Reexamining the Pooled Sampling Approach for Estimating Prevalence of Infected Insect Vectors

Our goal was to estimate seasonal changes in the proportion of Asian citrus psyllid, Diaphorina cart Kuwayama (Hemiptera: Psyllidae), carrying Candidatus. Liberibacter asiaticus. Our approach was to test Asian citrus psyllid by using pooled samples. The initial question was about pool size and the consequences of choosing poorly. Assuming no loss in sensitivity when diluting one infected individual with many healthy individuals, then it is recommend that a combination of all the published limits be used: keep the number of pools (n) above 20, the pool size (k) below 100, and the number of infected pools less than half the total number of pools. The most conservative approach to achieving the latter is to optimize pool size given an infection rate (p) such that k = In (0.5) / In (1 - p). Exceeding these limits increases the probability that all the pools will be infected. If this occurs, then that particular sample will be discarded. Use of multiple pool sizes can be used to manage this risk, but this approach may not always be practical. PooledInfRate is a good program for estimating prevalence, and it is available for free from the Centers for Disease Control and Prevention (CDC). The program provides corrected confidence intervals for prevalence estimates using one or multiple pool sizes. We used a randomization test approach as a contrasting methodology. The bias corrected CDC 95% confidence interval is an upper bound to the "true" 95% confidence interval, and we provide an estimate of the magnitude of the remaining bias in the estimate.