Studying Why and for Whom Intervention Works

Background: This presentation will provide a methodological overview of how research methods and statistical procedures can be used to identify (a) the defining characteristic of a subgroup for whom treatment works best (i.e., moderators) and (b) a mechanism by which treatment works (i.e., a mediator).

Objectives: Audience members will be able to identify the common misconceptions about how to identify mechanisms and characteristics of treatment responders. Audience members will be able to say why using the experimental design element in mediational and moderation analysis provides a better way to identify mechanisms and characteristics of treatment responders. Audience members will be able identify the key statistical test required to identify a treatment mechanism. Audience members will be able to identify the two key findings required to identify the characteristic of children who benefited most from a treatment.

Methods: Lecture format will be used to talk about the general issues. Two studies, each using a randomized between-group experimental design and samples of children with Down syndrome (N > 50), will be used to illustrate the issues.

Results: The first study confirmed the hypothesis that daily therapy sessions was more facilitative of spoken vocabulary than weekly therapy sessions because the former affect speech-like nonverbal vocal communication early in the treatment phase.  The second study confirmed the hypothesis that initial verbal imitation ability defined the subgroup of children in which speech recasts worked better than a common therapy method. The region of significance for “high verbal imitators” was .5 SD above the sample mean and included 26% of the sample.

Conclusions: Identifying correlates of change within a treated group is insufficient to identify characteristics of treatment responders. Fortunately, we have the methods to identify for whom treatment works and why: mediational and moderation analyses in the context of randomized between-group experiments.