Additional file 8 of Quantifying the potential value of antigen-detection rapid diagnostic tests for COVID-19: a modelling analysis

Additional file 8: Figure S5. Sensitivity analysis to varying Ag-RDT sensitivity and specificity. As a focal model output, all figures show the proportion of simulations in which an Ag-RDT was favourable, with different algorithms labelled by the different line colours. Panels A-C show the sensitivity of Ag-RDT being varied between 75 and 95% across the hospital and community settings, assuming specificity remained fixed at 98%. Panels D-F show the specificity of Ag-RDT being varied between 98 and 100%, assuming sensitivity remained fixed at 80%. Similar to the analysis presented in the main text, we assumed that all individuals were isolated whilst waiting for a NAT result in the hospital setting and that no one isolated whilst awaiting a test result in the community setting. Results illustrate that, in a community setting, increasing Ag-RDT sensitivity increased the favourability of the “Ag-RDT only” and “confirm Ag-RDT negative” strategies (panel C). For example, the favourability of an algorithm that confirms an Ag-RDT negative result increased from 79% to 83% when sensitivity increased from 75% to 95%. Increasing sensitivity had little impact on the “confirm Ag-RDT positive” strategy; since the only costs incurred under a community setting was the cost of a test, a NAT-only strategy was often cheaper and averted more infectious days than the “confirm Ag-RDT positive” strategy (the cost of testing with an Ag-RDT and confirming a positive result with a NAT test makes it costly, and by re-testing a positive result with a NAT, the sensitivity of the algorithm was lower due to the imperfect sensitivity of NAT). Similar to the hospital setting, specificity had little impact on an algorithm’s favourability (panel F).