Using Bayes theorem to estimate positive and negative predictive values for continuously and ordinally scaled...

Area of Science:

• Medical Diagnostics
• Biostatistics
• Psychometrics

Background:

• Positive predictive values (PPVs) and negative predictive values (NPVs) are crucial for interpreting diagnostic test accuracy and patient risk.
• Current methods often dichotomize continuous or ordinal test results, leading to a loss of valuable information.

Purpose of the Study:

• To extend the calculation of PPVs and NPVs using Bayesian theorem for continuously or ordinally scaled diagnostic tests.
• To enable a more nuanced interpretation of test results by avoiding unnecessary dichotomization.

Main Methods:

• Utilized Bayesian theorem to calculate the probability of disease given a continuously or ordinally scaled test outcome.
• Modeled probabilities of test results conditional on disease status within a Bayesian framework.
• Transformed these probabilities to disease status conditional on test results.

Main Results:

• Applied the method to publicly available data to estimate the probability of clinical depression given PROMIS Depression scores.
• Demonstrated that this approach allows for a more fine-grained interpretation of test scores compared to traditional PPV and NPV calculations based on dichotomized scores.

Conclusions:

• The proposed Bayesian method facilitates accurate and meaningful interpretation of diagnostic test results in clinical practice.
• Avoiding unnecessary dichotomization of test scores enhances the utility of diagnostic tests.