Exact confidence interval estimation for the difference in diagnostic accuracy with three ordinal diagnostic groups.

Area of Science:

• Medical Statistics
• Diagnostic Accuracy Research

Background:

• Traditional measures like Area Under Curve (AUC) and Partial Area Under Curve (PAUC) are extended to Volume Under Surface (VUS) and Partial Volume Under Surface (PVUS) for three ordinal diagnostic groups.
• Accurate confidence interval estimation for VUS and PVUS differences is crucial, particularly in studies with limited sample sizes.

Purpose of the Study:

• To propose a novel confidence interval estimation method for the difference in paired VUS and PVUS.
• To specifically address the challenges of small to moderate sample sizes in diagnostic accuracy studies.

Main Methods:

• Development of a new approach based on generalized inference concepts.
• A Monte Carlo simulation study was conducted to evaluate the performance of the proposed method.
• Comparison with existing parametric bootstrap and large sample approaches.

Main Results:

• The proposed generalized inference approach demonstrated reasonable coverage probabilities, even with small sample sizes.
• The method proved effective in providing reliable confidence intervals for VUS and PVUS differences.

Conclusions:

• The generalized inference approach offers a robust method for confidence interval estimation of VUS and PVUS differences, particularly beneficial for studies with small to moderate sample sizes.
• The approach was validated through simulation and illustrated with a real-world application in anemia patient blood test data.