Emulating Target Trials to Improve Causal Inference From Agent-Based Models.

Area of Science:

• Computational epidemiology
• Causal inference in public health
• Simulation modeling

Background:

• Agent-based models (ABMs) are crucial for studying emergent properties in population health, like infectious disease spread.
• Traditional causal inference methods, based on the potential outcomes framework, often assume no interference, which is violated in many health scenarios.
• A gap exists in understanding how to interpret simulation-based models, such as ABMs, within a causal inference context.

Purpose of the Study:

• To propose the integration of the target trial framework into the design of agent-based models.
• To clarify causal parameters of interest when estimating effects in the presence of interference or spillover.
• To make explicit the assumptions required for valid causal effect estimation using simulation models.

Main Methods:

• The study describes a methodological approach for designing agent-based models.
• It explicitly incorporates the target trial framework to guide model development.
• Focuses on scenarios where the 'no interference' assumption is violated.

Main Results:

• The target trial framework provides a structured approach to causal effect estimation in agent-based models.
• This integration clarifies the causal parameters being estimated.
• It enhances transparency regarding the assumptions underlying causal inference from simulation models.

Conclusions:

• Integrating the target trial framework into agent-based model design is essential for robust causal inference in population health.
• This approach addresses the challenge of interference in disease transmission and other health-related phenomena.
• It bridges the gap between simulation-based modeling and causal inference, improving the real-world applicability of model results.