Session: Incorporating Parent Participation Into Models Estimating Intervention Effects and Mediators (Society for Prevention Research 21st Annual Meeting)

This symposium presents three examples of incorporating parent participation levels in models estimating preventive intervention effects, as well as mediators, drawing on recent developments in causal inference models (e.g., Frangakis & Rubin, 2002; Jin & Rubin, 2009).  When participation in a preventive intervention is incomplete, intention-to-treat (ITT) analyses can underestimate intervention effects.  This problem is common when a target population is identified, an intervention is offered, and only a portion of the population participates or completes the full intervention.  Parent participation, in particular, is often far less than complete, due to numerous challenges of engaging and delivering interventions to parents who are balancing numerous roles and time-demands.  Growing evidence that strengthening parenting practices and parent-child relationships can contribute to preventing a range of mental, emotional, and behavioral problems makes the problem of accurately estimating effects of interventions involving parents of importance.   Causal inference approaches provide tools for identifying the optimal comparison group from the control group in order to compare with ‘intervention compliers’ in the intervention condition.  In addition to estimating intervention effects for those who participate, causal inference approaches can clarify the extent to which effects of multi-component interventions are accounted for by families who participate intensively (Connell & Dishion, in press).  This symposium supports the conference theme by attending to parents’ role in comprehensive strategies to promote children’s well being using innovative methods.

Bringing together researchers from multiple institutions, this symposium covers diverse preventive interventions involving parents that target a range of outcomes across childhood. 

The first paper, “Prevention of Problem Behavior Through Annual Family Check-Ups in Early Childhood” presents the use of Complier Average Causal Effect analyses (CACE; Imbens & Rubin, 1997) to evaluate impact of a multi-component intervention for high-risk parents of young children.

The second paper, “Incorporating Intervention Dosage Into Analysis of Family Interventions” presents two methods, one incorporating actual dosage for participants and predicted dosage for controls; and the second propensity scores. 

The third paper, “Multi-component Intervention Teaching Children Emotion Self-Regulation Skills” uses actual dosage for participants and imputed participation for parents in the control condition. 

 The discussant (George Howe) will summarize strengths and limitations of alternative causal methods and implications for identifying effective interventions.