AN EFFICIENT STRATEGY TO ESTIMATE INTENSITY AND PREVALENCE: SAMPLING METACERCARIAE IN FISHES

Accurate estimates of population-level parameters of parasites, such as prevalence and mean intensity, require large sample sizes. The processing of such samples becomes an overwhelming task when parasites are abundant, as with trematode metacercariae in fishes. In the present study, a subsampling method reduced processing time while maintaining an accurate estimation of metacercariae prevalence and intensity across 3 trematode species and 2 fish species. By double sampling, we generated regression models to predict total intensity from a combination of subsamples. The key to this approach lies in choosing the best strategy from a large number of potential subsampling routines. We selected the most efficient routine by weighing the costs and benefits of each. This approach, however, could not provide an estimate of parasite abundance when no parasites occurred in the initial subsample. To estimate prevalence accurately, our subsampling algorithm prescribed an additional sampling sequence using a new, optimal regression model. In addition, we optimized the technique to measure three parasite species infecting a single host simultaneously. This efficient subsampling procedure decreased the overall processing time per host by up to 91% while obtaining accurate (R 2 . 0.8) estimates for both prevalence and intensity. Individuals differ greatly in the number of parasites they har- bor. When parasites are numerous and widely distributed within the host, comprehensive counts can be extremely tedious, even after a few hosts. After counting tens of thousands of metacer- cariae from hundreds of fishes, we wondered if such counts could be obtained with less effort. Prevalence (proportion infected), mean intensity (parasites per infected individual), and mean abundance (parasites per in- dividuals examined) are 3 common population-level descriptors of parasite abundance (Bush et al., 1997). The accuracy and precision of these estimates increases with the number of hosts examined and with the quality of counts within each host. Such comprehensive examinations can be extremely time-consuming. One approach is to sample until the estimate converges on a stable mean within a predetermined confidence limit as deter- mined by bootstrapping, jackknifing, or parametric means. A way to further reduce the time spent per host is to examine specific tissues or organs (subsamples) rather than the entire host. An estimate of total intensity can then be extrapolated from the subsample. This technique sacrifices the accuracy of the count for individual hosts; however, it ultimately may in- crease the accuracy of population-level parameters by greatly increasing the number of hosts that can be sampled in a given time period. Estuarine fish that serve as the second intermediate host to trematode parasites provide an ideal system in which to apply a subsampling strategy. Several of these trematodes infect more than one tissue or organ in the fish; furthermore, these site generalists may be distributed throughout the host, with mean intensities of several hundred per fish. Difficulties in evaluating such high intensities are complicated further by the large sample of hosts needed for accurate measures of prevalence and mean intensity. To increase the efficiency of sample processing and, ultimately, the accuracy of popula- tion-level estimates, we developed a subsampling approach