Accelerated infection testing at scale: a proposal for inference with single test on multiple patients

Abstract

In pandemics or epidemics, public health authorities need to rapidly test a large number of individuals, both to determine the line of treatment as well as to know the spread of infection to plan containment, mitigation and future responses. However, the lack of adequate testing kits could be a bottleneck, especially in the case of unanticipated new diseases, such as COVID-19, where the testing technology, manufacturing capability, distribution, human skills and laboratories might be unavailable or in short supply. In addition, the cost of the standard PCR test is approximately USD 48, which is prohibitive for poorer patients and most governments. We address this bottleneck by proposing a test methodology that pools the sample from two (or more) patients in a single test. The key insight is that a single negative result from a pooled sample likely implies negative infection of all the individual patients. and It thereby rules out further tests for the patients. This protocol, therefore, requires significantly fewer tests. This may, however, result in somewhat increased false negatives. Our simulations show that combining samples from two patients with 7% underlying likelihood of infection implies that 36% fewer test kits are required, with 14% additional units of time for testing.