Earlier this year, when the coronavirus pandemic hit the United States, some mockers predicted scientific models of thousands of deaths.
Those models were unfortunately proved to be correct. They also provided a new warning that the recent rise in cases could mean that the number of deaths in the United States could almost double in the next four months.
Coronavirus modeler Alessandro Vespignani, director of the Institute of Network Science at Northeastern University, said: “If we go back to March, then we are saying that if we don’t deal with this problem carefully, we might Will cause 200,000 or 300,000 deaths.”
After defeating the initial wave of coronavirus infections, some European countries found themselves in a familiar situation: facing a surge in new cases and weighing which restrictions might help keep these numbers down. In the United States, after a brief decline at the beginning of this month, the number of new cases per day climbed again. According to NBC News statistics, since September 18, the average number of new Covid-19 cases in the country for 7 days has not been less than 40,000 per day.
For coronavirus modelers, this book is a foregone conclusion. Many people watch with horror and frustration as their predictions of the evolution of the pandemic and its potential death toll have come true.
Now, a widely cited model developed by the Institute of Health Measurement and Evaluation at the University of Washington suggests that by January, there may be a total of 378,000 deaths from the coronavirus in the United States.
But infectious disease modeling can be a tricky science, and it is vulnerable to criticism due to its uncertainty. Experts say that since the beginning of the pandemic, the coronavirus model has come a long way, so much so that some researchers are moving away from long-term predictions and focusing on more accurate predictions of Covid-19 in the next 6 weeks Forecast of internal trends. future.
Dr. Christopher Murray, director of the Institute and Professor of Health Index Sciences at the University of Washington, said that his team’s model has been improved a lot throughout the pandemic. Behavioral changes (for example, wearing a mask seriously) may lower their forecast for January, but he also fears that fatigue will gradually subside.
Murray added that this new trajectory has been seen in some European countries including Spain, France and the United Kingdom.
This is why modellers hope that people in the next few weeks will heed the warning that complacency and behavior changes related to the autumn and winter seasons may lead to a new wave of infections.
He said: “I think some people think that the worst has passed.” “The gradual decline in vigilance will promote the return of autumn and winter.”
The model of the University of Washington Institute is one of several models used by the Centers for Disease Control and Prevention to track pandemics. It has been criticized because it often contains a high degree of uncertainty and may lead to inaccurate predictions. In the early days, the model underestimated the number of deaths from Covid-19 nationwide, and predicted that by the end of August, there could be 60,415 deaths in the United States.
Nevertheless, the model will still be updated frequently, and will become more complete with the number of cases, hospitalizations and many other factors. The institute’s model estimates that by June, the number of deaths in the United States may reach 200,000 by October 1. This prediction will eventually be accurate within two weeks.
But infectious disease models will never be static, and there are many unknown factors that may greatly change existing predictions.
One of the factors is how seasonal changes affect the spread of the virus. There is no conclusive evidence that the coronavirus will spread more or less in autumn and winter. On the contrary, the impact of temperature drops on human behavior has attracted the attention of researchers, especially because cold weather may attract people indoors and make it difficult for people to socially alienate.
Associate researcher Sen Pei of Columbia University said: “In winter, people tend to stay indoors, which may be more likely to spread disease.” He completed a lot of modeling work for Covid-19. “But we still don’t know how the virus behaves in winter.”
Bei Ming said that the modeling of the new coronavirus faces huge challenges, but after nine months of data on the pandemic, his team’s predictions have become more complicated. However, one of the most difficult things to predict in the model is also one of the most important factors that may change the outcome of an outbreak: how humans respond to this situation.
Bei Ming said: “This is an unstable situation because people’s behavior will change over time, which is inherently unpredictable.”
This uncertainty is partly why Pei and other modelers avoid making long-term forecasts like the research model does, and instead focus on formulating short-term prospects for the next four to six weeks.
“No one really knows what will happen in the next few weeks.” said Gu Youyang, a data scientist who runs the coronavirus model (called Covid-19 Projections). Gu has no background in epidemiology or infectious disease modeling, so he designed a model that uses machine learning to “study” certain parameters that develop with the pandemic, such as the number of virus reproduction or R-naught, which represents The infectiousness of the disease.
“We don’t rely on any implicit assumptions,” Gu said. “We looked at the data and said: This is what we learned from what is happening.”
Gu said that his model only makes predictions, and it will not be able to predict until November. The surge in new cases in June and July will not lead to a surge in deaths, the same as the situation in March and April in the country.
Gu said: “We compared what happened in the United States with the situation in other parts of the world. These data do not support the rising rate of deaths.” “We eventually reached a peak of about 1,000 deaths per day, which is obviously still very important, but Less than many in the scientific community expected.”
Staying away from long-term forecasts is the common desire of other modelers, who believe that long-term forecasts are usually less accurate because they need to include various estimates to resolve uncertainties. For example, between now and January, travel bans, lockouts or other restrictions may be imposed, which will greatly change long-term forecasts.
Shaun Truelove, an assistant scientist and modeling expert at the Bloomberg School of Public Health at Johns Hopkins University, said: “This shift” allows us to move away from our initial project predictions and get closer to the goal of prediction. “Prediction is more based on what actually happened, rather than what might happen. “
Vespignani compares it to a weather forecast, which is more difficult to determine. He said he hopes that people will pay close attention to the coronavirus forecast, especially as the country prepares for an increase in new cases in the coming weeks and months.
He said: “We still have a long way to go.” “We must fight because the battle is not over yet.”