 48,325 were involved in one DUI offence each plus other traffic violations for a total of 140,316 traffic violations. Similarly, among the 21,103 second-time DUI offenders, 3,482 were convicted of two-dui offences each, the remaining 17,621 offenders being involved in 80,756 traffic violations including their 2-times DUI offences. In other words, the multiple DUI offenders (40,749) committed a total of 206,269 traffic violations in addition to their DUI offences

8 In comparison, it is clear that where the convicted first-time DUI offenders, on an average, were involved 2.6 times in traffic violations, the multiple DUI offenders were, on an average, involved 5.9 times in traffic violations. Therefore, it can be concluded that these multiple DUI offenders were not only highly involved in traffic violations, but their driving behavior were also highly dangerous for themselves as well as for their communities. Discussion Results of this study of convicted drinking-drivers in Tennessee coincide with earlier findings characterizing the alarming situation of the DUI offenders, especially the multiple-dui offenders (1-3,9,11,12). It should be noted that the DUI offences committed (182,219) by these convicted drinking-drivers were 46% of their total traffic violations (394,283) until Although the number of fatal accidents was not remarkably high, the accidents involved in property damage and bodily injuries (85,324) were accounted for about 22% of the total traffic violations committed by these DUI offenders. At the same time, minor traffic violations were not negligible, which made up about 32% of all traffic violations. Although we do not have the information about the pre-conditional factors in drinking-driving, and attitudes of these DUI offenders, the pattern of their traffic violations draw a clear picture of their negligent driving behavior, which is especially true for the multiple-dui offenders. The pattern of traffic violations further identifies the risk pattern of this DUI population. It is clear that whether it is one-time DUI offender or multiple-dui offender, none of them could drive without committing any traffic violations. All of the DUI offenders were high risk drinkingdrivers, but the multiple-dui offenders were the highest risk drinking-drivers. Hence, it can be concluded that there are no experienced DUI drivers, who can drive safely after being impaired by alcohol and/or other drug intake. It is true that this study was confined to Tennessee DUI drivers of 1999, so the implications of the results will be limited. However, it will not be unwise to conclude, based on these results, that the DUI offenders should not be treated lightly; they should be considered with more caution. The data indicated that on average the higher the number of DUI offences committed, the higher the number of total traffic violations committed per person. On the other hand, it can be suggested that these DUI offenders appear to be dangerous drivers, who did not show any responsibility to traffic regulations and/or human safety. Some programs should be developed to control their drinking-driving. References 1. Bailey JPM. Hard core offenders among drinking drivers in fatal accidents. In: Kloeden, CN and McLean, AJ. editors, Alcohol, Drugs and Traffic Safety, Proceedings of the 13 th International Conference, Adelaide, Australia, University of Australia, Adelaide, 1995, pp Begg DJ, Langley JD, Williams SM. Lifestyle factors as predictors of injury crashes among young adults in New Zealand: A longitudinal study. In: Mercier-Guyon, C. editor, Proceedings of the 14 th International Conference, Annecy, France, Centre d Etudes et de Reserches en Medecine du Traffic, Annecy, 1997, pp

9 3. Hedlund J, Fell J. Repeat offenders and persistent drinking drivers in the U.S. In: Kloeden, CN and McLean, AJ. editors, Alcohol, Drugs and Traffic Safety, Proceedings of the 13 th International Conference, Adelaide Australia, University of Australia, Adelaide, 1995, pp Klein JL, Anthenelli RM, Bacon NMK, Smith TL et al. Predictors of drinking and driving in healthy young men: A prospective study. Amer J Drug Alcohol Abuse 1994; 20(2): Kruger HP, Lobmann R. Factors predicting drunk driving self-reported behavior before and after raising a BAC limit. In: Mercier-Guyon, C. editor, Alcohol, Drugs and Traffic Satety, Proceedings of the 14 th International Conference, Annecy, France, Centre d Etudes et de Reserches en Medecine du Traffic, Annecy, 1997, pp MacDonald S, Mann B. Causes and correlates of drinking and driving. In: Utzelmann, HD, Berghaus, G and Kroj, G. editors, Alcohol, Drugs and Traffic Safety, Proceedings of the 12 th International Conference, Cologne, Germany, Verlag TUV Rheinland, Cologne, 1993, pp MacDonald S, Pederson LL. The characteristics of alcoholics in treatment for driving while impaired. Br J Addict 1990; 85: Oliver J. Alcohol, drugs and traffic deaths in the West of Scotland. In: Mercier-Guyon, C. editor, Alcohol, Drugs and Traffic Safety, Proceedings of the 14 th International Conference, Annecy, France, Centre d Etudes et de Reserches en Medecine du Traffic, Annecy, 1997, pp Simpson HM, Mayhew D. The hard core drinking driver. In: Utzelmann, HD, Berghaus, G and Kroj, G. editors, Alcohol, Drugs and Traffic Safety, Proceedings of the 12 th International Conference, Cologne, Germany, Verlag TUV Rheinland, Cologne, 1993, pp Stewart K, Cohen A, Taylor E, Sole C. Values and motivations of young drivers: Key components of impaired driving countermeasures. In: Kloeden, CN and McLean, AJ. editors, Alcohol, Drugs and Traffic Safety, Proceedings of the 13 th International Conference, Adelaide, Australia, University of Australia, Adelaide, 1995, pp Stoduto G, Adlaf E. Typological structure of adolescent drinking drivers. In: Mercier-Guyon, C. editor, Alcohol, Drugs and Traffic Safety, Proceedings of the 14 th International Conference, Annecy, France, Centre d Etudes et de Reserches en Medecine du Traffic, Annecy, 1997, pp Vingilis E. Driver characteristics: What have we learnt and what do we still need to know? In: Laurell H. editor, Alcohol, Drugs and Traffic Safety, Proceedings of the 15 th International Confernece, Stockholm, Sweden, 2000, pp

10

11 Drunk Driving in Injury and Fatal Accidents in France in Interaction between Alcohol and other Offences Filou, C. 2, Avenue du général Malleret Joinville, ARCUEIL Cedex France INRETS/DERA Abstract This article relates the French situation of alcohol in injury accidents and examines especially the relations between alcohol and other offences. Introduction The French statistics related to alcohol levels of drivers involved in injury accidents are incomplete. Indeed, they affect only the number of breath tests and positive tests performed by police after all accidents (injury and damage). The results of alcohol levels by checking after positive tests and by direct blood sample (impossible breath test) are not taken into account. Objectives and methods An exact representation of the drunk driving (legal and illegal) in injury accidents must take in consideration the alcohol levels of all accident-involved drivers. Two data were available in 2000: first, the national exhaustive injury accident file holding two variables according to "alcohol": method of detection (breath test, breathalyser, blood sample) and result (alcohol level) which is elaborated by police rapidly after the accident; secondly, the INRETS accident report file (1/50 of all reports) where the data and the practices are more detailed and precise which has been performed later. We made an interrogation of these two files. Results Two types of analysis may be introduced: analysis at individual driver and analysis at accident taking into account all drivers involved in the accident. Analysis at whole accident level is performed on the basis that an accident "without alcohol" corresponds to an accident in which no driver was over the legal limit and an accident is considered "with alcohol" if at least one driver was over the legal limit. Comprehensive file It contains 121,223 injury accidents (6,811 fatal). Some results (according to the number of vehicle involved, the category and the sex of driver) are available, More than one injury accident out of ten (11%) is an accident "with alcohol". This ratio increases to more than one out of three (35%) for fatal accidents and 45% for the single vehicle involved in fatal accidents

12 10,658 drivers (5.5%) had an illegal (over 0.49 g/l) alcohol rate (1,584 or 16.2% in fatal). For the pedestrian, the ratio is smaller in injury accidents (5.2%) but bigger in fatal one's (28.6%). Figure 1: Illegal alcohol levels in accidents 14% 12% 10% 8% 6% 4% 2% 0% g/l g/l >= 1.20 g/l Injury accident Fatal accident The proportion of drivers with "illegal alcohol" rate is more important for men (6.7%) than for women (1.8%) but the ratio is reduced in fatal accidents (18.2% vs. 6.3%). Illegal alcohol fluctuates according to the vehicle. Table 1: Drivers with "illegal alcohol" rate in accidents according to the vehicle Vehicles Fatal accidents Injury accidents Bicycle Moped Motorcycle Car Delivery truck Truck Bus Others 16.0% 28.3% 20.6% 18.0% 10.0% 1.5% 0% 11.6% 2.0% 4.3% 3.8% 6.3% 4.4% 1.1% 0.2% 6.8% Car drivers have the greatest proportion of "illegal alcohol" rate in injury accidents. But in fatal accident, the maximum concerns moped's drivers and the rate of motorcyclists is too greater than the one of car drivers. Whatever severity of the accident, bus and truck drivers are very respectful with the law on the alcohol

13 INRETS file 1,000 reports were examined. They involved 1,743 drivers and 138 pedestrians. Alcohol level is known in 80% accidents for 86% drivers. Figure 2: Alcohol investigation practices for injury accidents Drivers involved in injury accidents screening impossible or refused screening unknown screening NOT performed screening performed 8.2% 2.4% 8.0% 81.4% blood test blood test positive negative impossible result result 3.7% 4.5% 3.9% 77.5% result negative positive unknown result result 0.3% 2.2% 2.0% The search of alcohol was carried out for only 86% drivers (81.4% by screening test with confirmation if the result was positive, confirmation by breathalyser or blood test and 4.5% by direct blood test). We notes than when the screening had a positive result, first an illegal rate concerns only 93% drivers and then, the alcohol level is higher if the confirmation is measured by blood test than by breathalyser and those in spite of a longer delay for the measure. Table 2: Measures of the alcohol level after positive result by screening test Alcohol level (g/l) Blood test Breathalyser < >= % 8% 4% 42% 42% 9% 19% 14% 37% 21% Mean 1.81 g/l 0.71 mg/l Delay 1h 45mn 1h 20mn Variations are also observed between accidents "with alcohol" or "without alcohol" according to the localisation, the day and the time. The proportion of drivers with "illegal alcohol" rate fluctuates with the age. It is maximum for years old and minimum for younger

14 Figure 3: Drivers with illegal alcohol rate according to the age 8% 7% 6% 5% 4% 3% 2% 1% 0% 0 to 17 y. 18 to 24 y. 25 to 39 y. 40 to 54 y. over 55 y

15 Figure 4: Injury accident "with alcohol" according to the network 25% 20% 15% 10% 5% 0% Motorway Main road Secondary road Other road Injury accidents "with alcohol" were more frequent on rural than in urban area and this especially on main roads (that is new) but also on secondary roads. It is on the motorway that the legislation is the most respected. Table 3: Injury accident "with alcohol" according to the day Day of the week Monday Tuesday Wednesday Thursday Friday Saturday Sunday Bank holiday Injury accidents with alcohol 10% 8% 8% 15% 11% 19% 25% 14% The injury accidents "with alcohol" are more frequent during the night especially between midnight and 4h. The proportion of injury accidents "with alcohol" is bigger during the weekend especially on Sunday

16 Figure 5: Injury accident "with alcohol" according to the time 50% 45% 40% 35% 30% 25% 20% 15% 10% 5% 0% 0-4h 4-8h 8-12h 12-16h 16-20h 20-24h In the reports, police indicates the traffic offences committed by the drivers involved in injury accidents (speed, priority, manoeuvre, overtaking, stamped papers, others ). One or more "other" offence is founded among 47% drivers with "illegal alcohol" rate (vs. only 31% drivers with legal alcohol rate. Speed offence is correlated with alcohol infraction in 31% cases. Lake of priority, helmet or seat belt not wearing and no insurance are less frequent. The bigger proportion of drivers with legal alcohol rate who do not respect the priority must be explained: drivers with "illegal alcohol" rate are more involved in single accident out of intersection

17 Figure 6: Drivers with other offences in injury accidents 35% 30% 25% 20% 15% 10% 5% 0% Speed offence Lake of priority Helmet or sealt belt not wearing No insurance Illegal alcohol Legal alcohol Discussion In 1995, the alcohol legal limit was reduced to 0.5 g/l. It seems that this legislation has no influence on the number of drivers involved in injury accidents with "illegal alcohol" rate. Moreover, their mean alcohol rate did not decrease (2.15 g/l). In the French legislation, the search of alcohol is compulsory for all drivers involved in injury accident. In fact, this search is carried out for only 88%. So, the searchers are very anxious to know the results about the search of drug in fatal accidents, which is compulsory since October The link between illegal alcohol and speed offence in injury accidents is not surprising. However, the measures carried out on the road showed that the drivers with alcohol below the legal limit (between 0.01 and 0.24 g/l) exceed more the speed limits than those with "illegal alcohol" rate

18

19 Criminal Profiles of Drinking Drivers in Ontario S. Stewart 1, P. Boase 2, A. Reid 3 1 Ministry of the Attorney General, Toronto, Ontario, Canada. Ontario Court of Justice, Room 155, Old City Hall, 60 Queen Street West, Toronto, Ontario, Canada M5H 2M4, 2 Transport Canada, Ottawa, Ontario, Canada, 3 Ontario Provincial Police, Orillia, Ontario, Canada Keywords Accident, Alcohol Intoxication, Criminals, Driving Under the Influence, Conviction Abstract The paper reviews the criminal and driving history of a sub-sample of 99 drivers drawn randomly from drivers charged with criminal alcohol related driving offences in Toronto, Canada after the implementation of the immediate ninety- day administrative driver licence suspension (A.D.L.S.) in November, These data, referred to herein as the 1998 sample, and results are compared to the 1996 sample, which was reported on previously (1), to determine changes, if any, in the characteristics, as reflected in the criminal and driving records of those charged with drinking and driving related criminal offences in the face of the deterrent of the immediate ninety- day licence suspension. The paper reviews whether the immediate ninety day A.D.L.S. has deterred drivers with previous criminal or significant driver histories. Findings from the 1998 sub-sample that are distinct from the 1996 sub-sample findings are reported on. Some of the measurements in the 1998 sample are either the same as, or have little differential with the 1996 sub-sample and are not reported on. Introduction In Canada, driving while impaired and having a blood alcohol content (BAC) in excess of 80mgs, as well as refusing to provide a breath sample upon request of a police officer, are criminal offences under the Criminal Code of Canada. The criminal sanctions (minimum fine of $300 and a minimum three-month driving prohibition for a first conviction, with increasing penalties including minimum jail terms for subsequent convictions) were unchanged in 1998; however, the Province of Ontario implemented the immediate ninety- day A.D.L.S. Subsequent to the time of this sub-sample, the Criminal Code was amended with regard to penalties and sentencing. For example, as a result of the 1999 amendments, the minimum fine is $600 with a minimum 12 month driving prohibition and a BAC of over 160mgs became an aggravating factor in sentencing. Unlike those in the 1996 sample, the 1998 drivers faced an immediate and relatively significant consequence, upon detection, of losing their driver s licence for ninety days irrespective of the outcome of any criminal charges

20 The relevant document flow for a drinking and driving related criminal conviction in Ontario changed. The process remains the same, starting with the police charging the driver and serving the applicable notice related to the A.D.L.S. Charging documents are entered into the integrated court offences network and proceed through the judicial system. If a conviction is registered, for either a criminal or provincial driving related offence, the law requires the court office to notify the Registrar of Motor Vehicles. The notification process changed from a manual one to an automated one in 1998, with the exception of criminal conviction information from the Superior Court of Justice which is still transmitted manually. Very few drinking and driving charges are heard in this court; however those that are generally involve death or serious bodily harm. In order to ensure that the charges and dispositions appear on the driver s criminal record, the charging police service that receives the information from the court office must notify the Canadian Police Information Centre. Merging the records should provide a profile on the complete driving and criminal history of an individual. Materials and Method Information on 957 drivers who had been stopped and charged for a drinking and driving related criminal offence by the Toronto Police Service in 1997 and 1998 was obtained and compared to the 880 drivers in the 1996 sample. A sub-sample of 99 drivers was randomly selected from the 1998 sample and the criminal and driving records were manually retrieved and assessed, with the results compared to the 1996 sub-sample. Some of the information reported on for the 1996 sub-sample, such as the reason for the arrest and the blood alcohol concentrations is not available for the 1998 sample. Where possible, a similar analysis of the 1998 data was undertaken. The results were compared with the 1996 measurements and differences reported on. Of the 99 drivers in the sub-sample, 6% were female and 94% were male. They ranged in age from 18 to 79 years of age (22-73 in 1996), with the median age category being thirty to forty years old. Results Of the 99 drivers sampled, provincial driver records for all but one were located, a substantial improvement over the 1996 result of 15 unmatched records. A similar improvement was seen in the number of drivers without either a criminal or a provincial driver record match, which went from 7 in 1996 to only 1 in Conversely, the failure rate for the criminal record match increased from 2 to 9 drivers. Ten drivers (6 in 1996) had no record of the 1998 arrest on neither the criminal nor provincial driver record. Eliminating the drivers with record retrieval issues produces a sub-sample of 88 drivers with both criminal and provincial driver records unless otherwise noted. Of the 98 drivers with provincial driving records, 30 % (31% in 1996) were not licenced or suspended at the time of the driver record search (February, 2000) and an additional 3 drivers were never licenced. One driver was suspended for life as a result of a 1999 conviction (Ontario law was amended in 1998 to provide for escalating suspensions upon conviction including suspension for life on a fourth conviction. The record search period for calculating the convictions cannot go back further than September 30, 1993). Twenty-three percent (20% in 1996) of the suspended drivers were suspended as a result of an unpaid fine and 36% (50% in 1996) were under suspension because they were convicted of a

21 drinking and driving criminal offence. Three drivers were suspended for both a drinking and driving conviction and driving while disqualified or suspended and one for medical reasons. Five drivers (2 in 1996) remained unlicenced or suspended at the time of the search despite the fact that the reinstatement date preceded the search date. Three drivers who should have been unlicenced appeared as licenced. Of the 89 drivers in 1998 with matching criminal records, 8% refused to provide a sample on request (11% in 1996). In addition to the drinking and driving related criminal charges, one driver was charged at the same time with a drug-related offence, down significantly from the nine drivers in 1996, although 11 of the drivers had drug related matters on their criminal record (two with extensive drug related records). A number of other criminal or provincial charges were also laid arising out of the same incident including failing to stop for police (two drivers), fail to remain, drive while disqualified and care and control (one driver). As with the drug offences, this is markedly different from the earlier sub-sample in that fewer drivers faced additional criminal charges at the same time and the charges that they did face were not as assorted. The 1996 sample included criminal harassment, resisting arrest, obstruction and assault. In reviewing the criminal profile of the sampled drivers (89), just over 30% (45% in 1996) of the drivers were found to have other, non-drinking and driving related criminal charges or convictions by charge date, including parole violation, unlawfully at large, sexual assault, fail to appear, obstruction, assault and narcotic offences. One driver had been convicted on 18 prior occasions of one or more offences, but only one was a driving related criminal offence. Four of the drivers also had convictions for non-drinking and driving criminal offences post In reviewing the provincial driver records for the 98 drivers, 86% (58% in 1996) had non-alcohol related, provincial offence convictions such as careless driving, speeding, failing to report collision and drive with no insurance. The following tables show the incidence of selected convictions, as a percentage, on the driver records for the entire 1996 (721) and 1998 (748) driver samples respectively. The results seem similar for the two samples, except there appears to be more drivers with a previous alcohol infraction on record. This likely reflects improved automated record exchange related to the new processes associated with ADLS. Table 1: Convictions for Selected Charges Offence/Number 0 1 Multiple 0 1 Multiple Careless 84% 14% 2% 82% 14% 4% Driving While Suspended 86% 9% 4% 85% 9% 6% Impaired/Blow over.08 26% 42% 32% 20% 49% 31% Fail to Report Collision 90% 9% 2% 87% 11% 2% Speeding 30% 19% 52% 28% 17% 55% Other HTA Convictions 22% 19% 59% 24% 17% 59%

22 With regard to drinking and driving charges and dispositions, 34% (36% in 1996) of the drivers with criminal records available had drinking and driving related charges or convictions prior to the 1998 arrest. Thirty percent (10% of the total) had multiple priors, ranging from 5 drivers (12 in 1996) with 2 priors to 3 drivers with 3 priors to 1 driver (3 in 1996) with 5 priors. This represents a significant decline over the 1996 finding of 52 percent or 18% of the total drivers. The number of multiple priors per driver has also declined sharply over 1996; however, in contrast to the one driver in 1996, five drivers in 1998 had drinking and driving related convictions subsequent to the 1998 arrest. Three drivers were convicted of other criminal offences at the same court appearance as the drinking and driving charge, one each of break and enter & possession; theft and escape lawful custody. Of the number of drivers with criminal records, twelve (13%) had no indication or record of the 1998 sub-sample charges. This represents a decrease over the 1996 result of 23%. Three of the seven drivers charged with impaired and refuse to provide a breath sample were only convicted of impaired. One driver was convicted of impaired driving but had the charges of dangerous driving and fail to stop withdrawn. Eight drivers (3 in 1996) had the charges dismissed, stayed or withdrawn. The balance of the drivers (77%) were subsequently convicted of the drinking and driving related criminal charges either drive over, impaired or impaired and refuse-an increase from 1996 of over twenty percent. Eight-five of the drivers arrested in 1998 had a corresponding A.D.L.S. on the driver record. One of the drivers had a subsequent A.D.L.S. imposed while under a drinking and driving conviction related suspension. Four drivers who should have had an A.D.L.S. on the driver record did not, with four more drivers having no record of an A.D.L.S. or of the arrest. One driver charged with impaired and refused did not have the suspension on record even though the refusal qualifies the driver for the suspension. The balance of the drivers appear to have been charged with only impaired driving, which does not attract the suspension under provincial law. A review of the driver records indicates that 14 drivers (16% of those with a suspension) have had two administrative driver licence suspensions imposed between the start date of November 1996 and the record search date (February, 2000). A number of drivers had an A.D.L.S. on the provincial driver record but had no indication on the corresponding criminal record of the 1998 related charges. Nine drivers had an administrative driver licence suspension on the driver record but no subsequent record, criminal or provincial, of any related charges or outcomes. Seven of these drivers (one driver in 1996) had a conviction for careless driving on the driver record that appeared to correspond to the A.D.L.S. The records were reviewed to determine the number of months (to the nearest month) between the 1998 arrest and the disposition (conviction or dismissal) of the resulting charges. The range was from one day to 25 months, somewhat lower than in In 1998, 40% of the 76 drivers in the group had dispositions in six months or less; 47% in 12 months or less, with the balance taking from months. Two of the drivers appear to have plead guilty to secure release from custody after arrest as the dispositions were one and five days from the arrest date. Accused persons released from custody by the police would not appear in court for at least seven days. A person may be held in custody for a judicial interim release hearing before a judicial officer. Individuals who have little chance of release for whatever reason will sometimes plead guilty to the offence charged rather than remain in custody until trial, which could take three months or

23 more. Where a conviction resulted, the average time from charge to disposition was just over seven (9 in 1996) months (The two drivers convicted in one and five days were eliminated); for those not convicted it was ten months (11.6 months in 1996). The criminal and driver records of convicted drivers were reviewed to determine if the drivers received the appropriate criminal sentence as well as the appropriate administrative consequence. Of the convicted drivers with a criminal record, fifty had a prohibition order recorded as part of the sentence while twenty-one (just under 30%) drivers had no prohibition recorded. This is an increase over the 23% of convicted drivers with no corresponding prohibition in the 1996 sub-sample. Administratively, all drivers with the convictions on the driver record appear to have received the appropriate provincial suspension. Fore the full sample of drivers in the 1998 sample, by individual driver, 163 had no collision involvement on record, 224 had one collision, 152 had two collisions, 95 had 3 collisions and the remaining 114 drivers had more than 3 collisions on the record including 3 drivers with 12 or more on the record. This is almost identical to the 1996 data except that there was only one driver with 12 or more collisions during the time period. Discussion The results indicate that there has been a substantial improvement in record keeping with the introduction of automated transmittal between the court offices and the Registrar of Motor Vehicles. This improvement ensures that more convicted drinking drivers will receive the full scope of provincial, administrative consequences of any driving related criminal conviction as well as the corresponding treatment. Ontario has a remedial measures program, consisting of education and treatment units and will soon implement a post-conviction program of alcohol ignition interlock. Of concern, however, is the increased failure rate in matching drivers with criminal records. This could be reflective of a number of issues, such as fingerprinting or record follow up with the charging police agency or the Canadian Police Information Centre. It would appear that the driver record has become the best record of previous alcohol related driving convictions. This is not likely a serious issue, as long as the police check both records for previous conviction information. However, if only the criminal record is searched for nondriving related matters, such as at a bail hearing, then a drinking driver may benefit from the lack of a complete record. This situation is also likely to support the perception that drinking and driving criminal offences are not really criminal, but driving, in nature. Significantly fewer drivers were detected in the 1998 sub-sample with drugs and fewer drivers in the 1998 sub-sample refused to provide a sample on demand. Those in the later group, as in 1996, appeared to have a reasonable prospect of not being convicted of the refusal charge. The number of drivers who had the charges stayed, dismissed or withdrawn has increased (8 versus 3). Overall, the percentage of drivers convicted as a result of the arrest has increased by over 20% over The time taken to convict those drivers decreased by almost two months over Even for those who had the charges dismissed, the time to disposition decreased by a month and a half. The decrease in time from arrest to disposition is difficult to attribute directly to A.D.L.S. Fourteen drivers had the charges disposed of in two months or less, possibly because drivers under an A.D.L.S are permitted to have the suspension run concurrent (at the same time) as any conviction related suspension. The record information does not indicate whether a plea of guilty was entered or a trial held; however, it would be reasonable to infer that convictions entered

24 within three months of arrest are pleas of guilty as the criminal justice system in Ontario does not accommodate trials for those not in custody within three months. The introduction of A.D.L.S. and the corresponding entry onto the provincial driver record allows for charged drivers to be monitored irrespective of the outcome of the criminal charges. In the context of the Canadian criminal justice system, this allows for a fuller assessment of the number of drinking drivers who meet or exceed the criminal standard. Alcohol involvement on the part of drivers subsequently convicted of careless driving, with the same offence date, can be monitored for research purposes. However, pleading to careless is an advantage to the driver as the driver avoids the possibility of being identified as a repeat offender on a subsequent offence and also avoids the more immediate consequence of the onerous provincial sanctions that flow from conviction for a criminal driving offence. While the percentage of drivers with prior drinking and driving charges or convictions did not change appreciably, the percentage with multiple priors decreased over 20%. The number of drivers with subsequent arrests increased in the 1998 sub-sample. Five drivers, as opposed to the one in 1996, had post-1998 charges or convictions. Sixteen percent (14) of the drivers with an A.D.L.S. imposed have incurred two A.D.L.S. s subsequent to the implementation in November 1996 and the record search (February, 2000). Clearly, A.D.L.S. allows for the immediate, point in time identification of the drinking and driving behavior as opposed to the pre-a.d.l.s. environment. A.D.L.S. is only available when the driver is charged with drive over or refuse, not impaired driving or other criminal driving related offences. In conclusion, the comparison of the two sub-samples primarily showed an improvement. Fewer drivers faced other charges at the same time, fewer had a previous criminal record and fewer had prior drinking and driving charges. The fact that the number of drivers unlicenecd or suspended at the time of search remained the same may indicate that something happens to drivers charged with drinking and driving criminal offences. Some of these drivers should have been licenced, leaving open the possibility that they become disinterested in remaining in the driver licencing regime or that they subsequently leave the jurisdiction or willfully choose to be unlicenced. Acknowledgements The authors would like to acknowledge the assistance of Constable Jeff Patrick of the Toronto Police Service, the Ontario Provincial Police and the Royal Canadian Mounted Police for the considerable work in providing the samples and all necessary subsequent record retrieval. References 1. Stewart, S., Boase, P. and Lamble, R.W. (2000). Criminal Profiles of Drinking Drivers in Ontario. International Conference on Alcohol, Drugs and Traffic, Stockholm, Sweden

25 Driver Resistance Mark B. Johnson, James E. Lange, & Tara Kelley Baker Pacific Institute for Research and Evaluation Beltsville Drive, Suite 300, Calverton, Maryland, USA Keywords Heavy drinking, motivation, sex differences, designated driver, groups, normative influence Abstract Young people do the majority of their drinking within small groups. This research examines the intersection between individuals intentions to drink and the perceived group norm for drinking on actual drinking behavior. The paper suggests that individuals unique drinking motivations may be expressions of different group roles. The data show that the role of driver within groups may protect individuals from normative group pressure. The importance of understanding the processes of group construction is discussed. Introduction It has been widely acknowledged that young people do the majority of their drinking within groups of peers. Given the social nature of drinking, researchers have examined the influence of social factors such as group norms and group characteristics on the drinking behavior of individuals (e.g., Collins, Parks, & Marlatt [1]; Aitken [2]). Although evidence suggests that normative group drinking pressure does exist, the unique motivations and attitudes of individuals within groups clearly play a role in drinking behavior as well. For example, individuals intentions to get drunk, their belief that alcohol facilitates social interaction (3), and their belief that drinking was an important part of college (4) all predict drinking behavior. However, the characteristics of the group and the individual motivations of group members are not completely independent. Part of the unique individual variations within groups may be expressions of different group roles. Limited attention has been paid to the social construction of groups, and researchers have failed to examine closely how individual group members with their unique motivations and beliefs select other members to fill specific group roles. Social roles within naturally formed groups of friends may be functional, and although not necessarily stable or easy to define, they may have specific purposes for facilitating the group s goals. Within drinking groups, perhaps the role of driver is easiest to study because the role is salient, consciously known, functional, and usually acknowledged by all group members. Further, when it comes to drinking, the role of driver is predictive of very different drinking behaviors. Among groups of young people, the driver is more likely to refrain from drinking or to drink considerably less than their peers (e.g., Foss and Beirness [5]; Lange & Voas [6]). It is interesting that drivers

26 enter into the same drinking environment as their peers where normative pressures to drink should be at their peek and yet are able to maintain sobriety. The lower BACs typically found among drivers may occur (a) because situational factors that influence drinking (i.e., alcohol availability, loose regulation, normative environment that fosters party attitude) affect drivers differently than they do passengers, or (b) because individuals whose motivations and beliefs make them likely to drink less are selected by the group to be the driver. This research examines both possibilities. The paper first reports on role of the driver within groups and how this role relates to drinking behavior and drinking intentions; the paper then address how the role of driver may attenuate normative pressures to drink. Finally, the paper examines factors that may predispose individuals to be assigned to role of driver. Methods An opportunity to examine situational influences, individual intentions, and the interaction between the two became available from ongoing survey of drinkers at the U.S.-Mexico border. Beginning in the fall of 1997, researchers at the crossing between San Diego County and Tijuana, Mexico, have randomly sampled whole groups of U.S. residents (aged 18 30) en route to the bars and nightclubs south of the border. These participants were asked to provide basic demographic information about themselves, whether or not they drove to the border, information on their drinking history, their drinking intentions for the evening, and their perceptions of the drinking intentions of other group members. The researchers gave a hospital-style identification bracelet and took a breathalyzer test from each participant. These groups of participants were re-sampled upon their return across the border into San Diego, asked to provide information on their drinking behavior while in Tijuana, and given a second breath test. Between October 1997 and March 2000, entry and exit data were collected from 687 groups that contained at least one driver (some groups reach the border via taxi, trolley, or some other transportation). These groups comprised 2,369 participants. Because participants were clustered in naturally occurring groups, statistical assumptions of independent observations were tenuous. Therefore, most analyses were conducted using PROC MIXED in SAS (Version 8.0) to account for the clustered nature of the data. For purposes of analyses, the sample was split into two halves: one exploratory, the other confirmatory. Thus, attempts were made to replicate analyses in both samples. Although the results of analyses conducted on the confirmatory sample are not described below, only those findings from the exploratory sample that were replicated are presented. Results The first analysis was designed to test for differences in returning BACs between participants crossing the border as drivers and as passengers. As predicted, analysis of the first sample revealed that for both men and women, BACs of drivers were significantly lower than they were for passengers [F (1, 517) = 19.3, p<.01; F (1, 439) = 8.82, p<.01, for men and women respectively]. The average BAC of male drivers was.023, relative to.061 for male passengers. Similarly, the average BAC of female drivers was.020, relative to.050 for passengers. Next, we examined the extent to which participants drinking intentions (i.e., the extent to which they planned to drink before actually going to the bars) predicted their BACs. Participants indicated their intentions to Not Drink, Get a Slight Buzz, or Get Drunk. For both men and

27 women, analysis revealed a statistically significant relationship between drinking intentions before entering Tijuana and returning BACs [F (2, 517) = 53.2, p<.01; F (2, 439) = 35.7, p<.01]. Participants who intended to drink more heavily did indeed return with significantly higher BACs. Mean BACs were.016,.038, and.068; female BACs were.010,.040, and.067 for the three levels of drinking intentions, respectively. However, an analysis that was conducted to examine the interaction between drinking intentions and driving status failed to produce a statistically significant effect. Thus, both drivers and passengers tended to drink to the same extent as indicated by their drinking plans. A subsequent analysis, which treated drinking plans as a three-point scale variable, revealed that drivers planned to drink less than passengers, F (1, 1056) = 182.9, p<.01. The mean drinking-intentions score was 1.8 for drivers and 2.4 for passengers. There were no differences in drinking intentions between men and women. Because drinking intentions was shown to be predictive of returning BACs regardless of driver status, a series of analyses were conducted to examine one normative factor that might predict drinking intentions. It is possible that bar-goers base their drinking intentions in part on the Perceived Drinking Climate (PDC; i.e., on their expectations regarding how much alcohol others in their immediate social group plan to drink). Individuals who believe that their drinking companions will consume heavily may consume heavily as well. However, if the designated driver concept is used appropriately, we might predict that drivers drinking intentions will remain low regardless of the PDC. Analysis of group size revealed no statistically reliable results across the two samples. However, regressing drinking intentions onto driver status, PDC, and their interaction did reveal statistically significant effects, and these differed between men and women. For men, the main effect of PDC, F (2, 513) = 35.6, p<.01, was statistically significant; 1 those who perceived that others in their group would drink heavily planned to drink more heavily as well. The interaction between PDC and driver status was not significant (p<.09). However, a planned contrast was conducted to compare the drinking intentions of drivers who indicated that others in their group did not plan to drink with drivers who indicated that others in their group planned to get very drunk. The results revealed that for male drivers, PDC did indeed predict drinking intentions, F (1, 513) = 17.5, p<.01. This pattern is depicted in Figure 1a. For women, however, the results differed. As with men, PDC did predict drinking intentions, F (2, 434) = 18.2, p< However, in contrast to the men, the interaction was statistically significant as well, F (2, 434) = 9.2, p<.01. The nature of this interaction was revealed through the planned contrast. Although PDC was related to the drinking intentions of female passengers, the relationship between PDC and drinking intentions for female drivers was not statistically significant. This pattern is illustrated in Figure 1b. 1 The main effect for driver status was statistically significant as well, but this effect was produced and discussed in earlier analyses as well

28 Figure 1: Predicting Drinking Plans from Drinking Climate and Driver Status 1a. Men 1b. Women Drinking Plans Driver 1.4 Passenger Drinking Plans Driver 1.6 Passenger Not drink Slight buzz Get drunk Drinking Climate Not drink Slight buzz Get drunk Drinking Climate To examine factors that might predict who will be the driver after the group has been drinking, analyses examined two driver characteristics. Analyses were conducted to determine whether recent history of heavy drinking and implicit alcohol attitudes predicted driver status within groups. First, we hypothesized that individuals who did not have a recent history of heavy drinking (defined by consuming five or more drinks on a single occasion over the past 4 weeks) were more likely to be the driver. However, chi-square analysis (adjusting the expected frequencies according to group size) failed to find a statistically significant relationship between driver status and recent heavy drinking. Second, we examined drivers implicit alcohol attitudes. Implicit alcohol attitudes were measured using the Alcohol Association Scale (AAS; 7), a word association instrument designed to tap into implicit attitudes and mental representations of alcohol. Scores on this measure ranged from.00 to Participants with higher scores were thought to associate alcohol with aggressive concepts, whereas participants with lower scores were thought to associate alcohol with amiable experiences. We predicted that drivers would tend to have higher AAS scores (i.e., a more negative association with alcohol) than would passengers. For men, analysis revealed that drivers did indeed have higher (more negative) AAS scores than did passengers, F (1, 522) = 5.23, p<.05, driver M =.45, passenger M =.39. However, the pattern was not statistically reliable across samples from women. Discussion Individuals drinking intentions for that evening was a strong predictor of drinking behavior as measured by BACs. These intentions not only may reflect unique and personal motivations, but also could reflect role norms (e.g., being the designated driver) as well as group norms. One interpretation of the results presented herein is that group norms and role norms can be countervailing. The role of the driver amongst drinking groups dictates maintaining sobriety to ensure a safe drive home. However, the perception that the drinking climate for that evening will be heavy, moderate, light, and so on may implicitly establish a group norm; normative pressures may influence drinking behavior. The relationship between PDC and passenger drinking intentions is consistent with this contention

29 How does PDC relate to drinking intentions of drivers? The results suggest that women are better able to resist normative pressures to drink; the drinking intentions of female drivers remained low regardless of the PDC. Perhaps for women, the role and responsibility of being a driver was clearer and better defined, and this attenuated other normative influence. Conversely, male drivers planned to engage in heavier drinking to the extent that they perceived that other group members would drink heavily. It appears that male driver were less able to resist normative pressures to drink. The analysis of individual s implicit alcohol attitudes suggests that, at least for men, individuals who have a negative association with alcohol are more likely to be drivers within groups. This pattern was not found among women drivers. The results also failed to reveal a relationship between driver status and recent heavy drinking history. These results are consistent with the interpretation that role norms and larger group norms can compete for influence of drinking intentions. Perhaps, by more clearly defining the role and responsibilities of being a driver, drivers may be better able to resist normative pressures. Our understanding of drinking behavior may benefit from further research that addresses the way drinking groups are constructed and the nature of roles within groups. References 1. Collins RL, Parks GA, Marlatt GA. Social determinants of alcohol consumption: The effects of social interaction and model status on the self-administration of alcohol. J Consult Clin Psychol. 1985; 53: Aitken PP. An observational study of young adults drinking groups II. Drink purchasing procedures, group pressures and alcohol consumption by companions as predictors of alcohol consumption. Alcohol Alcohol. 1985; 20: Beck KH, Treiman KA. The relationship of social context of drinking, perceived social norms, and parental influence to various drinking patterns of adolescents. Addictive Behaviors. 1996; 21: Wechsler H, Davenport A, Dowdall G, Moeykens B, Castillo S. Health and behavioral consequences of binge drinking in college: A national survey of students at 140 campuses. JAMA. 1994; 272: Foss RD, Beirness DJ. Drinking passengers and their drivers: Roadside survey results. 40th Annual Proceedings of the Association for the Advancement of Automotive Medicine. Chicago: Association for the Advancement of Automotive Medicine; 1996: Lange JE, Voas RB. Youth escaping limits on drinking: Binging in Mexico. Addiction. 2000; 95: Lange JE. Alcohol's effect on aggression identification: A Two-channel Theory. Psychology of Addictive Behaviors. 2002; 16:

30

31 Social-cultural characteristics of DWI drivers compared with drug-alcohol-free drivers S. Zancaner, R. Giorgetti, G. Frison, Boscolo Mia, S.D. Ferrara Dipartimento di Medicina Ambientale e Sanità Pubblica, sede di Medicina Legale, Università degli Studi di Padova, via Falloppio n. 50, PADOVA, Italy. Abstract The study analyzes a population of drivers stopped during some highway police checks. The population has been divided into subgroups according to the presence of alcohol and drugs in the biological fluids. Social and cultural characteristics have been considered for each group. Introduction Some toxicological forensic tests have been carried out from 1997 to 2000 to prevent driving under the influence of alcohol or psychoactive substances, in collaboration with the Highway Police Department of Veneto and the Italian Red Cross. The tests have been done especially during weekend nights; the activity has been partially done within the project ROSITA (ROad SIde Testing Assessment), which is a multicentric European study. Car drivers have been stopped by the police, submitted to medical examinations, and samples of blood and/or urine have been taken. Toxicological tests have been done on biological samples to detect alcohol and psychoactive substances. The population studied has been divided into subgroups on the basis of the results of the toxicological analyses; the present work makes a comparison of social and cultural characteristics of car drivers belonging to each group. Materials and methods The population studied is made up of 1,977 subjects driving motor vehicles (1,894 males, 83 females); at least one sample of the biological fluids has been examined for each subject. Each subject has been informed about the methodology of the tests and the consequences of a possible refusal (punishment as if driving in a state of intoxication); then they have been submitted to a medical examination including collecting of general clinical details and samples of biological fluids; 1,748 blood samples and 1,702 urine samples have been taken. The following groups of subjects have been considered: BAC from 10 to 80 mg/100 ml; BAC over 80 mg/100 ml, sanctioned under Italian law for driving under the influence of alcohol (n.555);

32 subjects sanctioned under Italian law for driving under the influence of psychoactive substances (n.241); subjects drug addicts but who were not found in a state of acute poisoning at the moment of the tests (n.78); drivers intoxicated by cocaine (n.98); drivers intoxicated by cannabis (n.165); drivers with BAC < 10 mg/100 ml, selected as control sample (n.500); total population examined (n.1,977). The following parameters have been considered in each group: gender, age, driving licence category, driving experience, marital status, job, place of origin, education. Results Social and cultural characteristics of the car drivers are reported in the following tables, in relation to the presence of alcohol or psychoactive substances in the biological fluids. Table 1: lcohol and drugs in relation to the gender of drivers Gender Total Pop. NAND BAC BAC >80 DUID Drug users DUID cocaine DUID cannabis Male 95,8% 97,8% 94.48% 97,3% 97,1% 98,7% 97,96% 96,36% Female 4,2% 2,2% 5.52% 2,7% 2,9% 1,3% 2,04% 3,64% NAND: no alcohol, no drugs BAC: blood alcohol concentration DUID: driving under the influence of drugs Drug users: drugs in urine; no drugs in blood, no clinical signs of intoxication. Table 2: Alcohol and drugs in relation to the age of drivers Age (years) Total Pop. NAND BAC BAC >80 DUID Drug users DUID cocaine DUID cannabis <20 12,44% 14,8% 12,1% 8,2% 13,28% 7,69% 9,18% 16,97% ,65% 38,8% 39,7% 35,32% 46,47% 47,44% 47,96% 49,70% ,24% 22,2% 29,51% 28,83% 23,24% 26,92% 23,47% 21,21% ,35% 16,2% 10,62% 14,77% 13,28% 11,54% 15,31% 10,30% >35 9% 7,8% 7,86% 12,43% 3,32% 5,13% 4,08% 1,82% Unknown 0,30% 0,2% 0,21% 0,54% 0,41% 1,28% 0% 0%

33 Table 3: Alcohol and drugs in relation to the driving experience Driver Experience (years) Total Pop. NAND BAC BAC >80 DUID Drug users DUID Cocaine DUID Cannabis <1 5,97% 7,8% 5,73% 3,42% 7,47% 5,13% 3,06% 9,70% ,13% 36,4% 38,22% 29,19% 44,40% 41,02% 38,78% 52,12% ,23% 26,6% 32,27% 34,06% 25,31% 35,90% 27,55% 23,03% >10 25,95% 28,4% 21,87% 30,81% 20,75% 16,67% 27,55% 13,94% Unknown 1,72% 0,8% 1,91% 2,52% 2,07% 1,28% 3,06% 1,21% Table 4: Alcohol and drugs in relation to the driving licence Driver Licence Total Pop. NAND BAC BAC >80 DUID Drug users DUID cocaine DUID cannabis A motorcycle 0,35% 0,2% 0,42% 0.36% 0,41% 1,28% 0% 0,61% B car 83,8% 87,2% 83,23% 79,82% 87,97% 85,9% 88,77% 88,48% C little truck 7,19% 3,8% 7,22% 10,45% 7,06% 7,69% 7,15% 6,06% D Bus 3,04% 4% 3,61% 2,70% 2,08% 1,28% 2,04% 1,82% E Heavy truck 3,50% 3,2% 3,19% 4,69% 1,66% 1,28% 1,02% 2,42% None 0,76% 0,2% 0,85% 0,72% 0,41% 0% 1,02% 0% Other 0,45% 1% 0,42% 0,18% 0% 0% 0% 0% Unknown 0,91% 0,4% 1,06% 1,08% 0,41% 2,57% 0% 0,61%

34 Table 5: Alcohol and drugs in relation to the place of provenance Place of provenance Total Pop. NAND BAC BAC >80 DUID Drug users DUID cocaine DUID cannabis DUID amphet. Disco 37,13% 29,8% 43,31% 40,18% 35,69% 28,21% 41,84% 30,30% 68,19% Other public place 31,10% 28,4% 32,06% 36,58% 31,95% 26,92% 32,65% 32,73% 13,63% Private house 16,55% 18,4% 12,95% 12,61% 18,67% 29,49% 18,37% 20,61% 18,18% Other 14,56% 22,8% 10,62% 10,09% 13,69% 14,10% 7,14% 16,36% 0% Unknown 0,66% 0,6% 1,06% 0,54% 0% 1,28% 0% 0% 0% Table 6: Alcohol and drugs in relation to the marital status Marital Status Total Populat NAND BAC BAC >80 DUID Drug users DUID cocaine DUID cannabis Unmarried 86,9% 89% 85,57% 81,98% 92,12% 97,44% 92,86% 92,12% Married 9,20% 7,8% 10,19% 12,80% 6,22% 1,28% 4,08% 6,66% Legally separated 2,69% 2,2% 2,76% 3,06% 0,83% 0% 2,04% 0,61% Divorced 0,25% 0,2% 0,21% 1,08% 0% 0% 0% 0% Other 0,1% 0,2% 0,21% 0,36% 0,83% 0% 1,02% 0,61% Unknown 0,86% 0,6% 1,06% 0,72% 0% 1,28% 0% 0%

35 Table 7: Alcohol and drugs in relation to the employment status Employment Status Total Pop. NAND BAC BAC >80 DUID Drug users DUID cocaine DUID cannabis Employed 87,26% 86% 87,48% 89,55% 90,04% 75,64% 93,88% 87,27% Unemployed 4,55% 3% 5,10% 3,42% 4,15% 7,70% 2,04% 5,45% Student 6,38% 9% 5,52% 4,87% 4,57% 15,38% 2,04% 6,67% Pensioner 0,15% 0,2% 0,21% 0,18% 0% 0% 0% 0% Other 0,80% 1,2% 0,42% 0,72% 0,83% 0% 1,02% 0,61% Unknown 0,86% 0,6% 1,27% 1,26% 0,41% 1,28% 1,02% 0% Table 8: Alcohol and drugs in relation to the educational level Schooling Total Pop. NAND BAC BAC >80 DUID Drug users DUID Cocaine DUID cannabis None 0,15% 0% 0% 0,36% 0% 0% 0% 0% Primary school 3,60% 2,6% 1,91% 6,31% 4,56% 1,28% 6,12% 1,82% Middle school 46,49% 42% 46,92% 47,57% 55,19% 43,59% 52,04% 56,97% Professional School 17,35% 16,4% 18,05% 18,20% 19,09% 12,83% 20,41% 18,79% High school 27,57% 33,6% 27,60% 21,98% 18,67% 35,90% 19,39% 20% Parauniversity degree 0,30% 0,4% 0,42% 0,54% 0% 1,28% 0% 0% University degree 2,88% 3,2% 3,61% 3,42% 1,25% 2,56% 1,02% 1,22% Other 0,55% 0,6% 0,64% 0,36% 0,41% 1,28% 0% 0,60% Unknown 1,11% 1,2% 0,85% 1,26% 0,83% 1,28% 1,02% 0,60% Discussion and conclusion Nearly the whole population examined is composed of males (Table 1) who present intoxication by alcohol and psychoactive substances more often than females. Table n.2 describes the age parameter: there is a high prevalence of drug intoxication, in particular by cannabis, among the

36 youngest subjects, who have even less experience in driving. Six percent of the whole population and 9.7 % of the subjects poisoned by cannabis have their driving licence less than a year. In the oldest subjects and in particular in subjects over 35, intoxication by alcohol is found more frequently. Eighteen percent of the subjects punished for driving in a state of dtunkenness had a driving licence for heavy transport (table 4). There is no difference in the place of origin among the groups of subjects: the use of alcohol and drugs is shared equally in all the environmental settings; in any case, to analyze the typology of the identified substances is more meaningful: % of the subjects under the influence of cocaine and 70 % of the subjects positive for amphetamines (22) were coming from a disco. In the general population 37,13 % of subjects were coming from a disco. Thus the population studied presents a high prevalence of subjects using psychoactive substances in a disco setting. Most subjects submitting to the saturday-night tests were single; they were more frequently found intoxicated by drugs than by alcohol. Those subjects married or separated, on the other hand, were more frequently found under the influence of alcohol than of drugs. Data about the employment of subjects are not meaningful. The social and cultural level of the examined population is not very high. Most subjects found intoxicated by alcohol and drugs had the lowest social and cultural level. The study outlines two kinds of people: the first made up of subjects from 18 to 25 years old, characterized by a more frequent use of psychoactive substances, cannabis in particular; the second made up of drivers over 35; most of them driving in a state of intoxication by alcohol. Identification of social and cultural characteristics of drivers represents an attempt to prevent driving in a state of intoxication or under the influence of psychoactive substances: a higher level of education seems to be a protective factor; yet no one can be considered completely exempt from the phenomenon. Data collected identify discos as places where psychoactive substances are frequently used; this fact is quite worrying since we know that intoxication by drugs can increase the risk of being involved in road accidents. Even drivers with a driving licence for heavy transport frequently consume alcohol and drugs. Thus we can conclude that the rare Italian street surveys are not sufficient to discourage these atrisk behaviours. Data collected, in relation to the peculiar population studied, are very different to data collected in other works carried out in Quebec (4), New Zealand (5) end Denmark (6). The procedures are not standardized, so data are difficult to compare

37 In Italy street surveys are rare and not uniform in procedures and tests, so it is very difficult to collect data that really describe phenomenon incidence, social and cultural characteristics of the drivers etc.; reliable data would be important to plan efficient protocols to educate and survey drivers and at least to prevent a lot of road accidents. References 1. Zancaner S, Giorgetti R, Fenato F, Rossi A, Tedeschi L, Snenghi R, Frison G, Montisci M, Tagliaro F, Meroni M, Giron G, Marigo M, Ferrara SD. Psychoactive substances and driving disability: epidemiological roadside survey in north-east Italy. In: Kloeden CN, McLean AJ (Eds). Alcohol, Drugs and Traffic Safety. Proceedings of the 13 th International Conference, Adelaide, Australia, 1995, pp Zancaner S, Giorgetti R, Dal Pozzo C, Molinari G, Snenghi R, Ferrara SD. Driving under the influence of drugs. Correlation between clinical signs and type of intoxication. In: Mercier- Guyon C (Ed). Alcohol, Drugs and Traffic Safety, Proceedings of the 14 th International Conference, Annecy France, 1997, pp Ferrara SD. Alcol, droga, farmaci e incidenti stradali. CLEUP, Padova, Dussault C, Lemire AM, Bouchard J, Brault M. Drug use among Québec drivers: The 1999 roadside survey. In: Fred H, Smith RB (Ed). Alcohol, Drugs and Traffic Safety, Proceedings of the 15 th International Conference, Stockholm, Sweden, 2000, paper Keall MD, Frith WJ. Drink driving behaviour and its strategic implications in New Zealand. In: Mercier-Guyon C (Ed). Alcohol, Drugs and Traffic Safety. Proceedings of the 14 th International Conference, Annecy, France, 1997, pp Bernhoft IM. Characteristics of drunken drivers in Denmark. Alcohol, Drugs and Traffic Safety. In: Mercier-Guyon C (Ed). Alcohol, Drugs and Traffic Safety. Proceedings of the 14 th International Conference, Annecy France, 1997, pp Riley SC, James C, Gregory D, Dingle H, Cadger M. Patterns of recreational drug use at dance events in Edinburgh, Scotland. Addiction 96, ,

38

39 Identifying Possible Sources of Bias Introduced in Traffic Safety Research: Comparison of Blind Linkage with Volunteer Clinical Samples 1,2 R.E. Mann, 3 S. Macdonald, 2 M.L. Chipman, 1,2 E. Adlaf, 4 K. Anglin-Bodrug and 1,2 J. Zhao 1 Centre for Addiction and Mental Health, Toronto, Ontario; 2 Department of Public Health Sciences, University of Toronto; 3 Centre for Addiction and Mental Health, London, Ontario; 4 Department of Psychology, University of Western Ontario, London, Ontario. Abstract In research on alcohol, drugs and traffic safety, and on road safety in general, investigators have employed a variety of measures to construct samples for research. Differing sample construction procedures may introduce bias into the resulting samples, but this possibility has rarely if ever been assessed empirically. In this research we compare two samples of individuals who obtained treatment for a substance abuse problem. One sample was obtained by blind linkage procedures, that is, groups were identified in the clinical records. A second sample was identified in the clinical records, and then tracked, contacted, and asked to consent to participate in a research study. Comparisons of the two samples on psychosocial and problem indicators derived from clinical records revealed a large number of significant differences between the samples. In all cases, the differences indicated that the group tracked and asked for consent had higher levels of functioning than the blind linkage group. The possible reasons for these differences, and their implications for research and research policy, are discussed. Introduction In research on drinking and drug using drivers, and other areas in the road safety field, a variety of research approaches have been used over the years. Two of these approaches can be broadly characterized as those which involve complete samples of individuals drawn from records (e.g., 1-3), and those which involve samples of individuals who have been asked to provide their consent to participate in research (e.g., 4-6). Increasing concerns with ethical issues in research, among other factors, have meant that researchers are increasingly following the latter route. An important concern that is highlighted where researchers must choose between procedures in which complete samples are used, versus those in which incomplete samples are used due to the inability to find or obtain consent from all eligible participants, is the introduction of potentially serious bias into the latter type of study. This bias may occur for a variety of reasons. For example, in studies involving driving records of clients in treatment for a substance abuse problem, it can be speculated that individuals with the characteristics of most interest (e.g., those who use drugs most frequently, or who have the highest levels of dependence) will be the individuals most likely not to volunteer to participate, or who will be least likely to be found in order to seek agreement to participate. In assessing the effects of treatment for drug dependence on traffic safety measures, it is reasonable to suppose that individuals with the most serious drug

40 problems, or the most serious driving problems, will also be those who are most likely not to agree to participate or those who are most likely not to be found when sought. However, while this is a general concern in the research community, virtually no studies have compared the characteristics of the two types of samples. In a recent investigation of the collision experiences of individuals in treatment for alcohol, cannabis and/or cocaine problems (7,8), we had a unique opportunity to compare a complete samples of individuals with a sample that had to be tracked and from whom consent had to be obtained. We therefore compared the two groups on a variety of demographic and psychosocial measures, to test the hypothesis that the group from whom we had to obtain consent would be biased towards excluding individuals with more extreme characteristics and higher levels of dysfunction. Methods Participants Two samples were drawn from clinical records, involving clients who sought help for an alcohol or drug problem in 1994, 1995, 1996 or 1997 from the Centre for Addiction and Mental Health (CAMH - formerly the Clinical Research and Treatment Institute of the Addiction Research Foundation). All clients selected were at least 20 years of age at the time treatment began and lived in the greater Toronto area, thus eliminating confounding influences associated with place of residence. About 75% of the clients seen at CAMH meet this criterion. The substance abuse groups were composed of 7 sub-groups: alcohol only; cannabis only; cocaine only; alcohol and cannabis; alcohol and cocaine; cannabis and cocaine; and alcohol, cannabis, and cocaine. The first sample, for telephone interview purposes, was a random sample of 160 subjects in each of the seven sub-groups noted above, and was selected from patient records (projected N=1120). These people were targeted for a telephone interview. Patients must have been at least 20 years old at their assessment and have had addresses within the Greater Toronto area. It was found that many subjects did not have a valid address, phone number, lived in a group house, did not speak English, or did not provide a last name. These people also were excluded from the initial sampling frame. Patient records were first sampled from Sufficient numbers that met the selection criteria were obtained for the alcohol only and cocaine only groups. Subsequent random samples were taken from 1996 and 1997 to obtain the desired sample size of 160 patients per group. Insufficient numbers could not be obtained for the cannabis + cocaine and cannabis = cocaine + alcohol groups (i.e., 80 and 96 subject respectively were collected), and the final total number available for the telephone interview sample was 971. The second sample, drawn for blind linkage purposes, of 527 patients was randomly selected from each of the seven drug groups from We attempted to retrieve 80 patients in each of the 7 groups selected at random from the list of all clients seen initially for a substance abuse problem in 1994, for a projected total sample of 560. However, only 47 patients met the criteria of having a drug problem with all 3 substances. No interviews were conducted with this group. Data sources and measurement The data described here were obtained from clinical intake forms, patient records, and supplementary telephone interviews. When clients first arrive at CAMH they are all asked

41 routine questions (i.e., name, address, birth date, marital status and other demographic characteristics) and given a standardized drug use assessment interview from which their substance problems are identified. Over the course of a client's contact with CAMH, detailed clinical files are maintained. These files covered the following topic areas: presenting problem, alcohol and drug use history, family alcohol and drug use, social relationships, accommodation, educational attainment, financial, leisure, legal, physical, emotional concerns, previous treatment of any kind, crisis issues and treatment recommendations. Progress notes were also available and described each visit by the client, types of treatment received and information on the therapist s impression of the client's progress. In order to develop the coding scheme for the study, 40 patient files were reviewed and a coding form, the Detailed Client Coding Form, was developed and used for coding the information in the files. Following the development of the Detailed Client Coding Form, the sample files were reviewed and relevant information extracted onto the coding forms for data entry. Measures of severity of the substance abuse problem were drawn from the alcohol and drug use history chart, which was also useful for assessing whether other drugs were being used in addition to those reported in the face sheets. An interval level variable was constructed measuring intensity of treatment based on the number of hours in contact with the Clinical Institute. Type of program attended was treated as a nominal variable and could include: a youth program, an adult abstinent lifestyle program, individual counseling, group counseling, family counseling, guided self change, an Employee Assistance Program, low intensity outpatient program (i.e., one hour per week), and a high intensity day patient and residential program (i.e., 7 hours per day for 14 to 28 days). Since at least 5 years of follow-up data were available for each client, indicators of subsequent treatment and relapse were recorded. Additional variables extracted included suicide attempts, depression, anger, social supports and other emotional problems. Telephone interviews Tracking of individuals in the telephone interview sample began in the fall of The initial task was to obtain the telephone numbers of 971 patients, based on the first and last names of each subject as well as their phone number and address at the time they were initially assessed. Direct or likely telephone matches were made for 396 subjects (i.e., 40.8% of the initial sample) and these people, termed the telephone match sample, were telephoned for an interview. At least 20 phone call attempts were made for each telephone number in order to resolve their status. We were unable to make a positive identification in the telephone interview for 80 of these people and 3 were clearly the wrong person. For those where positive identification could be made 63 people were ineligible for the study because they did not have a drivers license (31), had died (8) or moved away from the Greater Toronto area (24). Another 13 people were identified by someone at the residence as being the correct person but were never able to be contacted. Finally, 63 people were successfully contacted but refused to participate in the study. The final response rate, comparing the participants with whom interviews were successfully completed (111) to the to the total participants and refusals among the posively identified group, was 64%, or 28% of the telephone match sample

42 Results We compared the telephone interview group (TI) with the blind linkage group (BL) on 58 demographic, substance use, and psychosocial problem measures derived from the Detailed Client Coding Form. Variables on which significant differences between groups were found are summarized in Table 1. It is clear that the samples differed substantially on a large number of variables. It is also clear that in every instance, the differences revealed that the TI group had lower levels of problems, or fewer difficulties, than the BI group. Discussion This study provides one of the first empirical assessments of the types of bias introduced by differing procedures used to obtain samples for road safety research. The differences in procedures were substantial, and it is important to note that not all comparisons between groups were significant. However, there is clear support for the general hypothesis that the way in which research samples are constructed can have a very large impact on the nature of the eventual sample obtained. All the significant differences observed support the specific hypothesis that sampling procedures requiring contact and consent to participate will act to exclude individuals who have more serious problems, or whose levels of dysfunction are higher that those who are eventually contacted and provide consent. The factors influencing or responsible for this biasing process cannot be identified specifically, but several could be operating. First, a substantial proportion of the potential TI sample was unable to be located for telephone calls. This could reflect a higher level of mobility over the follow-up interval, or insufficient resources to maintain residential stability or a telephone, among other factors. Additionally, although the participation rate among individuals who were successfully contacted (64%) is considered very good (9), it is possible that biasing factors were operating here as well. For example, individuals with higher levels of dysfunction may have been less likely to agree to the interview. Higher levels of substance abuse could be the direct or underlying cause for the various sources of bias identified here. Thus, since the interest of the research was in the impact of substance abuse on collisions and injuries, in addition to there being an indirect bias introduced by the TI interview procedures, the bias could be more direct in screening out individuals with the highest levels of exactly those characteristics or qualities we are interested in. However, although the factors mentioned here are all associated with higher levels of dysfunction or substance abuse, that need not necessarily be the case and additional work is needed to understand this biasing process. The likely impact of this bias on collisions and convictions can easily be hypothesized. We have observed, in general, that individuals with a substance abuse problem have increased collision risk (4-9). Thus, in a sample which is biased to exclude those who may have the highest levels of substance abuse problems, we would seem to be excluding those with the highest collision risk (analyses to assess this hypothesis are underway). If this is the case, it would seem to underscore the significance of differences found in studies using such biasing procedures. That is, the biasing procedures would make the comparisons between groups more conservative, and thus any significant differences that are found would likely be smaller that those that really exist. Thus, while our results highlight the bias that might be implemented by procedures which require tracking and obtaining consent, they do not necessarily imply that results from such studies are invalid. Instead, the results may be conservative and underestimate differences that really exist. However, this hypothesis requires more work to confirm

43 Table 1: Significant comparisons between Telephone Interview (TI) and Blind Linkage (BL) groups on measures derived from the Detailed Patient Coding Form Measure Test of Significance Direction of Result Problems with family χ 2 1 = 3.90, p<.05 BL group had higher levels of family problems members Likelihood of seeking help at another χ 2 1 = 6.79, p<.01 BL group were more likely to seek help at other agencies agency Alcohol consumption level t = 4.41, p<.001 BL group had higher alcohol consumption levels Emotional health χ 2 1 = 5.57, p<.05 BL group had more emotional health concerns concerns Likelihood of receiving outpatient counseling Likelihood of having legally imposed treatment Likelihood of having injected drugs Likelihood of having used cocaine in the past year Likelihood of having used other drugs in the past year. χ 2 1 = 4.82, p<.05 χ 2 1 = 5.96, p<.05 χ 2 1 = 8.81, p<.05 χ 2 1 = 15.35, p<.05 χ 2 1 = 3.90, p<.05 BL group was less likely to receive outpatient counseling, and more likely to be assigned to an inpatient program BL group was more likely to have legally imposed treatment BL group was more likely to have used injection drugs BL group was more likely to have used cocaine in the past year BL group was more likely to have used other drugs in the past year Total opioid use t = 2.62, p<.05 BL opioid users had higher total levels of opioid use Total other drug use t = 3.80, p<.01 BL other drug users had higher total levels of other drug use Our results also underscore the value of blind linkage procedures in research on this and related questions. Ethical issues in conducting research are never simple, and problems of consent and confidentiality in dealing with government, transportation and health datasets are particularly complex. One trend in recent years has been to deny access to these databanks for research purposes, or to introduce a blanket consent policy where any access requires individual consent. Our data demonstrate that such procedures can result in significant bias in the resulting sample. Thus, the solution to the consent and confidentiality problem of requiring individual consent in all instances will result in biased and inaccurate results. This may create problems that are worse than those a blanket consent policy is meant to solve. Acknowledgement This work was supported in part by a grant from the Canadian Institutes for Health Research

44 References 1. Chipman ML, Morgan P.P. The role of driver demerit points and age in the prediction of motor vehicle collisions. Brit J Prevent Social Med 1975; 29: Mann RE, Vingilis ER, Gavin D, Adlaf E, Anglin L. Sentence severity and the drinking driver: Relationships with traffic safety outcome. Accid Anal Prevention 1991; 23: Haberman PW. Alcohol and alcoholism in traffic and other accidental deaths. Amer J Drug Alcohol Abuse 1987; 13: Macdonald S. A comparison of the psychosocial characteristics of alcoholics responsible for impaired and nonimpaired collisions. Accid Anal Prevention 1989; 21: Mann RE, Anglin L, Vingilis ER, Larkin E. Self-reported driving risks in a clinical sample of substance users. In: Utzelmann, HD, Berghaus, G and Kroj, G. editors, Alcohol, drugs and traffic safety, Proceedings of the 12 th International Conference, Cologne, Germany, TUV Rheinland, 1993, pp Smart RG, Schmidt W, Bateman K. Psychoactive drugs and traffic accidents. J Safety Res 1969; 1: Chipman ML, Macdonald S, Mann RE. Driving, traffic crashes and addiction: Clients in addiction treatment programs. Amer J Epid 2001; 153: Macdonald S, Mann RE, Chipman ML, Adlaf E, Wells S. Collisions and traffic violations of treatment groups for alcohol, cocaine and cannabis abuse, and a matched control group. Presented at the 24th Alcohol Epidemiology Symposium of the Kettil Bruun Society, Toronto, May 28, Aday, LA. Designing and conducting health surveys: A comprehensive guide (second edition). Jossey-Bass Publishers, San Francisco,

45 The Influence of MDMA on Cognition and Psychomotor Function, and the Importance for Driving Capacity 1 CTJ Lamers, 1 JG Ramaekers, 2 N Samyn, 3 NL Read and 1 WJ Riedel 1 Experimental Psychopharmacology Unit, Brain and Behaviour Institute, Dep. of Neurocognition, University of Maastricht, Maastricht, The Netherlands, 2 National Institute of Criminalistics and Criminology, section Toxicology, Brussels, Belgium, 3 Institute for Transport Studies, University of Leeds, United Kingdom Keywords MDMA, alcohol, driving capacity, task performance Abstract In this study the influence of single doses of MDMA 75 mg and alcohol 0.5 g/kg on cognitive and psychomotor performance was assessed in 12 healthy recreational MDMA users. A single recreational dose of MDMA improved tracking performance under single and double task conditions. Movement speed in the choice RT paradigm improved increased after MDMA, while the ability to estimate time to collision was impaired after MDMA. Alcohol impaired tracking performance 1 and 2 hrs after alcohol but returned to baseline levels 3 and 5 hrs after drinking. While MDMA produced both stimulating and impairing effects on skills related to driving, it is concluded that further research employing real driving paradigms and interaction studies with alcohol and possibly THC are needed. Introduction (±)-3,4-Methylenedioxymethamphetamine (MDMA; ecstasy), is a commonly used psychoactive recreational drug related to mescaline and stimulant drugs such as amphetamine (1,2). Surveys among young people visiting raves, indicate that as much as 64-96% of visitors of raves and big parties uses ecstasy (3,4). Up to 60% of the people that planned to drive a motorized vehicle after a rave were under the influence of ecstasy (3). Not much is known about the influence of MDMA on driving performance but case reports seem to justify concern about traffic safety behavior and involvement in traffic accidents after the use of MDMA (5). Driving performance consists of a complex combination of cognitive and psychomotor functions. MDMA is often associated with poor cognitive function. Cognitive impairment, such as memory impairment in abstinent users has been demonstrated in many studies (6,7) but only a few studies addressed the acute effects of MDMA on cognition and psychomotor function. When acute effects of ecstasy were studied under uncontrolled circumstances at a party, a significant reduction in verbal recall and visual scanning was observed (8). In studies conducted under controlled conditions a single dose of MDMA (range mg/kg) did not affect immediate recall, digit repetition and selective attention. One prior placebo-controlled study addressed the

46 acute effects of MDMA on psychomotor performance. No effect of a single dose of MDMA (75 or 125 mg) on simple reaction time was observed (9). The current study investigated the acute effects of a recreational dose of MDMA (75mg) on psychomotor and cognitive performance, vital signs, mood and cortisol concentrations in recreational ecstasy users under experimentally controlled conditions. The tests described in this abstract comprise a variety of tasks measuring skills related to driving. They have been previously shown to be sensitive to a variety of psychoactive drugs (10). Furthermore an effort is made to investigate the relation between MDMA concentrations and task performance, physiology and mood. Alcohol treatment was implemented in the design as an active control. Methods Design & procedure The study was conducted according to a placebo controlled, 3-way crossover, double blind, and double dummy design. Subjects underwent all 3 treatment conditions on three separate days, spaced at least two weeks apart, receiving placebo, alcohol or MDMA. Twelve healthy recreational MDMA and alcohol users, with no current or history of physical or psychiatric disease, completed the study. Drug use was prohibited throughout the study. Before entering a test day subjects were tested for recent drug use and females were also tested for pregnancy using a urine drugs screen and urine pregnancy test. Before treatment administration subjects consumed 2 sandwiches for lunch. Throughout the day subjects had free access to water, isotonic drinks, orange juice and sugar free chewing gum. At 1.15 p.m. subjects received study treatments. Placebo was administered as 400 ml orange juice and 25 ml bitter orange syrup. MDMA was administered by replacing part of the bitter orange syrup by MDMA 75 mg. During alcohol administration part of the orange juice was replaced by pure alcohol (.5mg/kg body weight). Driving-related task performance: The Critical Tracking Test (CTT) measures the ability to control an inherently unstable error signal in a 1 st -order compensatory tracking task. Subjects attempt to keep a cursor centered in the middle of a display using a joystick while the cursor tends to move away from the center. The point where the subjects losses control is defined as the 'critical frequency' or lambda-c (λ c ) expressed in radians per second (rad/sec). Theoretically, λ c is the reciprocal of the operating delay lag in human closed-loop control. The CTT was conducted at 0, 1, 2, 3 and 5 h post treatment. The Divided Attention Task (DAT) assesses the subjects ability to divide attention between tracking and monitoring tasks performed simultaneously. The tracking subtask is similar to the CTT, but the error signal velocity is fixed at 50% of individual s optimal performance during training (λ c /2). Tracking error is measured as the average absolute distance (mm) between the cursor's position and display center. The other subtask consists of monitoring 24 peripheral LED displays fixed to the corners of the screen, each presenting numerals, 0-9, which change asynchronously every 5 seconds. The subject removes his foot from a pedal as quickly as possible after detecting the target numeral, "2". Median correct reaction time (msec) to targets is the second response measure. The DAT was conducted at 1 and 3 h post treatment

47 The Motor Choice Reaction Time (MCRT) is a test in which reaction time (RT) is studied as a function of task complexity. Reaction time is divided into initiation time (time between target and onset of response) and movement time (time of movement execution). The MCRT was conducted at 1 and 3 h post treatment. The Object Movement Estimation under Divided Attention (OMEDA) task is in essence a task to estimate time to contact (TTC) of a moving object to a specific location. The subject is seated in front of a computer screen on which a yellow circle occludes the center. The circle varies in size per trial (2, 100 or 200 pixels). From one of the corners, a red dot (target) travels towards the center of the screen and travels underneath the yellow circle and will no longer be visible. The subject estimates when exactly the target reaches the center of the screen by pressing a foot pedal. During the trial, 5 geometrical shapes appear; one on top of the occlusion circle and one in each of the corners. For the secondary task the subject has to press a button in case the geometric shape at the occlusion circle matches one of the others. Absolute TTC error and the number of correct responses to the geometric targets (divided attention) were combined using Z-scores. The Signal Detection Task (SDT) is a visual search task. Small white squares are presented in a pseudo-random fashion on a computer screen. Twenty squares are randomly assigned being 2.5 cm apart. Squares move to a different location on the screen and subjects are required to respond to the target stimuli, defined as a set of four stimuli forming a square of 2.5 x 2.5 cm, by pressing a button as fast as possible. Sensitivity (A ) defined as the non-parametric proportion of correctly identified targets corrected for false positives is the dependent measure. Vital signs Starting from baseline, pulse rate, blood pressure and body temperature were assessed in a relaxed sitting position every 30 min. until 5.5 h after drug intake, using an automated vital signs monitor (Dinamap 1800 BP; Critikon Inc., Tampa FL and an in-ear thermometer, respectively). Pharmacokinetics MDMA was determined in plasma, saliva, sweat and urine, at baseline, and every hour until 5 hours after drug intake. All samples were frozen at -20 C until analyzed. Detailed information about the collection and analyses of samples are reported elsewhere (11). Subject s breath alcohol concentration (BAC) was assessed every 30 min., starting at baseline until 5.5 h after drug intake, by means of a Lion SD-4 Breath Alcohol Analyzer. Statistics Dependent variables representing task performance were tested for the main effect and interactions between Treatment and Time using a multivariate repeated measures analysis of variance (General Linear Model). Exceptions occurred for Initiation Time in the MCRT where Task Complexity was added as an extra within subjects Factor. Analyses of performance on the OMEDA task were accomplished by entering both dependent variables (TTC error and divided attention error) in a bivariate 3 x 3 (treatment x occlusion diameter) repeated measures design. A polynomial contrast was used for Time and Complexity, while a simple contrast was used for Treatment for univariate comparisons of all drug-placebo differences

48 Correlations are analyzed using the intra-subject correlations. Pearson s r of all individual correlations between two factors are averaged and tested for significance using an independent one-sample t-test. The _-probability criterion for determining the significance of mean differences and correlation was defined as (p <.05). Results 8 Male and 4 female subjects completed the study (mean age 23.5, range 21-30; mean weight 65.9 kg, range kg). All subjects used ecstasy (mean lifetime use: 39, range 5-125), marijuana (mean lifetime use: 760, range ) and alcohol (mean units/week: 17, range 4-50). Mean plasma C max of MDMA was 178 ng/ml (range ng/ml) at 3 hours after drug intake. Mean BAC peak concentration was reached 60 minutes after alcohol intake (.31 mg/ml) and during the second repetition of the test battery BAC had dropped to 0.01 g/ml. During last assessments the BAC s had dropped to 0 in all subjects (12). Driving-related task performance: Tracking performance in the CTT improved after MDMA as compared to placebo, resulting in higher λ c. Alcohol had no main effect on λ c. There was an interaction of Treatment by Time between alcohol and placebo caused by the fact that λ c was lower at 1 and 2 h after alcohol relative to placebo, while λ c was higher than post placebo 3 and 5 h after alcohol (p=.001). In the DAT, MDMA improved sub-critical tracking performance, represented by tracking error, as compared to placebo. Alcohol did not affect tracking error in the DAT. Movement Time improved after MDMA as compared to placebo (100 and 108 msec respectively while movement time after alcohol (111 msec) did not differ from placebo. There was no effect of MDMA or alcohol on initiation time. In the OMEDA task, there was a trend towards an impairing effect of MDMA on performance. MDMA also tended to impair TTC estimation more when occlusion diameter was larger (p=.066). Alcohol did not affect TTC error compared to placebo. Alcohol or MDMA did not affect visual search performance on the SDT. Table 1: Mean (sd) performance averaged over all assessments after treatment of the CTT, DAT, MCRT and OMEDA and the outcome of the univariate analyses under the influence of placebo, alcohol and MDMA (_ = improvement, _= impairment) Dependent variable Placebo Alcohol Mdma Placebo Alcohol Placebo Mdma Tracking (CTT) 4.52 (.14) 4.54 (.15) 4.87 (.16) p=ns p=.03 _ Tracking error (DAT (1.25) (.94) (1.14) p=ns p=.02 _ Movement Time (MCRT) (5.47) (7.27) (4.87) p=ns p=.01 _ TTC error (Z-transformated Units; Omeda) (.08) (.06).103 (.13) p=ns p=.055. _ Vital signs Vital signs, presented by pulse rate, body temperature and blood pressure all rose as a function of MDMA as compared to placebo. Alcohol had no effect of vital signs. Correlations Intra-subject correlations between the MDMA levels in blood plasma, urine, saliva, sweat and physiology, and performance are presented in table

49 Table 2: The intra-subject correlations between blood plasma, urine, saliva and sweat with body temperature, pulse rate, blood pressure and task performance over all assessments after placebo and MDMA treatment. (Significance: * p<.05, ** p<.001). Variable Blood plasma Urine Saliva Sweat Physiology Body temperature.24*.30** Pulse rate.40**.39**.49**.27* Diastolic blood pressure.37**.39**.35**.29** Systolic blood pressure.53**.49**.51**.39** Performance Critical tracking.32*.33**.34*.27* Tracking error -.37* -.35* -.48** -.37* Movement time -.57** ** -.66** Time to contact error Discussion The purpose of the study was to investigate the effect of a single dose of MDMA 75 mg and a small dose of alcohol.5mg/kg psychomotor performance, cognition, physiology and mood. An important part of the study was to investigate the influence of MDMA and alcohol on driving related behavior and to make a first attempt to analyze the correlations between MDMA levels and task performance and vital signs. Although there was no main effect of alcohol on λ c and tracking error, there seemed to be an effect of alcohol on tracking performance relative to placebo as indicated by a significant treatment x time interaction. In the CTT λ c was lower after alcohol as compared to placebo1 and 2 h post treatment, indicating impaired performance. When the test was repeated 3 and 5 h post treatment performance on the CTT returned to baseline after alcohol, while performance after placebo seemed slightly impaired. Alcohol had no effect on task performance of the MCRT. A single dose of MDMA improved psychomotor performance. Critical tracking performance, i.e. λ c, increased by % as compared to placebo. Tracking error in the Divided Attention Task decreased by 11.4% after MDMA as compared to placebo. In the MCRT, movement time increased after MDMA use, while initiation time the time between presentation of the target and the onset of responds- was not affected by MDMA. A single dose of MDMA impaired performance on the primary OMEDA task, the time to collision (TTC) task. The unique component of the OMEDA task is the perception and correct estimation of object movement. After MDMA the error between estimated and actual TTC increased and subjects had more difficulty predicting the TTC as the occlusion of the center of the screen increased in size relative to placebo. The increment in TTC error after MDMA may reflect a disturbance in perception of time and space, also observed by other researchers (13,14). The decreased ability to estimate and predict movement can result in impaired estimation of other traffic movements at crossroads, leading to acceptance of smaller gap between vehicles, indicating increased risk taking behavior. Dangerous driving and accidents after MDMA use have already been reported in the past (5,15). Furthermore, impaired co-ordination, difficulty concentrating and hallucinations have been observed in people while under the influence of MDMA (13-15). These factors, sometimes combined with exhaustion after a night of dancing, have already led to involvement in -even fatal- car accidents of MDMA users in the past (5,16). Furthermore, MDMA dosages may be higher than in the present study and are often used in combination with sedative psychoactive substances such as marijuana and alcohol. Often combining 2 or more psychoactive substances, e.g. marijuana and alcohol, often increases the impairment caused by a drug when taken alone

50 (17). The current study was part of a larger experiment assessing aspects of driving and simulated driving under the influence of MDMA with and without other drugs and/or alcohol. The study assessing simulated driving confirms these findings. While MDMA alone only had minor effects on driving performance, MDMA combined with other drugs (mostly THC) and/or alcohol increased driving speed and smaller gap acceptance (18). Alcohol had no effect on vital signs while MDMA increased body temperature, pulse rate and blood pressure. The increased vital signs correlated with MDMA levels in most body fluids. MDMA levels also correlated with improved psychomotor function. MDMA levels did not significantly correlate with impaired task performance observed in this study. The analyses of correlations in this study were a first attempt to get more insight in the relation between MDMA levels and task performance. However, duplication of these findings in a larger study sample is necessary. Based on current and future results, saliva may proof itself a good and easy collectable alternative for blood samples in off-road screening. In sum we can conclude that certain aspects of vehicle control, e.g. tracking capacity improved after MDMA. Although tracking capacity is an important aspect of driving, improved tracking by no means automatically indicate increased driving safety. The increased risk taking and the users decreased ability to estimate and predict movement can result in impaired estimation of other traffic movements at crossroads, leading to acceptance of shorter gaps, especially when MDMA is used in combination with other drugs. This form of increased risk taking makes traffic safety under the influence of MDMA questionable. References 1. de Man, R.A., Morbiditeit en sterfte als gevolg van ecstacygebruik. Ned Tijdschr Geneeskd, (37): p Morgan, M.J., Ecstasy (MDMA): a review of its possible persistent psychological effects. Psychopharmacology (Berl), (3): p Spruit, I.P., XTC in Nederland. 1997, Trimbos-Instituut in opdracht van ministerie van Volksgezondheid, Welzijn en Sport: Den Haag. p Winstock, A.R., P. Griffiths, and D. Stewart, Drugs and the dance music scene: a survey of current drug use patterns among a sample of dance music enthusiasts in the UK. Drug and alcohol dependence, (1): p Henry, J.A., K.J. Jeffreys, and S. Dawling, Toxicity and deaths from 3,4- methylenedioxymethamphetamine ("ecstasy") [see comments]. Lancet, (8816): p McCann, U.D., Mertl, M. Eligulashvili, V. and Ricaurte, G.A., Cognitive performance in (+-) 3,4-methylenedioxymethampetamine (MDMA, 'ecstacy') users: a controlled study. Psychopharmacology, : p Morgan, M.J., Memory deficits associated with recreational use of "ecstasy" (MDMA). Psychopharmacology Berl, (1): p Parrott, A.C. and J. Lasky, Ecstasy (MDMA) effects upon mood and cognition: before, during and after a Saterday night dance. Psychopharmacology, : p

51 9. Cami, J., Farre, M. Mas, M., Roset P.N., et al., Human pharmacology of 3,4- methylenedioxymethamphetamine ("ecstasy"): psychomotor performance and subjective effects. J Clin Psychopharmacol, (4): p Ramaekers, J.G., Muntjewerff, N.D., Uiterwijk, M.M.C. van Veggel, L.M.A. et al., A study of the pharmacodynamic interaction between befloxatone and ethanol on performance and mood in healthy volunteers. Journal of Psychopharmacology, (4): p Samyn, N., De Boeck, G., Wood, M., Lamers, C.T.J., et al., Plasma, oral fluid and sweat wipe ecstasy concentrations in controlled and real life conditions. Forensic Sci. Int., 2002, in press 12. Lamers, C.T.J., Ramaekers, J.G., Muntjewerff, N.D. Sikkema, K. et al., Dissociable effects of a single dose of MDMA on psychomotor skills and attentional performance. Submitted. 13. Downing, J., The psychological and physiological effects of MDMA on normal volunteers. J Psychoactive Drugs, (4): p Vollenweider, F.X., Gamma, A. Liechti, M, Huber, T. et al., Psychological and cardiovascular effects and short-term sequelae of MDMA ("Ecstasy") in MDMA-naieve healthy volunteers. Neuropsychopharmacology, (4): p Schifano, F., Dangerous driving and MDMA ('Ecstacy') abuse. Journal of Serotonin Research, : p Morland, J., Toxicity of drug abuse--amphetamine designer drugs (ecstasy): mental effects and consequences of single dose use. Toxicol Lett, : p Lamers, C.T.J. and J.G. Ramaekers, Visual search and urban city driving under the influence of marijuana and alcohol. Human Psychopharmacology Clinical and Experimental, (5): p Brookhuis, K.A., D. De Waard, and L.M.C. Pernot, A driving simulator study on driving performance and traffic safety after multiple drug use, consisting of MDMA (Ecstasy) and various other psychoactive compounds., in T 2000 Alcohol, Drugs and Traffic Safety., I.H. Laurell, Editor. 2000: Stockholm, Sweden

52

53 Alcohol-Impaired Driving Recidivism Among First Offenders More Closely Resembles That of Multiple Offenders W.J. Rauch 1, P.L. Zador 1, E.M. Ahlin 1, H. Baum 1, D. Duncan 1, R. Raleigh 2, J. Joyce 2 and N. Gretsinger 2 1 Center for Studies on Alcohol, Westat, Rockville, Maryland, USA, 2 Maryland Motor Vehicle Administration, Glen Burnie, Maryland, USA. Abstract Alcohol-impaired driving legislation and sanctions have historically been aimed at the offender with multiple driving while intoxicated (DWI) convictions, with little or no attention paid to the first-time offender or to alcohol-related events other than DWI [such as administrative per se (APS) violations involving breath alcohol content (BrAC) of 0.10 or more, APS breath test refusal and probation before judgment (PBJ)]. It is a widely held belief among the legislature and judicial branches of state government that first offenders criminally convicted of an alcoholrelated traffic law are drivers with a single and isolated alcohol-related violation that results in arrest. This finding is inconsistent with published estimates that a person can drive while impaired by alcohol 200 to 2,000 times before being arrested once for alcohol-impaired driving (1-6). Moreover, some drivers manage to have their records expunged under certain conditions, and many state motor vehicle administration (MVA) offices routinely purge driving records after a set number of years. Therefore, it is reasonable to assume that the typical so-called first-time offender will have had an extensive history of alcohol-impaired driving by the time he or she makes it into the MVA s record system. The current research examines the relative risk of alcohol-related recidivism among drivers with one, two and three or more alcohol-related events (not just convictions) and expands prior research (7) using an updated data set. Our findings suggest that first-time alcohol-related traffic offenders are at a high and significant risk of recidivating even after one alcohol-related event and that alcohol-impaired driving recidivism among first offenders more closely resembles that of multiple alcohol offenders. The results demonstrate that any alcohol-related traffic event (APS BrAC of 0.10 or more, APS breath refusal, and PBJ), not just convictions, should be perceived by the courts, the MVA, and physicians as a marker for future alcohol-related recidivism. The results also suggest that relative risk among females is similar to the risk among males once females have had one alcohol-related event. Introduction The purpose of this study was to determine the statewide alcohol-impaired recidivism rate among Maryland drivers with no, one, two and three or more prior alcohol-related traffic violations (events). In the State of Maryland, an alcohol-related traffic event may result in administrative penalties mandated under APS regulations for failing or refusing the breath alcohol test and

54 criminal penalties mandated for a conviction. In addition, under certain conditions, an offender can be placed on PBJ following a conviction or nolo contendere plea. A closer look at recidivism rates among drivers with so-called first, second and third or more alcohol-related traffic events is warranted because of four factors: 1) the low probability of arrest for alcohol-impaired driving; 2) the practice of expunging and/or purging driver records; 3) our finding in a companion paper (8) that any alcohol-related event (whether administrative in nature or through criminal sanctions and/or diversion programs) significantly and substantially increases the future risk of an alcohol-related event and therefore should be considered a marker for future recidivism; and 4) the perceived leniency with which state legislative and judicial systems handle so-called first-time offenders. Methods We considered all alcohol-related traffic events (DWI/DUI, PBJ and two APS events: BAC of 0.10 or more and breath test refusal). Data were extracted from the Maryland MVA s driver record database after personal identifying information had been deleted from drivers records. Since PBJs are maintained by the MVA as a segregated record, those records were also extracted and linked for analysis. Drivers who had died or moved from the state were excluded. Records were also removed if the driver s license had expired 6 months before December 31, For many driver records, multiple APS, conviction and PBJ records were found for the same date of an alcohol-related event, and these duplicates were removed from the database. Analyses were restricted to drivers who were included in the Maryland Driver License Record or Segregated files and events with dispositions between January 1, 1994, and December 31, All alcoholrelated events from 1973 to 1999 were counted among the prior alcohol-related events. Data on alcohol-related traffic events occurring between 1973 and 1999 were analyzed. Possibly because of administrative and/or procedural factors, conviction counts prior to 1973 were small; therefore, disregarding alcohol-related events prior to 1973, which occurred 20 or more years before the start of the study period, was unlikely to affect any of the estimates. More than 55 million driver records were screened for the study period , representing more than 23 million drivers of interest. Research was conducted using a Maryland driver record database extract from March 2001 and updates prior estimates (7). It can take up to 12 months or more for cases to work their way through the administrative and judicial systems and reach a final adjudication and there is an additional lag period before the outcomes appear on a driver record so using this extract more accurately reflects alcohol-related events occurring during Results In Tables 1-4, estimates based on all Maryland drivers are tabulated by prior alcohol-related event and year, total number of prior alcohol-related events by number of prior events (0, 1, 2, 3+) and year, rate of alcohol-related events per 1,000 Maryland drivers by number of prior events (0, 1, 2, 3+) and year, and proportion of female drivers by number of prior events and year. The tables present summary statistics for the number of drivers and sex at the middle of each year (June 30). Disposition counts are reported for the whole calendar year, and disposition rates are reported as the ratios of disposition counts to the corresponding mid-year driver counts. Table 1 displays the count of drivers by the number of prior alcohol-related events and year. The number of drivers in Maryland increased during this period from about 3.75 million in 1994 to about 4.03 million in In 1999, more than 3.75 million drivers had no prior alcohol-related

55 traffic events; 188,769 drivers had committed one event; 56,971 different drivers had committed two events; and 34,451 different drivers had committed three or more events. Table 1: Number of Drivers by Prior Alcohol-Related Events and Year Number Year of Prior Events All 0 3,533,047 3,553,554 3,584,750 3,625,251 3,678,828 3,752,584 21,728, , , , , , ,769 1,030, ,063 46,673 49,272 51,749 54,396 56, , ,105 25,275 27,490 29,854 32,106 34, ,281 All 3,755,623 3,787,078 3,829,487 3,881,554 3,946,902 4,032,775 23,233,419 Table 2 displays the number of alcohol-related events by number of prior events and year. The number of alcohol-related traffic arrests that resulted in a disposition of any type (APS, criminal, PBJ) increased from 18,628 in 1994 to 21,568 in In 1999, 12,236 events were for first-time offenders; 4,966 were for drivers who had one prior alcohol-related event; 2,279 were for drivers who had two prior events; and 2,087 were for drivers with three or more prior events. Table 2: Number of Alcohol-Related Events by Number of Prior Events and Year Number Year of Prior Events All 0 10,487 11,310 11,140 11,733 12,110 12,236 69, ,579 4,828 4,805 4,892 4,903 4,966 28, ,003 2,218 2,299 2,331 2,329 2,279 13, ,559 1,758 1,715 1,969 1,955 2,087 11,043 All 18,628 20,114 19,959 20,925 21,297 21, ,491 The number of drivers in Maryland increased by 7% from 1994 to 1999 (Table 1) while the number of alcohol-related events increased 14% (Table 2). Thus the number of alcohol-related events in Maryland is increasing twice as fast as the number of drivers. Table 3 displays a rate analysis of the number of alcohol-related events per 1,000 drivers by number of prior events and year. Among drivers with no prior events, the average annual rate of alcohol-related first-time offenders was 3 per 1,000 drivers. Among drivers with one, two and three or more prior alcohol-related events, the average annual rates were, respectively, 28, 44 and 64 per 1,000 drivers

56 Table 3: Rate of Alcohol-Related Events Per 1,000 Drivers by Number of Prior Alcohol- Related Events and Year Number of Year Prior Events All All Remarkably, the average rate of alcohol-related events was increased almost 10-fold (933 percent) by the first prior event, by 1,467 percent by the second prior event and 2,133 percent by the third or additional prior event, relative to drivers with no prior events (Figure 1). However, the rate of a subsequent alcohol-related event only increased by about 230% for drivers with three or more prior events relative to drivers with 1 prior event. Thus the first alcohol-related event results in a larger change in risk for subsequent events than multiple offenses. Furthermore, the rates of subsequent events for first offenders is more similar to that of multiple offenders than drivers with no prior events. Overall, the rates among drivers with no prior events receiving their first event changed relatively little during the study period. However, among drivers with one or more prior alcohol-related event, annual rates declined by about 10 percent between 1994 and Figure 1: Percent Increase of Alcohol-Related Events Relative to Drivers with No Prior Events 2500% 2,133% 2000% 1500% 1000% 500% 933% 1,467% 0% 1 Event 2 Events 3+ Events Number of Prior Alcohol-Related Events

57 Table 4 displays the proportion of female drivers by the number of prior alcohol-related events and year. Females accounted for 50 percent of all drivers during the study period. Females represented 52 percent of drivers with no prior alcohol-related event, but only 17 percent of drivers with one prior event, 11 percent of drivers with two prior events and 7 percent of drivers with three or more prior events. Table 4: Proportion of Female Drivers by Number of Prior Alcohol-Related Events and Year Number of YEAR Prior Events All All The rate of alcohol-related events was examined by sex. The relative risk of a first offense (i.e., an individual had no prior alcohol-related events) is 4.3 times higher for males than for females. However, once a first offense has occurred, the relative risk is almost the same for males and females (1.2 for one prior event, 1.1 for two prior events and 1.1 for three or more prior events). Males are four times more likely than females to have a first alcohol-related event, but once a first event has occurred, males and females are at a similar relative risk of recidivating. Discussion Our research, based on driver data from the entire State of Maryland, produced four major findings: (1) during the study period , the first alcohol-related event proved to be a more powerful statistical risk factor for recidivism than did subsequent alcohol-related events; (2) recidivism among first offenders more closely resembles that of multiple offenders than among drivers with no priors; (3) any alcohol-related traffic event, not just convictions, should be perceived by the courts, the MVA, and physicians as a marker for future alcohol-related recidivism; and (4) the relative risk among females is similar to risk among males--once females have had one alcohol-related event. These findings call into question the use of segregated records, expungement, and the purging of any driver history of alcohol-related violations. It would be unconscionable for a woman with a history of breast cancer to exclude this information from her attending physician because she had gone for a period of years without a related health problem, or to purge or expunge her medical history. However, the traditional approach to deterring alcohol-impaired driving not only allows for expungement of a history of alcohol-impaired driving, or the purging of driver records, but routinely focuses only on convictions for determining highway safety risk. This analysis (and companion papers) definitively demonstrate that, like other diseases, no history of an alcoholimpaired driving event, whether handled through administrative procedures, the criminal justice system, or a diversion program, should be expunged, purged, or segregated from a driver s

58 record. Alcohol-impaired driving is a national health problem and if we are to decrease injuries and deaths on the nation s highways, public health policy should classify first offenders using a broader definition of alcohol-related events [as defined in this paper and a companion paper (8)] instead of the legal criminal definition used by state licensing agencies, state legislative and judicial branches of government, physicians, and public health policy analysts. Once first offenders are properly identified, public policy needs to focus on early intervention and treatment of these first offenders. Acknowledgement This work was supported by Grant R01 AA11897 from the National Institute on Alcohol Abuse and Alcoholism. References 1. Borkenstein RF. Problems of enforcement, adjudication and sanctioning. Proceedings of the Sixth International Council on Alcohol, Drugs and Traffic Safety. Toronto, Ontario, Canada, Beitel GA, Sharp MC, Glauz WD. Probability of arrest while driving under the influence of alcohol. J Stud Alcohol 1975; 36(1): Anda RF, Remington PL, Williamson DF. A sobering perspective on a lower blood alcohol limit (letter to the editor). JAMA 1986; 256(23): Voas RB, Hause JM. Deterring the drinking driver: The Stockton experience. Accident Anal Prev 1987; 19(2): Hingson R. Environmental strategies to reduce chronic driving while intoxicated. Transport Res Circ 1995; (437): National Highway Traffic Safety Administration. Crash Course on Impaired Driving: Maryland Collegiate Conference. Timonium, Maryland, Rauch WJ, Zador PL, Ahlin EM, Raleigh R, et al. Alcohol-impaired driving recidivism among first offenders more closely resembles that of multiple offenders. Alcoholism Clin Exp Res 2001; 25(5):150A. 8. Rauch WJ, Zador PL, Ahlin EM, Raleigh R, et al. Any first alcohol-impaired driving event is a significant and substantial predictor of future recidivism. Proceedings of the 16 th International Conference on Alcohol, Drugs and Traffic Safety. Montréal, Quebec, Canada,

59 A Geographic Analysis of DWI Offenders W.F. Wieczorek and A.Y. Naumov Center for Health and Social Research, Buffalo State College, Buffalo, New York, USA Keywords Alcohol, DWI, cluster analysis, geography Abstract There are almost no studies of the geographic distribution of DWI offenders. Basic information such as whether DWI offenders are randomly distributed in the population or tend to come from specific neighborhoods could have important implications for DWI prevention and interventions. If geographic clusters are identified, anti-dwi efforts can be targeted at specific areas, whereas this type of geographic targeting would not be appropriate if the DWI population is randomly distributed. The objective of this study is to determine whether the home locations of DWI offenders are spatially clustered by using appropriate spatial analytic methodology. All DWI offenders (i.e., any drinking-driving conviction) from in Erie County, New York form the database for this study. Over 15,500 DWI offender home addresses were geocoded and allocated to census tracts and block groups. A spatial scan methodology based on a case-control approach was used to determine whether census tracts or block groups formed significant geographic clusters. Results based on the analysis of DWI offenders at the census tract level and block group level identified a number of statistically significant spatial clusters. The geographic analysis found that specific high and low rate areas could be identified based on official DWI conviction information. The clusters based on block groups provided a refinement of the clusters found at the tract level. The geographic distribution of DWI offenders is clustered and not random, which could be used to target intervention programs. Introduction Drinking and driving offenses (DWI) occur in a number of geographic contexts, which include the road system, places where alcohol is consumed, locations of crashes, locations of arrests, and the home locations where the DWI offenders reside. Despite the various geographic links with DWI, there are relatively few studies that examine these geographic factors. Gruenewald and associates have examined the role of alcohol availability in drinking and driving (1). Wieczorek and Coyle (2) used a spatial regression technique to identify factors associated with a DWI rate in census tracts such as non-skilled occupations, high school education level, percent male population, and white ethnicity (2). In addition, simple spatial cluster analysis was used to suggest that DWI offender home locations were not randomly distributed (3); however, the technique was not able to identify specific clusters of either high or low rate locations. There is great potential for advanced geographic cluster analysis to provide guidance for DWI prevention by identifying specific low and high DWI rate neighborhoods. DWI prevention has not usually

60 been focused on areas that generate a disproportionate number of DWI offenders. The purpose of this study is to determine whether the home locations of DWI offenders are geographically clustered by utilizing advanced spatial cluster analysis. Methods The study was conducted in Erie County, New York, with a population of about 968,000 at the time of the study. The county includes a large urban area (Buffalo), suburban areas and semirural towns. The home address information for all persons (15,551) convicted of DWI (any drinking and driving offense) from in Erie County form the database for the study. The home addresses of the DWI offenders were geocoded using TIGER line files and Erie County tax parcel information. After geocoding, the individuals were allocated to census tracts and block groups. Census population data were used to create DWI rates for the tracts and block groups. The block groups and tracts were analyzed separately to assess the impact of the geographic unit used for analysis and to determine if the cluster solutions showed convergent results. The spatial cluster analytic techniques used was Kulldorff s (4) SatScan algorithm. This is a spatial scan method that uses a scan window to move over the tracts and block groups to sum the number of cases (i.e., DWIs) in the overlapping scan windows to identify specific clusters. The method is a case-control approach that compares the expected number of cases in an area with the number of controls to identify the clusters. The method uses a likelihood ratio statistic that accounts for the multiple comparisons made during the spatial scan to identify statistically significant clusters. SatScan identifies clusters of any shape or size, the specific members of each cluster, and both high and low rate areas. This case-control cluster method controls for the underlying population distribution so that clusters are a not merely a reflection of population density. Results The two main cluster analyses (one at the block group level, the other at the tract level) both found significant clusters of DWI offenders. It is important to compare the simple DWI rate maps with the cluster analysis results. Figure 1 shows the DWI rate map for DWIs at the census tract level. The mean DWI rate at the tract level is /10,000 persons. Visual inspection of this map shows that identifiable groups of tracts have similar DWI rates. Note that the natural breaks method was used to create the intervals on this map (and in Figure 3). The natural breaks method utilizes the variance of the rates to create the specified number of categories by using standard deviation units to define the intervals. This creates intervals with meaningful break points, but with a different number of tracts in each interval group, as compared to quantile interval methods that place the same number of tracts in each interval group. Despite the visual impression of the rates shown in Figure 1, there is no way to know which groups, if any, of the tracts actually form statistically significant clusters

61 Figure 1: DWI conviction rate for census tracts in Erie County, New York Figure 2: DWI clusters at the census tract level

62 The significant clusters identified at the tract level are shown in Figure 2. Cluster 1 is a low rate cluster located in the central urban area (p=.001, 1,123 DWI cases found, 1,812 cases expected). A small geographic area is typical for census tracts in highly urbanized areas, such as those in cluster 1. Cluster 2 is another high rate cluster that includes a large portion of the major suburban areas of the county (5,265 DWI cases found, 4,461 cases expected). Cluster 2 also includes some rural tracts (recognized by their large geographic areas). Cluster 3 is a high rate cluster found in a suburban-urban transition area with substantial heavy industry (p=.001, 367 DWI cases found, 246 cases expected). A visual comparison between Figures 1 and 2 show that the significant clusters overlap with the simple rate map, but are also substantially different. Figure 3 shows the DWI rates at the block group level. The mean DWI rate at the block group level is /10,000 persons. Note that block groups are smaller geographic units than census tracts; census tracts (population range 2,500-8,000) are usually composed of fewer than ten block groups. Greater geographic variability is found at the scale of block groups as compared to tracts (see Figures 1 and 3). More nuances are present in the geographic distribution because estimates of DWI rates for smaller populations are more variable. Visualization of clusters among the block groups is much more difficult than for the tract-level map. The results of the spatial cluster analysis are shown in Figure 4. The finer-grained information from the block group-level analysis resulted in substantially more clusters in comparison to the tract-level analysis (7 clusters vs. 3 clusters). All of the block group clusters were statistically significant at the p=.001 level. At the block group-level analysis, cluster 1 coincides well with the first cluster of the tract-level analysis. They are located in the same urban area and are low rate clusters. Cluster 1 had an expected number of 2,879 DWI cases, whereas only 2,042 were found, which indicates that this cluster includes a larger population than the same cluster at the tract level. Clusters 2 (4,995 DWI cases vs. 4,132 expected) and 5 (787 DWI cases vs. 575 expected) are high rate clusters that show a refinement of the large suburban high rate cluster at the tract-level into two distinct groups. Cluster 5 added some areas that were not included at the tract level. Cluster 2 extended to additional areas towards the urban core, whereas some rural tracts that were included in the high rate tract cluster are no longer in any cluster in the block group analysis. In figure 4, Cluster 3 is a low rate group in a suburban area (273 DWI cases vs. 522 expected) that does not have a comparable cluster at the tract level. Cluster 4 is a low rate group that is an artifact of data reporting because it includes an Indian reservation for which reliable addresses for DWI convictions are not available. At the tract level, this artifact was not identified as a cluster with a possible explanation that the single tract was divided into multiple block groups, which then were identified as a cluster. Cluster 6 is a high rate group (471 DWI cases vs. 321 expected) that coincides with the third cluster at the tract level. Cluster 7 is a small low rate group (18 DWI cases vs. 54 expected) in the central urban area that does not have a corresponding group at the tract level. One noteworthy difference between the two cluster analyses is that the block group clusters tend to include more cases and population than do the comparable clusters at the tract level. Even when the large high-rate suburban tract cluster was split into two clusters of block groups, the total number of cases for the two block-group clusters was greater ((5,782 vs. 5,265)

63 Figure 3: DWI conviction rate for block groups in Erie County Figure 4: DWI clusters at the block group level

64 Discussion The results at the tract and block levels of analysis provide strong evidence that the spatial distribution of DWI offenders is clustered, with some areas having significantly higher rates and other areas marked by lower rates. These clusters are not an artifact of population density because the case-control methodology used for the analysis controls for density and makes appropriate statistical tests to identify the specific cluster members. The results of the analyses at two geographic scales (tracts and block groups) resulted in convergent and complementary findings. The analysis with smaller geographic units provided a greater number of significant clusters, most of which coincided with the results at the tract level. The results suggest that a greater amount of geographic information is available from smaller geographic units, with point level data (i.e., exact home addresses) being the most preferred for spatial analysis. DWI clusters were found in urban, suburban, and rural areas of the county, indicating that clusters are possible within any type of residential area. Although specific analysis of socioeconomic status variables such as income and poverty were not conducted for this study, the cluster results tend to reflect socioeconomic patterns. The low DWI rate urban cluster is found in the poorest areas of the city, while the suburban higher rate clusters are in working class and middle class towns. The results provide a basis to make rational decisions on targeting of resources for DWI prevention and interventions, especially if those interventions focus on areas that generate a disproportionate number of DWI offenders. Primary prevention can be focused at these areas to reduce general alcohol consumption and to provide safe-ride alternatives. Public service announcements (e.g., billboards) should also be targeted toward the high rate cluster areas. Enforcement also could be targeted to the higher rate locations. Although the results strongly support the existence of DWI clusters, there are a number of limitations and issues that require future research. It is possible that enforcement practices may be differential and could cause DWI clusters; however, the data are from a five-year period, which minimizes the impact of short-term DWI enforcement blitzes, and many of the clusters cross police jurisdictions. Future analyses need to examine the relationships between location of the DWIs, alcohol-related crashes, and alcohol outlets with the home locations of the offenders. In additional, spatial cluster analysis that controls for ethnicity, gender, and age are necessary to provide a more complete view of the geographic distribution of DWI offenders. References 1. Gruenewald, P.J., Millar, A.B., Treno, A.J., Yang, Z., et al.. The geography of availability and driving after drinking. Addiction 1996; 91(7): Wieczorek, W.F. & Coyle, J.J. Targeting DWI prevention. Journal of Prevention and Intervention in the Community 1998; 17(1): Wieczorek, W.F. & Hanson, C.E. New modeling methods: Geographic information systems and spatial analysis. Alcohol Health & Research World, 1997; 21(4): Kulldorff, M., Rand, K., Gherman, G., Williams, G., and DeFrancesco, D. SaTScan v2.1: Software for the Spatial and Space-Time Scan Statistics. National Cancer Institute, Bethesda, MD

65 Road Users Behaviour Monitoring in Estonia Antov, D., Rõivas, T. STRATUM OÜ Abstract Rapid increases in motorization have raised concerns regarding accidents and fatalities internationally. Even nations experiencing declining numbers of fatalities, such as Estonia, have made it a major public policy goal to decrease fatalities. Traffic fatalities declined from 491 in 1991 to 204 in 2000 in Estonia. Nonetheless, the Federal Government is intent on another 50% decrease by This paper studies two steps necessary to achieve this decrease, (1) understanding the perceptions of road-use behavior and (2) field observations of driving and pedestrian practices. In May and June subjects were surveyed to assess their perceptions of road-use behavior. Drunken driving, lack of seat-belt use in the rear seat and speeding on rural roads were all perceived to be problems. Generational differences accounted for the greatest differences in perception. The young gave higher scores on most road-use activities, suggesting that better driver education may be needed. The number of fatalities and field observation indicates that treatment of pedestrians is an area in need of attention, especially since the perception is that it is not a major problem. Pedestrians accounted for 39% of the motor-vehicle related fatalities in Estonia in And 70% of drivers were observed to be in violation of the law requiring them to yield to pedestrians. Pedestrians themselves are not without blame, with 26% of pedestrians observed to be in violation at signalized crossings. Finally, speeding continues to be a perceived and observed problem

66