Robust likelihood inference for diagnostic accuracy measures for paired organs.

Area of Science:

• Ophthalmology
• Biostatistics
• Medical Diagnostics

Background:

• Paired eye data are common in ophthalmology for disease diagnosis.
• Current correlation models assume equal accuracy (exchangeability) between fellow eyes, which may not always hold.
• Accurate assessment of diagnostic test performance is crucial for clinical decision-making.

Purpose of the Study:

• To propose a parametric robust likelihood approach for testing the equality of accuracy measures between fellow eyes.
• To provide a method that does not require modeling the correlation between fellow eyes.
• To offer a flexible procedure for inferring diagnostic accuracy in general paired designs.

Main Methods:

• Developed a parametric robust likelihood method.
• Focused on testing the equality of accuracy measures (sensitivity/specificity) without correlation modeling.
• Utilized simulations and real-world ophthalmology data for validation.

Main Results:

• The proposed robust likelihood procedure effectively tests the equality of accuracy measures between fellow eyes.
• The method demonstrates applicability in general paired diagnostic designs.
• Simulations and data analysis confirmed the procedure's effectiveness.

Conclusions:

• The parametric robust likelihood approach provides a valuable tool for analyzing paired eye data in ophthalmology.
• This method allows for testing accuracy equality without relying on correlation models.
• The procedure enhances the inference of diagnostic accuracy in paired designs.