Amid all the brouhaha over COVID-19 – the biggest respiratory virus threat globally since the 1918 influenza pandemic – confusion reigns over exactly what epidemiological models of the disease are predicting. That’s important as the world begins restricting everyday activities and effectively shutting down national economies, based on model predictions.

In this and subsequent blog posts, I’ll examine some of the models being used to simulate the spread of COVID-19 within a population. As readers will know, I’ve commented at length in this blog on the shortcomings of computer climate models and their failure to accurately predict the magnitude of global warming. 

Epidemiological models, however, are far simpler than climate models and involve far fewer assumptions. The propagation of disease from person to person is much better understood than the vagaries of global climate. A well-designed disease model can help predict the likely course of an epidemic, and can be used to evaluate the most realistic strategies for containing it.

Following the initial coronavirus episode that began in Wuhan, China, various attempts have been made to model the outbreak. One of the most comprehensive studies is a report published last week, by a research team at Imperial College in London, that models the effect of mitigation and suppression control measures on the pandemic spreading in the UK and U.S.

Mitigation focuses on slowing the insidious spread of COVID-19, by taking steps such as requiring home quarantine of infected individuals and their families, and imposing social distancing of the elderly; suppression aims to stop the epidemic in its tracks, by adding more drastic measures such as social distancing of everyone and the closing of nonessential businesses and schools. Both tactics are currently being used not only in the UK and U.S., but also in many other countries – especially in Italy, hit hard by the epidemic.

The model results for the UK are illustrated in the figure below, which shows how the different strategies are expected to affect demand for critical care beds in UK hospitals over the next few months. You can see the much-cited “flattening of the curve,” referring to the bell-shaped curve that portrays the peaking of critical care cases, and related deaths, as the disease progresses. The Imperial College model assumes that 50% of those in critical care will die, based on expert clinical opinion. In the U.S., the epidemic is predicted to be more widespread than in the UK and to peak slightly later.

What set alarm bells ringing was the model’s conclusion that, without any intervention at all, approximately 0.5 million people would die from COVID-19 in the UK and 2.2 million in the more populous U.S. But these numbers could be halved (to 250,000 and 1.1-1.2 million deaths, respectively) if all the proposed mitigation and suppression measures are put into effect, say the researchers.

Nevertheless, the question then arises of how long such interventions can or should be maintained. The blue shading in the figure above shows the 3-month period during which the interventions are assumed to be enforced. But because there is no cure for the disease at present, it’s possible that a second wave of infection will occur once interventions are lifted. This is depicted in the next figure, assuming a somewhat longer 5-month period of initial intervention.

The advantage of such a delayed peaking of the disease’s impact would be a lessening of pressure on an overloaded healthcare system, allowing more time to build up necessary supplies of equipment and reducing critical care demand – in turn reducing overall mortality. In addition, stretching out the timeline for a sufficiently long time could help bolster herd immunity. Herd immunity from an infectious disease results when enough people become immune to the disease through either recovery or vaccination, both of which reduce disease transmission. A vaccine, however, probably won’t be available until 2021, even with the currently accelerated pace of development.

Whether the assumptions behind the Imperial College model are accurate is an issue we’ll look at in a later post. The model is highly granular, reaching down to the level of the individual and based on high-resolution population data, including census data, data from school districts, and data on the distribution of workplace size and commuting distance. Contacts between people are examined within a household, at school, at work and in social settings.

The dilemma posed by the model’s predictions is obvious. It’s necessary to balance minimizing the death rate from COVID-19 with the social and economic disruption caused by the various interventions, and with the likely period over which the interventions can be maintained.

Next: Coronavirus Epidemiological Models: (2) How Completely Different the Models Can Be