Analysis of a large registry-based study of prescription drugs and road traffic crashes
Abstract
Drugs (alcohol, illicit, prescription or over-the-counter) have a potential effect on the skills needed for driving, a task that involves a wide range of cognitive, perceptual and psychomotor activities.
In what concerns medicinal drugs, disentangling their impact on road traffic crashes is a complex issue for several reasons: 1/ the large variety of pharmaceutical classes, with various prevalence of use in the general population; 2/ the confounding underlying health conditions; 3/ the potential medicinal benefits of drugs that may lead to improved rather than impaired driving ability; 4/ the adaptive behaviors: whilst on medication, people may pay more attention to compensate for changes in perceived risk; 5/ the dose, cumulative dose and duration of drug consumption (prevalent, intermittent, incident); 6/ the co-consumption and interaction of drugs,… [Elvik, 2012]. As a result, there is a relatively small epidemiological
literature examining the associations between medicinal drugs and impaired driving. A major approach to this complex subject relies on the use of population-registries data such as those conducted in UK [Barbone et al., 1998; Gibson et al., 2009], Norway [Engeland et al., 2007], Finland [Karjalainen et al., 2012] or France [Orriols et al., 2010]. An advantage of registry-based research is the potential to study associations between rare exposures and outcomes in a population large enough to provide sufficient precision. On the other hand, these studies can lead to high-dimensional and/or large-scale datasets, whose analysis needs to be addressed using appropriate statistical and computational techniques.
In this work, we report the use of shrinkage techniques [Tibshirani, 1996] in exploratory analysis of the large study of prescription drugs and road traffic crashes described in Orriols and colleagues, 2010; Avalos and colleagues, 2012, 2013. We insist on practical aspects, by detailing model selection and stability problems, implementation issues and discussing the obstacles for the use of these techniques in epidemiological studies.