Not Every Model Can be a Supermodel

Infectious disease modeling is an important tool that can help predict the course of an outbreak and how specific interventions might fare. However, what is often lost in the course of the press reports of the model’s results and the simplified explanations given to policymakers, is the fact that this is a model and not in any way authoritative. All models are based on specific assumptions, which correspond to reality in varying degrees, as well as predictions about microbial transmission, human behavior, and other factors. All of these aspects of models lead to a degree of uncertainty in their predictions.

During the peak of the Ebola panic in the US, various modeling numbers, including one suggesting that there could be up to 1.4 million cases of Ebola, were in the press continually without the appropriate caveats such as the assumptions that formed the basis for the model that gave that result. Another important aspect of a model is whether it is deterministic and static or dynamic (stochastic). This is a vital distinction, as we know that during the course of an outbreak, things change. People modify their behavior, transmissibility changes, and countermeasures are deployed. Such facts severely limit the utility of static models as they are often unable to fully incorporate such dynamism. By contrast, stochastic models allow much more dynamism and therefore can be much more useful.

A new, somewhat technical, paper published in the Proceedings of the Royal Society B makes these important points using the Ebola outbreak as an example. In this paper, King et al. compared the results of using deterministic and stochastic models showing that deterministic models lead to an underestimation of uncertainty in prediction.

The points addressed in this paper are important to emphasize because in the midst of the complex decision-making that characterizes an outbreak response, understanding how uncertain or certain a possible scenario is can be vital. As such, highlighting the salient aspects of a model to policymakers and journalists can help ensure these high-consequence decisions are completely informed.