Efficient semiparametric inference for two-phase studies with outcome and covariate measurement errors

Area of Science:

• Biostatistics
• Epidemiology
• Health Data Science

Background:

• Observational studies using routinely collected data (e.g., electronic health records) often suffer from measurement errors in both outcomes and covariates.
• Errors in outcomes and covariates can be correlated, complicating analysis.
• Existing two-phase designs typically address covariate errors only or assume simple random sampling for validation, limiting their applicability.

Purpose of the Study:

• To propose a novel semiparametric approach for general two-phase measurement error problems with quantitative outcomes.
• To accommodate correlated errors in both outcomes and covariates.
• To allow for arbitrary selection schemes in the second phase (validation subsample).

Main Methods:

• Development of a semiparametric statistical framework.
• Implementation of a computationally efficient and numerically stable expectation-maximization (EM) algorithm.
• Maximization of the nonparametric likelihood function to obtain estimators.

Main Results:

• The proposed semiparametric method effectively handles correlated measurement errors in outcomes and covariates within a two-phase design.
• The developed EM algorithm provides stable and efficient estimation.
• Simulation studies demonstrate the superiority of the proposed methods over existing approaches.

Conclusions:

• The novel semiparametric approach offers a robust solution for measurement error problems in observational studies with complex error structures.
• The method is applicable to general two-phase designs with arbitrary validation sample selection.
• The approach was successfully illustrated in an observational HIV study, highlighting its practical utility in health research.