Overcoming limitations of the Rogan-Gladen correction: A closed-form solution to a simplified Bayesian method for...

Area of Science:

• Epidemiology
• Biostatistics
• Medical Diagnostics

Background:

• Prevalence estimation is crucial in epidemiology but often inaccurate due to diagnostic misclassifications.
• The frequentist Rogan-Gladen method corrects for misclassification but can yield implausible results, especially at extreme prevalence values.
• Traditional Bayesian methods offer statistical advantages but are often too complex for general use.

Purpose of the Study:

• To develop a simplified Bayesian prevalence correction method that is accessible and mitigates drawbacks of existing techniques.
• To offer a practical alternative that balances ease of use with statistical rigor.
• To compare the performance of the simplified Bayesian method against traditional frequentist and Bayesian approaches.

Main Methods:

• Developed a simplified Bayesian model by assuming non-informative priors for true prevalence and neglecting uncertainties in sensitivity and specificity.
• Conducted a large-scale simulation study to compare the simplified Bayesian method with the Rogan-Gladen method and traditional Bayesian methods.
• Evaluated methods based on accuracy, boundary behavior, and coverage of credible intervals.

Main Results:

• The simplified Bayesian method demonstrated improved boundary behavior and coverage compared to the Rogan-Gladen method, particularly at low prevalences.
• Traditional Bayesian methods consistently achieved near 95% coverage, outperforming both simplified Bayes and Rogan-Gladen under high uncertainty.
• Simplified Bayes and Rogan-Gladen methods generally exhibited undercoverage, especially when sensitivity and specificity had significant uncertainty.

Conclusions:

• The simplified Bayesian method provides a practical compromise, enhancing usability over traditional Bayesian approaches while improving upon the Rogan-Gladen method's limitations.
• This new method is reliable when uncertainty in diagnostic test characteristics (sensitivity and specificity) is limited.
• It offers a more accessible yet statistically sound option for correcting apparent prevalence estimates in epidemiological studies.