Confidence intervals for the difference in paired Youden indices

Area of Science:

• Biostatistics
• Medical Diagnostics
• Statistical Inference

Background:

• Comparing diagnostic test accuracy is crucial in medicine.
• The Youden index is a common metric for diagnostic test evaluation.
• Inference for the difference in paired Youden indices remains underexplored.

Purpose of the Study:

• To propose an exact confidence interval for the difference in paired Youden indices.
• To compare the performance of the proposed interval against existing methods.
• To provide a reliable statistical tool for diagnostic test accuracy comparison.

Main Methods:

• Development of an exact confidence interval using generalized pivotal quantities.
• Comparison with maximum likelihood estimate-based and bootstrap-based intervals.
• Simulation studies evaluating coverage probability and average interval length.

Main Results:

• The proposed exact confidence interval demonstrates superior performance.
• Outperformance is evident even with small sample sizes under normal distributions.
• The interval provides accurate and reliable estimation for the difference in paired Youden indices.

Conclusions:

• The proposed exact confidence interval is a valuable tool for comparing paired diagnostic tests.
• It offers improved accuracy and reliability over existing methods.
• The method is applicable in real-world diagnostic accuracy assessments.