Wildlife disease surveillance under uncertainty: an adaptive search-theoretic framework for early detection of transboundary animal diseases

Abstract

Rapid detection is critical for successful management of transboundary animal disease incursions in wild host populations. However, decisions about how best to allocate wildlife disease surveillance effort must be made under high uncertainty. Risk-based surveillance can improve efficiency but approaches that focus surveillance too narrowly on expected high risk areas could have low power to detect unexpected events. We developed and field-tested an adaptive, search-theoretic surveillance framework for detecting transboundary animal disease incursions in wild ungulates in New South Wales, Australia. Key principles that guided the frameworks development included accommodating uncertainty, regularly updating search priorities based on expected risk and spatial coverage, and a flexible structure that allows the system to respond to changing information or conditions over time. We created a coarse state-wide risk map that served as a weakly informative prior describing expected variability in disease incursion risk, loosely focused on foot and mouth disease virus (FMDv). Risk and search values were updated every three months based on realised surveillance effort and estimated detection probabilities over the preceding 12 months, meaning that areas of persistently high risk could nonetheless have low search value if they had recently been intensively searched. Surveillance activities collected blood and swab samples from 1,964 wild pigs (Sus scrofa) during 110 sampling occasions over a two-year evaluation and refinement period. Activities sought to simulate FMDv surveillance operations, but FMDv serological tests were not available at the time. Effort was consistently concentrated in areas of high search value, with at least 74% of sampled cells in the highest risk class. Estimated surveillance system sensitivity ranged from 0.86 to 0.93 over five successive updating cycles and increased as operational procedures were refined. Although the surveillance program was based on FMDv incursion risk, it also fulfilled its secondary objective of detecting unexpected events, including detecting Japanese encephalitis virus in wild pigs before detections in humans and domestic animals. By combining risk-based surveillance with adaptive updating of search priorities in a modular structure, the framework provided a flexible and generalisable approach for early detection of transboundary and emerging animal disease incursions in wildlife populations under high uncertainty.