The Québec Alcohol Ignition Interlock Program: Impact on Recidivism and Crashes L. Vézina Highway Safety Research and Strategy Société de l assurance automobile du Québec Keywords Alcohol, ignition interlock, conviction, recidivism, crashes Abstract The alcohol ignition interlock program began operating in Québec in December 1997. Someone convicted of driving while impaired (DWI) is allowed under the program to drive a vehicle on condition it be equipped with an ignition interlock, and the individual's voluntary participation allows for a shortening of the period of licence revocation. This evaluation seeks to measure the program s impact on both DWI recidivism and crashes. The methodology is a cohort study comparing a group comprised of individuals convicted of DWI who took part in the ignition interlock program, and a control group made up of drivers convicted during the same period who did not participate. The data concerning first-time DWI offenders and that pertaining to repeat offenders is presented separately. Participation in the interlock program resulted in a reduction in the repeat DWI rate by 80% during the first 12 months for first-time offenders and by 74% during the first 24 months among repeat offenders. Figures for night crashes involving no more than one vehicle, when compared to overall crash data, also tend to show the effectiveness of the program in lowering the incidence of impaired-driving mishaps. In both cases, however, the effect tends to disappear after removal of the interlock, which leads to the conclusion that it serves more as a restraining device than a tool for driver rehabilitation. This suggests that extending the period of interlock use should be considered, at least for drivers who are likely to become repeat DWI offenders. Introduction The alcohol ignition interlock program was implemented in Québec in December 1997. Other countermeasures were introduced simultaneously included administrative license suspension, vehicle impoundment for driving while disqualified and mandatory evaluation for DWI repeat offenders. Enrollment in the Québec ignition interlock program is on a voluntary basis, and allows a motorist sentenced for impaired driving to obtain a restricted licence with authorization to
operate a vehicle equipped with an interlock device. When the ignition interlock program was implemented, it was aimed at all drivers sentenced for DWI. In fact, it was aimed as much at drivers who had been sentenced for the first offence as those who had repeated the offence, since the main criteria for eligibility were having served the entire driving prohibition period as provided for under the Canadian Criminal Code (CCC) and not having other active driving suspensions on record. Originally, a driver could therefore, upon finishing the driving prohibition of a 3, 6 or 12-month period (either a first, second or third conviction within five years), obtain a restricted licence for the time remaining in the penalty period of a total duration of one, two or three years. However, because of amendments made to the CCC, the program was solely available to drivers who have been sentenced for an initial offence within the preceding five years during the period between July 1999 and December 2001. To date, studies carried out on the interlock device tend to show that this is an effective means for reducing DWI recidivism (Morse and Elliot 1, Beirness et al. 2, Tippetts and Voas 3, Beck et al. 4 ). Accordingly, the lower rates of repeat offences of this kind observed among the group of drivers using an interlock device, compared to a control group, vary from 60% to 75% during the period in which the device is installed. However, most of these studies also show that the effect seems to wane after the device is removed. In addition, the impact on crashes has not been well documented to date as few studies have broached the topic. A preliminary evaluation of the Québec program, based on first-year results, indicated a decrease in the rate of impaired driving offences as well as a decline in the rate of crashes involving bodily injuries further to the installation of an interlock device (Dussault and Gendreau 5 ). Although this initial evaluation has revealed encouraging findings, the preliminary assessment of the before-during-after period remains incomplete due to the recent time aspect of the data and due to the absence of a control group. The objective of this study is to evaluate the impact, after a three-year period, of the interlock device on repeat DWI offences by differentiating the effects among first-time offenders from repeat offenders. It also aims to shed a certain amount of light on the impact this program has on the entire range of motor vehicle crashes and more specifically on alcohol-related crashes. Methods The methodology is a cohort study comparing a group comprised of individuals convicted of DWI who took part in the ignition interlock program (study group), and a control group made up of drivers convicted during the same period who did not participate. The study group is comprised of drivers sentenced for impaired driving after December 1997 who enrolled in the ignition interlock program, whereas the control group is made up of drivers sentenced during the same period who did not participate in the program. For every individual meeting this criterion, the first sentencing as of that date was used as a benchmark penalty for calculating repeat impaired driving behaviour. The study group is composed of 42,563 drivers sentenced for impaired driving between December 1, 1997, and January 26, 2001. Of that number, 81% (34,405/42,563) had been sentenced for the first time in a five-year period and 19% were repeat offenders (8,158/42,563). The latter group comprised
drivers who had been convicted for a first time but who were sentenced for more than one incident at the same time, giving rise to a penalty of two years or more. In that same period, 9,896 drivers obtained a restricted licence (8,846 first-time offenders and 1,050 repeat offenders) that allowed them to have an interlock device installed; this group accounted for 26% of the first-time offenders and 13% of repeat offenders. The percentage of repeat offenders with an interlock device is lower since that group was not eligible for the program during the period between July 1999 and December 2001. Drivers who display repeat behaviour by continuing to drink and drive are selected according to criminal sentencing or an immediate suspension of their licence, as related to an offence committed after the effective date of the benchmark penalty. Single-vehicle nighttime crashes (occurring between 9 p.m. and 6 a.m.) were used as a surrogate measure for alcohol-related crashes. The primary analysis technique used in this study is the survival analysis. This analysis allows us to take into account the length of time elapsed between the beginning of the penalty period and the occurrence of a first event (repeat offence or crash), compared to other analysis techniques that limit consideration of the rate of incidence. It also provides the advantage of being able to process censored data. Thus, the survival curves, according to the Kaplan-Meier method, are used to illustrate the probabilities of non-repeat behaviour in both groups at different times (t), based on the beginning of the penalty period and the Cox s model of proportionate rates allows us to introduce co-variables such as the characteristics of drivers belonging to either group. However, to ensure the validity of the Cox model, the impact of the factor we are seeking to measure must be constant over time. If that condition is not met, we must perform a stratified analysis over time, making the survival for each period conditional to the survival observed during the preceding period. In our case at hand, the effect of the interlock device program varies over time, so that stratification is necessary to ensure a certain degree of homogeneity. The stratification retained is different, depending on whether we consider first-time offenders or repeat offenders but the cutoff is rather clear in both cases: 12 months for the first group, and 24 months for the second. These periods closely correspond to the period with the interlock device (including the waiting period) and the post-interlock device period. Results - Demographic characteristics Drivers sentenced for impaired driving are mainly men (89 %) with an average age of 36.7 years. However, there are a few statistically significant differences between the interlock program participants and the drivers making up the control group. Among the group of first-time offenders, we find a greater proportion of men enrolled in the ignition interlock program and of a slightly older age, whereas among repeat offenders, those participating in the program are on average younger than the drivers without an interlock device.
The tracking period of all drivers in the study lasted 622.5 days on average. The length was greater for repeat offenders and drivers who had an interlock device. The average length of participation in the program was 209.9 days for first-time offenders and 348.8 for repeat offenders. - Impact on DWI recidivism First, the survival curves were traced for both groups according to the Kaplan-Meier method (figure 1). The differences observed between the group with interlock devices and the control group were statistically significant and indicate a lower recidivism rate for the group so equipped for the entire period covered. However, we noted changes to the appearance of the survival curves subsequent to 12 months of follow-up in the case of first-time offenders and after 24 months among repeat offenders, as the curves tend to come together beyond those two points. Incidentally, this confirms the stratified analysis according to the Cox model (table 1). Thus, among first-time offenders, the risk ratio of DWI recidivism for drivers with an interlock device as opposed to the control group was 0.201 for the first 12 months, then climbing to 1.218; both results are statistically significant. Therefore, the risk of falling into repeat drunk-driving behaviour for drivers with an interlock device is significantly lower during the first 12 months, or during the period when the device can be installed and that, thereafter, it becomes significantly higher than that observed among drivers in the control group. Among the group of repeat offenders, the risk of recidivism (RR=0.260) for the first 24 months among drivers with an interlock device was significantly lower while after that period it was basically the same in both groups, even though the risk ratio is reported as 1.908. Nonetheless, it is important to note that the gains observed at the outset of the period broadly offset the rises that occur afterward, resulting in significant gains for the entire period. This applies both to first-time offenders and repeat offenders. Table 1: Stratified analyses over time with the Cox model for DWI recidivism Risk ratio* P-value First-time offenders 12 months > 12 months Repeat offenders 24 months > 24 months 0.201 1.218 0.260 1.908 * adjusted for gender and age <.0001 0.0059 <.0001 0.0870
First-Offender Group Repeat-Offender Group 1 1 Proportion of non-repeat DWI 0,98 0,96 0,94 0,92 Proportion of non-repeat DWI 0,98 0,96 0,94 0,92 0,9 0 3 6 9 12 15 18 21 24 27 30 33 36 39 month 0,9 0 3 6 9 12 15 18 21 24 27 30 33 36 39 month Figure 1: interlock no-interlock interlock no-interlock Survival curves according to Kaplan-Meier for DWI recidivism - Impact on crashes The survival analysis on single-vehicle nighttime crashes conducted using the Kaplan-Meier model (Figure 2) gives rise to non-significant differences between the groups of drivers with or without interlock devices, both as regards first-time offenders and repeat offenders. After we stratify the results, however, certain differences come to light (see Table 2). Accordingly, we noted that the risk of crash is not significantly different during the first few months after the penalty period begins and that, thereafter, the risk among drivers with an interlock device becomes greater than that of drivers belonging to the control group. Risk ratios were also computed for all crashes and we observed that the risk of crashes among drivers with an interlock device is greater in all cases. Table 2: Stratified analyses over time with the Cox model for Single-vehicle nighttime crashes (and all crashes)* Risk ratio** P-value First-time offenders 12 months > 12 months Repeat offenders 24 months > 24 months 1.053 (3.557) 1.468 (1.652) 0.464 (2.160) 4.941 (3.867) * Results for all crashes appear in parentheses ** Adjusted for gender and age 0.8495 (<.0001) 0.0163 (<.0001) 0.1406 (<.0001) 0.0116 (<.0001)
First-Offender Group Repeat-Offender Group 1 1 Proportion without crashes 0,995 0,99 0,985 Proportion without crashes 0,995 0,99 0,985 0,98 0 3 6 9 12 15 18 21 24 27 30 33 36 39 month 0,98 0 3 6 9 12 15 18 21 24 27 30 33 36 39 month interlock no-interlock interlock no-interlock Figure 2: Survival curves according to Kaplan-Meier for Single-vehicle nighttime crashes Discussion The findings of this study have confirmed the effectiveness of the interlock program to reduce DWI recidivism while the device is in place. However, we noted that the rate of recidivism observed after the penalty period runs out is either significantly higher among first-time offenders, or the same among repeat offenders who had an interlock device, compared to the control group. These last results incidentally are similar to observations made by Tippetts and Voas who noted an increase in repeat behaviour of this type among first-time offenders and repeat offenders, after the removal of the device, and those of Beck et al. who found no significant differences among repeat offenders beyond the first 12 months. Therefore, at this stage, an interlock device appears to be more of an issue of contention than a means of rehabilitation since the gains seem to dissipate when the device is removed from the vehicle. However, it has been noted that the gains realized early in the period are sufficiently great to show that, despite all, a reduction in recidivism is observed over the entire period. In terms of the impact of the program on alcohol-related crashes, the findings reported as to single-vehicle nighttime crashes seem, at first glance, to be less conclusive since the risk ratios calculated during the first few months (1.053 among first-time offenders, and 0.464 among repeat offenders) are not statistically significant. Nevertheless, these findings merit special attention since it is important to consider the fact that this indicator could be contaminated by the effect of crashes not related to alcohol and that it is also very sensitive to a difference in exposure between the two groups. Although nighttime crashes involving just one vehicle represent the best approximation available for crashes attributable to alcohol consumption, this measure is not perfect as it also takes into account a certain number of crashes not related to alcohol. That is why these results become more interesting when compared with the results obtained for all crashes combined.
Furthermore, the factor of exposure can also allow us to qualify the findings obtained for crashes since the fact of being involved in a crash reflects of course the conduct of an individual but also reflects to a great extent the distance travelled by that individual. First, we can advance the hypothesis that drivers who chose an interlock device have a greater need for their vehicle and that they generally travel a higher number of kilometres. This was observed by Morse and Elliot in a previous study. Next comes the fact that the drivers subjected to a conventional penalty simply do not have the right to drive and, even though a substantial number continue to drive anyway, they tend to cover small distances. This would explain the decrease in the risk ratio of total crashes for first-time offenders with an interlock device once the penalty period is over, or when the drivers of the control group can regain their licence. Therefore, once distinctions have been made as to the limits of the indicator chosen and the measurement of exposure, the results obtained in terms of crashes involving just one vehicle at night also tend to show the effectiveness of the ignition interlock program over the period during which it is installed. The voluntary aspect of participation in the program also raises the question of a potential selfselection bias. Indeed, it is possible that the drivers taking part in an ignition interlock program are people more inclined to comply with set rules and that this would be the major reason for the relative decrease observed among that group. However, as pointed out by Tippetts and Voas, it is not very likely that the self-selection bias could alone account for the decrease, given the greater rate of recidivism observed after the device s removal. The fact that the risk ratios of DWI recidivism and crashes potentially attributed to alcohol of the participants in an interlock program are significantly higher following the removal of the device also raises a few questions, such as the absence of the rehabilitation effect. However, it is important to once again consider the question of relative exposure. When we compare the risks after a penalty period, our analysis presupposes that the exposure of the interlock group is relatively comparable to that of the control group, even though this might not be entirely the case. We do know that a high number of drivers opt not to regain their driver s licence after a penalty period, and we could suppose that this might affect the drivers in the control group to a greater extent, since the participants in the interlock program had expressed their desire to regain their licence beforehand. The proportion of non-reintegration is perhaps smaller in Québec due to the vehicle-seizing policy in effect for driving while disqualified, but could reach 50% according to Tippetts and Voas. A way in which to solve the problem would be to take into consideration solely those drivers who recovered their driver s licence at the end of their penalty period, and conduct a new study of this subgroup to ensure all analyses are conducted on the same basis of comparison. This would be a fairer way to gauge the level of rehabilitation attained by the device. Findings from this second assessment of the Québec ignition interlock program confirmed its effectiveness in reducing the recidivism in driving while impaired (DWI). The findings also
suggest that the impact on alcohol-related crashes ought to be evaluated more accurately through additional studies. However, in both cases we noted that the effects tend to diminish beyond the periods covered by the interlock device. Therefore, this infers we should extend the interlock device use period, at the very least, among drivers showing a major recidivism risk. References 1 Morse, B.J. and Elliot, D.S. Effects of Ignition Interlock Devices on DWI Recidivism: Findings From a Longitudinal Study in Hamilton County, Ohio. Crime & Delinquency, Vol. 38, No. 2; pp. 131-157, 1992. 2 Beirness, D., Marques, P.R., Voas, R.B. and Tippetts, S. Evaluation of the Alberta Ignition Interlock Program: Preliminary Results in C. Mercier-Guyon (Editor). Alcohol, Drugs and Traffic Safety T '97: Proceedings of the 14 th International Conference on Alcohol, Drugs and Traffic Safety. Annecy, France; pp.193-199, 1997. 3 Tippetts, A.S. and Voas, R.B. The Effectiveness of the West Virginia Interlock Program. Journal of Traffic Medicine, Vol. 26 (1-2), pp. 19-24, 1998. 4 Beck, K.H., Rauch, W.J., Baker, E.A. and Williams, A.F. Effects of Ignition Interlock License Restrictions on Drivers With Multiple Alcohol Offences: A Randomized Trial in Maryland. American Journal of Public Health 89(11), pp. 1696-1700, 1999. 5 Dussault, C and Gendreau, M. Alcohol ignition interlock: One-year s experience in Québec. In Proceedings of the 15 th International Conference on Alcohol, Drugs and Traffic Safety, May 22-26, 2000, Stockholm, Sweden, 5 p, 2000.