What are the effects of an age appropriate, time varying intervention on substance use? What are the effects of an individually tailored, need-based intervention on substance use? How do we evaluate program effectiveness when there are variations in implementation fidelity? If we were to increase the level of parental monitoring, how would substance use change? These important prevention research questions require estimating the effect of a time varying intervention on a response. In all four cases, there are common correlates of both the selection of the treatment or independent variable, and the outcome. For example, in the case of an individually tailored, need-based intervention, individuals at highest risk of substance use receive the most intensive intervention. Thus risk of substance use is correlated with both the amount of intervention received and the outcome. Although this is an excellent prevention strategy, it presents major problems in the assessment of treatment effects. In fact, the widely accepted traditional methods for assessing treatment effects in this context can lead to highly misleading results, even making a successful intervention appear harmful. In this study we will develop methods for assessing treatment effects for time varying preventive interventions. The approach will be based on statistical adjustments that incorporate into the analysis all collected information related to the selection of the intervention level. We will develop such adjustments in three general types of analyses used extensively in drug abuse prevention research: regression analysis, survival analysis, and growth curve modeling. In addition, we will extend these methods to situations where the data have a multilevel structure, for example, data from classroom-based interventions. The new procedures will be applied in empirical prevention data, primarily data from the FAST Track intervention and the Baltimore study. Software will be developed to make it convenient for researchers to apply these methods. The end result will be much more realistic estimates of drug abuse prevention treatment effects.
Showing the most recent 10 out of 443 publications
