Studies
A Longitudinal Survival Analysis of Drivers with Multiple Alcohol-Related Traffic Offenses: Fifth Year Follow-Up of a Randomized Ignition Interlock License Restriction Trial in Maryland.
Alcohol-impaired driving continues to be a major public health problem in the United States. During 2000, alcohol was involved in 40% of all traffic fatalities (1). One promising countermeasure to alcohol-impaired driving is the ignition interlock license restriction program. Ignition interlock devices are designed to prevent an alcohol-impaired driver from starting and operating a motor vehicle. License restriction means that drivers are approved for license reinstatement on the condition that they agree to a license restriction prohibiting them from operating a vehicle without an ignition interlock device. The conditional interlock license restriction is prominently displayed on the driver’s license of each program participant. Ignition interlock license restrictions have shown potential for reducing recidivism among drivers with multiple alcohol-related traffic offenses (2). Later analyses of this same cohort examined recidivism rates 3 and 4 years after study entry (3, 4). In the current paper, new recidivism rates for the one year interlock license restriction program and second (i.e., first year post-interlock) through fifth year (i.e., fourth year post-interlock) of follow-up are presented using newly developed electronic data abstraction methods. There was no statistically significant difference in the recidivism rates for the interlock license restriction (4.0%) and control groups (4.8%) in the fifth year after study entry.
Introduction
A statewide-randomized trial of ignition interlock license restrictions was performed in Maryland
and the results of that program evaluation were reported in the American Journal of Public Health
(2). Participants were 1,387 drivers with multiple alcohol-related traffic offenses whose driver’s
licenses were either suspended or revoked for multiple alcohol offenses and who were eligible for
license reinstatement after undergoing a variety of treatment programs. Drivers were randomly
assigned to participate in an ignition interlock license restriction program or the usual postlicensure
treatment program (control group) lasting 12 months (2-4). Participants in the interlock program had a
restriction on their drivers’ licenses indicating they could drive only a vehicle equipped with an
ignition interlock. Drivers who owned a car had to have the device installed within 45 days; those
who did not own a car had to sign a waiver that they would not drive a car unless it was equipped
with an interlock. The interlock license restriction and customary treatment programs were
administered by the MVA rather than the courts. This approach ensured greater consistency of
case management and handling of license restrictions and allowed the MVA to monitor and
enforce compliance with the license restrictions.
The initial results in this cohort (2) were reproduced using a more standardized data abstraction method (3, 4). Although the recidivism rates for both the interlock and control groups were higher than those in the initial report, the analysis confirmed the earlier results. The current study differs from the previous analyses by using a standardized electronic data abstraction method, counting alcohol-related “violations” rather than “arrests,” and including all alcohol-related recidivism events by a repeat offender (i.e., not just the first event) in the analysis. All data reported in this paper uses the revised methodology.
The purpose of the current study was to determine the effectiveness of the ignition interlock license restriction program in reducing alcohol-related traffic recidivism 5 years after program implementation (i.e., in the fourth year after the ignition interlock license restriction program ended). In particular, we were interested in determining whether the convergence in alcoholrelated traffic recidivism between the interlock and control groups, observed during the second, third and fourth years, would be maintained during the fifth year. We were also interested in determining whether the recidivism rates of interlock and control group participants would increase significantly beyond the levels observed during the first 4 years.
Methods
Data for this study were provided electronically by the Maryland Motor Vehicle Administration
(MVA) after personal identifying information had been deleted from drivers’ records. Data from
this file were matched to our earlier data file, an Access database into which data had been
manually abstracted and double key-entered. We then compared the recidivism rates calculated
using the electronic and manual abstraction methods. The differences were minor and did not
change any of the conclusions previously reported. An additional manual comparison of
violation codes revealed no systematic biases. We also used alcohol-related “violations” rather
than “arrests,” and included all alcohol-related recidivism events by a repeat offender (i.e., not
just the first event) in the analysis. Given the randomized nature of the design, all participants
assigned to the interlock condition were included in the analysis whether or not they had the
device installed, as appropriate under the intention-to-treat design.
Thirty-one drivers died and 43 moved out of state during the study period. There were no statistically significant differences between the interlock and control groups in the proportions of drivers who died (Chi-Square = 0.90, d.f. = 4, p = 0.92) or moved out of state (Chi-Square = 7.37, d.f. = 4, p = 0.12). The analysis did not control for drivers who died or moved out of state during the study period, which may result in a slight overestimation of survival rates.
Results
During the first year of the study, 26 of 693 offenders in the interlock license restriction program
(3.8%) and 62 of 692 offenders in the control group (9.0%) committed an alcohol-related traffic
violation. This difference was statistically significant, with a relative risk of 0.40 (95% CI 0.25,
0.63). Being in the interlock program reduced a driver’s risk of committing an alcohol-related
traffic violation by about 60 percent. During the second year, when the ignition interlock license
restrictions were removed, the alcohol-related traffic recidivism rates for anyone recidivating in
year 2 in the two groups converged and were not statistically significantly different. During the
second year, the ignition interlock group had a slightly lower recidivism rate than the control group
(6.9% vs. 8.1%), although the recidivism rates of the two groups were not statistically significantly
different. During year 3, the rates were higher for the interlock group (7.8%) than for the control
group (6.5%) but the difference was not statistically significant. In year 4, the rates were also
higher for the interlock group (8.8%) than for the control group (7.1%) but the difference was not
statistically significant. Thus, the reduction in alcohol-related traffic recidivism for the interlock
group was limited to the first year of the program, when the interlock license restriction program
was in effect, and there was no evidence that the first-year benefits extended into subsequent years.
Being in the interlock license restriction program significantly reduced a driver’s risk of
committing an alcohol-related traffic violation by about 60 percent during the first year when the
program was in effect (Chi-Square = 15.78, d.f. = 1, p < 0.0001). Contrary to first year findings,
second, third, fourth, and fifth year recidivism rates of drivers in the interlock program and
standard treatment (control) group converged and were not statistically significantly different
(Table 1).
| Table 1: | Recidivism Rates in the Interlock (n = 693) and Control Groups (n = 692) by Year of Follow-Up, Using Electronic Abstraction and Replacement | ||
| Year | Interlock | Control | |
| 1* | 0.038 | 0.090 | |
| 2++ | 0.069 | 0.081 | |
| 3++ | 0.078 | 0.065 | |
| 4++ | 0.088 | 0.071 | |
| 5++ | 0.040 | 0.048 | |
| *Chi-Square = 15.78, d.f. = 1, p < 0.0001 ++Not statistically significant (95% Confidence) |
|||
During year 5, 4.0% of the 693 drivers in the interlock group (n = 28) and 4.8% of the 692 drivers in the control group (n = 33) recidivated (Table 2). This difference was not statistically significant (Chi-Square = 0.44, d.f. = 1, p = 0.51). After 4 years of increasing recidivism rates in the interlock group, the rates for year 5 are much lower for both the interlock and control groups.
| Table 2: | Recidivism Among Drivers in the Interlock and Control Groups at the Fifth-Year Follow-up | |||||||
| Group | Drivers (N) | Drivers with Violation in Year 5 | % of N+ | |||||
| Interlock | 693 | 28 | 4.0 | |||||
| Control | 692 | 33 | 4.8 | |||||
| Total | 1,385 | 61 | 4.4 | |||||
| +Chi-Square = 0.44, d.f. = 1, p = 0.51; not statistically significant | ||||||||
Figure 2 displays cumulative survival curves for the 5-year study period for all drivers who committed an alcohol-related traffic violation, including drivers with a prior alcohol-related traffic offense during years 1, 2, 3, and/or 4. It should be noted that at the time of randomization there were no statistically significant differences in the number of previous alcohol-related traffic violations between the interlock group (mean = 3.66, SD = 1.45) and the control group (mean = 3.65, SD = 1.28). Five year survival effect of the interlock license restriction program was also analyzed using an Anderson-Gill Multiplicative Hazards Model, stratified on year. The interlock license restriction program was found to reduce the hazard ratio by about 12 percent. However, this reduction was not statistically significant (Chi-Square = 2.24, d.f. = 1, p = 0.135).
In addition to overall recidivism rates, we also examined first-time recidivism (i.e., drivers who did not recidivate in years 1, 2, 3 and 4) among drivers in the interlock and control groups during year 5. Four percent of drivers in the interlock group (n = 21) and 4.5% of drivers in the control group (n = 23) recidivated for the first time since study enrollment during year 5 (Table 3). There were no statistically significant differences between interlock and control group drivers who recidivated for the first time during year five (Chi-Square = 0.15, d.f. = 1, p= 0.70). The positive effects of the interlock program do not extend over the 5-year period of study. At the end of 5 years, the overall effects of the interlock license restriction program are not statistically significant (Chi-Square = 0.63, d.f. = 1, p = 0.43).
Figure 2: Survival Until Recidivism, Years 1-5
| Table 3: | First-Time Recidivism Among Drivers in the Interlock and Control Groups During the Fifth-Year of Follow-up | ||||||||||
| Group | Drivers (N) | Drivers Without Violation in Years 1-4 (N) | % of N | Drivers with 1st Offense in Year 5 | % of N+ | ||||||
| Interlock | 693 | 528 | 76.2 | 21 | 4.0 | ||||||
| Control | 692 | 516 | 74.6 | 23 | 4.5 | ||||||
| Total | 1,385 | 1,044 | 75.4 | 44 | 4.2 | ||||||
| +Chi-Square = 0.15, d.f. = 1, p = 0.70; not statistically significant | |||||||||||