The RAPIDD Ebola forecasting challenge: Model description and synthetic data generation

Abstract

The Ebola forecasting challenge organized by the Research and Policy for Infectious Disease Dynamics (RAPIDD) program of the Fogarty International Center relies on synthetic disease datasets generated by numerical simulations of a highly detailed spatially-structured agent-based model. We discuss here the architecture and technical steps of the challenge, leading to datasets that mimic as much as possible the data collection, reporting, and communication process experienced in the 2014-2015 West African Ebola outbreak. We provide a detailed discussion of the model's definition, the epidemiological scenarios' construction, synthetic patient database generation and the data communication platform used during the challenge. Finally we offer a number of considerations and takeaways concerning the extension and scalability of synthetic challenges to other infectious diseases.

Keywords: Computational modeling; Ebola; Forecast.

Figure 1

A Map of Liberia with county capitals, major hospitals and population. B Ebola transmission dynamic scheme. The values and the distribution of the transition times are reported in the Supplementary Information.

Figure 2

Stochastic simulation output. 95% CI, 50% CI, median simulation and the simulation selected as representative for the scenario are depicted. A Scenario 1. B Scenario 2. C Scenario 3. D Scenario 4.

Figure 3

Cumulative number of cases as a function of time in each county of Liberia for the four scenarios considered in the challenge. Each epidemic is based on the stochastic simulation selected in Fig. 2.

Figure 4

Snapshot of the landing web page providing the challenge participants access to the database and the Supplementary Information regarding each scenario.

Figure 5

Weely number of cases in the original model output and as provided for the challenge after adding noise and underreporting. A Scenario 1. B Scenario 2. C Scenario 3. D Scenario 4. Vertical lines indicate the 5 data release dates of each scenario, corresponding to 5 prediction times points.

Figure 6

A visualization of a branch of the EVD transmission tree in Scenario 1. Nodes of various colors and shapes denote different type of EVD patients. Links denote the transmission routes among cases.