Compartmental Model Diagrams as Causal Representations in Relation to DAGs

Area of Science:

• Epidemiology
• Causal Inference
• Mathematical Modeling

Background:

• Compartmental models have a century-long history in infectious disease epidemiology.
• Causal directed acyclic graphs (DAGs) have been widely adopted in epidemiology since the 1990s.
• Both methods aim to represent causal relationships in public health.

Purpose of the Study:

• To illustrate how compartmental model diagrams can represent concepts similar to causal DAGs.
• To demonstrate the utility of compartmental models in depicting interactions and feedback cycles.
• To bridge the gap between researchers using different causal inference frameworks.

Main Methods:

• Utilized a chronic disease example (type 2 diabetes and dementia incidence).
• Compared the representational capabilities of compartmental models and DAGs for causation, mediation, confounding, and collider bias.
• Highlighted the parametric nature of compartmental models for describing state changes.

Main Results:

• Compartmental models can represent the same causal concepts as DAGs.
• Compartmental models explicitly depict interactions and feedback cycles.
• Compartmental models offer parametric or semiparametric descriptions of biological processes.

Conclusions:

• Compartmental model diagrams are a valuable tool for causal inference in epidemiology.
• These diagrams can parametrically express concepts also represented by DAGs.
• Recognizing the links between compartmental models and DAGs can enhance communication and facilitate quantitative analyses.