Quantifying transmission of emerging zoonoses: Using mathematical models to maximize the value of surveillance data

Monique R. Ambrose; Adam J. Kucharski; Pierre Formenty; Jean-Jacques Muyembe-Tamfum; Anne W. Rimoin; James O. Lloyd-Smith

To test whether a high rate of spillover would inundate the system with so many cases 1296 that the temporal clustering patterns resulting from human-to-human transmission could be 1297 obscured, we simulated datasets with spillover rates up to 0.1. This value corresponds with an 1298 expected 59,312.5 spillover events during the five year simulation, which corresponds to an 1299 . CC-BY 4.0 International license is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/677021 doi: bioRxiv preprint average of 36.5 per year in each locality. At this rate of spillover, there is an average of only ten 1300 days between spillover events, a shorter period than the mean generation time for human-to-1301 human transmission events, which was sixteen days. Across the range of spillover rates tested, 1302 the method did very well at both point estimates and capturing the true parameter values within 1303 the 95% CI (an average of 94.3% of CIs included the true value of R and 94.9% included the true 1304 value of λ z ; S7 Fig, S2 Table) . As the spillover rate increased from 0.0001 to 0.1, estimates of R 1305 tended to improve (posterior means closer to true value and smaller CIs). While the absolute 1306 error on estimates of λ z increased as spillover rate increased, the relative error tended to decrease. 1307