Causal Mediation Analysis

Definition and scope

Estimation of pathway-specific causal contrasts through a post-treatment mediator, with explicit direct/indirect estimands and assumptions beyond those needed for the total treatment effect. The target population, use, and analysis remain protocol decisions; this card explains the choices and their consequences rather than approving a study.

Operational specification

Define treatment, mediator, outcome, their timing, baseline confounders, exposure-induced mediator-outcome confounders, target population, and controlled, natural, or interventional direct/indirect contrasts. Natural effects involve cross-world counterfactual structure; alternative interventional effects can target different questions under different identification conditions.

Analysis and diagnostics

Use an identification argument and estimator compatible with mediator/outcome types and the assumed confounding structure. Check positivity, measurement and temporal ordering, treatment-mediator interaction, and sensitivity to unmeasured mediator-outcome confounding. Explain the effect scale before decomposing a total effect or calculating a proportion mediated.

When NOT to use and common errors

Do not call the coefficient change after adjusting for a mediator a causal indirect effect, condition on an exposure-induced collider without justification, assume randomized treatment randomizes the mediator, or report a proportion mediated without checking scale, interaction, and near-zero/opposite-signed effects.

Worked example and implementation boundary

Under a synthetic additive decomposition already justified by a causal model, direct risk difference is 0.03 and indirect risk difference is 0.02. Sum them; the arithmetic does not establish identification. The stated additive total is 0.05. Natural or interventional decomposition assumptions must be justified separately. The code performs that explicitly stated calculation with synthetic inputs. It is not a complete fitted analysis and does not demonstrate the identification assumptions, estimate sampling uncertainty, or certify application in another population.

Data-source and reporting requirements

Preserve the provenance, dates, units, and observation rule for each input. Claims capture services and transactions under observable benefits; absence of a record is not proof that an event or resource use did not occur. EHR and registry measurements depend on the collection process and may miss care outside the system. Linked sources add reconciliation and selection questions. Primary collection can improve a missing domain but still requires valid measurement and follow-up. Report which of these limitations affects the stated use, which evidence supports the chosen inputs, and what remains uncertain.

Worked example

Scenario

Under a synthetic additive decomposition already justified by a causal model, direct risk difference is 0.03 and indirect risk difference is 0.02. Sum them; the arithmetic does not establish identification.

Dataset

Synthetic inputs with units defined in the scenario; no patient data.

Steps

Result

The stated additive total is 0.05. Natural or interventional decomposition assumptions must be justified separately.

Trade-offs

Code example

direct_rd = 0.03
indirect_rd = 0.02
result = direct_rd+indirect_rd
assert abs(result - 0.05) < 1e-9
print(result)

direct_rd <- 0.03
indirect_rd <- 0.02
result <- direct_rd+indirect_rd
stopifnot(abs(result - 0.05) < 1e-9)
print(result)

data worked_example;
  direct_rd = 0.03;
  indirect_rd = 0.02;
  result = direct_rd+indirect_rd;
  put result=;
run;

Citations

1. [1]Valeri L, et al. Mediation analysis allowing for exposure–mediator interactions and causal interpretation: Theoretical assumptions and implementation with SAS and SPSS macros.. Psychological Methods. 2013.