Specifying and Testing Mechanisms in Social Work Science Using Modern Mediation Analysis

As social work science develops, answering questions about mechanisms, or the causal processes that explain how two constructs are related to each other, is vital to better understanding and changing our social world. Moving beyond questions of whether interventions work, or whether constructs are related, to how and why those relationships exist, generates critical knowledge about which social work interventions to use in which contexts, how to streamline and enhance the impacts of social work interventions, and how we might most effectively (and efficiently) intervene to improve our world. Such questions can be addressed using mediation analysis; however, the application of mediation analysis methods is often problematized in social work research where interventions are frequently complex and occur in multilevel contexts with samples of varying sizes at each level. In addition to these complexities, there have been major developments in mediation analysis during the last 20 years which have made older approaches obsolete and call for mastery of state-of-the-art methods.

The goal of this workshop is to equip social work researchers with practical strategies and tools for specifying and testing mechanisms between antecedent (including complex interventions) and outcome variables using mediation analysis in a multilevel context. The workshop will outline a series of research steps and provide a practical guide to developing a research aim that evaluates mechanisms. The guide will be particularly helpful for social work researchers responding to the National Institutes of Health experimental therapeutics paradigm, which requires specification and testing of mechanisms. The steps include: (1) specifying mechanistic relationships and hypotheses using causal pathway models, (2) selecting a sample, measures, and design to quantitatively test mechanistic hypotheses, and (3) selecting a mediation analysis approach aligned with the mechanistic questions and study design. Special attention will be given to the use of statistical mediation analysis to quantitatively test mechanisms in a multilevel context with data collected over time. Two practical examples of mediation analyses completed in the context of NIH-funded studies focused on improving the implementation of evidence-based mental health interventions will be presented. The session will include structured questions and time for attendees to (a) apply the steps to developing a research aim of their own, and (b) receive feedback on ideas from other attendees and the presenter. In light of the conference theme of strengthening social impact through collaborative research, the workshop will highlight methods for engaging community partners in specifying and testing mechanisms of complex interventions, with special emphasis on articulating how this type of research can optimize community impact.