Modelling Ebola in West Africa

What did the study set out to achieve?

The study developed a number of mathematical models of the spread of Ebola in West Africa and fitted these to emerging data using Bayesian methods. These models were used to inform a variety of decisions in real-time, and to evaluate the effectiveness of various interventions. The models and analyses, which were distributed widely to partners, were also used to help plan clinical trials more efficiently (in particular vaccine trials).

A series of district-level models were also developed (both deterministic and stochastic) to estimate the impact of Ebola treatment centres (ETCs) and Ebola Community Units (also known as Community Care Centres). The project further developed the model to assess the impact of the expansion of treatment beds in both Liberia and Sierra Leone.

In doing developing these models, this work helped track changes in the epidemiology; analyse the impact of different countermeasures; and estimated the impact of different approaches to treat patients and control the epidemic.

What were the key findings?

• Impact of Ebola Community Units – Real-time models were used to help plan for how many treatment unit and community care centres might be required to limit the spread of the epidemic.
• Evaluate changes in incidence and Case Fatality Rate (CFR) – These were tracked over time and across different geographical and organisational stratifications (e.g. by treatment unit). The study also correlated CFRs with underlying characteristics of the patients, e.g. age, sex, delay between onset and hospitalisations.
• Develop and parameterise a real-time spatially explicit meta-population model – The research developed and parameterised a real-time spatially explicit meta-population model in which individuals could spread infection to others within their district (=prefectures or counties) or neighbouring districts. The model was fitted to the weekly incidence data on EVD and took account of under-reporting of cases and changes to the reproduction number over time.
• Analyse individual-level data on previous EVD outbreaks from DRC – Data was compiled and analysed from all previous EVD outbreaks recorded from the Democratic Republic of Congo. These analyses and the full anonymised datasets were published and made freely available
• Impact of Socio-cultural factors on disease spread – the study attempted to collect data on operational and behavioural variables, in order to correlate these with changes in the EVD reproduction number.

Key outputs:

• Weekly reports and projections: http://ntncmch.github.io/ebola/
• Two-day modelling symposium held at LSHTM: https://sites.google.com/site/ebolamodelling/
• 11 peer-reviewed publications (See full details below)