Optimally Pooled Viral Testing

Ben-Amotz, D.

It has long been known that pooling samples may be used to minimize the total number of tests required in order to identify each infected individual in a population. Pooling is most advantageous in populations with low infection probability, but remains better than non-pooled testing up to an infection probability of 30%. The present predictions imply that optimal pooling may be used to extremely efficiently test populations with infection percentages down to 0.1%, in which case a single round of optimal pooling with an average of as few as 6 tests may be sufficient to uniquely identify every infected individual in a population of 100 (and increases to 20 tests when at 1% infection). Additional testing efficiency may be realized by performing a second round of pooled testing, thus reducing the average number of tests required to test a population with 1% infection from 20 to 14 out of 100 (and from 6 to 4 with 0.1% infections). These best case predictions, obtained assuming perfect test accuracy and specificity, provide a quantitative measure of the optimal pool size and expected testing efficiency gains in populations with infection probabilities ranging from 0.1% to 30%, and are supported by recent COVID-19 detection sensitivity and optimized pool size experiments.