Machine learning-driven COVID-19 early triage and large-scale testing strategies based on the 2021 Costa Rican Actualidades survey

Abstract

The COVID-19 pandemic underscored the importance of mass testing in mitigating the spread of the virus. This study presents mass testing strategies developed through machine learning models, which predict the risk of COVID-19 contagion based on health determinants. Using the data from the 2021 “Actualidades” survey in Costa Rica, we trained models to classify individuals by contagion risk. After theorize four possible strategies, we evaluated these using Monte Carlo simulations, analyzing the distribution functions for the number of tests, positive cases detected, tests per person, and total costs. Additionally, we introduced the metrics, efficiency and stock capacity, to assess the performance of different strategies. Our classifier achieved an AUC-ROC of 0.80 and an AUC-PR of 0.59, considering a disease prevalence of 0.26. The fourth strategy, which integrates RT-qPCR, antigen, and RT-LAMP tests, emerged as a cost-effective approach for mass testing, offering insights into scalable and adaptable testing mechanisms for pandemic response.