Improved confidence intervals of a small probability from pooled testing with misclassification

Area of Science:

• Biostatistics
• Epidemiology
• Medical Diagnostics

Background:

• Rare disease prevalence estimation is crucial for public health.
• Dorfman's pooled testing offers efficiency but requires careful statistical analysis.
• Imperfect biomarkers introduce classification error, complicating prevalence estimation.

Purpose of the Study:

• To construct accurate confidence intervals for rare disease prevalence using Dorfman's pooled testing.
• To evaluate the performance of confidence intervals when disease status is determined by an imperfect biomarker.
• To compare the efficiency of standard methods versus proposed alternatives.

Main Methods:

• Utilized Dorfman's pooled testing procedure for group testing.
• Derived confidence intervals by converting intervals for the probability of a positive test within a pool.
• Investigated Wald confidence intervals based on normal approximation.
• Proposed and analyzed alternative confidence interval methods.

Main Results:

• Normal approximation-based Wald confidence intervals demonstrate inefficiency in coverage probability.
• The inefficiency persists even with a large number of tested pools.
• Proposed alternative methods show improved performance regarding coverage probability and interval length.

Conclusions:

• Standard confidence intervals are inadequate for rare disease prevalence estimation with pooled testing and imperfect biomarkers.
• Alternative statistical methods are necessary for reliable estimation in such scenarios.
• The choice of confidence interval method significantly impacts the accuracy of rare disease prevalence estimates.