Test sensitivity is secondary to frequency and turnaround time for COVID-19 surveillance.
Main result
The study shows that frequent tests and those for which results are communicated rapidly are more important than very sensitive tests in controlling the epidemic (assuming that those who test positive self-isolate effectively). Note, however, that these conclusions are valid only for tests used for epidemiological surveillance and not for individual diagnostic purposes.
Takeaways
In order to control the epidemic, it is more important to perform regular virological tests and to provide results rapidly than to have extremely sensitive tests. It is, however, necessary to test regularly.
Strength of evidence Undetermined
The model suggests that, for epidemiological surveillance, frequent tests are more important than sensitive tests, which could have implications in terms of the testing strategies. The model assumes that positive cases self-isolate effectively, which is not always the case. Furthermore, the number of tests necessary to control the epidemic is substantial (daily tests of every inhabitant of New York would require certain logistics and lots of money!), which could have been discussed. A secondary aim of the article seems to be reaffirmation of LAMP tests, sometimes criticized in the United States for their false-negative rate.
Objectives
The authors examine systematic screening strategies – that is, ones where an entire population is tested and not simply contact cases. The authors compare different strategies according to: the frequency of tests (from daily to bimonthly), the rapidity of receiving results (and thus the length of time between testing and quarantine of positive cases), and the sensitivity of the tests (minimum viral load for detection).
Method
The testing strategy is simulated using two agent-based models. One of the two models represents a well-mixed population, without specific contact structure; the other model represents transmission in the city of New York (following a previously published model), according to demographic data of the city (household size) and specific contact matrices.
In these models, individuals are either healthy, infected, recovered, isolated (after receiving a positive test), or self-isolated if they are symptomatic (20% of infections). The model assumes that each individual in the population is tested every D days using a test with a limit of detection LOD (10^3 or 10^6 viral copies/mL) and receives a result after T days.
The authors model the viral load over the course of each individual infection; this viral load affects the infectivity of the individuals, as well as the probability that a virological test will be positive. Viral load as a function of time is represented by a linear piecewise function parameterized by three points controlled by random variables: 1) the time between infection and a viral load of 10^3 copies/mL, 2) the highest viral load and the time it is reached, 3) the time for the viral load to decrease below 10^6 copies/mL. The parameters are chosen from clinical studies on the viral load of SARS-CoV-2 over time.
In the model presented in the main text, infectivity of an individual is correlated to the log10 of their viral load, and the authors test other relationships between infectivity and viral load in the appendix.
bibliovid.org and its content are bibliovid property.