Highly Pathogenic Avian Influenza (HPAI) threatens wild and domestic birds, mammals, and humans. The global spread of HPAI through wild birds requires timely, spatially accurate detection for enhanced preventive measures. However, detecting outbreaks in wildlife is challenging due to reliance on opportunistic sampling and testing of small numbers of wild animals, resulting in insufficient data on the temporal and geographical distribution of infections. We aimed to create a HPAI hotspot prediction model for the Netherlands, using wild bird mortality reports and confirmed HPAI incidents from 2016–2022. Variables for human, wild bird, and sampling density were used as statistical adjusters in various combinations. The Bayesian binomial model employed a case-crossover design, where HPAI incidents in wild birds were cases, and controls used the same date and location as the case but transposed to 2019, when no HPAI was reported in the Netherlands. We tested 24 models in a ten-fold cross-validation design. The best model had an area under the curve score of 0.68, a sensitivity of 0.47, and a specificity of 0.79, and included wild bird mortality and density. The yearly spatial distribution of predicted wild bird HPAI outbreak areas generally matched laboratory-confirmed HPAI cases, especially from 2020 onward, except for an atypically intensively sampled area west of Amsterdam before 2020. Most HPAI outbreak predictions occurred from October through March. Our model highlights potential HPAI outbreak areas over time and can be used to direct sampling efforts, potentially increasing both effectiveness and timeliness.
Petie et al. (Thu,) studied this question.