Online algorithm for assignment of specimens to pooled or individual testing using risk models provides a practical way to increase testing capacity

Abstract

Background: To improve throughput in diagnostic and screening testing for infectious diseases, I developed a straight-forward algorithm that uses individual risk to optimize the decision about pooled or individual testing. Methods: The online greedy algorithm provides an recommendation for filling pooled testing queue for optimal testing in pools of variable size. Observational data from Medical University of South Carolina COVID-19 diagnostic testing was used to estimate capacity gains under this algorithm versus optimal fixed pooling based on population prevalence. Results: The online pooling recommendations based on this algorithm resulted in statistically better capacity gains than optimal pools of fixed size (P-value 0.003 and 0.002, for pools of 5 or 6, respectively). This is especially significant since the underlying individual-level risk prediction model attained only a moderate predictive accuracy. Conclusions: This result suggests that when combined with a better risk prediction and integrated in an appropriate informatics ecosystem this approach cab offers an opportunity for resilient pooled testing strategies for pathogens while incorporating relevant operational constraints of pathology laboratories.