Distinguishing causes and correlates of drinking and driving

Abstract

The overall framework for this review For many social problems it is extremely difficult to determine cause-effect relationships with any degree of certainty. Drinking and driving is one such social problem where causes are difficult to delineate. A major distinction exists between causes and correlates of driving while impaired (DWI). For a correlation,* only one condition needs to be satisfied-i.e., there is a relationship between the variable in question and impaired driving. However, in order to infer causation of drinking and driving, three conditions should be satisfied: (1) a relationship between the variable in question and impaired driving must exist; (2) the suspected cause must precede the effect; (3) alternative explanations for an empirical relationship must be ruled out. Delineating probable causes of drinking and driving is important for the development and implementation of effective countermeasures. Without a better understanding of causal variables, resources may be wasted by addressing behaviors that do not translate into reducing traffic collisions caused by impaired driving. If a factor is a correlate but not a cause, it may be useful for targeting high-risk groups for interventions, such as educational or enforcement programs. However, attempts to modify a factor that is a correlate but not a cause could be unsuccessful in reducing subsequent drinking and driving behavior. Undoubtedly multiple causes exist for drinking and driving. Human behavior is complex, and for a given individual there are likely to be several reasons for engaging in any particular activity. For a group of individuals, reasons for a behavior can increase substantially. The orientation of this review is to apply the principle of parsimony to the interpretation of the results. The law of parsimony suggests that complicated or more abstract scientific explanations should not be hypothesized until simpler explanations have been ruled out.' Causes can be either direct or indirect. A parsimonious approach delineates that direct paths are more likely to be causal than indirect ones. Although it is theoretically possible that an indirect variable (a variable more distal in the causal chain) is more causal than a direct one (a variable more proximal in the causal chain), the likelihood of this occurrence is low. Also, consistent with a parsimonious approach, the strength of statistical relationships is considered in terms of deducing causation. The stronger the statistical relationship, the stronger the likelihood that the antecedent variable is causal.2 Confounders and intermediate variables In this paper a confounder is defined as a variable that wholly or partially accounts for, or masks, an association with a third variable.2 In the case of drinking and driving, a confounder is a correlate of DWI that is associated with other predictors but is not a consequence of those predictors. Throughout this review, consistent with an epidemiological approach, sociodemographic variables are treated as confounders.2 Age and gender are likely the two most important confounders.2 While males and younger people are much more likely to drive while impaired, these two variables have little etiological importance. It is difficult to imagine that being male or within a certain age group causes one to drink and drive. Rather, direct characteristics that are more predominant among males than females (perhaps the propensity to take risks or to drink larger quantities of alcohol) likely better explain DWI occurrences. An additional reason for treating demographic variables as confounders is that they are either impossible or very unlikely to be altered through interventions.2 Although sociodemographic variables are treated as confounders from an etiological perspective, this does not mean that studies on demographic correlates of DWI are not valuable in terms of developing intervention strategies. In fact, interventions targeting certain socio-demographic groups can be quite successful, such as raising the drinking age in order to reduce DWI offenses and alcohol-related collisions among younger people. …