# Prediction model for spread of coronavirus

A previous thread on Bioinformatics StackExchange citing an article by the Economist which references a cartoon of predictive outcomes by the CDC (Economist article on coronavirus) suggests a model in epidemiology using the "basic reproductive rate", commonly known as R0. Although I'm no bioinformatician, I imagine that one could tweak a model based on similar infections like SARS and HDN1. Yet, I found no models predicting or giving confidence intervals for the spread of coronavirus. I wonder if the the data is still too sparse, e.g. we don't know if it's seasonal (mentioned in the above thread) and we cannot infer the virulence.

Would such a model be reliable? If not, how many months of data would we need to reliably use a model predicting the spread?

## 1 Answer

The original R0 calculation was developed in the early 1900s to model malaria transmission and is known as the Ross-MacDonald model so your idea is over 100 years old. The model was very famous for identifying the optimal strategy for malarial control by targeting the adult mosquito rather than the mosquito larvae. What it said was reducing the mosquito population and reducing malaria transmission were two different issues. So the bednet strategy is based in this idea and particularly spraying the inside of houses with insecticide.

R0 was famously used in the 1980s to predict the HIV pandemic when WHO thought it would remain localised in Africa.

The problem with R0 and Coronaviruses is we don't know the asymptomatic population who could also transmit the virus.

This population is potentially very large. Epidemiologists will attempt extrapolation from cruise ship quarrentine, eg the Diamond Princess where everyone is tested to understand this transmission parameter and project this onto the rest of the planet. Clearly this could have limitations or inaccuracies. The other issue I am baffled by is seasonal transmission and whether this will occur with COVID-19.

One thing R0 of COVID-19 will identify is the latent period transmission which occurs before the symptoms, i.e. the incubation period. Essentially, people with cough/mild symptoms before they self isolate and become really ill with a lung infection will transmit the virus. In this context R0 will be/was very influential for understanding when to move between quarrentine and containment phases of the pandemic.

• Have other models improved on this 100 year old model? It is now around 40 days since the quarantine of the Diamond Princess and around 2 weeks since they all disembarked; when would we know the asymptomatic population on that ship? Commented Mar 13, 2020 at 8:47
• The 100 year old model is for malaria. The fundamental idea of a (R0) compartmental model which is solved by integration is very much alive.
– M__
Commented Mar 13, 2020 at 9:09
• The social stuff on the Diamond Princess I don't know about, I just know it is an ideal population to help solve the asymptomatic/transmission issue.
– M__
Commented Mar 13, 2020 at 9:10
• @miguelmorin there are many studies on this actually, but as Michael explains, the R0 values are all estimates and vary wildly. Just to give you an idea, this study, which is being used to inform the planning of the UK government (and probably others) posits an R0 of 2.4, while this one (which has not been peer reviewed, and which makes some pretty enormous assumptions; I wouldn't trust it much, personally) suggests R0 could be as high as 26.5.
– terdon
Commented Mar 26, 2020 at 13:21
• To illustrate just how different those values are, if we assume R0=2.4, then if I am infected, I will infect an average of 2.4 other people. If they then do the same, after 10 steps, I will have ultimately infected 2.4^10 = ~6340 people. If R0=26.5, then after 10 steps, I will have infected 26.5^10 = ~1.7e10^14. That's 17 followed by 13 0s, one hundred and seventy trillion people.
– terdon
Commented Mar 26, 2020 at 13:32