COVID 19: Where should we head next?
Being stuck at home for over a month, you must have figured out your daily routine by now. And I bet there’s at least 15 mins of Facebook scrolling amid all the work we have. It’s completely okay and it’s normal!
So likewise, I’m used to scroll through my news feed once in a while in the middle of work. No wonder most of it is about COVID 19 these days. I notice most people play around with numbers and the love for statistics has become real over the past couple of weeks. I saw a plenty of dashboards, visualizations and models coming up , which I personally believe is a good thing. BUT, the million dollar question, “ Is this data correctly visualized/modeled ?”
Being a person who comes from a statistical background, what I have learnt since the inception is that there’s some sort of “UNCERTAINTY” associated with statistical models. In other words, instead of saying something is happening for sure, we are always looking at a probability of something happening.
Some COVID 19 epidemiological models scare both you and me saying there will be millions of deaths worldwide, its not going to end too soon etc etc. But, do we really need to be scared ? Where should we head next with these existing Models ? When I couldn’t answer these questions it just made me guilty for not taking the epidemiology course module in uni and made me curious to find out the basis of these epidemiological models. And here’s what I found:
Typically the models are created by epidemiologists to predict the progression of an infectious disease. These models are based on several epidemiological theories. However there have been scenarios in the past where these theories have been overridden by practical consequences. One fine example is Dr. John Snows Cholera map.
John Snow’s Cholera Map
This famous illustration drives us back in time to the year 1854 ,where one of the most terrifying disease outbreaks in the history of the western world occurred in London, England. Within a single week, 10% of the population of Soho (an area of the city of Westminster) had caught with Cholera.
In the world of the 1850 s, cholera was believed to be spread by miasma in the air. But the germs were not properly understood by then and hence this serious and sudden outbreak of cholera in London’s Soho was a mystery which is quite similar to the COVID 19 situation we face today.
Among these several epidemiological theories suggested to control the outbreak, one man, Dr. John Snow, had a different perspective. He believed that the danger was in the water but not in air. He initially mapped the cases as given in the image below and found that clusters of cholera cases occurred around a public water pump and later identified that this water pump was the source of the spread of cholera. This map proved, for the first time, this terrible outbreak was spreading through crystal-clear fresh water that came out of pumps but not by inhalation of city’s air. Dr. John Snow’s Cholera map drastically changed how people look at the disease outbreak.
COVID 19 Model and Change in Predictions
Several models and predictions evolved during the past month from different sources. But in most cases these models became invalid with time and certain measures. In the U.K , initially they had almost no isolation measures in place and the idea was to create “herd immunity.” But thereafter an epidemiological model from Imperial College London projected that without drastic interventions, more than half a million Britons would die from COVID-19.
This led things to change direction, shutting down public life and ordering the population to stay at home.But a few days after imposition of the lock down ,the scientist who led the Imperial College team, testified before Parliament that, with these measures they could see the eventual death toll cut to ‘substantially less’ than 20,000 which is a remarkable turn from the initial prediction. This turn in the prediction made by the scientist was widely criticized over several leading websites in the U.K.
Tricky Parts of Epidemiological Models
So as mentioned above there’s always a chance for the predictions to change , in other words these models are UNCERTAIN. Why ?
Variety of Potential Outcomes
The epidemics grow exponentially. Modeling an exponential process necessarily produce a wide range of outcomes. These outcomes are not what exactly you’ll see in the future. But it’s always a range of possibilities, a range of possibilities that are highly susceptible to change with our actions.So once an epidemiological model is believed and acted upon , the model predictions will look false because epidemics are especially sensitive to initial inputs and timing.
Lot of Unknowns
Typically in model equations we need a set of variables to plug in, to make the prediction. These variables are chosen based on the data that we have. Since this is a novel virus, the model makers might be missing on a lot of unknowns. Most of the data are still in the process of being experimented. What is the number of people who get infected within an exposed group ? Do people who recover have immunity? What are the false positive and false negative rates of our tests? Likewise the list can go on and on.
BUT , We need Epidemiological Models !
So now you must be thinking , if they are so uncertain and mislead us to a different dimension do we really need them ? Yes we do and here’s why:
Epidemiology gives something more important other than the predictions, It is instrumental in identifying and assessing our activities towards a better future by considering the disastrous possibilities that lie ahead of us. An epidemiological model could have a range of possible outcomes in a wide probability spectrum. While the more probable outcomes locate in the middle of the range, the extreme ends of this range of possibilities represent fairly optimistic and fairly pessimistic, but less likely, outcomes.
In the context of COVID 19, an Optimistic possibility could be that a lot of people might have already been infected and recovered, and are now immune, meaning we are putting ourselves through a too-intense quarantine whereas the pessimistic side of could say millions of people are dying and this is going to continue for years.
Where should we head with these Models?
Literally, we have to ignore all those optimistic possibilities and focus on those that represent the worst of the outcomes and try to overcome them as we can. For instance if social distancing reduces transmission and reduce the spread, we do it to overcome the worst future that we see. By trying to overcome those worst possibilities, we sort of unintentionally shape the parameters of a future model.
So, even though the predictions of a model might involve a lot of uncertainty , error bars and is inapplicable with the underlying data,it’s always vital for us to look at the catastrophic consequences of the model and try to act to overcome them.
End of the day it might seem that we got scared and have overreacted on the evil part of the model, but that just means We won!
