A two-stage estimation for screening studies using two diagnostic tests with binary disease status verified in test...

Area of Science:

• Biostatistics
• Epidemiological Methods
• Medical Diagnostics

Background:

• Screening studies often use multiple binary tests to detect diseases.
• A key challenge is non-identifiability of disease rates for individuals with negative results on all tests.
• Existing methods rely on homogeneous association models and numerical solutions.

Purpose of the Study:

• To propose a novel statistical framework for estimation and inference in two-binary-test screening studies.
• To address the non-identifiability problem using a constrained maximum likelihood estimation (MLE) approach.
• To develop a unified two-stage estimation method for practical application.

Main Methods:

• Formulating the screening problem as a constrained maximum likelihood estimation (MLE) problem.
• Developing a unified two-stage estimation approach to solve the MLE.
• Proposing an association-ratio plot for visualizing and comparing homogeneous association models.

Main Results:

• The proposed constrained MLE approach yields a closed-form solution, simplifying estimation.
• The unified two-stage method provides an efficient way to solve the estimation problem.
• The association-ratio plot serves as an effective tool for model selection and comparison.

Conclusions:

• The constrained MLE method offers a statistically sound and computationally efficient solution for two-binary-test screening studies.
• The proposed visualization tool aids in understanding model fit and selecting appropriate association models.
• This work advances statistical methodologies for diagnostic test evaluation in public health.