Computer models help us predict spread of pandemic

In August, over 500,000 deaths from COVID-19 will be registered. This prediction was made by the Imperial College in London in March, shocking Britain and forcing Prime Minister Boris Johnson to impose quarantine measures in the country. In the U.S., predictions were even more frightening than about 2 million deaths unless the government went into effect.

These estimates are based on the SIR model, which analyzes the number of people infected with infectious diseases in a specific population within a time frame. The model divides the population into three categories - endangered, infected, and healed. Scientists then add another formula that determines the speed of spread from one category to another. This helps to predict the number of those who will get infected and those who die:

“I hope people understand that this is a simplified version of what happens in reality. If we want to analyze the situation in more detail, consider, for example, how many of the infected will need to be admitted to the hospital to predict how much the health system will be burdened, says epidemiologist Helen Yankees.

The factor that changes in these models is quarantine measures. Scientists use the data for new cases of infection before and after these measures to produce accurate predictions.

Models help identify so-called reproduction figures a factor that calculates the number of people infected by a patient who is sick during the transition period. If there's more than one infected, the reproduction rate multiplys.

When we started calculating in February or March, most of the models that were created were used for previous epidemics of forecasts. We decided to use the coronary epidemic as a base,” says Massachusetts Institute for Technology professor George Barbastathis.

The main challenge of the model was lack of data on the actual number of coronary infected people, especially since many of the infected do not show symptoms. In response to the problem, scientists at Washington University developed a pattern based on the number of patients who died from the disease.

We knew we didn't have enough tests to diagnose, and we didn't know how many people were infected. In order to make valid predictions, we decided to create calculations after the death toll in a locality or a place to reach at least 50 people.

We thought this methodology offered more stability in predictions. We were calculating growth rates or decline,” says Ali Mokdad with the University of Washington.

This model remained the most optimistic until the death toll reported significant increases in early May. Today scientists predict 140 thousand dead by the beginning of August, or twice as many as the initial predictions. This increase is seen as a result of the easing of isolation measures.

When we started to see an increase in population activity, two weeks later an increase in the spread of the virus was observed. The impact immediately begins on increasing the pace of death,” says Mokdad.

Tens of different models use mathematical algorithms, but all models have a common one: they warn not to remove quarantine measures faster than it is recommended.