Individual and temporal variation in pathogen load predicts long-term impacts of an emerging infectious disease.

Abstract

Emerging infectious diseases increasingly threaten wildlife populations. Most studies focus on managing short-term epidemic properties, such as controlling early outbreaks. Predicting long-term endemic characteristics with limited retrospective data is more challenging. We used individual-based modeling informed by individual variation in pathogen load and transmissibility to predict long-term impacts of a lethal, transmissible cancer on Tasmanian devil (Sarcophilus harrisii) populations. For this, we employed approximate Bayesian computation to identify model scenarios that best matched known epidemiological and demographic system properties derived from 10 yr of data after disease emergence, enabling us to forecast future system dynamics. We show that the dramatic devil population declines observed thus far are likely attributable to transient dynamics (initial dynamics after disease emergence). Only 21% of matching scenarios led to devil extinction within 100yr following devil facial tumor disease (DFTD) introduction, whereas DFTD faded out in 57% of simulations. In the remaining 22% of simulations, disease and host coexisted for at least 100yr, usually with long-period oscillations. Our findings show that pathogen extirpation or host-pathogen coexistence are much more likely than the DFTD-induced devil extinction, with crucial management ramifications. Accounting for individual-level disease progression and the long-term outcome of devil-DFTD interactions at the population-level, our findings suggest that immediate management interventions are unlikely to be necessary to ensure the persistence of Tasmanian devil populations. This is because strong population declines of devils after disease emergence do not necessarily translate into long-term population declines at equilibria. Our modeling approach is widely applicable to other host-pathogen systems to predict disease impact beyond transient dynamics.