Simulating the diffusion of PPR-like disease to improve surveillance

Abstract

Animal mobility is a founding element of pastoral culture and fundamental to the economy in West Africa. Herds’ movement, involving mainly cattle, sheep, and goats, is vital for adapting to climate fluctuations, optimizing natural resources, and managing risks in livestock production, as well as livestock trade and exchanges. However, these movements contribute to the spread of transboundary animal diseases such as Peste des petits ruminants (PPR). Nigeria experienced its first PPR outbreak in the 1960s-1970s, and outbreaks have regularly occurred since then. Yet, no adequate surveillance system has been put in place, and the absence of proper animal movement tracking remains a critical shortcoming in addressing this ongoing threat. Because of this, we rely onad-hocactivities, like mobility surveys, to collect this information. However, these data could be partial and limited in time and space, hindering the capacity to identify suitable areas for monitoring disease circulation (sentinel nodes). Market survey data from three northern states are collected once to reconstruct the small ruminant mobility network. A group of areas with a high potential to infect each other were identified (Contagion Cluster). A stochastic Susceptible-Infected-Recovered (SIR) was used to simulate the spread of a PPR-like disease through movements. Sentinel nodes and their key characteristics were identified using Random Forest classification. The network (missing movement) was predicted with a hierarchical random graph (HRG), and uncertainty analysis assessed the effects of missing movements on the epidemic extent and identity and characteristics of sentinel nodes. The number of sentinel nodes varied with the epidemic’s severity, but their characteristics remained consistent. The uncertainty analysis results showed that adding 1% of the most probable missing links did not affect the final size of the epidemic. However, a significant difference appears when adding more than 3% of the missing links, causing a gradual fluctuation in the epidemic’s final size. The final size of the epidemic stabilized when adding more than 50% of the most probable links. The challenge posed by incomplete animal movement data underscores the need for further research and data collection efforts. However, predicting missing links presents a promising method for enhancing the reliability of epidemic prediction and sentinel node identification. This potential to improve disease surveillance and control strategies is a significant implication of our study.