the effect that would be obtained by targeting all youths with a universal program. Nothing is known about these possible trade-offs.3 The pieces of evidence available are insufficient to support a conclusion about the optimal targeting strategy for prevention efforts. Clearly, additional research is needed.
For example, it would be useful to identify a subpopulation at elevated risk for serious and chronic drug use, randomly assign this population to receive prevention as usual, high-quality universal prevention programming, or intensive targeted high-quality programming. The study would follow these subjects for several years, measuring drug use initiation, frequency and quantity of use, age at cessation of use from each category of drug, and problems related to use. Such a study would provide invaluable evidence about the relative merits of targeted versus universal prevention for high-risk populations. Even more informative would be a study that, in addition to the above, applied the same conditions to a general population of youths not at elevated risk for drug problems.
The targeting issue is closely related to the issue of how best to measure the effectiveness of prevention programs. The success of universal programs is most often measured by reductions in the prevalence of use in the general population. The success of programs targeting higher-risk populations could focus on the quantity of use or the problems related to use. This issue is discussed next.
This report is about data and research needs for policy on illegal drug use. The committee has chosen to focus its attention primarily on illegal drugs, whose use is very costly to both individuals and society. Studies of the effectiveness of prevention programs generally do not measure the effects of use of illegal drugs, such as cocaine and heroin, primarily because such use has not been frequent in the school-attending populations and school districts where most of this research has been conducted. This disconnect is not troublesome for most prevention researchers and policy makers because, as noted earlier, a major assumption in this field is that early use of cigarettes, alcohol, and marijuana lead to later use of more harmful substances.
It may be possible to infer something about the size of the diffusion effect by comparing the effect sizes of studies that randomly assign subjects within a social unit, such as a school, with studies that randomly assign the social units. Diffusion effects should weaken the effect in the within-unit design more than in the between-unit design. Everything else being equal, the magnitude of the difference in effects for these two designs would be a measure of diffusion effects.