Evaluation of the English bovine brucellosis surveillance system considering probability of disease introduction and non-random sampling.

The English surveillance system for bovine brucellosis was evaluated. The confidence in detecting at least one infected herd in the local population (surveillance system sensitivity or SSe), and the confidence in freedom from disease (PFree) adjusted (PFreeAdj) for the probability of disease introduction from abroad by imported animals (PIntro), were estimated for quarterly surveillance periods of 2016; because dairy herds were tested quarterly on bulk tank milk (BTM) with an antibody indirect ELISA. A stochastic model was developed and six surveillance components (representing also the local population strata), were evaluated. All English herds and their relative risk (RRs) of infection within each stratum were considered. The importance of each component was assessed using actual national data, which reflected non-random sampling. The contribution of the abortions testing was assessed with particular focus, because a decline in statutory submissions was observed in recent years. Beef herds without submissions (B-NoTest herds) at the laboratories were still considered as a population stratum, where infected cattle could be imported. Additionally, we evaluated the importance of different hypothetical design between-herds prevalence (Ph) values, at which the country could be classified as "infected". The potential negative effect on SSe due to the dilution of antibodies when individual samples are pooled within BTM and tested by the milk iELISA, was also investigated. The quarterly median SSe and PFreeAdj were both ≥ 95 % if at least four (0.008 %) herds were infected in the country due to independent import events. The system appeared able to substantiate Official Brucellosis Free (OBF) status frequently (on quarterly basis) using Ph=0.2 % (EU legislation). The component based only on BTM testing (M herds) showed the highest sensitivity; while the surveillance components based on abortions or post import calving (PIC) testing, had very low sensitivity at the (considered) Ph values lower than 0.2 %. In contrast, at Ph = 0.2 %, components based on abortion testing had median sensitivity between 91.3 % and 99.9 %, and the dilution effect on BTM testing did not change remarkably the SSe and PFreeAdj. When Ph was set to 1-2 infected herds (0.002-0.004 %), these were usually allocated by the model within the B-NoTest stratum (the largest stratum) and SSe reduced. Thus, if policy considers necessary increasing the SSe for low Phs (system's optimization as an early warning system); the cost efficiency of active risk based surveillance in beef herds (considering imports) could be investigated in the future.
Competing Interests: Declaration of Competing Interest None.
(Crown Copyright © 2020. Published by Elsevier B.V. All rights reserved.)