Group testing regression models with dilution submodels.

Area of Science:

• Biostatistics
• Epidemiology
• Infectious Disease Modeling

Background:

• Group testing is essential for screening sexually transmitted diseases, but often suffers from increased false negatives.
• Dilution of positive samples by negative ones in pooled testing can cause positive pools to test negative, leading to biased regression models.
• Existing methods to address dilution often rely on external biomarker data, which may not be available.

Purpose of the Study:

• To develop a novel regression model for group testing that accounts for dilution effects.
• To provide a method for formally testing the presence of dilution using only group testing data.
• To improve the accuracy of population-level regression models in the presence of dilution.

Main Methods:

• Augmented existing group testing regression models with a parametric dilution submodel.
• Estimated all parameters using maximum likelihood estimation.
• Validated the approach through simulation studies and application to infectious disease data.

Main Results:

• The proposed method effectively adjusts for dilution bias in group testing regression.
• The framework allows for statistical testing of dilution presence.
• The methods were successfully applied to real-world infectious disease datasets, demonstrating practical utility.

Conclusions:

• The new approach provides a robust method for analyzing group testing data with dilution, enhancing regression model accuracy.
• This method is advantageous as it does not require external biomarker data, making it suitable for surveillance studies.
• The ability to formally test for dilution is a significant advancement for understanding and correcting bias in group testing.