Transporting results in an observational epidemiology setting: purposes, methods, and applied example

Area of Science:

• Epidemiology
• Biostatistics
• Causal Inference

Background:

• Medical research often prioritizes internal validity over external validity (generalizability).
• Limited methods exist to test the generalizability of findings from observational studies across different populations.
• Transporting effect estimates is crucial for applying knowledge from study populations to target populations.

Purpose of the Study:

• To describe the conceptual framework and assumptions for transporting results from a population-based study to a target population in an observational setting.
• To address the challenge of minimizing biases in observational studies and accounting for population differences when transporting estimates.
• To illustrate the application of these methods in life-course epidemiology.

Main Methods:

• Combines methods for causal inference with methods for transporting effect estimates.
• Utilizes the targeted maximum likelihood estimator (TMLE) for estimation.
• Applies the framework to a life-course epidemiology example for illustrative purposes.

Main Results:

• Demonstrates a method to assess and enhance the external validity of observational study findings.
• Provides a structured approach to combining causal inference with transportability methods.
• Successfully applied the targeted maximum likelihood estimator in a life-course epidemiology context.

Conclusions:

• Testing external validity is essential for the real-world application of medical research.
• The proposed framework and methods facilitate the transport of effect estimates from observational studies.
• Targeted maximum likelihood estimation is a viable tool for addressing transportability challenges in epidemiology.