Mediation effects that emulate a target randomised trial: Simulation-based evaluation of ill-defined interventions on multiple mediators

TL;DR

This paper proposes a simulation-based framework to evaluate hypothetical interventions on mediators, emulating a target trial, especially when actual interventions are ill-defined and data are limited.

Contribution

It introduces novel interventional effects and a g-computation approach for evaluating hypothetical mediator interventions using observational data.

Findings

Demonstrates the method with adolescent self-harm data

Shows how to emulate shifts in mediator distributions

Provides a framework for policy-relevant intervention evaluation

Abstract

Many epidemiological questions concern potential interventions to alter the pathways presumed to mediate an association. For example, we consider a study that investigates the benefit of interventions in young adulthood for ameliorating the poorer mid-life psychosocial outcomes of adolescent self-harmers relative to their healthy peers. Two methodological challenges arise. Firstly, mediation methods have hitherto mostly focused on the elusive task of discovering pathways, rather than on the evaluation of mediator interventions. Secondly, the complexity of such questions is invariably such that there are no existing data on well-defined interventions (i.e. actual treatments, programs, etc.) capturing the populations, outcomes and time-spans of interest. Instead, researchers must rely on exposure (non-intervention) data to address these questions, such as self-reported substance use and employment. We address the resulting challenges by specifying a target trial addressing three policy-relevant questions, regarding the impacts of hypothetical (rather than actual) interventions that would shift the mediators' distributions (separately, jointly or sequentially) to user-specified distributions that can be emulated with the observed data. We then define novel interventional effects that map to this trial, emulating shifts by setting mediators to random draws from those distributions. We show that estimation using a g-computation method is possible under an expanded set of causal assumptions relative to inference with well-defined interventions. These expanded assumptions reflect the lower level of evidence that is inevitable with ill-defined interventions. Application to the self-harm example using data from the Victorian Adolescent Health Cohort Study illustrates the value of our proposal for informing the design and evaluation of actual interventions in the future.