Diagnostic test accuracy and prevalence inferences based on joint and sequential testing with finite population...

Area of Science:

• Biostatistics
• Epidemiology
• Diagnostic Test Evaluation

Background:

• Traditional two-test models assume binomial sampling, which is inaccurate for large or complete population samples.
• Accurate estimation of diagnostic test performance and disease prevalence is crucial in public health and veterinary medicine.

Purpose of the Study:

• To develop and present statistical methods for evaluating diagnostic test accuracy and prevalence estimation using finite population data without a gold standard.
• To analyze joint (simultaneous and sequential) testing strategies and their inference.
• To compare finite population (hypergeometric) and infinite population (binomial) inference methods.

Main Methods:

• Development of statistical methods for diagnostic test accuracy and prevalence estimation under finite population sampling.
• Application of methods to real and simulated datasets.
• Comparison of hypergeometric and binomial-based inference models.

Main Results:

• Finite population sampling resulted in smaller posterior standard deviations for prevalence estimates compared to infinite population sampling.
• Sensitivity and specificity estimates were not significantly different between finite and infinite population models.
• The study provides recommendations on sample size relative to population size for appropriate model selection.

Conclusions:

• The developed methods offer a more accurate approach to diagnostic test evaluation and prevalence estimation in situations with finite populations.
• Finite population corrections are important for accurate prevalence estimation, particularly with large sample sizes.
• Understanding the limitations of the binomial assumption is key for reliable disease surveillance and test performance assessment.