Impacts of environmental and socio-economic factors on emergence and epidemic potential of Ebola in Africa

Abstract

Recent outbreaks of animal-borne emerging infectious diseases have likely been precipitated by a complex interplay of changing ecological, epidemiological and socio-economic factors. Here, we develop modelling methods that capture elements of each of these factors, to predict the risk of Ebola virus disease (EVD) across time and space. Our modelling results match previously-observed outbreak patterns with high accuracy, and suggest further outbreaks could occur across most of West and Central Africa. Trends in the underlying drivers of EVD risk suggest a 1.75 to 3.2-fold increase in the endemic rate of animal-human viral spill-overs in Africa by 2070, given current modes of healthcare intervention. Future global change scenarios with higher human population growth and lower rates of socio-economic development yield a fourfold higher likelihood of epidemics occurring as a result of spill-over events. Our modelling framework can be used to target interventions designed to reduce epidemic risk for many zoonotic diseases.

Fig. 1

System-dynamics model of zoonotic disease transmission. Letters a–f indicate major system components, arrows showing links, and key sub-components in smaller font. Within the global physical environment (a), both the host niche (b) and infected host niche (c) are nested subsets, which all vary over a relatively slow time-scale. Endemic human populations are nested within the global human population (e), with human socio-economic factors (g) affecting all human populations. Spill-overs happen in the fast-moving spatial and temporal interface of these two nested systems (d), where both infected hosts, susceptible people and spill-over specific factors occur, resulting in infected human populations (f)

Fig. 2

Environmental Mechanistic Model (EMM) EBOV Simulation Schematic. Within 0.0416° (5.6 km at equator) grid cells across the globe (a), we used a SEIFR (Susceptible, Exposed, Infectious, Funeral and Removed) disease compartmental model (b), to estimate the number of people in each compartment. S–E transmission rate was determined for each grid cell by calculating the force of zoonotic infection (between hosts and humans) λ_z, and within human populations λ (see Methods). Travel of exposed or infectious individuals between grid cells occurred across existing road and flight transport networks, with transmission rate ε_fr. Mean transition rates used as the starting parameters for simulations were as follows: α for E–I was calculated as the reciprocal of incubation time in days (α = 1/7), γ_σ (I–F transition rate) was the product of the probability of the reciprocal of days infectious (γ = 1/9.6) and maximum poverty-weighted case fatality rate (σ = 0.78), γ_1-σ (I–R transition rate) was the product of the probability of the reciprocal of days infectious (γ = 1/9.6) and probability of recovering (1-σ), and γ_F (F–R transition rate) was the reciprocal of the burial time of 2 days. Each simulation was run 2500 times for 365 days, only including grid cells containing a human population. The total number of people in each compartment per grid cell, per day from each simulation was then used to calculate the total number of index and secondary cases and mapped spatially (c). Bioclimatic, land use and demographic conditions were then changed to predicted values for 2070 to estimate changes to λ and λ_z, and the simulations repeated to investigate impacts of global change on disease (d)

Fig. 3

Present day risk of EVD cases caused by Zaire Ebola virus (EBOV) from EMM simulations. Maps represent the mean number of EVD-EBOV cases between zero (dark blue) and 0.3 (yellow) per grid cell (0.0416°—5.6 km at equator) across 2500, 365-day simulation runs for the present day, where (a) shows all cases (both index and secondary), (b) index cases only, and (c) index cases from epidemics (1500+ cases). White crosses in (a) represent log outbreak size. White symbols in (b) represent all locations of known EVD index cases from different viral strains, where circles represent Zaire (EBOV), square Sudan (SUDV), triangles Taï Forest (TAFV), and tetrahedrons Bundibugyo (BDBV). Diagonal white cross in (c) represents location of index case from 2014–2016 epidemics, white stars the locations of Ebola outbreaks that have occurred since the modelling was run

Fig. 4

Most common country locations for importation of EBOV infected individuals. Map shows the number of importations per simulation that countries outside Africa received via airline flights. Countries with the most EBOV infected individuals imported are represented in yellow with darker green, then blue, coloured countries having proportional fewer importations and dark blue showing zero importations and the EVD endemic area. Data come from 2500 simulations of EVD outbreaks under present data climate, land-use, demographic and transportation conditions

Fig. 5

Change in future risk of EVD cases caused by Zaire Ebola virus (EBOV) for 2070. Maps represent mean change in per grid cell (0.0416°—5.6 km at equator) EVD case probability from zero (yellow) to −0.06 (green) and 0.06 (red), aggregated at the country level with data from EMM simulations for 2070. Rows and columns show all reasonable combinations of the different scenarios of global change (GCAM-RCP4.5, AIM-RCP6.0, MESSAGE-RCP8.5 and SSP1 to 3)

Fig. 6

Comparison of 2070 EMM simulation scenarios by EVD-EBOV final outbreak size. Circles represents standardized residuals from a chi-squared test of association (χ = 466.27, df = 10, p < 0.001) between simulation scenario and final outbreak size category. More orange/red colours show a greater than expected (vs. randomly allocated) number of outbreaks for any given combination of scenario and final outbreak size, with more blue colours representing fewer than expected outbreaks. Size of circle also indicates the overall quantity different to expected, with large circles contributing more to the overall chi-value compare to smaller circles