Optimizing early detection of avian influenza H5N1 in backyard and free-range poultry production systems in Thailand

Abstract

For infectious diseases such as highly pathogenic avian influenza caused by the H5N1 virus (A/H5N1 HP), early warning system is essential. Evaluating the sensitivity of surveillance is a necessary step in ensuring an efficient and sustainable system. Stochastic scenario tree modeling was used here to assess the sensitivity of the A/H5N1 HP surveillance system in backyard and free-grazing duck farms in Thailand. The whole surveillance system for disease detection was modeled with all components and the sensitivity of each component and of the overall system was estimated. Scenarios were tested according to selection of high-risk areas, inclusion of components and sampling procedure, were tested. Nationwide passive surveillance (SSC1) and risk-based clinical X-ray (SSC2) showed a similar sensitivity level, with a median sensitivity ratio of 0.96 (95% CI 0.40–15.00). They both provide higher sensitivity than the X-ray laboratory component (SSC3). With the current surveillance design, the sensitivity of detection of the overall surveillance system when the three components are implemented, was equal to 100% for a farm level prevalence of 0.05% and 82% (95% CI 71–89%) for a level of infection of 3 farms. Findings from this study illustrate the usefulness of scenario-tree modeling to document freedom from diseases in developing countries.