Fig. 2: Efficiency and sensitivity of pooled testing during a simulated epidemic.
From: Group testing via hypergraph factorization applied to COVID-19
Average values of efficiency (relative to individual testing) and sensitivity of a variety of pooling designs are shown for each day, with results averaged across 200,000 random trials. For sensitivity, raw averages are shown as dots with degree-8 polynomial fits overlaid as curves; the curves for efficiency depict raw averages. During the days 40–90 (highlighted), the prevalence grows exponentially from 0.03% to 2.46%. a, b Comparison of HYPER with alternative methods that use n = 96 individuals per batch (a) or n = 384 individuals per batch (b). HYPER designs with q = 2 splits were chosen to have the same maximum pool sizes (nq/m = 12 for H96,16,2; nq/m = 24 for H384,32,2) as the array designs. Dorfman designs (i.e., HYPER designs with q = 1) with matching pool sizes are also included. Sensitivity (bottom panels) depends heavily on pool sizes, due to dilution of viral loads. c, d HYPER evaluated with varying numbers of pools (c, m = 32, 16, 12) and numbers of splits (d, q = 1, 2, 3). The designs are affected by the increasing prevalence over time to varying degrees. As prevalence increases, efficiency decreases (as more stage 2 tests become necessary), while sensitivity increases (as larger viral loads begin to rescue small viral loads that would have been missed). More efficient designs tend to be less sensitive, creating a trade-off.
