Early detection of emerging swine pathogens: Comparative evaluation of online change-point methods

Timely detection of pathogen outbreaks in the U.S. swine population is crucial for protecting animal health and mitigating disruptions to food production. This study evaluates the efficacy of several state-of-the-art online change-point detection methods, including BEAST, Strucchange, a weighted CUSUM method, and several versions of BFastmonitor, in detecting outbreaks in pathogen case time series quickly and reliably. Time series data are simulated from state-level pathogen surveillance of Influenza A virus (IAV), Mycoplasma hyopneumoniae (MHP), Porcine deltacoronavirus (PDCoV), Porcine epidemic diarrhea virus (PEDV), and Porcine reproductive and respiratory syndrome virus (PRRSV) in the largest U.S. pork-producing states. Our results indicate that methods incorporating explicit modeling of trend and seasonality deliver more robust outbreak detection, particularly for time series exhibiting complex temporal and seasonal patterns. These methods generally outperform others in both timeliness and reliability, underscoring their operational value. Furthermore, we examine practical aspects of long-term monitoring, including the influence of stable period length, spacing between outbreaks, nonstationarity, and dependence structures. To demonstrate real-world applicability, we provide a detailed case study of BFastmonitor applied to an eight-year historical swine pathogen time series, highlighting its feasibility for routine surveillance.