Epidemic surveillance systems and containment strategies in complex networks

Healthcare systems encounter significant challenges in detecting and controlling the spread of infectious diseases while preventing healthcare resource overburden. One of these challenges is the implementation of surveillance strategies to locate hot spots of transmission in a population (individuals or groups of individuals) that might have a high impact on transmission dynamics. In the context of contact network epidemiology, it is known that central nodes (nodes with high degree, eigenvector centrality, strength, and betweenness centrality, among others) can be used as sensors to detect the early emergence of outbreaks. This study addresses the question: What is the most informative subset of susceptible individuals for early epidemic detection and control? We evaluate four novel surveillance strategies in complex directed-weighted networks using an SEIR epidemic model across four networks with varying complexities and disease intensities (R0=[min,max]). We observed that the optimal surveillance strategy was highly dependent on both network structure and infectiousness of the disease. Results show that in low-assortativity, strongly connected networks, the dissimilarity strategy best predicts peak transmission, independent of disease intensity and structure. In highly connected, low-density scale-free networks, the ensemble strategy correlates strongly with the epidemic curve and provides early warnings under moderate and high disease intensities.