By Siân Jenkins & Martyn Bull
The spread of coronavirus (COVID-19) is causing huge alarm, but there is clear evidence that it can be slowed and even stopped by effective public health measures, informed by robust disease modelling.
In South Korea, this has been demonstrated most effectively. Central to the country’s success has been a wide reaching testing programme, supplemented by swift moves to quarantine the affected and isolate their close contacts.
In the UK, confusion about the realities of the virus has remained. To address this, the Prime Minister introduced new measures on Monday to enforce social distancing, including police powers to fine and disperse the noncompliant.
However, the UK’s lockdown and broader coronavirus mitigation strategy still relies significantly on public consent. For example, shopping infrequently and leaving the house only once a day to exercise.
So, if the disease is to have the minimum possible impact on society, key stakeholders and the population must have a clear understanding of the strategy being adopted to control the disease. They must also be clear about its intended effect and the roles they have to play.
Epidemic models are key to understanding how fast a disease is spreading, what is likely to happen in the future and how this can be influenced by control measures.
What is an SEIR model?
An SEIR model is a predictive model of disease dynamics that has been in existence for many years, and still forms a core tool of epidemiology.
An SEIR compartmental model represents the flow of people between different disease stages:
- Susceptible are people that are vulnerable to infection.
- Exposed are people who have come into contact with an infection but aren’t infectious.
- Infected are people who are infectious following exposure.
- Recovered are people who have recovered from the disease.
- Dead are people who have died from the disease.
The rate at which the susceptible people become exposed is directly related to:
- the number of susceptible people;
- the number of infected people;
- the probability of catching the virus when you come into contact with it.
These parameters all feed into the basic reproductive number for a disease, denoted R0. This represents the average number of newly exposed per infected individual.
- R0<1: Any outbreak is expected to die out naturally.
- R0>1: The outbreak is expected to increase.
The parameters described are factors we can control through public health intervention, acting to reduce R0.
This type of model is currently being used by global stakeholders as a basis for coronavirus mitigation. By modelling coronavirus and its mitigation strategies, stakeholders have a better understanding of disease spread. They also have the opportunity to strategically apply the science and achieve positive results.
How can disease modelling help to stop the spread of coronavirus?
The likely progression of the virus within the population can be explored through disease modelling. Augmenting the previous SEIR model with preventative measures allows scientists to explore their impact on the spread of the disease.
The aim of these measures is to disrupt the spread of the disease. This can be achieved by either reducing the population of different disease states or by altering the rate that individuals transition between them.
The SEIR model above represents the incorporation of multiple public health intervention strategies. The aim is to reduce the number becoming infected and as a consequence, reduce the number of disease-related deaths.
Social isolation reduces the number of susceptibles, reducing the spread of the disease. However, when people violate their isolation, they open themselves to infection again and re-enter the susceptible state, potentially leading to further spread of the disease.
Only those in the infected state, who are not quarantining, can spread the disease. Quarantine reduces the number that can infect the susceptible population, reducing the number of new infections. However, if individuals leave quarantine too early, they remain capable of infecting others.
There are many factors involved in the probability of becoming exposed once an individual comes into contact with the virus. These can include the infectiousness of the disease and whether the virus is still alive when encountered on a surface. By washing your hands more frequently and taking extra care when cleaning, this probability can be reduced.
Development of a vaccine is of great importance. It would remove people from the general susceptible population, both those socially isolating and those not, providing them with the immunity required. However, until this is developed, disobeying Government instructions runs the risk of infecting more susceptibles and significantly escalating the crisis.
Based on disease models such as the one above, it is possible to determine which public health measures minimise the spread of infection and to what extent they need to be applied. After reporting a peak of over 900 new cases daily in late February, South Korea is now reporting figures in the double digits, thanks to successful application of public health measures.
Can the UK emulate successful coronavirus mitigation strategies?
Effective disease modelling will allow experts to determine which public health strategies best fit the UK. For example, mitigation strategies that avoid placing unnecessary strain on the NHS.
The graph above shows the infected population, generated by a simple SEIR model, over time. The yellow curve shows the scale of infection when the basic reproductive number is 3.3. Here a large proportion of the population is infected simultaneously. The number of cases falls back towards zero once most people have caught the disease and acquired a degree of immunity, although this immunity may not last forever.
The other curves show the effect of reducing the basic reproductive number, which could be the result of introducing public health measures. The peak of the curves diminish as the basic reproductive number decreases. This is the effect that governments refer to as flattening the curve. It reduces the strain on the healthcare system by preventing an overwhelming number of people becoming seriously ill simultaneously.
Current public health measures implement a multi-pronged attack on the spread of the virus, aiming to reduce the parameters associated with its spread and so reduce the basic reproductive number of the virus. Without these, we risk prolonging the impact of coronavirus.
Effective disease models, like SEIRs, require accurate inputs, such as initial conditions and parameters. South Korea has been largely successful in preventing the spread of coronavirus because their testing programme gave them high quality data at scale, providing situational awareness and estimation of these inputs.
As a result of the large amount of data collected by countries such as South Korea and Singapore, parameters for an SEIR type model for coronavirus were well understood before the virus reached the UK. Scientists are able to utilise these parameters in pre-existing SEIR models that also incorporate UK specific population dynamics, to forecast the impact of the outbreak.
However, the current scale of the outbreak is required to make best use of these models. Boris Johnson has announced that more coronavirus tests will be conducted on UK patients in line with WHO guidelines. If the scale of testing increases significantly, this will provide the UK with the situational awareness required to make best use of disease modelling for planning public health strategies.
How can we manage the long-term impacts of coronavirus?
Approaches designed to mitigate coronavirus’s spread and impacts have varying degrees of success. A key contributing factor to this is that people don’t always behave as expected. This means that model results aren’t achieved in reality.
Regardless of this, it is likely to be many months, if not years, before this problem is solved anywhere. That means there is an onus on all key stakeholders to fulfil roles that prevent the pandemic from becoming uncontrollable.
Riskaware is conducting research in information management systems that will help improve the long-term management of coronavirus and its impacts. Modelling that assesses the effectiveness of disease interventions, including control measures and the uncertainty in model outputs will be crucial.
We are also working on better methods to estimate the parameters and current status of an epidemic from the data.
By achieving these goals, which will be assisted greatly by improved levels of testing, we aim to better arm governments and public health experts in their mission to keep the public safe now and in the future.
Get in touch to discuss our modelling solutions here