Bayesian nonparametric inference for the three-class Youden index and its associated optimal cutoff points

Area of Science:

• Biostatistics
• Medical Diagnostics
• Machine Learning

Background:

• The three-class Youden index is crucial for evaluating medical test accuracy and selecting optimal cutoffs for disease classification.
• Classifying subjects into ordinal disease categories (e.g., no disease, mild, advanced) requires robust statistical methods.

Purpose of the Study:

• To present a novel Bayesian nonparametric approach for estimating the three-class Youden index.
• To determine optimal cutoff values for three-class disease classification using this new method.
• To apply the method to real-world data for assessing cognitive impairment.

Main Methods:

• Utilizing Dirichlet process mixtures, a robust Bayesian nonparametric modeling technique.
• Developing an approach to estimate the three-class Youden index and optimal cutoffs.
• Conducting a simulation study to validate the method's performance.
• Applying the methodology to Trail Making Test data in Parkinson's disease patients.

Main Results:

• The Bayesian nonparametric approach effectively estimates the three-class Youden index.
• Optimal cutoff values were successfully determined for disease classification.
• The method demonstrated robustness in handling complex data distributions.
• Application to Parkinson's disease data provided insights into cognitive impairment assessment.

Conclusions:

• The proposed Bayesian nonparametric method offers a powerful tool for medical test accuracy evaluation.
• This approach enhances the ability to classify subjects into ordinal disease categories.
• The methodology is applicable to complex datasets and real-world clinical scenarios, such as Parkinson's disease research.