Using real-time modelling to optimise an outbreak response: Insights from the 2017 Ebola outbreak in the Democratic Republic of the Congo

Abstract Important questions for policy makers during infections disease outbreaks include: i) How effective are public health measures?; ii) When can resource-intensive and restrictive interventions be removed? We used mathematical modelling to address these questions during the 2017 outbreak of Ebola virus disease in Likati Health Zone, Democratic Republic of the Congo. The index case developed symptoms on 27th March 2017, and eight cases occurred in total prior to the arrival of the Ebola Response Team (ERT) on 15th May 2017. We used a branching process transmission model to estimate that before the arrival of the ERT, the reproduction number was R=1.49 (95% credible interval (0.67,2.81)). Based on the full distributional estimate of R, the risk of further cases occurring if the ERT had not been deployed was estimated to be 0.97 (i.e., there was a 97% chance of additional cases in the absence of the ERT). Following the arrival of the ERT, no further cases arose, suggesting that interventions implemented by the ERT were effective. We then used the same transmission model to estimate in real-time when the ERT could be withdrawn. By the time of the end-of-outbreak declaration and withdrawal of the ERT (2nd July 2017), the risk of future cases in the absence of the ERT was only 0.01, indicating that the decision to withdraw the ERT was safe. We also evaluated the sensitivity of our modelling results to the estimated value of R, and considered different criteria for determining when the ERT could be withdrawn. As well as providing insights into interventions during the 2017 EVD outbreak, this research provides a modelling framework that can be used during future infectious disease outbreaks to determine the effectiveness of control measures and to guide when to relax or remove interventions.