Impact of low testing numbers on chronic wasting disease apparent prevalence.

Abstract

Chronic wasting disease (CWD) is a fatal, neurodegenerative disease of cervids, and its management heavily relies on diagnostic testing. Test results are commonly used to calculate 'apparent prevalence' (AP) - the percent of animals tested for CWD (CWD tests) with CWD-positive test results (CWD cases) - but this obscures how tests and cases individually contribute to this statistic. This is most relevant when CWD testing is limited because when few animals are tested, detection of even a single infected deer can result in a high AP that poorly reflects reality. We hypothesized that when CWD testing is limited, AP is negatively driven by testing - rather than cases - with more tests corresponding to lower APs. Graphed CWD surveillance data from townships in Illinois and Wisconsin, USA, indicate that CWD AP values ≥50% were only observed when <23 deer were tested. We used Bayesian multilevel zero-inflated Beta regression to model AP as a function of CWD tests, CWD cases and nonlinear transformations of these two terms separately for each state. The best-fit models of both identified a statistically significant negative relationship between AP and testing numbers that was modified by a positive nonlinear test covariate. This means adding tests when testing is low can have a big impact on decreasing the AP, but this relationship weakens as testing increases. We urge treating apparent prevalences ≥50% with caution and emphasize the importance of increasing the test results when initial surveillance has yielded <23 tests.