Estimating the performance of mass testing strategies for COVID-19: a case study for Costa Rica

Devising effective mass testing strategies to control and suppress COVID-19 pandemic waves make up a complex sociotechnical challenge. It requires a trade-off between performing detection technologies in terms of specificity and sensitivity, and the availability and cost of individual tests per technology. Overcoming this trade-off requires first predicting the level of risk of exposure across the population available. Then selecting testing strategies that match resources to maximize positive case detection and optimize the number of tests and their total cost during sustained mass testing campaigns. In this article, we derive the behavior of four different mass testing strategies, grounded in guidelines and public health policies issued by the Costa Rican public healthcare system. We assume a (privacy-preserving) pre-classifier applied to patient data, Capable of partitioning suspected individuals into low-risk and high-risk groups. We consider the impact of three testing technologies, RT-qPCR, antigen-based testing and saliva-based testing (RT-LAMP). When available, we introduced a category of essential workers. Numerical simulation results confirm that strategies using only RT-qPCR tests cannot achieve sufficient stock capacity to provide efficient detection regardless of prevalence, sensitivity, or specificity. Strategies that harness the power of both pooling and RT-LAMP either maximize stock capacity or detection, efficiency, or both. Our work reveals that investing both in data quality and classification accuracy can improve the odds of achieving pandemic control and mitigation. Future work will concentrate, based on our findings, on constructing representative synthetic data through agent-based modeling and studying the properties of specific pre-classifiers under various scenarios.