New confidence intervals for the difference between two sensitivities at a fixed level of specificity

Area of Science:

• Medical Statistics
• Diagnostic Test Evaluation

Background:

• Comparing diagnostic tests requires evaluating their performance metrics, such as sensitivity and specificity.
• Assessing sensitivity at a fixed specificity is crucial for selecting optimal diagnostic tools.

Purpose of the Study:

• To propose and evaluate new confidence intervals for the difference between sensitivities of two continuous-scale diagnostic tests at a fixed specificity level.
• To compare the performance of these novel intervals against existing methods.

Main Methods:

• Development of three new confidence intervals for sensitivity difference.
• Conducting simulation studies to assess interval performance.
• Comparison with the normal-approximation-based interval by Wieand et al.

Main Results:

• The proposed intervals are computationally straightforward.
• Simulation results indicate the new intervals offer better coverage accuracy.
• The newly proposed intervals also provide improved interval length compared to the existing method.

Conclusions:

• The newly developed confidence intervals are recommended for comparing diagnostic test sensitivities at a fixed specificity.
• These intervals offer practical advantages in terms of ease of computation and statistical performance.