As the COVID-19 pandemic continues worldwide, there is an urgent need to detect infected patients as quickly and accurately as possible. Group testing proposed by Technion [2] could improve efficiency greatly. However, the false negative rate (FNR) would be doubled. Using USA as an example, group testing would have over 70,000 false negatives, compared to 35,000 false negatives by individual testing. In this paper, we propose a Flexible, Accurate and Speedy Test (FAST), which is faster and more accurate than any existing tests. FAST first forms small close contact subgroups, e.g. families and friends. It then pools subgroups to form larger groups before RT-PCR test is done. FAST needs a similar number of tests to Technion's method, but could sharply reduce the FNR to a negligible level. Again taking USA as example, FAST reduces the number of false negatives to just 2000 while it is seven times faster than individual testing.
Flexibility of FAST strategy
With appropriate settings including prevalence and within-close-contact-subgroup-infected rate, FAST can generate the optimal pooling size with the corresponding best efficiency to achieve. Under the guidance of FAST, the falsenegative rate and the false positive rate can be maintained at a dramatically low level with good performance in efficiency in some range of parameters.
"Predictive” group testing In some areas, government would propose universal testing among all the population,where millions of individuals could be involved. In order to perform a more efficient testing, it is prioritized to test people who have high probability of getting infected. Before testing, we can make prediction of the risk level for someone based on their contact information, symptoms and residential areas. Such risk prediction is usually non-uniform among different groups of people. In the future, we hope to develop methods to predict the risk of people and build testing strategies based on different predictions.
Reference & Source information: https://www.medrxiv.org/
Read More on :