An empirical analysis of the Ebola outbreak in West Africa

TL;DR

This paper analyzes the 2014-2016 West African Ebola outbreak using an agent-based SIR model on a Euclidean network, fitting the data well and estimating key outbreak parameters.

Contribution

It introduces a novel application of a Euclidean network SIR model to fit Ebola outbreak data and estimate susceptible populations.

Findings

The model fits the outbreak data well.

Estimated peak infection times match observed data.

Reproduces data qualitatively with appropriate parameters.

Abstract

The data for the Ebola outbreak that occurred in 2014-2016 in three countries of West Africa are analysed within a common framework. The analysis is made using the results of an agent based Susceptible-Infected-Removed (SIR) model on a Euclidean network, where nodes at a distance $l$ are connected with probability $P(l) \propto l^{-\delta }$, $\delta$ determining the range of the interaction, in addition to nearest neighbors. The cumulative (total) density of infected population here has the form $R(t) = \frac{a\exp(t/T)}{1+c\exp(t/T)}$, where the parameters depend on $\delta$ and the infection probability $q$. This form is seen to fit well with the data. Using the best fitting parameters, the time at which the peak is reached is estimated and is shown to be consistent with the data. We also show that in the Euclidean model, one can choose $\delta$ and $q$ values which reproduce the data for the three countries qualitatively. These choices are correlated with population density, control schemes and other factors. Comparing the real data and the results from the model one can also estimate the size of the actual population susceptible to the disease. Rescaling the real data a reasonably good quantitative agreement with the simulation results is obtained.