Any learning task, particularly a complex one such as driving an automobile, needs time and experience to arrive at a good performance. New licensees have higher crash rates than experienced ones (Laberge-Nadeau et at., 1992) at any age. Young licensees, 16-24 years old and particularly those 16-19, are overrepresented in road crashes. In Quebec in 1992, the young licensees 16-24 years old, were involved in 23% of injury car crashes although they represented only 13% of all licensees and 12% of the Quebec population (Letendre, 1995). Young men are at a 2.64 higher risk than 25-year-olds and older.
Materials and method
This study is population-based covering two periods, two years before and after the reform, concentrating particularly on the involvement in crashes of new licensees as drivers. Straightforward descriptive analyses will be followed by statistical models to evaluate the pre and post-periods.
A special file was created by Pichette and Bisson (1994) from the Societe de I’assurance automobile du Quebec (SAAQ). The SAAQ is a public corporation that insures all Quebecers for motor vehicle injuries; it also regulates and administers access to driving licenses. This special file contained all persons who started the process of obtaining a learner’s permit for the first time for class 5 (private car) of the Province of Quebec between March 1, 1989, and February 28, 1993, a population of about 400 000 learners of all ages. Before the reform, this process started by obtaining a learner’s permit, and after the reform by attempting to pass the theory exam.
The population we have studied was limited to new licensees whose learning period was 270 days or less and for whom a full year of crash records was available after they obtained the driving license. Before the reform 48.6% of the men obtained the license within 90 days, 22.8% between 91 and 180 days, 9.4% between 181 and 270 days, and 19.2% took longer than 270 days. After the reform, 84.5% of the men obtained the license between 91 and 180 days, 9.8% between 181 and 270 days, and 5.7% took longer than 270 days.
Model and Variables
For dichotomy variables Xii’ the value of exp(~) is the odds ratio of the crash event for those licensees characterized by x.. = 1 compared to those with x.. = O. For continuous lj lj variables exp(~) is the factor of change in the odds when the explanatory variable increases by one unit. In order to adjust for the panel effect, i.e. for possible within-subjects correlation, the generalized estimation equations (GEE: Liang & Zeger, 1986; Sager et at., 1988) technique was ~ applied when estimating the parameters of the logistic regression model. The ratio ~ I std(~) was used to testing whether ~ = 0 or not with the standard Nona distribution, i.e. the asymptotic approximation.
The explanatory variables can be grouped into two sets, time-independent and time-dependant variables. These variables, except for the economic indicators, are the ones that are available in the files of the SAAQ for every new licensee in the Province of Quebec. The first set contains only dichotomous variables, namely.