Incorporating the dilution effect in group testing regression

Area of Science:

• Biostatistics
• Epidemiology
• Infectious Disease Screening

Background:

• Group testing is a cost-effective method for infectious disease screening, testing pooled specimens instead of individual ones.
• A key challenge in group testing is the "dilution effect," where positive signals are obscured in large negative pools.

Purpose of the Study:

• To propose a novel statistical framework for group testing data that integrates estimation and case identification.
• To address the dilution effect by analyzing continuous biomarker levels in pooled samples.

Main Methods:

• Developed a framework analyzing continuous biomarker levels from pooled samples.
• Estimated a binary regression model for disease probability and biomarker distributions for cases and controls.
• Utilized biomarker distribution estimates to optimize pool-specific diagnostic thresholds for enhanced case identification.

Main Results:

• The proposed framework effectively merges estimation and case identification, improving accuracy.
• Pool-by-pool threshold selection based on biomarker distributions enhances diagnostic precision.
• Demonstrated utility using hepatitis B virus data from an Irish prison population.

Conclusions:

• The new statistical approach offers a more accurate and efficient method for infectious disease screening using group testing.
• This framework mitigates the dilution effect, leading to improved identification of infectious individuals.