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

Area of Science:

• Epidemiology
• Causal Inference
• Public Health Interventions

Background:

• Epidemiological research often faces challenges in evaluating interventions that modify mediating pathways.
• Existing mediation methods primarily focus on pathway discovery, not mediator intervention evaluation.
• Lack of well-defined interventions and data hinders the assessment of hypothetical treatments for complex health issues.

Purpose of the Study:

• To propose a novel framework for evaluating hypothetical interventions targeting multiple mediators.
• To define mediation effects that map to a target trial simulating intervention effects on mediators.
• To address the methodological challenges in evaluating ill-defined mediator interventions in epidemiological studies.

Main Methods:

• Developed a framework defining mediation effects for a target trial of hypothetical interventions.
• Specified a target trial to assess impacts of hypothetical interventions shifting mediator distributions.
• Simulated shifts by setting mediators to random draws from user-specified distributions, using g-computation for estimation.

Main Results:

• The proposed framework enables the evaluation of hypothetical interventions targeting multiple mediators.
• G-computation allows estimation under expanded causal assumptions suitable for ill-defined interventions.
• The method was applied to adolescent self-harm data, demonstrating its value for future intervention design.

Conclusions:

• The novel framework provides a method for evaluating hypothetical mediator interventions in epidemiology.
• This approach is valuable for informing the design and evaluation of actual public health interventions.
• The study highlights the utility of simulating intervention effects on mediators when real-world interventions are not well-defined.