A Bayesian method for synthesizing multiple diagnostic outcomes of COVID-19 tests

Area of Science:

• Epidemiology
• Biostatistics
• Infectious Diseases

Background:

• The global spread of COVID-19 necessitates accurate and timely diagnosis.
• Diagnostic tests for SARS-CoV-2 exhibit variable sensitivity and specificity, leading to potential false negatives.
• Inconsistent results from multiple tests pose challenges for clinical decision-making.

Purpose of the Study:

• To develop a Bayesian probabilistic framework for synthesizing results from multiple COVID-19 diagnostic tests.
• To enhance the accuracy of SARS-CoV-2 infection diagnosis by integrating diverse test outcomes.
• To provide a user-friendly tool for clinicians to interpret integrated diagnostic results.

Main Methods:

• Proposed a Bayesian probabilistic framework utilizing the sensitivity and specificity of individual diagnostic methods.
• Developed a web application for real-time integration of two or more COVID-19 test results.
• Conducted simulation-based assessments to compare the model's accuracy and precision against single-test outcomes.

Main Results:

• The synthesized posterior probability of infection outperformed individual test results.
• The Bayesian framework demonstrated higher accuracy and precision compared to single-test outcomes.
• The developed web application effectively integrates multiple test results for clinical evaluation.

Conclusions:

• The Bayesian probabilistic framework offers a superior method for diagnosing COVID-19 when multiple tests are employed.
• The accessible web application aids physicians in clinical evaluations by integrating diverse diagnostic data.
• Accurate integration of multiple diagnostic tests is crucial for effective pandemic control and patient management.