Innovative methods for examining mediators in randomized controlled intervention trials

Abstract

Summary: Researchers postulate that theory-based interventions influence behavior by changing theoretical constructs believed to be important for behavior change. These theoretical constructs are referred to as “mediators” because it is believed that the constructs mediate the relationship between the intervention and behavior change. A majority of studies have used the Baron and Kenny (1986) method to examine mediators; however, this method has been criticized for having low power. The purpose of this presentation is to provide an overview of various mediation models that can be used in intervention research. Various statistical approaches to mediation analyses will be described and the pros and cons of each method will be discussed. To demonstrate the use of various mediation models, the three presenters will apply various mediation statistical methods in randomized controlled trials examining physical activity behavior change. Specifically, Dr. Beth Lewis will present mediator data from the Jump Start study (n= 150) and Healthy for Life (n=448), both randomized controlled print-based physical activity trials examining several potential mediators. Dr. Papandonatos will present multiple mediation models for Project STRIDE (n=239), a randomized clinical trial comparing print- and telephone-based interventions to a delayed-treatment control. Dr. Ester Cerin will compare the results of four methods of mediation analyses applied to a small-scale print- and telephone-based physical activity trial (n=52). The discussant, Dr. Bess Marcus, will provide her perspective and discuss ideas for future directions in mediation research.