Post-test diagnostic accuracy measures under tree ordering of disease classes.

Area of Science:

• Medical Diagnostics
• Biostatistics
• Health Outcomes Research

Background:

• Traditional diagnostic accuracy measures include predictive values (PPV, NPV) and likelihood ratios (LR+, LR-).
• Predictive values assess post-test probability in populations with varying disease prevalence.
• Likelihood ratios link pre-test and post-test probabilities for individual patients.

Purpose of the Study:

• To introduce and analyze generalized predictive values and likelihood ratios.
• To evaluate a novel approach using a tree ordering of disease classes for diagnostic accuracy.
• To demonstrate the application of these methods with real-world lung cancer data.

Main Methods:

• Development and analysis of generalized predictive values and likelihood ratios.
• Utilizing a tree ordering of disease classes.
• Conducting simulation studies to assess method effectiveness.
• Applying methods to real lung cancer diagnostic data.

Main Results:

• The proposed generalized methods offer a flexible framework for diagnostic accuracy evaluation.
• Simulation studies demonstrate the utility and robustness of the new approaches.
• Application to lung cancer data illustrates practical implementation and potential benefits.

Conclusions:

• Generalized predictive values and likelihood ratios provide a valuable extension to existing diagnostic accuracy metrics.
• The tree ordering approach enhances the analysis of complex disease classifications.
• These methods hold promise for improving diagnostic decision-making in clinical practice.