Inference about binocular sensitivity and specificity of screening tests for paired organs

Area of Science:

• Biostatistics
• Ophthalmic diagnostics
• Medical statistics

Background:

• Perera et al. introduced novel binocular accuracy measures for paired organ diagnostic tests.
• Existing methods utilize Gaussian copula models to address inter-eye correlation.

Purpose of the Study:

• To propose an alternative, simpler inference approach for the new binocular accuracy measures.
• To evaluate the suitability of binomial models for independent patient data.

Main Methods:

• Development of a binomial model-based inference framework.
• Numerical simulations to assess model performance and robustness.
• Analysis of the Perera et al. dataset for validation.

Main Results:

• Binomial models provide a simple and direct method for inference when patient data are independent.
• The proposed approach was illustrated using the Perera et al. data.
• Robustness of Gaussian copula and binomial models under misspecification was investigated.

Conclusions:

• Binomial models offer a practical and efficient alternative for inferring binocular accuracy measures in independent patient cohorts.
• The findings support a simplified statistical approach in diagnostic test evaluation for paired organs.