Leveraging External Validation Data: The Challenges of Transporting Measurement Error Parameters.

Area of Science:

• Epidemiology
• Biostatistics
• Health Research Methods

Background:

• Measurement error is a common challenge in epidemiological studies.
• Validation data are crucial for estimating measurement error parameters like sensitivity and specificity.
• Acquiring validation data is expensive, making secondary data use attractive but requiring methods to address systematic differences.

Purpose of the Study:

• To derive estimators for risk and risk difference using external validation data to address outcome misclassification.
• To develop methods for transporting misclassification parameters when covariates are differentially distributed between study and validation samples.
• To compare approaches for handling covariates that may induce bias when transporting misclassification parameters.

Main Methods:

• Derivation of estimators for risk and risk difference leveraging external validation data.
• Development of two covariate adjustment strategies: standardization and iterative outcome modeling.
• Proof of identification, parametric model estimation, and simulation studies to assess performance.

Main Results:

• The study provides methods to account for outcome misclassification using external validation data.
• Two approaches are presented to handle differential covariate distributions, with one (iterative modeling) avoiding bias induced by M-bias.
• Simulations demonstrate the performance of the proposed methods.

Conclusions:

• External validation data, when used with transportability methods, can effectively address outcome misclassification.
• The choice of covariate adjustment method is critical to avoid bias in causal effect estimation.
• The methods are illustrated with an application to preterm birth risk and the effect of maternal HIV infection.