The hazards of bedside Bayes

Area of Science:

• Medical Diagnostics
• Clinical Decision Support
• Health Services Research

Background:

• Bayesian and decision-analysis approaches in clinical problem-solving often rely on readily available sensitivity and specificity data for diagnostic tests.
• Currently, comprehensive and reliable data on the performance of many common diagnostic tests is limited.
• This lack of data poses a challenge for accurate quantitative medical decision-making.

Purpose of the Study:

• To critically review the existing literature on the sensitivities and specificities of commonly used diagnostic tests.
• To assess the variability in reported performance metrics for these tests.
• To inform clinicians, quantitative decision-making advocates, and researchers about the reliability of diagnostic test data.

Main Methods:

• A critical literature review was conducted.
• Data on sensitivities and specificities were extracted for seven commonly used diagnostic tests.
• Statistical analysis was performed to identify significant variability in reported results.

Main Results:

• Significant variability was observed among the reported sensitivities and specificities for five of the seven evaluated diagnostic tests.
• This variability was unexpected and highlights a critical issue for quantitative approaches.
• Two specific tests, the rapid-sequence excretory urogram and the thallous chloride TI 201 cardiac stress test, demonstrated consistent reported sensitivities and specificities.

Conclusions:

• The reliability of sensitivity and specificity data for many common diagnostic tests is questionable due to significant inter-study variability.
• This variability must be considered in the application of Bayesian and decision-analysis methods in clinical practice.
• Further research is needed to establish standardized and reliable performance metrics for diagnostic tests to improve clinical diagnosis and decision support.