Optimal pooled testing design for prevalence estimation under resource constraints

Abstract

Accurate estimation of disease prevalence is essential for mitigation efforts. Due to limited testing resources, prevalence estimation is often conducted via pooled testing, in which multiple specimens are combined and tested via a single test. The pool design, i.e., the number and sizes of testing pools, has a substantial impact on estimation accuracy. Determining an optimal pool design is challenging, especially for emerging or seasonal diseases for which information on the status of the disease is unreliable or unavailable prior to testing. We develop novel optimization models for testing pool design under uncertainty and limited resources, and characterize structural properties of optimal pool designs. We apply our models to estimate the prevalence of West Nile virus in mosquitoes (the main vector of transmission to humans). Our findings suggest that estimation accuracy can be substantially improved over the status quo through the proposed optimal pool designs.