Read on Twitter
Why do COVID-19 modelling groups typically produce ‘scenarios’ rather than long-term forecasts when exploring possible epidemic dynamics? A short thread... 1/
Coverage of modelling is often framed as if epidemics were weather - you make a prediction and then it happens or it doesn’t. But COVID-19 isn’t a storm. Behaviour and policy can change its path... 2/
This means that long-term COVID forecasts don’t really make sense, because it’s equivalent of treating future policy & behaviour like something to be predicted from afar (more in this piece by @reichlab & @cmyeaton: https://www.washingtonpost.com/outlook/2020/09/15/scientists-want-predict-covid-19s-long-term-trajectory-heres-why-they-cant/). 3/
But of course, policy can change rapidly, informed by available evidence. Pointing out ‘a storm is coming’ won’t stop a storm, but pointing out a growing COVID outbreak could result in efforts to curb transmission. 4/
That’s why COVID models generally use scenarios, to illustrate epidemiological consequence of available options (including doing nothing). In this sense, models are a tool to aid decision making, not one to make weather-style predictions. 5/
That being said, it can be possible to make short-term forecasts of things like COVID hospitalisations and deaths, because these are impacts of events (i.e. infections) that have already happened, and hence should be more predictable. 6/
It’s worth noting there are some infections where longer term forecasting does make more sense. E.g. flu dynamics depend on immunity and seasonality but not on reactive control, which makes forecasting theoretically (if not practically) easier: https://www.pnas.org/content/116/8/3146 7/
For endemic childhood infections - like rotavirus - cycles of outbreaks driven by build up of immunity then new susceptibility among children have made it possible to predict long-term effects of introducing vaccination.
https://science.sciencemag.org/content/325/5938/290.abstract 8/
https://science.sciencemag.org/content/325/5938/290.abstract 8/
More on how to evaluate forecasts, and discussion of challenges of forecasting infections like Ebola, where control measures also curbed transmission ( https://journals.plos.org/ploscompbiol/article?id=10.1371/journal.pcbi.1006785). 9/9
