Using Digital Surveillance Tools for Near Real-Time Mapping of the Risk of International Infectious Disease Spread: Ebola as a Case Study

Sangeeta Bhatia; Britta Lassmann; Emily Cohn; Malwina Carrion; Moritz U.G. Kraemer; Mark Herringer; John Brownstein; Larry Madoff; Anne Cori; Pierre Nouvellet

In this section, we present additional analyses carried out on predicting the spatial spread of the epidemic. Classification of alerts raised 1, 2, 3, and 4 weeks ahead are shown in Fig 47. We also assessed the sensitivity of the model when the analysis was restricted to countries other than Guinea, Liberia and Sierra Leone (Fig 48) . At 93% threshold, the model exhibited high specificity (83.3%) and sensitivity (82.0%) in predicting presence of cases in weeks following a week with no observed cases in each country (Fig 49) . In predicting presence of cases in countries with no or low incidence, or in a week following a week in which no cases were observed, the sensitivity improved at higher thresholds with a reasonably low false alert rate (Fig 50) . Finally, we find that the model is able to attain a high sensitivity and specificity relatively early in the epidemic using all three data sources (Fig 51) . Since WHO data used here were only available for Guinea, Liberia and Sierra Leone, we affixed data from ProMED for all countries other than these three to WHO data for the purpose of classifying alerts.