Design and Analysis Issues in Community-Based Drug Abuse Prevention

Abstract

This article reviews the design and analysis issues that face community-based drug abuse prevention trials. Such trials allocate intact social groups to study conditions and fall within the general class of studies called “community trials.” Special concerns for community trials include selection bias, differential mortality and maturation, contamination, poorly defined constructs, weak interventions, violation of assumptions underlying analysis methods, and low power. Potential solutions include careful selection of comparison units, tight control over distribution of intervention materials, matching, randomization, adding more units to each condition or more frequent observations in each unit, monitoring exposure to intervention-like activities in all sites, regression adjustment for covariates, modeling time, and selection of an appropriate analysis model. Four approaches to analysis are illustrated in an example based on an adolescent tobacco use prevention study.