Clarifying causal mediation analysis for the applied researcher: Defining effects based on what we want to learn

Abstract

The incorporation of causal inference in mediation analysis has led to theoretical and methodological advancements-effect definitions with causal interpretation, clarification of assumptions required for effect identification, and an expanding array of options for effect estimation. However, the literature on these results is fast-growing and complex, which may be confusing to researchers unfamiliar with causal inference or unfamiliar with mediation. The goal of this article is to help ease the understanding and adoption of causal mediation analysis. It starts by highlighting a key difference between the causal inference and traditional approaches to mediation analysis and making a case for the need for explicit causal thinking and the causal inference approach in mediation analysis. It then explains in as-plain-as-possible language existing effect types, paying special attention to motivating these effects with different types of research questions, and using concrete examples for illustration. This presentation differentiates 2 perspectives (or purposes of analysis): the explanatory perspective (aiming to explain the total effect) and the interventional perspective (asking questions about hypothetical interventions on the exposure and mediator, or hypothetically modified exposures). For the latter perspective, the article proposes tapping into a general class of interventional effects that contains as special cases most of the usual effect types-interventional direct and indirect effects, controlled direct effects and also a generalized interventional direct effect type, as well as the total effect and overall effect. This general class allows flexible effect definitions which better match many research questions than the standard interventional direct and indirect effects. (PsycInfo Database Record (c) 2021 APA, all rights reserved).

Figure 1.

An increasing trend in the popularity of mediation analysis in scholarly research The raw counts in the top panel are counts reported by Google Scholar on two searches for articles (excluding patents and citations) with “mediation analysis” in the title, and for those with the same phrase anywhere in the text; the adjusted counts are adjusted for the fact that the volume of all Google Scholar entries varies in size from year to year, using 2015 as the standard year. In the bottom panel, the counts are reported by PsycINFO on the same two searches. These searches were conducted on 20/12/2018.

Figure 2.

Traditional approach: effects defined as (functions of) regression coefficients

Figure 3.

The total effect

Figure 4.

Natural (in)direct effects – depicted in the special case with no intermediate confounder

Figure 5.

Interventional (in)direct effects – depicted in the general case with an intermediate confounder Figure note: An equivalent representation of Y(a, $\mathcal{M}$(a′|C)) is Y(a, L(a), $\mathcal{M}$(a′|C)).

Figure 6.

Controlled direct effect