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

The ability of the model to robustly predict future outbreak trajectory was limited and depended on the data source (Fig 2) as well as on the time window used for inference (calibration window) and the forecast horizon. Results using a 2-week calibration window and a 4-week forecast horizon using ProMED data are presented in the main text (see Figs 6 and 7 for other forecast horizons and calibration windows). Overall, 48.7% of weekly observed incidence across all three countries were included in the 95% forecast interval (49.3% and 57.5% for HealthMap and WHO respectively, SI Table 1 ). Typically, model forecasts were 0.5 times lower or higher than the observed incidence (95% CrI 0.0 -32.0) based on the median relative mean absolute error (Fig 3D) , see Methods for details). We found no evidence of systematic bias in any week of the forecast horizon (median bias 0.12, Fig 3A) .