Bayesian spatial modelling of Ebola outbreaks in Democratic Republic of Congo through the INLA-SPDE approach.

Abstract

Ebola virus (EBV) disease is a globally acknowledged public health emergency, endemic in the west and equatorial Africa. To understand the epidemiology especially the dynamic pattern of EBV disease, we analyse the EBV case notification data for confirmed cases and reported deaths of the ongoing outbreak in the Democratic Republic of Congo (DRC) between 2018 and 2019, and examined the impact of reported violence on the spread of the virus. Using fully Bayesian geo-statistical analysis through stochastic partial differential equations (SPDE) allows us to quantify the spatial patterns at every point of the spatial domain. Parameter estimation was based on the integrated nested Laplace approximation (INLA). Our findings revealed a positive association between violent events in the affected areas and the reported EBV cases (posterior mean=0.024, 95% CI: 0.005, 0.045) and deaths (posterior mean=0.022, 95% CI: 0.005, 0.041). Translating to an increase of 2.4% and 2.2% in the relative risks of EBV cases and deaths associated with a unit increase in violent events (one additional Ebola case is associated with an average of 45 violent events). We also observed clusters of EBV cases and deaths spread to neighbouring locations in similar manners. Findings from the study are therefore useful for hot spot identification, location-specific disease surveillance and intervention.