A mixed-effects Bayesian regression model for multivariate group testing data

Area of Science:

• Statistics
• Epidemiology
• Biostatistics

Background:

• Group testing and multiplex assays enhance infectious disease screening efficiency.
• Complex data structures from these methods can impede public health surveillance.
• A statistical framework is needed to manage this complexity.

Purpose of the Study:

• To develop a general Bayesian framework for analyzing multiplex group testing data.
• To address challenges in infectious disease surveillance posed by complex data structures.
• To enable accurate estimation of disease prevalence and correlations.

Main Methods:

• A mixed multivariate probit model was developed for group testing data.
• The framework incorporates correlations between disease statuses and population subgroup heterogeneity.
• Spike and slab priors were used for automated variable selection.
• A posterior sampling algorithm was created for model fitting.

Main Results:

• The Bayesian framework successfully estimates disease prevalence from multiplex group testing data.
• The model accounts for complex dependencies and population heterogeneity.
• Numerical studies and real-world data analysis (chlamydia, gonorrhea) demonstrate methodology effectiveness.

Conclusions:

• The proposed Bayesian framework offers a robust solution for analyzing multiplex group testing data.
• This approach enhances the efficiency and accuracy of infectious disease surveillance.
• The methodology is adaptable to various group testing protocols and multiplex assay designs.