Hierarchical Clustering on Principal Components Analysis to Detect Clusters of Highly Pathogenic Avian Influenza Subtype H5N6 Epidemic across South Korean Poultry Farms

Abstract

Several outbreaks of highly pathogenic avian influenza (HPAI) in poultry have already been documented across the world, causing major economic losses. Research on diverse perspectives for future HPAI outbreaks’ prevention is desperately needed. It is critical to determine high-risk areas for HPAI outbreaks in order to develop high-level biosecurity in all such areas. The aim of this study is to identify high-risk areas as hotspots for high rates of birds’ infection and mortality and culling. We used “hierarchical clustering on principal components” (HCPC) to classify infected poultry farms in South Korea based on the point prevalence rate, infections, and deaths in susceptible birds. The linear combination of the original predictors was determined using “principal component analysis (PCA)”. Based on PCA, we applied the hierarchical clustering algorithm, which divided the data into four clusters based on the dissimilarity matrix. These four groups of poultry farms were identified on the basis of five variables. According to the findings based on the HCPC method, poultry farms in “cluster 4” had significantly higher average bird infections with high mortality when compared to other clusters. Similarly, the poultry farms in “cluster 2” had robust average bird culling in place to limit bird infectivity and mortality due to a high number of susceptible birds. The poultry farms belonging to “cluster 3” had a significantly higher average point prevalence rate of HPAI H5N6 cases than the rest of the clusters. Based on this study, it is recommended that poultry farms with a high number of infections and mortality in susceptible birds should implement proper biosecurity management to control HPAI infections while avoiding the culling of a large number of birds.