One prevalent argument in support of the utility of environmental preventive interventions to reduce alcohol-related problems rests upon the observation that these problems are broadly represented in the drinking population. While risks related to drinking at the individual level do increase with increased drinking, most aggregate-level drinking problems are attributable to ‘moderate’ and ‘light’ drinkers. Hence, the ‘prevention paradox’. While treatment programs reach only indicated problem drinkers (e.g. those entering treatment programs) and educational preventive interventions reach only selected subpopulations of users (e.g. young people in school), environmental prevention programs reach everybody, including those at a low individual-level risk who are responsible for the many drinking problems. This is a splendid story but for several major problems. First, no one seems to know what attributes signify a ‘light’, ‘moderate’ or ‘heavy’ drinker. Second, attempts to solve this problem using other even more problematic concepts related to individual experiences with alcohol, such as ‘intoxication’, replace externally defined objective criteria with reports about internally defined subjective states. Third, and most problematic, problems related to alcohol, even among ‘light’ or ‘moderate’ drinkers, may have nothing to do with ‘light’ or ‘moderate’ drinking. ‘Light’ and ‘moderate’ drinkers may only have problems when drinking ‘heavily’. Thus, one can see why progress towards understanding the prevention paradox has stalled. The central methodological problems with these studies are not only respondent attrition and subjective bias, noted by Rossow & Romelsjö[1], but also a category error reflecting our persistent focus upon individual drinkers rather than their drinking [2, p. 43]. While characteristics of drinkers, for example frequency and quantity of use, represent dispositions to drink in certain ways, patterns of drinking, actual amounts consumed at specific times and places are the cause of drinking problems. Against this background, the excellent work by Rossow & Romelsjo has many strengths, not the least of which is dealing directly with the issue of subjectivity in assessments of problem outcomes; so our comments reflect not on their work, but on the larger scientific frameworks in which this work is conducted. Standing on the shoulders of Strauss & Bacon [3], alcohol epidemiology has failed to ask two central questions: ‘How does drinking affect problems?’ and ‘How do problems affect drinking behavior?’ Despite many years of study of statistical associations between measures of drinking and problems [4-7], these questions have remained unanswered. The general impression given by this literature is that it is enough to correlate harms related to drinking with characteristics of drinkers (e.g. their frequencies and quantities of use). However, because harms related to drinking shape drinking behaviors, and vice versa—a negative feedback loop—estimates of these correlations are biased [8]. These biases are not ameliorated with increased sample sizes or refinements of measurement procedures. Rather, they require detailed attention to (a) the theoretical and mathematical foundations of models that explicitly relate drinking to problem outcomes and (b) statistical issues in the assessment of dynamic systems relating drinking-to-problems-to-drinking (i.e. non-recursive equation models), a point that has not seeped into epidemiological argument [9, 10]. The question implied by the prevention paradox is not ‘What drinkers produce the bulk of alcohol problems?’, but ‘What drinking produces the bulk of alcohol problems?’. A well-developed theoretical and mathematical treatment of the relationships between drinking and problems, and suitable data, can be used to answer the second question, teasing out specific drinking acts (e.g. occasions of use at different levels) and their relationships with problems (e.g. dose–response models). This is not accomplished by correlating standard self-report measures of drinking (e.g. frequency and quantity) with problem outcomes. These measures represent the dispositions of drinkers (i.e. how much each is likely to drink), not drinking (e.g. number of days drinking at specific levels). Nor do such correlations constitute a theoretical model of the manner in which drinking leads to problems. In order to understand drinking and problems we must study drinking, not drinkers. In order to properly characterize the prevention paradox we need to know, at the population level, how drinking contributes to harmful outcomes.
Search from 250 M+ research papersSearch from 250 M+ research papers
Open Access
Apr 1, 2009
Peter Miller + 2
