On Monday, Imperial College London’s COVID-19 response team released a report detailing its efforts to create a model of how different monitoring efforts could change the trajectory of the coronavirus pandemic in the US and UK. The report is published by a World Health Organization publication, but it was provided to the governments of these countries, where it is said to influence their responses.
The report paints a grim picture of millions of deaths in the two countries if nothing is done, as infections significantly exceed the capacity of their hospital systems to treat patients. But it does believe that aggressive measures could contain the virus until an effective vaccine can be developed, although that would mean shutting down many aspects of society for more than a year.
Before this causes panic, however, it is important to emphasize that these results are based on a model that, due to the incomplete information we currently possess, is imperfect and must rely on a number of assumptions. We’ll run through all of these limitations before going into the conclusions the Imperial College team drew from the model.
Epidemiology is the study of diseases at the population level. One of the tools in this area is epidemiological models, which try to represent how infectious diseases can spread within a population. So to understand this study, you need to understand how the model handles virtual infections.
In this research, the model goes back to the level of individuals. They are distributed in a country – in this case the UK or the US – based on high-resolution population data. Contacts with other individuals, enabling the spread of an infectious agent, are established in a number of locations: within a household, at school, at work, and in social situations such as shopping. The researchers used census data to get typical household sizes, as well as the age distribution of local populations. Mean class size data was used to handle school contacts, while workplace size and mean commute distances were used to model workplace contacts.
Overall, this is an incredibly detailed model and has already been used for peer-reviewed publications.
With the population in place, you need to add the coronavirus. Here the researchers are dealing with imperfect information, as all of our data on the virus is necessarily incomplete. So while we know that each infected individual tends to infect about two others, we don’t have a precise measure of its infectivity, and even a small difference in this value will be amplified when expanded to a national level.
While the authors do their best to base their assumptions here on real-world data, any values describing coronavirus behavior are uncertain. They used an incubation period of five days and assumed that the transmission starts 12 hours before the onset of symptoms. They also assumed things like the percentage of hospitalized cases that would require intensive care facilities (30 percent), the overall death rate (half of those in intensive care, biased toward the elderly), etc. The values they used reproduce the limited information we have so far about the spread of the pandemic, but the erratic testing in most countries almost certainly means some of them are not going through.
For both the US and UK, the authors modeled infectious individuals entering the population in early January, with the number of new arrivals doubling every five days as the pandemic spread worldwide. The total number of new entrants was calibrated using mortality data from the two countries.
With a working model of the pandemic in the UK and US, the researchers were able to try out different scenarios, from doing nothing to an increasingly draconian combination of restrictions on population movements. These can be roughly divided into two categories. One is mitigation, an effort to slow the spread so that the country’s health care system is not overwhelmed before previously infected people begin to provide herd immunity. And the second is suppression, which uses various forms of isolation to try to slow the transmission rate to the point where each infected person has less than an equal chance of passing it on.
However, there is no hard boundary between these categories. Closing schools would be a mitigating tactic in its own right, but could be part of a package of measures being used to quell the virus. Taken together, the researchers are considering combinations of the following methods:
- Case isolation: infected persons stay at home for a week.
- Home quarantine: all family members of an infected person stay at home for two weeks.
- Social distancing of the elderly: the over-70s reduce social contacts (assuming 75 percent compliance).
- Social distancing from everyone: everyone except students reduces social contact (assuming 75 percent compliance).
- School closure: 75 percent of schools are closed.
Some of these have mixed effects. For example, social distancing efforts are believed to increase the number of transmission opportunities within a household. And closing schools would increase contact between people in a neighborhood, because children might play together more. This isn’t entirely realistic — social distancing compliance would of course improve as news reports of the hospital system’s collapse roll in — but they aren’t unreasonable numbers for a model.
So what’s happening?
Bad things, if nothing is done. In the UK, the epidemic peaks in mid-May, reaching levels well beyond what the country’s critical care infrastructure can handle, resulting in about 22 deaths per day for every 100,000 people. That equates to about 14,000 people dying a day for a few weeks, with a total of half a million dead. In the United States, a younger and more diffuse population means that the peak comes a little later and does not rise as high per capita, but the larger population means that more than 2 million people will die.
Note that this does not include the additional deaths that will occur from people requiring critical care for reasons other than the coronavirus.
When looking at intensive care facilities, most mitigation strategies don’t work particularly well. While they all flatten the curve to varying degrees, closing schools or isolating infected people still leaves hundreds of thousands of people in need of hospitalization without anywhere to go. Even a combination of measures (case isolation, home quarantine and social distancing of the elderly) completely floods the medical system.
To have a real impact in the UK or US, general social isolation is needed. Combined with domestic quarantine and case isolation, the need for intensive care hospital beds will be exceeded for several months, but the exceedance will be only minor and may be manageable with emergency measures. The combination of school closures, case isolation and general social distancing keeps peak infections low enough to ensure intensive care capacity is not exceeded, and the problems will gradually ease after the April peak.
The problem with this is that the virus continues to circulate at a low level among the population in this scenario. The authoring model releases the restrictions in September, and in October a spike appears that rivals that caused by doing nothing to begin with. (Although less effective forms of suppression have a lower peak because so many people became infected in the intervening months.)
On their own, the results would suggest that the best way to minimize the death rate is to maintain social distancing indefinitely, until a vaccine is developed that provides adequate protection for the herd’s immunity.
The biggest test of this model is that some countries that have been aggressive on testing and suppression efforts have successfully contained the spread of the coronavirus and are beginning to lift restrictions. If the authors’ model is correct, countries like China and South Korea are at risk of seeing a resurgence in cases.
But the researchers also see the possibility that the initial suppression strategy will give these countries a long-term edge. With increased vigilance and a well-developed testing and isolation infrastructure, it should be possible for them to identify newly arrived cases and situations where the virus continues to circulate in the background. And through contact tracing and aggressive testing, it should become possible to control the coronavirus, a process South Korea is now beginning to try.
While the US and UK have missed the boat on preventing the pandemic from settling in their countries, they may not need to take aggressive suppression measures until a vaccine is developed. If successful, South Korean efforts could provide a model for easing the constraints required for oppression. And add another wrinkle to all epidemiological models of this coronavirus outbreak.
Correction: Clarification of the derivation of one of the percentages.