The Impact of Host-Based Early Warning on Disease Outbreaks

Mark Hernandez; Lauren E Milechin; Shakti K Davis; Rich DeLaura; Kajal T Claypool; Albert Swiston

The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/2020.03.06.20029793 doi: medRxiv preprint quarantine only on alert when the system detects pre-symptomatic signs of exposure. We integrated the early warning performance metrics from our PRESAGED algorithm-populationwide probability of early detection, daily false-alarm rate, and early detection time-into a base SEIR model. As shown in Figure 7 , a host-based early warning system would trigger exposed individuals to transition into quarantine at a rate characterized by the product of the parameters λ and PD. The parameter λ is an early detection rate defined as the inverse of the system's average early detection time for a given pathogen. The parameter PD is the population-wide probability of early detection, i.e., the fraction of the exposed population that will present early detections of pathogen exposure prior to becoming infectious. Missed detection cases, when the system fails to produce an early warning, occur with a probability of 1 -PD across the exposed population. For this policy, false alarms of the host-based early warning system would cause healthy individuals to be incorrectly quarantined. Our model takes this notion into account by adding a transition rate from the susceptible to the quarantined compartment, characterized by a daily false-alarm rate κ, the probability of an individual's early warning system presenting a false alarm on a given day. . The quarantine-on-alert policy has effectively reduced the additional disease cases and has largely limited those in quarantine to those individuals who were initially exposed.